Workshop Report

Final Report of the Technical Requirements forImage-Guided Spine Procedures Workshop*

April 17–20, 1999, Ellicott City, Maryland, USA

Report Editor: Kevin Cleary, Ph.D., Imaging Science and Information Systems (ISIS) Center,Department of Radiology, Georgetown University Medical Center

Organizing Committee: James Anderson, Ph.D., Johns Hopkins Medical Institutions;Michael Brazaitis, M.D., Walter Reed Army Medical Center; Gilbert Devey, B.S., Georgetown

University Medical Center; Anthony DiGioia, M.D., Shadyside Hospital;Matthew Freedman, M.D., M.B.A, Georgetown University Medical Center;

Dietrich Gronemeyer, M.D., Witten/Herdecke University; Corinna Lathan, Ph.D., TheCatholic University of America; Heinz Lemke, Ph.D., Technical University of Berlin;

Don Long, M.D., Ph.D., Johns Hopkins Medical Institutions; Seong K. Mun, Ph.D., GeorgetownUniversity Medical Center; Russell Taylor, Ph.D., Johns Hopkins University

SECTION 1: WORKSHOP OVERVIEW

INTRODUCTION

The “Technical Requirements for Image-GuidedSpine Procedures” Workshop was held April 17–20, 1999, in Ellicott City, Maryland. The generalobjective of the Workshop was to determine thetechnical requirements for image-guided proce-dures in the spinal column, the spinal cord, and theparaspinal region. While the Workshop title indi-cated a focus on image-guided procedures, theWorkshop’s participants were encouraged to thinkmore broadly and to include computer-assisted androbotically assisted spine procedures in their re-view. The workshop was by invitation only, andapproximately 70 experts, about two-thirds of

whom were Ph.D.s and one-third M.D.s, partici-pated.

This document is organized as follows. Afterthis introductory section, Sections 2–7 contain thereports of the six working groups. Each workinggroup report consists of an overview and segmentsdevoted to clinical needs, technical requirements,and research priorities. Section 8 is a workshopsummary and Section 9 covers special sessions.There are three appendices: Appendix A is theworkshop program, Appendix B lists the workshopparticipants, and Appendix C is the report bibliog-raphy.

ACKNOWLEDGMENTS

A special debt of gratitude is due to the workshopsponsors:

* This report can also be found on the World Wide Web by starting at:

http://www.isis.georgetown.edu

and following the links to conferences and the Spine Workshop. Bound copies of the workshop report are available for $30to U.S. addresses and $40 to other countries due to higher shipping costs. The report is 119 pages and includes 13 pages ofcolor photos. Ordering instructions are on the web site, or the report editor can be contacted at: cleary@isis.imac.georgetown.edu

Computer Aided Surgery 5:180–215 (2000)

● National Science Foundation (BES-9804700)● U.S. Army Medical Research and Materiel

Command (DAMD17-99-1-9532)● The National Institutes of Health

(1R13CA81313-01)● The Whitaker Foundation● Picker International and DePuy Motech

AcroMed

This report is the product of many authors, and thecontents do not necessarily reflect the position orpolicy of any of the sponsors. We would also liketo thank Audrey Kinsella, M.S., for her editingexpertise and Barbara Hum, M.D., for her assis-tance in preparing the report.

COMMON THEMES AND RECOMMENDATIONS

From the Working Group reports, the following sixthemes have been identified:

1. Spinal disorders are a major public healthproblem and a potentially correctable sourceof disability. Surgical treatment, when indi-cated, produces variable outcomes that maybe improved by less-invasive, image-guidedprocedures.

2. Modeling, segmentation, and registrationare fundamental technical tools that still re-quire major advances to be more clinicallyuseful. These technical problems are impor-tant to many areas in image guidance, notjust in the spine, but also in other clinicalspecialties. Validation is a significant issuehere as well. It is interesting to note that,while progress has been made in addressingthese problems and commercial systems in-corporating some of this technology haveappeared, many of the technical issues thatdominated the discussion at this workshopare the same as those mentioned in previousworkshops4-6 on computer-assisted surgery.

3. Improved image processing and displayis critical to advancing image-guided proce-dures of the spine and image-guided proce-dures in general. Several Working Groupscommented that real-time image acquisitionand display, in particular real-time three-dimensional (3D) rendering and fast, intra-operative, 3D imaging systems, would beextremely valuable in this respect.

4. There is a significant communication andknowledge gapbetween technical and clin-ical personnel. Each faction has its own

vocabulary and specialized knowledge.While more people are becoming conver-sant with both areas, how to best bridge thisgap and foster collaborative efforts is animportant issue for further advancement ofthe field.

5. Clinical outcomes studiesare important tohelp determine if these technological ad-vances improve patient outcomes. Eco-nomic issues also need to be considered.While it is acknowledged that outcomesstudies are difficult to design and carry out,they should be pursued and funding shouldbe made available for them.

6. Infrastructure issues, such as reimburse-ment, liability concerns, and conflicts be-tween specialties, may be as important astechnical issues in advancing the field.Therefore, these issues must be addressed inaddition to a focus on needed technical de-velopments.

From these six common themes, as well as othersmentioned in the Working Group Reports (Sections2–7) and the Summary Presentation (Section 8), thefollowing recommendations were made:

1. To hasten the development ofclinicallyuseful applications of modeling, segmen-tation, and registration, additional re-sources for research should be made avail-able in these areas. These resources shouldbe directed towards the development ofmedically relevant techniques, which mayalso require fundamental scientific ad-vances.

2. A common and open standard infrastruc-ture is needed for the next generation ofimage-guided operating rooms or interven-tional suites, to be used both for spine pro-cedures and for all procedures in general. Arequest for proposals (RFP) should be is-sued to define this open standard, identifycommon elements, suggest possible archi-tectures, and develop appropriate user inter-faces. As part of this effort, NIH and otherfederal agencies should encourage partner-ships between medical researchers and med-ical equipment designers and manufacturersto develop common elements for image-guided and minimally invasive surgery thatinclude a research interface.

3. Application testbedsare needed to ensureclinical relevance, identify potential pitfalls,

Workshop Report 181

and facilitate collaboration between techni-cal and clinical personnel. The developmentof these testbeds is not supported by thecurrent NIH R01 hypothesis-driven fundingmechanism. Other funding mechanisms,such as the phased innovation award mech-anism designed to encourage technology de-velopment and used in the recent request forapplications on prostate cancer from the Na-tional Cancer Institute, should be created tofund these testbeds.

4. Specific equipment and instrumentationneeds that are required to advance the fieldshould be supported. For example, high-fidelity haptic interfaces, modular systemsfor spinal work and fast 3D visualization,and robotic instrumentation for surgery andtherapy are prerequisites for advancing thefield.

5. Multidisciplinary training and educationshould be supported, including programsthat allow engineers and scientists to gainclinical knowledge and physicians to gaintechnical knowledge.

6. A follow-up spine workshop should beheld in two or three years to track progressand re-evaluate the state of the field.

WORKING GROUPS

The Workshop consisted of plenary sessions andWorking Group meetings. The plenary sessionswere aimed at providing background for both clin-ical and technical areas. The Working Group meet-ings were to focus on the specific technical areas.Each Working Group had a technical leader (Ph.D.)and a clinical leader (M.D.). These leaders and eachof the six groups’ participants are listed in theWorking Group reports.

The Working Groups were each charged withinvestigating a specific technology area. A generalsummary of the Working Groups’ emphases are asfollows:

Working Group 1: Operative Planning andSurgical Simulators. This group focused onpreoperative planning, which will be increas-ingly used to define the best approach to theanatomy of interest, simulate the results of asurgical intervention, and evaluate the conse-quences of different approaches. The group alsodiscussed surgical simulation for training andeducation as well as for preoperative planning.

Working Group 2: Intraprocedural Imagingand Endoscopy. This group discussed all of theimaging modalities that may be used duringprocedures, including the intraprocedural use ofCT, MR, ultrasound, and fluoroscopy. As intra-procedural imaging becomes more common, thequestion of identifying the modality most appro-priate for particular procedures will continue toarise. The trade-offs between cost, accuracy, andinformation provided were discussed. Thisgroup also considered the use of endoscopicimages in spine procedures, and the potential forfusing endoscopic images with the 3D imagingcapability of CT or MRI.

Working Group 3: Registration and Segmen-tation. This group focused on all aspects ofregistration including 3D/3D registration (suchas CT to MRI), 3D/2D registration (CT to fluo-roscopy), and registration for instrument track-ing. While there has been a great deal of workdone in registration and segmentation, the devel-opment of easy-to-use, robust, and automaticregistration and segmentation algorithms is stillan elusive goal.

Working Group 4: Anatomical and Physio-logical Modeling. This group discussed anatom-ical and physiological modeling as well as soft-tissue modeling, such as deformable models.The use of modeling in image-guided proce-dures is still in its infancy, and fundamentalissues concerning the creation, use, and valida-tion of models remain. Accurate and reliablemodels are key to advancing the state of the artin surgical simulation and operative planning,among other areas.

Working Group 5: Surgical Instrumentation,Tooling, and Robotics. This group consideredsurgical instrumentation, including cages andother devices for fusing the spine. Tooling in-cludes the special-purpose devices needed toaccess the spine through percutaneous or mini-mally invasive techniques. In the future, roboticsystems may be used to assist in these proce-dures. These robotic systems may include pas-sive, semi-active, and active systems.

Working Group 6: Systems Architecture, In-tegration, and User Interfaces. The role of thisgroup was to define the systems architecture forthe image-guided spine procedure systems of thefuture. For example, how should the varioustechnologies such as registration, tracking, and3D visualization be integrated into a system thatthe clinician can use? What is the appropriate

182 Workshop Report

user interface for such a system (3D mouse,heads-up display, touch screen, voice-operated,eye tracker, etc.)? This group also discussedvarious technologies that were not covered byother groups, including image-guided surgerysystems.

WORKSHOP RATIONALE

When we first starting planning for the Workshopin the fall of 1997, the question arose as to whyanother workshop was needed. The reason is sim-ple: workshops develop infrastructure and help laythe groundwork for the development of the field.For example, early workshops on image-guidedtherapies in 19915 and computer-assisted surgery in19936 helped set research directions for the field.The 1993 computer-assisted surgery workshop wasfollowed by an NSF-sponsored workshop on robot-ics and computer-assisted medical interventions in1996.4 At this point, research activity in the field isbeginning to increase noticeably, as evidenced byspecialty conferences and the appearance of dedi-cated journals. While these earlier workshops weregeneral and included all clinical areas, we are nowseeing the emergence of specialty workshops, suchas the spine workshop which is the subject of thisreport. Other specialty workshops include the pros-tate cancer workshop in June 1999§ and the severalimage-guided workshops convened by the NationalCancer Institute and the Office of Women’s Healthin the Spring of 1999.

IMAGE-GUIDED PROCEDURES

Image guidance has been used in one form oranother in various medical procedures since theintroduction of X-rays. Recently, however, therehas been a marked increase in interest in this field,which can be largely attributed to developments involumetric imaging and increased computer pow-er.7 Volumetric imaging includes computed tomog-raphy (CT), magnetic resonance imaging (MRI),and 3D ultrasound, which are capable of producinga 3D representation of the human body. The mem-ory capacity, processing capability, and relativelylow cost of present-day computers enable the rapidanalysis of these 3D data sets.

Image guidance, in the form of framelessstereotaxy, has been driven primarily by the neu-rosurgery community. Neurosurgery in the brainrequires precise navigation through an anatomi-cally complex and delicate organ. For neurosur-

gery, computer-assisted surgery (including imageguidance) is an enabling technology that allowsnew techniques to be employed.2 However, imageguidance as currently used in the spine is primarilya safety measure for preventing iatrogenic injuries.The Workshop organizers hope that this meetingwill be a first step in expanding the use of imageguidance in spine procedures by identifying therelevant clinical areas, defining the technical prob-lems, and proposing potential solutions.

SECTION 2: OPERATIVE PLANNINGAND SURGICAL SIMULATORSThe Report of Working Group 1

AUTHORS

Frank Tendick, Ph.D., University of California SanFrancisco (Technical Leader)

David Polly, M.D., Walter Reed Army Medical Cen-ter (Clinical Leader)

Daniel Blezek, Ph.D., Mayo ClinicJames Burgess, M.D., Inova Fairfax HospitalCraig Carignan, Ph.D., University of MarylandGerald Higgins, Ph.D., Ciemed TechnologiesCorinna Lathan, Ph.D., The Catholic University of

AmericaKarl Reinig, Ph.D., University of Colorado

OVERVIEW: DEFINITIONS AND STATE OF THE

ART

This Working Group explored the requirements forpre- and intraoperative planning and simulationtechnologies that can be used in image-guided sur-gery of the spine. The first step was to define theterms: simulator and planner. Using the broadestdefinitions possible was thought to be important soas to avoid biased preconceptions toward the valueof certain technologies currently available andknown to our participants.

SIMULATORS AND PLANNERS: DEFINITIONS

A Simulator is defined here as an interactive vir-tual environment used to improve human perfor-mance. Note that this definition does not requirethat a simulator be computer based. A simulator isvirtual in the sense that it behaves in some ways

§ http://www.amainc.com/admetech/admetech.html

Workshop Report 183

equivalently to the real patient, but is not the realpatient. Thus “sawbones,” or plastic models ofanatomy, would qualify as simulators by our defi-nition. In addition, the simulator must permit inter-action, because interactivity is necessary for learn-ing and practicing perceptual motor skills. We alsodo not confine the role of simulators only to train-ing, because they could also be used for planning.

A Planner uses tools, including simulation,to improve human performance on the patient-spe-cific task at hand. There is overlap between simu-lation and planning, but neither is inherently asubset of the other. The essence of a planner is toprovide assistance in performing a procedure on aspecific patient.

STATE OF THE ART: ADVANCES IN SURGICAL

SIMULATORS TO DATE

Clinically, the state of the art in simulators fortraining is still at a primitive stage, and uses cadav-ers and sawbones for teaching purposes. The stateof the art in planning takes several forms, includingmulti-modality radiological studies and interven-tional magnetic resonance imaging (MRI). Com-mercial image-guidance systems are available, buttheir utility in the spine is limited. Currently, thesesystems are used for pedicle screw sizing and basictrajectory planning.

The state of the art in computer simulationhas advanced to a level of prototype partial tasktraining. Specifically, only parts of research orcommercially-oriented procedures can be dem-onstrated via computer simulation. For example,3D computer graphics workstations and PCs per-mit surface models of moderate complexity, onthe order of tens of thousands of polygons, atinteractive update rates of 15 Hz or more. An

example of 3D visualization is shown in Figure2-1. Haptic interfaces with force feedback arecommercially available. Most operate using threedegrees of freedom (DOF), although six-DOFdevices have recently been introduced, as shownin Figure 2-2. The physical modeling methodsused to simulate tissue behavior and generateforces for the haptic display are still primitive,however. Methods in the literature are typicallybased on mass-spring-damper meshes, linearelastic finite elements, or a variety of non-phys-ically based, ad hoc methods.

CLINICAL NEEDS: COMMON TASKS AND

PROCEDURE-SPECIFIC NEEDS

Our Working Group separated image-guided spineprocedures into two tasks that were common to allmajor procedures and others that were procedure-specific.

COMMON TASKS AND NEEDS

IN IMAGE-GUIDED SURGERY

The first major common task of image-guided spineprocedures is to identify the optimal trajectory forthe procedure, which is a function of the anatomicallevel of the spine on which the procedure willfocus. This task includes defining the starting andgoal points of the intervention and identifying aworking corridor that provides adequate accesswhile minimizing the risk of damage to fragiletissues. The planner must provide the clinician withthe ability to identify structures, and then to deter-mine relationships on the global scale from whichto plan a surgical approach and, on the fine scale, toverify adequate clearance between structures. Im-

Figure 2-1: Three-dimensional visualization for surgicalsimulation (Courtesy of Daniel Blezek, Ph.D., and RichardRobb, Ph.D., Mayo Clinic).

Figure 2-2: Six-degree-of-freedom force feedback de-vice (Courtesy of SensAble Technologies).

184 Workshop Report

ages are obtained well before the operation, whichallows for a planning time-frame of anywhere fromhours to weeks of off-line computations such assegmentation. However, intraoperative planningshould provide sufficient speed and interactivity forperformance of these tasks to be accomplished rap-idly within a few minutes.

The second common task involves obtain-ing adequate perception of the anatomy of thespine. This means providing appropriate infor-mation to the clinician to support clinical deci-sion-making in real time during the procedure.Two major elements are required to completethis task. The first is tissue discrimination, oridentifying the type and characteristics of tissuesso that damage to fragile structures can beavoided. Intraoperative imaging and/or registra-tion with previously obtained and segmentedimages may provide this needed informa-tion. However, it is important in this instance tohave precise information on the position of rel-evant structures which are relative to the currentlocation. The second element focuses primarilyon “location”, or knowing where one is relativeto the desired trajectory. This assessment re-quires obtaining global as well as local informa-tion.

PROCEDURE-SPECIFIC TASKS AND NEEDS IN

IMAGE-GUIDED SURGERY

This Working Group also identified procedure-spe-cific needs for decompression, stabilization, anddeformity correction procedures. The major need indecompression is to provide sufficient soft-tissueresolution to enable the surgeon to remove theminimum amount necessary while avoiding neuraldamage. In stabilization, better models and plan-ning are needed to enhance the placement of in-strumentation to achieve the optimal biomechanicalperformance of implants. Biomechanical modelsalso need to be developed and integrated into plan-ners to aid in deformity correction procedures.These models would be useful for analysis andprediction of the response of the tissue and implantto the procedure, including loads, deformation, andfatigue.

TECHNICAL REQUIREMENTS FOR PLANNING

AND SIMULATION TOOLS

This Working Group organized the technical re-quirements based on the clinical tasks described inthe previous section. These requirements relate totrajectory planning needs, interactive simulationduring spinal surgery, and increased need for hu-

man factors research into the efficacy of tasks com-pleted during image-guided surgeries.

PREOPERATIVE REQUIREMENTS FOR PLANNING

To plan a trajectory necessary for a procedure,accuracy of about 1 mm is needed to distinguishimportant structures. The level of resolution in asimulation would ideally be about 10 times higher(or 0.1 mm) to achieve sufficient fidelity in geo-metrical and physical models. It would thus bedesirable to have high-resolution data sets availablefor incorporation into simulators for later use inintraoperative planning.

The ability to segment soft tissues to distin-guish bone from nerve structures and vessels iscritical for both planning and simulation. Preoper-atively, there is time to run segmentation off-lineover a period of several hours up to weeks, but tobe of use intraoperatively, the period must be lessthan about 5 minutes. Achieving these precisionand time requirements is a high priority for en-abling effective trajectory planning. Intelligent as-sistants to aid the clinician in planning, as well asintelligent tutors for simulation, will be useful, butdepend first on the achievement of the resolutionquality- and time-related priorities described above,and so are not as critically immediate an issue.

REQUIREMENTS FOR INTRAOPERATIVE

SIMULATION

Issues Related to Perception and Visualization.There are additional technical requirements forachieving adequate anatomical perception during aprocedure. The clinician must be able to distinguishtissue type and determine its current location withinthe anatomy. Three-dimensional image-to-patient/instrument registration, discussed in greater detailby Working Group 3, is very important in thisregard. The information provided by preoperativeand intraoperative imaging modalities must be in-tegrated in an interactive manner that allows theclinician to readily alter viewpoints and edit plans.There is a great need for integrated modeling aswell, so that the surgeon can predict the effect oftreatment. This development needs to include me-chanical models and supporting data to predict theeffects of

● instrumentation in deformity correction,● treatments on the courses of nerves and the

resulting strain, and● the interaction between bone and implants.

Workshop Report 185

In simulation, haptic interfaces are still rela-tively in their infancy. There is a need for high-fidelity, six-degree-of-freedom devices with forcefeedback and physical models of instrument-tissueinteraction to simulate full contact. Relevant dataon soft-tissue viscoelastic properties must be ob-tained to support these models.

Issues Related to Cognitive and Human Fac-tors. A subtle but important aspect of image-guided surgery is the manner in which surgicalinformation is displayed and how the clinicianinteracts with these data. From the informationprovided to the surgeon, he or she must constructa 3D mental model of anatomical space and thenuse this model to plan a procedure. Completingthe process effectively can be challenging. Evenwhen a 3D data set is available, every viewpointof the data provides different information. Theclinician must therefore integrate multiple viewsto perform a complex and often highly preciseaction, such as the placement of pedicle screws.Enabling effective processing of, and interactionwith, anatomical data received by surgeons dur-ing image-guided surgery is an area requiringmuch further study.

Although a fair amount is known about howpeople construct and represent knowledge of spa-tial information, there is still little known about theintegration of spatial skills in solving complexproblems. Human factors experts should study therole of spatial cognition in surgery, driven by taskanalyses, to determine the optimal means of pre-senting information if image-guided surgery is tomeet its potential.

Need for Predictive Biomechanical Models. Inaddition to the general need for biomechanicalmodeling discussed above, there are specifictechnical development needs for effectively com-pleting/improving specific procedures which useimage-guided surgery. For example, the biome-chanical effect of instrumentation in stabilizationprocedures for instability, deformation, and frac-tures is still poorly understood. Development ofmodels of the interaction between the disk andnerves would improve the performance of diskherniation and disk/nerve root decompressionprocedures. Initially, gathering and collating em-pirical data from the experience of multiple clin-ical groups may help to predict the response ofanatomical structures within the spine to loads on

typical instrumentation configurations. Fully pre-dictive biomechanical models would eventuallyaid in the placement of instrumentation. Finally,intraoperative imaging should be capable of dis-tinguishing changes in soft tissue for the purposeof gauging the progress of a procedure. An ex-ample is checking the adequacy of tumor resec-tion to ensure that all of the tumor has beenremoved.

RESEARCH PRIORITIES

Table 2-1 summarizes and prioritizes the technicalrequirements needed for effective planning andsimulation of image-guided surgery of the spinedescribed in the previous sections.

Table 2-1. Research Priorities for Planningand Simulation Needs for Image-GuidedSpinal SurgeryHigh Priority• (*) Task analysis and cognitive modeling of human

performance by human factors experts, with specialemphasis on the role of spatial cognition in image-guided surgical spine procedures.

• (*) Development of high-fidelity haptic interfaces whichcan simulate anatomical models of varying complexity.

• (*) Development of visualization and interactionalgorithms and modes to allow the clinician to alterviewpoints and interactively plan the procedure.

• Imaging tools with accuracy of 1 mm resolution fordiscrimination of structures needed for planningpurposes.

• Image data sets with 0.1 mm resolution for simulationpurposes.

• Segmentation algorithms for distinguishing bone fromneural structures and vessels.

• Dynamic registration methods for intraoperativeplanning.

• Biomechanical models and data for deformitycorrection, effects on nerve location and strain, andbone-implant interaction.

Medium Priority• (*) Prediction of spine and instrumentation response to

loads, based on an empirical data library.• (*) Development of automated aids for corrective

instrumentation placement.Lower Priority• (*) Intelligent assistance for planning needs.• (*) Intelligent tutoring for simulation purposes.• Biomechanical models for predicting disk-nerve

interaction.(*) Asterisk identifies research priorities of special importance to planningand simulation. Other priorities are shared with one or more of the otherWorking Groups.

186 Workshop Report

SECTION 3: INTRAPROCEDURALIMAGING AND ENDOSCOPYThe Report of Working Group 2*

AUTHORS

Jeffrey L. Duerk, Ph.D., Case Western Reserve Uni-versity (Technical Leader)

Dietrich Gronemeyer, M.D., Witten/Herdecke Uni-versity (Clinical Leader)

Benedicte Bascle, Ph.D., Siemens Corporate ResearchLaurence Clarke, Ph.D., NIH Office of Imaging

TechnologyMartin Deli, B.S., Witten/Herdecke UniversityGilbert Devey, B.S., Georgetown University Medical

CenterWilliam Herman, B.S., Food and Drug AdministrationBarbara Hum, M.D., Georgetown University Medical

CenterYongmin Kim, Ph.D., University of WashingtonArthur Rosenbaum, M.D.Vance Watson, M.D., Georgetown University Med-

ical CenterS. James Zinreich, M.D., Johns Hopkins Medical

Institutions

OVERVIEW

Five imaging modalities that can be used to guidesurgical and interventional procedures in the spineare discussed: computed tomography (CT), mag-netic resonance imaging (MRI), X-ray fluoroscopy,ultrasound, and endoscopy. Each of these tools isdescribed in terms of capabilities for real-time cap-ture and display as well as in terms of relative costs.

CLINICAL NEEDS

A range of clinical/pathological conditions that arejudged by this Working Group to be amenable toimage-guided procedures of the spine (such as pro-cedures of the ilio-sacral joint and vertebral frac-tures) are identified and ranked in importance, interms of immediate impact and numbers of pa-tients. Types and adequacy of imaging modalitiescurrently used for spinal interventions are noted inthree tables [not reproduced here].

TECHNICAL REQUIREMENTS

No single imaging modality currently meets the clin-ical and technical needs for both diagnostic and ther-apeutic procedures of the spine. In this Working

Group’s Future System Requirements for Image-Guided Spine Procedures, five phases of develop-ment for requirements of a future, more inclusive,multi-applicable system for image-guided proceduresare identified and described. Areas focused on includepreoperative imaging needs, improved virtual naviga-tion, and verification of tissue status, among others.RESEARCH PRIORITIES

Two aspects of priorities defined by this WorkingGroup are:

1. Those related to technical development is-sues. The authors call for development ofopen and modular imaging systems, amongother design needs.

2. Those which call for changing the infra-structure that is currently in place amongthose working in the area of spine proce-dures. The authors focus on needed changesin tasks and roles, particularly in the trainingneeds of team members involved in image-guided procedures.

SECTION 4: REGISTRATION ANDSEGMENTATIONThe Report of Working Group 3

AUTHORS

Benjamin Kimia, Ph.D., Brown University (Techni-cal Leader)

Elizabeth Bullitt, M.D., University of North Carolina(Clinical Leader)

Lou Arata, Ph.D., Picker International & DePuy Mo-tech AcroMed

Gene Gregerson, M.S., Visualization Technology, Inc.Alan Liu, Ph.D., National Institutes of HealthYanxi Liu, Ph.D., Carnegie Mellon UniversityMurray Loew, Ph.D., George Washington UniversityNassir Navab, Ph.D., Siemens Corporate ResearchY. Raja Rampersaud, M.D., University of TorontoJoseph Wang, Ph.D., The Catholic University of

AmericaWilliam Wells, Ph.D., Harvard Medical School and

Brigham and Women’s HospitalTerry Yoo, Ph.D., National Library of MedicineJianchao Zeng, Ph.D., Georgetown University Med-

ical CenterQinfen Zheng, University of Maryland

* Editor’s note: This chapter has been accepted for publication in Academic Radiology, and therefore only an executivesummary is included here.

Workshop Report 187

OVERVIEW: THE USE OF IMAGING IN

REGISTRATION AND SEGMENTATION

Image-guided surgery is widely accepted as thestandard of practice for many intracranial proce-dures. For a number of reasons, some related totechnical difficulties, image-guided surgery is notyet generally used for spine procedures, eventhough such guidance could greatly enhance thecurrent standard of practice. The purpose of thischapter is to analyze the technical requirements thatare needed for manipulation of patient image datato help define criteria by which image-guided sur-gery can be used effectively for spine procedures.

Image-guided surgery may employ imagedata in any of three ways:

● First, preoperatively acquired images may beused for surgical planning, surgical simula-tion, or model creation. This procedure oftenextracts particular structures of interest fromthe image data.

● Second, images of the patient may be obtaineddirectly during an operation for the purpose ofhelping to guide the procedure. These intra-operatively acquired images are often of lowerquality or informational content than thoseobtained preoperatively, however.

● The third method of employing image datacombines intraoperatively acquired images oflower informational content with higher qual-ity, preoperatively obtained images. This pro-cedure of combining images requires that theintraoperatively acquired images be placedwithin the same coordinate system as the pre-operatively acquired images. Both sets of im-ages must also be placed within the patient’scoordinate system in the operating room, inaddition to the coordinates of the surgicalinstruments used. For imaging proceduressuch as these to be of clinical utility, theymust meet certain constraints of both time andaccuracy. Furthermore, for imaging proce-dures such as these toobtain clinical accep-tance, they must also undergo rigorous tests ofvalidity.

This Working Group’s report analyzes the techni-cal requirements forsegmentationand registra-tion of medical images as applied to the particularproblems associated with spine procedures.

SEGMENTATION AND REGISTRATION:DEFINITIONS

Segmentation is defined as the delineation andlabeling of image regions as distinct structures.

Segmentation is required to extract and define ob-jects of interest from image data for anatomic dif-ferentiation, to create models, and to implementsome forms of registration.

Registration is defined as the mapping of coordi-nates between any two spaces specifying volumet-ric images, the patient, or the instruments. Regis-tration is required to map one image to another, andto map any image to the patient.

CLINICAL NEEDS: ISSUES IN THE USE OF

IMAGE-GUIDED SPINE SURGERIES

Several of the approaches to surgery of the spinethat are discussed below are applicable to the cur-rent level of technology. However, there is also astrong need for the development of practical, clin-ically useful, intraoperative 3D imaging systems,which the authors believe to be feasible in the nextfive to ten years. A number of research issuesrequire close attention, as indicated below.

ISSUES OF ACCURACY AND SPEED

For the clinician, image guidance with an accuracyof 1–2 mm is required in order to avoid injuring thespinal cord while undertaking surgical procedures.Clinical requirements of registration speed varyaccording to the procedure performed. For someprocedures, such as pedicle screw placement, itmay be acceptable to wait 5 minutes until registra-tion is undertaken. For other procedures, such asthose performed under endoscopic guidance, regis-tration must be performed within 10–20 seconds toallow the procedure to continue smoothly. In gen-eral, for intraoperative procedures, because delay isdetrimental to the patient’s welfare, the upperbound on allowable delay for technical processingof image data depends on the perceived clinicalcontribution of the information. The delay time-frame is usually in the range of seconds to a fewminutes.

FOUR CATEGORIES OF SPINE PROCEDURES

We define four categories of spine procedures forwhich the use of image-guided surgery appearspromising in improving patient health outcomes.These categories are:

1. Instrumentation and percutaneous proce-dures

2. Resection of tumors and arteriovenous mal-formations (AVMs)

3. Treatment of spinal instability4. Treatment of disc disease

188 Workshop Report

Some instrumentation procedures, such as correc-tion of scoliosis, would benefit both from directimage guidance as well as from creation of a pre-operative model of the spine and tracking of inter-segmental motion for predicting tension upon thespinal cord. Other types of instrumentation, such aspedicle screw placement, would substantially ben-efit from direct image guidance.

1. Instrumentation and percutaneous pro-ceduresMany percutaneous and almost all instrumen-tation procedures are currently performed ei-ther in the CT scanner or (most commonly)under fluoroscopic guidance. X-rays of a typ-ical instrumented patient are shown in Figure4-1. A high-speed, image-guided method ofregistering the therapeutic instrument with thepatient and of accurately determining the in-strument’s trajectory with reduction of radia-tion exposure to the patient would be benefi-cial. Such advancements would affect largenumbers of patients.

2. Tumor resectionRemoval of the majority of spinal tumorsprobably does not require special imageguidance. However, image-guided surgerymay be important for the removal of somelarge tumors which have extended into thechest or pelvis, and it is likely to be impor-

tant to the treatment of almost all AVMs.For the latter, as well as for any highlyvascular tumor, segmentation and symbolicdescription of the blood supply to the lesionand to the normal spinal cord would addsignificantly to current therapeutic stan-dards, although only a relatively small num-ber of patients would be affected. A typicalspinal tumor is shown in Figure 4-2.

3. Treatment of spinal instabilitySpinal instability is a common problem.When instability occurs below the level ofC2, surgical intervention is almost alwaysrequired. For many patients, such as the oneshown in Figure 4-3, image-guided surgerywith registration of preoperative images tothe patient would be beneficial for the samereasons that image guidance of instrumen-tation procedures would be useful.

It should be noted, however, that somepatients with unstable spines may exhibitabrupt translations of spinal segments dur-ing operative positioning or even during theprocedure. Such pathological movement isdifficult to model and predict. High-speed,3D intraoperative imaging would providethe best method of managing such prob-lems.

Figure 4-1: Surgical instrumentation (Courtesy of Elizabeth Bullitt, M.D., University of North Carolina). The patient hasundergone both anterior and posterior cervical plating. Even minor errors in the angle of screw insertion can produce patientinjury.

Workshop Report 189

4. Disc removalStandard, open operative methods of discremoval do not require special image guid-ance. However, new methods of endoscopicor percutaneous disc removal do. It is notyet clear, however, that these new methodsare superior or equal to standard operativemethods. Segmentation of disc from scarand of scar from nerve root would be highlyvaluable during disc removal by anymethod, however, in order to reduce thechance of nerve root injury. Intraopera-tively, it is often difficult to find a discfragment under a layer of scar tissue that isadherent both to the disc fragment and to thenerve root. Precise knowledge of the loca-tions of both the disc fragment and of thenerve root would reduce the amount of ex-ploration required and the possibility ofnerve root injury.

TECHNICAL REQUIREMENTS: VALIDATION,REGISTRATION, AND SEGMENTATION

As in other medical imaging applications, the clin-ical needs involving image-guided surgeries of thespine, as described above, do not directly map to awell-defined engineering problem. The challengefor our Working Group was to address clinicalneeds in the context of well-formulated technicalproblems from which improvements would provideclinical benefits and identify realistic boundariesfor the amounts of accuracy and speed required ofsuch procedures. These technical problems includesegmentation, registration, and a component-wiseand overall system validation, as described belowin order of perceived priority.

VALIDATION

In the course of the Workshop, as well as duringthis Working Group’s meetings, serious concernswere expressed about the need to validate existingimaging systems. Estimates regarding the accuracywith which the surgical instrument can be placedwere varied. Furthermore, procedures for the over-all validation of the system or its components werenot unambiguously defined. Thus, research effortsto address validation issues in the spine are ofhighest priority. Specifically, the measurements of

Figure 4-2: Spinal tumor (Courtesy of Elizabeth Bullitt,M.D., University of North Carolina). Note the associatedmass of blood vessels similar to an arteriovenous malfor-mation (AVM). Some of these blood vessels also supply theconus of the spinal cord (arrow).

Figure 4-3: Spinal instability (Courtesy of ElizabethBullitt, M.D., University of North Carolina). This patienthas a typical thoracic compression fracture (arrow) withkyphosis and angulation of the spine. Surgery was required.

190 Workshop Report

accuracy/precision, robustness/stability, reliability/reproducibility, and, finally, clinical utility (e.g., asmeasured by time required, extent of user interac-tion, surgical value, etc.) are most significant.

Furthermore, there is a need for the compila-tion of a database which includes a variety ofmeasurements, including an absolute standard forvalidation assessment. From this database, existingand future registration algorithms can be comparedeffectively. The generation of this database is anitem of high priority, and it should consist of diag-nostic and therapeutic images with embedded fidu-cials as the “gold standard”.

REGISTRATION

Registration involves aligning distinct coordinatesystems that are available from volumetric, preop-erative images (typically CT, or CT and MRI),patients, instruments, and intraoperative images.Our Working Group classified registration tech-niques into three categories, as follows:

1. 2D image-3D image registration

2. 3D image-patient/instrument registration

3. 3D image-3D image registration.

From a clinical perspective, preoperative planningusually involves 3D image-3D image registration,while intraoperative procedures may rely on either2D image-3D image and 3D image-patient/instru-ment registration. If interventional MRI/CT isavailable, then intraoperative procedures can relyon 3D-3D imaging attained by registering intraop-erative lower-quality images with higher qualitypreoperative images.

Each of these registration techniques is de-scribed in our Working Group’s perceived order ofsignificance.

2D Image-3D Image Registration. A typical exam-ple of 2D image-3D image registration is the regis-tration of intraoperative fluoroscopic images (2D)with preoperative CT data sets (3D). Since 3D intra-operative imaging is not currently widely applicable(nor believed to be so in the near future), we view 2Dimage-3D image registration as a high-priority re-search area. In addition, for certain procedures, thedirect registration of preoperative fluoroscopic imageswith intraoperative MR images would alleviate therequirement of acquiring CT data.

Our Working Group identified three areaswhere technical improvements in this image regis-tration category were needed: accuracy, speed, andease of integration in a clinical protocol.

3D Image-Patient/Instrument Registration. The3D image-patient/instrument registration procedurebrings the coordinate system of the patient, as mea-sured by the instrument, in registration with thecoordinates of the patient in the 3D preoperativeimage. Currently, this procedure is accomplishedby “point pair” matching, in which anatomicallandmarks are selected interactively and the twocoordinates are registered and constrained by thematched landmarks. Since the “landmarks” (exam-ples being spinous processes and medial edge offacet) are not defined by pinpoint accuracy, butrather have finite extent, the accuracy of the regis-trations that are obtained via this method is limited.

Alternatively, some systems use measure-ments from the surface of the bone to generate a“cloud of points”, which are then registered tosurfaces extracted from preoperative images. Un-fortunately, the variations in the distribution ofgenerated clouds of points lead to inaccuracies inthe measurements. There is, however, the potentialto generate the uniformly distributed cloud ofpoints via laser, ultrasound, video, or video/stereosensor technologies to achieve better accuracy.

A second drawback of the systems based on“cloud of points” is that, because only the accessi-ble/visible portion of the bone is measured, smallinaccuracies in matching this portion to extractedsurfaces lead to large inaccuracies in the “blind” orinaccessible portion of the vertebrae. The clinicalimplications related to these substantial inaccura-cies are obvious.

Our Working Group suggested the use ofultrasound (US) as the modality having the greatestpotential to address this particular problem. Exam-ples might include placing US patches on the bellyof the patient, or even using the bone itself as theUS transmitter! It was also suggested that con-structing a surface model from the cloud of pointsfirst and then matching the two surfaces will incor-porate more of the geometrical structure in thematching process, thus constraining it and leadingto more accurate registrations.

3D Image-3D Image Registration. The 3D im-age-3D image registration is valuable when, inseveral of the types of spine procedures currentlyundertaken, both CT and MR images are acquired.The intraoperative use of MR images in conjunc-tion with fluoroscopic images requires that there bea preoperative registration of these two modalities.In addition, this registration process, when com-bined with fusion, leads to better presentation of thedata needed for preoperative planning. In the fu-

Workshop Report 191

ture, when interventional 3D imaging becomeswidely available, 3D image-3D image registrationwill be needed to augment the lower-quality intra-operative image with the higher-quality preopera-tive image(s).

Particular technical requirements of this pro-cess are assessed as follows:

● Requirements for speed are not as stringentfor preoperative registration as for intraoper-ative ones; however, delays need to be inminutes and not hours for practical reasons.

● The extent of user interaction should not ex-ceed more than a few minutes.

● Accuracy is, as with other registration types, asignificant concern.

SEGMENTATION

The generally shared view of this Working Groupwas that segmentation of preoperative and intraoper-ative images is typically required primarily as a meansfor providing surface-based registration methods.However, in several distinct areas, segmentation rep-resents an important “stand-alone” problem.

First, in some applications, anatomical struc-tures need to be differentiated, as in preoperativesurgical planning to correct an arteriovenous mal-formation (AVM) or in differentiating disc fromscar tissue. Second, segmentation is required forbuilding anatomical and physiological modelsneeded for biomechanical modeling, a topic ad-dressed by Working Group 4. These models wouldthen be used for simulation and training purposes.Third, segmentation is needed as a step towardsbuilding digital, or electronic, atlases of the spinewhich depict not only typical spinal anatomy, butalso its relative geometry and alignment, as well astypical variations in anatomy.

For spinal surgery, segmentation will be mostcommonly useful when applied to bony structures.Other structures are also significant in some cases,however. Examples include definition of the spinalcord during scoliosis surgery, vascular structuresduring resection of AVMs, and some tumors. Fig-ure 4-2 showed a complex case of an AVM andtumor involving the spinal cord. Segmentation ofthe various structures with definition of the bloodsupply of the cord and tumor would have been ofgreat help intraoperatively during this procedure.

In summary, segmentation seems likely to beuseful in the following clinical areas:

1. In definition of the boundaries of bony sur-faces in order to help guide instrumentationprocedures such as pedicle screw placementor scoliosis surgery. Accurate segmentationcombined with registration of the patient tothe preoperative CT scan could, in suchcases, prevent mis-insertion of a screw intoneural structures.

2. In definition of the vascular territories ofvessels feeding highly vascular tumors orAVMs. Knowledge of the structures sup-plied by an individual vessel could helpprevent interruption of an artery that, un-known to the surgeon, supplies the spinalcord as well as the lesion.

3. In the delineation of disc, scar, and neuraltissue in order to reduce the amount of ex-ploration required and the chance of tearingthe dura during “redo” disc operations.

4. In definition of structures used for both3D-3D and 2D-3D registration.

5. In the creation of biomechanical spinalmodels and atlases of spinal anatomy.

It also should be noted that segmentation isneither required nor the best approach for severalother types of clinical problems. For example, themajority of tumor removals and “first-time” discremovals by open operation require neither imageguidance nor segmentation. Although surgery on agrossly unstable spine would benefit from imageguidance, such guidance would probably best beapproached through direct, 3D intraoperative imag-ing. Nevertheless, the number of procedures thatwould benefit from segmentation either directly orindirectly (through use of segmentation as a pre-lude to registration) is significant.

RESEARCH PRIORITIES

This list summarizes the research priorities weview as important to image-guided spine surgery.We view all items in this list as important.

1. Long-term goalsDevelopment of intraoperative, fast, 3Dimaging systems of reasonable cost thatallow easy patient access with preserva-tion of a sterile field; that can cover a largevolume while providing high detail; andthat limit the current problems of radiation(CT) or fringe field (MR).

192 Workshop Report

2. Shorter-term goalsa) Emphasis on validation of methods, with

establishment of accepted criteria for eval-uation of registration methods, creationand use of a standard database with em-bedded fiducials, and measurements of ac-curacy/precision, robustness/stability, reli-ability/reproducibility as well as of surgicalutility (the time and user interaction re-quired).

b) Development of accurate, intraoperative3D-2D image registration (e.g., registra-tion of intraoperative fluoroscopic imageswith a preoperatively acquired CT scan, orregistration of endoscopic images with apreoperative MR scan). Deformable reg-istration will be required in many cases.

c) 3D image-patient/instrument registration.Ultrasound may have potential in this area,possibly by placing patches on the abdomenor even by using the bone itself as the ultra-sound transmitter.

d) 3D image-3D image registration, particu-larly in regard to CT-MR registration. Asthe patient position may be different dur-ing each procedure, deformable registra-tion may be required. Issues of speed andthe extent of user interaction that is re-quired are important.

e) Segmentation for delineation of bony sur-faces during instrumentation procedures,differentiation of tissues (e.g., disc versusscar), biomechanical model building, andthe creation of atlases of spinal anatomywhich depict elative geometry and align-ment.

SUMMARY

Spinal surgical procedures can significantly benefitfrom image-guided surgery, which is currentlywidely accepted for intracranial procedures. OurWorking Group addressed the technical require-ments for the use of image-guided procedures inthe context of clinical needs in surgery of the spine.The highest priority item is the development ofprocedures for the evaluation of an overall systemand its components. The development of widelyaccepted clinical systems requires improvements inaccuracy, speed, extent of user interaction, and easeof integration in a clinical protocol, which in turndemands the design of technical innovations forregistration.

SECTION 5: ANATOMICAL ANDPHYSIOLOGICAL MODELINGThe Report of Working Group 4

AUTHORS

James Anderson, Ph.D., Johns Hopkins UniversityC.I.S.S.T. Engineering Research Center (Techni-cal Leader)

Ron Kikinis, M.D., Brigham and Women’s Hospital(Clinical Leader)

Andrew Bzostek, M.S., Johns Hopkins UniversityC.I.S.S.T. Engineering Research Center

Ed Chao, Ph.D., Johns Hopkins UniversityChristos Davatzikos, Ph.D., Johns Hopkins Univer-

sity C.I.S.S.T. Engineering Research CenterMatthew Freedman, M.D., M.B.A., Georgetown Uni-

versity Medical CenterMartha Gray, Ph.D., Massachusetts Institute of Tech-

nologyNoshir Langrana, Ph.D., Rutgers UniversityGreg Mogel, M.D., U.S. Army Medical Research and

Materiel CommandThomas Whitesides, M.D., Emory University

OVERVIEW: THE ANATOMICAL AND

PHYSIOLOGICAL MODELING PROCESS

The anatomical and physiological modeling devel-opment process includes

● model representation,● image segmentation and registration (both to

atlases and real-time adaptations),● model construction,● visualization and image display,● simulation,● plan optimization, and● validation and adaptations to the systems.

Integration of each of these processes will play acritical role in the development of comprehensivemodels.

Innovative, computationally efficient meth-odologies must be developed which integrate rigid-body modeling with deformable modeling and re-construct (redefine) the model owing to the effectsof these external influences. Physiological model-ing of the interface between the soft and hardstructures that are present in the spine is anotherimportant task. Eventually, all of these models needto be patient-specific models. This construction willrequire successful mapping from models to patient-specific data sets. For initial model development,

Workshop Report 193

research should focus on soft-tissue modeling, seg-mentation of heterogeneous tissue components, ba-sic biomechanical properties, and proper alignmentand positioning of component parts.

MODELING: AN OVERVIEW

All image-guided spine procedures require someform of pretreatment planning. The planning pro-cess aims toward translating, integrating, and cou-pling preoperative computer-constructed modelsand therapy plans with intraoperative actions. Thedesired goal of this planning is to better understandboth normal and disease or injury processes and tooptimize care and management of the patient.

The degree of complexity of the pretreatmentplanning process can vary considerably. Planningefforts can range from creating simple visualiza-tions of image data sets to developing highly so-phisticated models and execution plans that mayrequire augmentation of the surgeon’s eye and/orfacilitating physician interaction with multiple datasets and the use of patient-specific simulations.Close assessment of anatomical and biomechanicalor physiological models of the patients is central tothe planning process.

MODELING AND IMAGE-GUIDED THERAPY

The term “modeling” has many different meaningswith respect to image-guided therapy. For the pur-poses of this chapter, modeling will focus on thedevelopment and/or use of anatomical/physiologi-cal and/or biomechanical data sets that provideopportunities to predict, evaluate, simulate, vali-date, develop, and enhance the outcomes of surgi-cal or other therapeutic image-guided spine proce-dures. The modeling process may involveintegration of multiple forms of both anatomicaland functional image data sets with anatomicalatlases, biomechanical data, and computational al-gorithms. In addition, this interactivity requires ob-taining and integrating information regarding thephysical properties of surgical and therapeutic in-struments, and of sensor input data that are neededto predict the interactions of the surgeon and in-struments with various tissues.

Modeling of the spine and paraspinal regionfor the above applications, such as prediction andvalidation, is a formidable task and one that is in itsinfancy of development. A true understanding ofhow such models can be used either in training orpretreatment planning cannot ignore the complexinteractions of biomechanics and physiology orpathophysiology prior to, during, and following thetherapeutic intervention. Our understanding of out-

comes requires that we first understand the basicsof modeling the normal spine and paraspinal re-gion. The development of useful models will re-quire validation of the visual and physical param-eters as well as the acceptance of and value to thephysician end-user.

THE REGISTRATION PROCESS OF MODELING

A fundamental issue that arises when using anatom-ical models is that these models must somehow beadapted to the individual anatomy of a patient. Infor-mation that is associated with these models is thenautomatically transferred to the patient’s images. Thisprocess is achieved via registration, which in its sim-pler form is rigid, and in a more complex form isdeformable, perhaps even incorporating physicalproperties of anatomical structures.

Figure 5-1: Elastic registration of spine images (Cour-tesy ofChristos Davatzikos, Ph.D., Johns Hopkins Univer-sity). The images in the top row are midsagittal MR sectionsfrom two different individuals. The bottom left shows anelastic deformation of the model (top left) to the target (topright). The bottom right shows an overlay of the bottom leftimage with an outline of the target, indicating a goodregistration at all levels of the spine.

194 Workshop Report

Registration models should initially focus ontransforming image data to rigid or nondeformablemodels, with deformable models and soft tissue mod-eling being of second priority. Real-time registrationof intraoperative images with atlases and preoperativemodels are necessary for performing effective imageguidance and obtaining accurate intraoperative data.Initial work in the modeling process should concen-trate on those preliminary procedures that requiresimpler registration methods, such as rigid bodiesinterfaced with elastic spring modeling approaches(the discrete element analysis technique). Rigid-bodymodeling involving registration and fusion of data ismuch better established than soft-tissue modeling.Second-generation work should concentrate on moreadvanced procedures that involve articulating flexibleor deformable tissues such as the intervertebral discand paraspinal ligaments, and nervous system softtissue.

An example of elastic registration of spineimages is shown in Figure 5-1.

The Deformable Modeling Process. Deformablemodeling is a complicated process owing to thedifficulties of adequately representing the deforma-tions of different soft tissues. Deformable registra-tion is still in a preliminary development stage. Itinvolves shape modeling and reliance on deform-able atlases using physical models and statistical

shape models. Many methods for undertaking de-formable modeling have been developed during thepast several years, including snakes, which useenergy functions to represent the static shapes ofcontours (or surfaces in 3D) and which deformuntil they reach their minimum energy. Dynamicdeformable models represent both shapes and mo-tions of contours. When both internal and externalforces reach a balance, the contours (or surfaces)come to rest at their final locations.

Finite element methods (FEMs) that are capableof handling large deformations are one of the mostcommonly used approaches in the computation ofphysically based representations of deformable mod-eling. In addition, probabilistic deformable modelscombine the characteristics of both prior and sensormodels in terms of probability distributions.

An example of a finite element simulation oftumor growth in the brain is shown in Figure 5-2.The same techniques could be applied to the spine.

The Physiological Modeling Process. Physiolog-ical modeling includes the dynamic functional as-pect of the deformation of soft tissues. These mod-els also apply to sensor interaction with tissues,tissue resistance and other properties, and func-tional imaging registration, e.g., positron emissiontomography (PET) with anatomical imaging suchas computed tomography (CT) or magnetic reso-

Figure 5-2: Finite element simulation of tumor growth in the brain (Courtesy of Christos Davatzikos, Ph.D., Johns HopkinsUniversity). Left: MR image of a normal subject. Right: Simulation of the soft tissue deformation.

Workshop Report 195

nance imaging (MRI). Physiological informationcould include electromyelogram (EMG), MRI tag-ging, nuclear magnetic resonance (NMR) metabo-lism spectroscopy, and Doppler ultrasound imaging.

Compared to anatomical modeling, physio-logical and biomechanical modeling have not beenas extensively studied. Modeling these functionalparameters is even more challenging than that ofanatomical deformable modeling because, in mostcases, it is possible to acquire only indirect mea-surements of a physiological process. A physiolog-ical activity usually varies over time, which makesquantification difficult. As such, a challenging issuein physiological modeling is the accurate acquisi-tion and identification of biomechanical informa-tion, which includes mechanical properties of softtissue and its interaction with surrounding tissues.

Finally, dynamic anatomical and physiologi-cal modeling should include the influence of mus-cle contractions and respiratory and vascular pul-sations on the spinal structures. Innovative,computationally efficient methodologies must bedeveloped which integrate rigid-body modelingwith deformable modeling and reconstruct (rede-fine) the model owing to these external influences.This development will make the virtual modelphysiologically and functionally accurate.

CLINICAL NEEDS

The importance of anatomical and physiologicalmodeling becomes more apparent as increased at-tention is being given to the development of less-invasive procedures that reduce health care costsand do not sacrifice quality of health care delivery.Advances in modeling will rely not so much onmolecular approaches but instead on basic integra-tion of image data sets and physiological/biome-chanical data of both the bone and soft-tissue com-ponents. We are just in the beginning of thisprocess and all of the correct questions and needshave not been clearly defined. However, the ulti-mate clinical requirement must be that outcomes orimprovements in treatment be predicted before thetherapy is provided.

This Working Group felt strongly that theproper use of anatomical and physiological or bio-mechanical models could result in improved out-comes. Improvements can be accomplished by us-ing preoperative models to guide intraoperativeactions that will minimize tissue damage or enablemore specific interventions. The most importantclinical need is improving the ability to achieveincreased realism in the models and simulations.

To meet this need, there has to be a muchbetter understanding of pathogenesis and analysis

of factors affecting loads on the spine and connec-tive tissue in both normal and pathological tissues.New information gained from the use of these newmodels must complement information derived fromclinical cases and provide information about thebiomechanics of surgical planning. We may knowvery little about modeling soft-tissue organs suchas the brain, but we know even less about an area ascomplex as the spine.

To advance modeling efforts we need to:

● Gather a vast amount of anatomical and phys-iological information, particularly about thespine.

● Compile information regarding adequate bio-mechanical models of muscles under normaland abnormal stress.

● Develop physiological models and modelingof the interface between soft and hard tissuesthat are present in the spine.

● Gather more data regarding the effects ofloading on the spine and basic informationabout muscle functions.

● Develop models that quantify the relationshipbetween spinal damage and clinical symp-toms, which in some cases is poorly understood.

Six research focuses for physiological mod-eling were identified by this Working Group. Theprocess of modeling the spine and paraspinal re-gions must start with simple models of normalanatomy and physiology, but the long-term goalsshould be the design of:

1. Patient-specific models.2. Practical implementation and realistic ap-

proximation methods.3. Successful mapping from models to data sets.4. Validation parameters defined by both cli-

nicians and engineers, a goal which is es-sential at every step in model development.

5. An accurate model incorporating phenom-ena such as spine motion dynamics.

6. Computational efficiency and validationmeasurements and parameters.

Although several potential clinical applicationswere discussed by this Working Group, includingspinal fusion and fixation procedures, vertebro-plasty, and discectomy, the group felt that the mostcommon procedures, and possibly those most ame-nable to modeling, were related to spinal stabiliza-tion applications and correcting spinal deformitythat is either idiopathic or post-traumatic in nature.Three areas of immediate clinical need include:

196 Workshop Report

1. Positioning of components.2. Biomechanical modeling of bone-ligament-

muscle components, including modeling ofthe material properties of bone, mineral con-tent, and structure and fatigue strength ofthe elements.

3. Consideration of wear patterns, aging, rangeof motion analyses, remodeling, and dis-ease-related factors.

The models must be modular, interchangeable, andpatient specific. In all cases it is important to vali-date and determine error margins in the developedmodels. The Working Group also gave high prior-ity to training-based models that included visual-ization components and measurements, outcomesanalyses, and testing of physicians’ skill level andexperience. The Working Group recommended thatthe biomechanical models be multi-segmental,cover the entire spine, and include significant soft-tissue components of ligaments and muscle forces andthe relative physiological parameters. Finally, thegroup recognized that the models must reflect theeffects of surgery, including modification with instru-mentation, bone removal, and fusion procedures.

TECHNICAL REQUIREMENTS

Seven technical requirements for advancing workin anatomical and physiological modeling of thespine were identified by this Working Group:

1. Further research into spinal modeling devel-opment needs. Technical requirements for modelingof spine procedures are not dissimilar from thoseassociated with other areas of the body, except that inmany cases they are more difficult. Modeling of thespine and paraspinal soft tissue introduces problemsrelated to segmentation of multiple, heterogeneoustissue components, including bone, muscle, liga-ments, vascular structures, and neural components.The anatomical relationships between these compo-nents are complex and poorly understood with respectto model development.

2. Further understanding of physiological andbiomechanical properties related to the spine. Ofeven greater difficulty than understanding complexanatomical relationships within the spine is themodeling of muscle physiology and biomechanicalproperties related to the spine. Currently, little in-formation is available regarding the constraints ofsoft-tissue components in the paraspinal region.The influence of functional parameters such asgravity, abdominal muscular support, age, varia-tions in intradiscal pressures, kinematics, and var-ious loading and weight-bearing parameters need to

be considered. Modeling of the interactions of theseparameters will be extremely difficult.

3. Multi-modality imaging registration tech-niques for spinal surgical procedures. Functionalimaging studies related to muscle strain and stressusing MRI tagging may be of value, but littleresearch has been done in most areas of the body,with the exception of the heart. Automatic imagesegmentation techniques need to be developed fordiscriminating between the heterogeneous soft-tis-sue components. Multi-modality image registrationtechniques need to be implemented for enablingregistration of preoperative images with real-timeintraoperative images. Finally, respiratory and evenvascular pulsation motion-related issues need to beaddressed for enabling registration of preoperativeand intraoperative image data.

4. Identification of technical requirementsneeded for developing models for image-guidedsurgery. Steps within this development processwere identified by our Working Group, as listed inFigure 5-3.

In addition to addressing each of the stepssuggested in Figure 5-3, our group recommendedthat there also be a hierarchical organization ofproblems that can be addressed for each stage oftechnical development. Currently, a limited numberof developments are underway in model simulationof areas of the human body. One of these is ortho-pedic/arthroscopy simulator systems which are be-ing developed for studying interventions of theknee and shoulder. This Working Group was notaware of any major developments in simulationsystems for use in spinal interventions.

5. Development of algorithms to track tissuedeformation. Tissue deformation is a major techni-cal problem in surgery of the spine. Other technicalareas in need of development include real-time per-

Figure 5-3: Required technical components of the modeldevelopment process.

Workshop Report 197

formance and research into both non-linear deforma-tion and identifying characteristics of anisotropic ma-terials. Some of these issues are closely related to thedevelopment of physiological and biomechanicalmodeling of soft tissues. A compromise between fi-nite element mesh resolution and the achievable com-plexity of current biomechanical models is unavoid-able due to the demanding computational resources.More efficient algorithms need to be developed forbetter understanding the deformation of non-linearviscoelastic tissue models, collision detection be-tween deformable bodies, and computation of contactforces or pressures between deformable bodies. Inte-gration of both anatomical and physiological model-ing will become a key issue in this field. Validation ofall of the technical developments is critical as eachprogresses.

6. Development of algorithms to compute muscleactivity and roles in spinal stability. Muscle con-tractions and co-activation is yet another majorresearch issue that needs to be developed. Musclegroups provide active control and dynamic forcesto the paraspinal regions, which provide spinalstability. Several optimization algorithms havebeen developed to compute the roles of differentmuscles in static postures, but have had very lim-ited success in practice (if any). Innovative ap-proaches are needed to address muscle co-activa-tion and the roles of muscles, soft tissues, andvertebral bones in stabilization of the spine. Inte-gration of the above research findings will play acrucial role in the design of the comprehensive mod-el(s) needed for image-guided spine procedures.

7. Additional requirements of models for image-guided spinal surgical systems. Models that aredeveloped for work on the spine should be general-ized, but they should also be individualized and adapt-able to individual patient factors such as age, sex,history, and patient-specific anatomy. The systemsmust be practical and include real-time performancestandards. In addition, these imaging systems must bedesigned with a hierarchical organization of problemsthat can be addressed at each stage of the technolog-ical development. Tissue mechanical properties mustalso be included in model development.

RESEARCH PRIORITIES

The following research priorities were suggestedby this Working Group:

1. Initial model development should focus onclinically relevant problems of deformityand spine stabilization.

2. Anatomical models should focus on shape

construction and the proper alignment andpositioning of component parts.

3. Biomechanical properties are critical andconsiderable work needs to be done to betterunderstand the interaction of heterogeneoussoft-tissue components and bony structures.

4. Initial model development must incorporatedata on wear patterns, age, stress, and loadbearing.

5. Initial model development must reflect theeffects of surgery or other interventions.

For initial model development, research should focuson soft-tissue modeling, segmentation of heteroge-neous tissue components, basic biomechanical infor-mation such as kinematics, forces, and tissue stresses,as well as the proper alignment and positioning ofcomponent parts. Physician interaction and validationstudies must be a part of the evolution of the modelsat every stage of development.

SECTION 6: SURGICALINSTRUMENTATION, TOOLING,AND ROBOTICSThe Report of Working Group 5

AUTHORS

Michael Peshkin, Ph.D., Northwestern University(Technical Leader)

John Mathis, M.D., Lewis Gale Medical Center(Clinical Co-leader)

John Kostuik, M.D., Johns Hopkins Medical Institu-tions (Clinical Co-leader)

Fred Barrick, M.D., Inova Fairfax HospitalNorman Caplan, M.S., Johns Hopkins University

C.I.S.S.T. Engineering Research CenterNeil Glossop, Ph.D., Traxtal TechnologiesRandy Goldberg, M.S., Johns Hopkins University

C.I.S.S.T. Engineering Research CenterNobuhiko Hata, Ph.D., Brigham and Women’s Hos-

pitalMichael Loser, Ph.D., Siemens Medical EngineeringMichael Murphy, Ph.D., Louisiana State UniversityRussell Taylor, Ph.D., Johns Hopkins University

C.I.S.S.T. Engineering Research Center

PREVENTIVE CARE OF THE SPINE:AN OVERVIEW

Our Working Group particularly considered theconsequences of an aging U.S. population, whichwe believe has significant implications for care ofthe spine in the near future. Currently, the single

198 Workshop Report

largest presenting complaint leading to spinal in-terventions is lower back pain. Direct costs mightbe estimated at $14 billion annually, with the ad-ditional annual cost of failed surgeries at perhaps$5 billion. These costs may be expected to rise.

As our population ages,preventiveprogramswill require large-scale delivery of certain proce-dures, particularly injections. Diagnosis and pre-vention will be considered together here as therelevant procedures, with similar technical require-ments. Efficient and extremely safe delivery ofthese procedures is needed; otherwise, preventivecare will not be appealing to those who need it.

CLINICAL NEEDS

Our Working Group identified and assessed threeimage-guided spinal interventions which can effectimproved outcomes for patients with lower backpain. These include needle procedures for nerveroot decompression; better visualization for inter-ventions focused on compression fractures; andminimally invasive techniques to destroy tumor inthe spine.

USE OF IMAGE-GUIDED SURGERY IN NEEDLE

PROCEDURES FOR NERVE ROOT

DECOMPRESSION

Compression of the nerve roots or spinal cord is acommon problem. It can be congenital or the resultof Paget’s disease, degenerative disease, spondylo-sis, ligamental ossification, fractures, tumors, andother causes. Compression is a painful conditionthat may require intervention, or “bony decompres-sion”. The current standard of treatment is an opendecompression procedure. Currently, less-invasivetreatments have not proven effective.

Image-guided surgery (IGS) or robotic tech-niques can, however, contribute to both the effi-ciency and safety with which needle proceduresused for diagnosis or treatment of compression maybe carried out. Decompression of the nerve root orcord is accomplished by removing tissue thatplaces pressure on the neural element. Accuratetargeting reduces the (small) chance that sensitivestructures can be inadvertently damaged. It can alsoimprove the speed with which a needle procedurecan be accomplished.

Examples of robotic systems that might beapplied to image-guided spine procedures areshown in Figures 6-1 and 6-2. The first figureshows a new-generation remote center-of-motionrobot developed at the Johns Hopkins University.This robot is designed for “steady hand” microsur-gery to extend human ability to perform micro-manipulation. For image-guided spine procedures,

a device like this might be used to assist in needleplacement while ensuring that dangerous regionssuch as the spinal canal are avoided. The secondfigure depicts a robotic system for precise needleinsertion under radiological guidance. The systemhas been applied to kidney biopsy and presents amodular structure comprising a global positioningmodule, a miniature robotic module, and a radiolu-cent needle driver module.

Accurate targeting, which can be facilitatedby IGS, is perhaps even more important for achiev-ing accurate diagnostic results. A primary tool ofdiagnosis for the cause of pain is the injection ofanesthetic or steroids in or near (within about 1 mmof) a sensory nerve. For a variety of reasons, notleast the placebo effect, at least two — and often

Figure 6-1: Steady hand robot (Courtesy of Russell Tay-lor, Ph.D., Johns Hopkins University).

Figure 6-2: PAKY needle insertion robot (Courtesy ofDan Stoianovici, Ph.D., Johns Hopkins Medical Institu-tions).

Workshop Report 199

more — injections are needed. Reliable diagnosisthus depends on reliable targeting of the injectedmaterial. Diagnosis is greatly facilitated if one cancount on successive injections being delivered tothe same location on the nerve.

In order for significant societal investment tobe made in preventive programs for spinal pain,outcomes validation studies are required. Con-trolled studies are needed to determine the accuracyand efficacy of these needle delivery programs andto determine their preventive significance.

ANOTHER APPLICATION FOR IMAGE-GUIDED

SURGERY: COMPRESSION DISK FRACTURES

Compression of the nerve root is also a widespreadproblem, and can result from herniated, prolapsedor protruded, extruded, or sequestered discs. Com-pression fractures number some 500,000/year, andprobably are orders of magnitude larger in numberif we consider cases in prevention as well. A largefraction will require operative intervention. Clinicalissues which can benefit from IGS techniques arethose which assist in identifying how much cementto use, where to put it, and how to control where itgoes.

The state of the art in discectomy includesmicrodiscectomy, nucleotomy, micro-endoscopicdiscectomy, and laser ablation. These interventionsare effective for many types of disc procedures andcan be performed almost on an outpatient basis.

ANOTHER APPLICATION: TUMOR REDUCTION

Here one wishes to remove the tumor in order tohelp the body to maintain its immunological effort.The idea is to destroy (or “munch”) most of atumor, deposit a tumoricidal agent, and do this withminimally invasive technology. Spinal tumors arealmost always located in the vertebral bodies, andthe tumors are generally of soft material. This ap-plication represents somewhat more sophisticatedtechniques than does a needle procedure. What isneeded is a minimally invasive “muncher” guidedby IGS. Such a tool will be discussed at greaterlength below.

ONE AREA OF POSSIBLE IMPROVEMENT USING

IGS: STABILIZATION AND FUSION

Stabilization involves the use of metallic implantsand is performed to eliminate motion, usually forfusion of segments in the spinal region. Stabiliza-tion can be required due to incidence trauma, fortumor removal, or to assist with fusion. Fusion istypically done as follows:

1. Removal of disc and/or facets and/or bonyend-plates.

2. Addition of grafting material.

3. Stabilization using a mechanical constructsuch as a cage, rods, or plates fixed to thevertebrae with wires, plates, cortical bonescrews, pedicle bone screws or hooks, or acombination of these.

The state of the art in stabilization and fusionrequires the invasive introduction of screws andother hardware. It is simply much easier to intro-duce plates and screws and to fasten them in anopen operation, although some clinicians have beeninvestigating ways to accomplish these tasks in aminimally invasive manner.

Most attempts at minimally invasive implan-tation of screws and needles for stabilization cur-rently require frequent use of fluoroscopy, whichuses ionizing radiation. Sophisticated new instru-mentation and techniques such as computer-aidedsurgery (CAS) will have to be developed for per-cutaneous stabilization. This can be accomplished,it is thought, by using CAS or new imaging meth-ods such as the open CT or open MR.

Fusion has been made easier and much lessinvasive owing to the introduction of cages, butlong-term results are not yet known. Cages are notappropriate in all cases as they do not provide thesame degree of stabilization that is provided byconventional fusion procedures.

TECHNICAL REQUIREMENTS

IMAGING

The state of the art in guiding needle procedures isrepresented by manual fluoroscopic and CT guid-ance. We noted that interventional radiologists per-form biopsies with CT guidance, doing so itera-tively by positioning a needle and sliding thepatient in and out of the CT scanner. This processis labor-intensive and slow. We considered a num-ber of alternative IGS technologies to facilitateneedle procedures (and similar interventions suchas the “muncher”).

Optically tracked tools correlated with CTdata sets (of which Sofamor Danek’s StealthSta-tiont is an example) duplicate an open proceduremore slowly and, perhaps, more accurately. How-ever, we did not find significant benefit in thistechnology for facilitating needle procedures. Whatis needed is a minimally invasive technique, andone for which the registration process is rapid,convenient, and accurate.

200 Workshop Report

Ultrasound has been available for use in spinetrauma interventions for 15 years or so. It has neverbeen accurate enough and, we believe, has beenmore or less abandoned. It is labor-intensive to use;it needs a dedicated technologist to produce goodimages, because ultrasound imaging often has a lotof variability; and its use is not straightforward.Ultrasound pictures are also difficult to interpret.While this may change with recent developments inultrasound, it is currently not used very much in thespine.

Fluoroscopy has many advantages, allowingintraoperative imaging and intraoperative registra-tion. IGS techniques (as opposed to manual fluo-roscopy) would also minimize radiation exposurefor both patient and surgeon. A disadvantage offluoroscopy, however, is that image quality can beproblematic, especially in cases of low bone-den-sity and of obesity. These problems might be alle-viated by a fluoroscopic overlay on preoperativeCT images.

The use of IGS in, for instance, compressionprocedures will require the development of moreeffective imaging. It is extremely difficult to workin tight recesses of the spine without having theadvantage of high-resolution, unambiguous im-ages. Our Working Group examined several op-tions, including use of the intervascular MR coil,frameless stereotaxy (which was dismissed as notproviding needed accuracy or up-to-date images),and foraminoscopy.

GUIDANCE

We polled the three clinicians in our Working Groupon their preferred mode of guidance: Should an IGSsystem (1) simply indicate the current target of en-tirely hand-held tools (real-time video overlay), orshould it (2) involve a positioner for guidance (“robotline-up”), or should it (3) more actively perform theinsertion? One clinician opted for #1 (video overlay)while two clinicians preferred physical guidance (“ro-bot line-up”). No one expressed any interest in a robotmore actively involved than that. The clinician thatopted for video overlay expressed a lack of trust inrobotic positioners, which one could imagine mightbe allayed over time and with advancements in thefield.

ENDOSCOPIC TOOLS: DISK REMOVAL

“MUNCHER”What is currently available for endoscopic disk re-moval is a rigid tool known as a “muncher.” This isinherently a linear tool: its path can be obstructed bybone or anatomical structures. The actual surgery to

access the tunnel that the nerve traverses is extremelytight and instrumentation required to perform nerveprocedures must be extremely dexterous.

Improvements are therefore needed in boththe visualization of the nerve and tools which cannavigate around obstacles (i.e., bite away the bone).The tool needs to have a mobile tip once it ispositioned correctly. Specifications for such a toolare as follows:

● 1-cm range of motion● capability for suction: volume of material re-

moved is 1 to 10 cm3

● grasping forceps● perhaps a drill or burr for taking out pieces of

bone● low force-level requirements● CT-like navigation down to the foramen is

required, then visual guidance is needed● needed tool tip angles may be 30 degrees for

disk removal, but 90 degrees would provideadditional capabilities

● tool body could be up to 10 mm in diameter,narrowing to 4 mm for the part that wouldenter a disk, and narrowing further to about 2mm at the tip.

Improved instruments for spinal procedures, nota-bly in endoscopic visualization, will be necessaryto achieve advancements in spinal interventions.

RESEARCH PRIORITIES

Although bone morphogenetic proteins (BMPs) andgene therapy will become more and more importantin treating the spine over the next several years, theywill never obviate the need for intervention. Thesematerials will precipitate rapid and stable fusion, butprecise instrumentation will be required to deliver thematerials to the appropriate locations in the spine.These biologically active materials will require min-imally invasive intervention (similar to that beingenvisioned for vertebroplasty) for delivering theagents to the appropriate locations.

There was a definite consensus in our WorkingGroup that the future lay in a marriage of biologicaltreatments and minimally invasive systems to deliverthe agents to the location accurately. One examplemight be a development like an “injectable bonescrew”, which could solve a problem by enabling thesurgeon to locate a screw percutaneously and inject anagent rather than introduce one.

However, in terms of determining biomedicalresearch priorities, we feel that advancement of the

Workshop Report 201

instrumentation, biologically active materials, andIGS for delivery all need to be studied simulta-neously. The following priorities are critical:

1. To achieve this advancement, infrastructureneeds include making the visualization andregistration systems, and data fusion, be-come standard procedures in the OR. Weneed to undergo a sizeable shift in focusfrom surgical “carpentry” to information-intensive surgery.

2. For these technologies to be deployed, NIHand other innovative research/support insti-tutions need to do much more focused workon systems research and development. ThisWorking Group believes that if we want tocouple IGS to biologicals, we have to investin the delivery systems. Merely to be able tostudy the effect of the biologicals, we needto do controlled delivery of the image-guided procedures.

SECTION 7: SYSTEMARCHITECTURE, INTEGRATION,AND USER INTERFACESThe Report of Working Group 6

AUTHORS

Robert Galloway, Ph.D., Vanderbilt University(Technical Leader)

Richard Bucholz, M.D., St. Louis University (Clini-cal Leader)

Jae Jeong Choi, M.S., Georgetown University Med-ical Center

Kevin Cleary, Ph.D., Georgetown University Medi-cal Center

Sarah Graham, M.S., Johns Hopkins UniversityC.I.S.S.T. Engineering Research Center

Heinz Lemke, Ph.D., Technical University of BerlinRichard North, M.D., Johns Hopkins Medical Insti-

tutionsMonish Rajpal, M.S., Johns Hopkins University

C.I.S.S.T. Engineering Research CenterRamin Shahidi, Ph.D., Stanford UniversityAnn Sieber, R.N., Johns Hopkins Medical InstitutionsLaura Traynor, B.S., University of UtahMichael Vannier, M.D., University of Iowa

IMAGE-GUIDED SURGICAL SYSTEMS FOR THE

SPINE: CURRENT LIMITATIONS AND

CHALLENGES

Image-guided surgery has, to this point, been pre-dominantly applied to intracranial procedures.

Most cranial guidance systems consider the skulland brain to be rigid structures for which a single,rigid transform can suffice for registration of imagedata. Even for cranial interventions, this assump-tion is incorrect, as the brain is neither homoge-neous nor rigid. In addition, few guidance systemsallow intraoperative information retrieval and mod-ification of the preoperative images based on thisintraoperatively acquired information. These issuesmust be addressed for image-guided surgery of thespine to be accurate.

The spine can be considered (as a first ap-proximation) as a series of rigid bones connectedby flexible structures. Unless the patient is immo-bilized within a rigid constraint during imaging andinterventions, the relative positions of the vertebraecannot be assumed to be the same as when anypreoperative scans were obtained. This variabilitynecessitates the capability to either capture addi-tional data intraoperatively or to perform multiplespatial and temporal registrations.

Beyond issues of lesion targeting, image-guidedsystems for spinal procedures must be developed toallow for treatment of a variety of structural disorders.These procedures will range from simple disk re-moval to correction of gross deformities which re-quire exposure of several spinal levels. Image-guidedspinal systems must either track multiple objects anddepict their relative position and angulation, or permitintraoperative imaging of these structures to producean accurate model of the spine as it exists within theoperating room.

CLINICAL ISSUES: DESIGNING AND BUILDING

FUTURE IMAGE-GUIDED SYSTEMS

The overall goal of this Working Group was toenvision effective tools for image-guided surgeryof the spine. The group felt that a valuable ap-proach would be to develop an architecture thatcould incorporate advances in localization, regis-tration, display techniques and targets, and trajec-tory definition into systems which have demon-strated their clinical usefulness and significance.Mechanics of this image-guided system involvecapturing information about the spine during theintervention that is required for accurate interven-tion, and presenting that information in a timeinterval and manner that is appropriate for the in-tervention. These interventions may be disk proce-dures, spinal instrumentation, procedures for thebiopsy and ablation of cancer, surgery for grossdeformity, or needle procedures. A typical image-guided surgery system used in the operating roomfor cranial interventions is shown in Figure 7-1.

202 Workshop Report

The user interface for this system, which presentsaxial, sagittal, coronal, and 3D views, is shown inFigure 7-2. A similar system can be used for spineinterventions, such as the placement of pediclescrews.

There are a variety of specific clinical needsthat must be addressed in order to advance thedevelopment of image-guided surgery of the spine.Among these clinical needs are issues related toregistration procedures and input of data; networkrequirements; graphical user interface (GUI) archi-tecture; acquisition and classification of surgical-related information sources and data; and accumu-lation of valid data for comparative patient healthoutcomes studies. Each of these clinical issues willnow be described briefly.

Registration Procedures and Input of Data. Aspecific clinical need for improved flexibility inregistration procedures and for the input of data(measured signals, 2D images, and 3D image sets)was identified by the Working Group. To satisfac-torily meet this clinical requirement, the design ofsurgical image guidance systems will have to bemore open. They will need to be connected to adata network and be able to transfer data fromhospital databases and from diverse informationsources directly to the operating room.

Network Requirements. The process of image-guided surgery is an example of a mission-criticalsystem, just like the internal network of a modernairplane. The pilot relies upon the plane’s networkto provide information on a real-time basis aboutthe position of control surfaces, engine functions,plane location, and guidance information. Simi-larly, patient safety and positive surgical outcomesdepend on rapid, secure, and stable data transfer to

and from the interventionalist. The systems shouldbe able to initiate and terminate individual datastreams and set bandwidth and communication pri-orities for individual data streams.

Adjustable Graphical User Interfaces (GUIs).One difficulty in defining system architecture, in-tegration, and graphical user interfaces (GUIs)stems from the fact that the development of thesystem must be closely coupled to the preferencesof the operating surgeon. Advances in technologywill require new surgical techniques, just as newsurgical techniques place demands on existingtechnology, which promotes the development ofnew technology. Against the backdrop of this rapiddevelopmental cycle, surgeons must be comfort-able in their mastery and control of the devicesused during a procedure. As surgeons vary greatlyin their approach to technological innovation, eachdevelopment may be accepted and/or used quitedifferently by different surgeons and specialists.Some surgeons desire greater technical control overthe system and some want to use it as a “point andshoot” mechanism.

In addition, systems may be used for differentfunctions, and indeed not all are meant to be mul-tifunctional. A system used for discectomies willnot require the same functions as one used for thecorrection of gross deformity. Although there can-not be specialized systems for each type of proce-dure, a system should instead have a selection

Figure 7-1: Image-guided surgery system in operatingroom (Courtesy of Richard Bucholz, M.D., St. Louis Uni-versity).

Figure 7-2: Image-guided surgery system user interface(Courtesy of Richard Bucholz, M.D., St. Louis University).

Workshop Report 203

mechanism from which the surgeon can alter theamount and type of information displayed and beable to tune in and out specific data streams asneeded.

While there is considerable appeal in allow-ing flexibility of system function to enable surgicaltechnique and to support differing desires of sur-geons, there is also a danger in aiming toward “tooflexible” a system. Such flexibility makes it ex-tremely difficult to posit valid comparisons acrossprocedures and between surgeons. As these sys-tems develop, clinical study design needs to ad-dress the issue of controlling systems’ flexibilityin order for valid comparisons to be made acrosssites.

Information Sources: Acquisition and Classifi-cation Systems for Surgical-Related Data. Thereare two classes of input to image-guided surgerysystems — preoperative and intraoperative. Preop-erative data are traditionally comprised of tomo-graphic image sets; however, surgical plans andhistoric data on prior cases should also be part ofthe input data stream. Intraoperative data shouldinclude intraoperative images (both three- and two-dimensional) as well as electrophysiological andmechanical information. Other data sources —such as biomechanical studies, comparisons of sur-gical instrumentation characteristics, and patient-specific data (for example, a history of smoking,concomitant disease, and other factors) — shouldalso be incorporated and available to make thesystem a true information appliance. Mechanicaldata, such as how the spine responds to specificforces, are very important in determining what canbe achieved with surgery and how to bestachieve it.

Accumulation of Outcomes Study Data. With thebroad acceptance of image-guided surgery systemdevelopment, there is considerable anecdotal evi-dence that such systems improve surgical interven-tions by reducing morbidity and allowing morecomplete resections. However, careful prospectivecomparisons with conventional surgical techniqueshave not been made. Concomitant with the devel-opment of image-guided spinal surgery techniques,methods for assessing process effectiveness shouldbe developed to provide a mature and establishedmethodology to demonstrate clinical efficacy. As aresult of creating an established baseline for com-parisons, truly valid outcomes studies would thenbe possible.

TECHNICAL REQUIREMENTS FOR IMAGING

SYSTEMS’ DESIGN AND INTEGRATION

PROCESSES

INTRAOPERATIVE IMAGING MODALITIES AND

DEVELOPMENT NEEDS

Using the four types of imaging systems that arecurrently employed intraoperatively during spinalsurgeries — namely, ultrasound, endoscopy, fluo-roscopy, and intraoperative tomography — ourWorking Group focused on the technical require-ments for enhancing these current technologies andthe procedures used for image-guided spinal sur-gery.

1. Ultrasound is a real-time (30 frames persecond, 1 frame latency), gray-scale imag-ing stream. If color Doppler is employed toexamine blood flow, then the system mustbe capable of displaying color. By providing3D localization and trajectory information,localization data from the ultrasound imagescan be extracted noninvasively. However,given the reflective nature of ultrasound im-ages, basic research is needed on methods tomake this registration process feasible.

2. Endoscopy is also performed in real time.It, too, is inherently true color, which effec-tively triples the bandwidth requirements ofthis imaging technology. If the endoscopicdata are to be used quantitatively, image-guided surgery systems must allow for on-line distortion correction.

3. Fluoroscopyprovides real-time, gray-scale,highly resolved, and therefore large images.However, since the distortion inherent tofluoroscopic devices is position dependent, theguidance system must be able to correct theresulting distorted images. Recently, a com-mercial fluoroscopy-based image-guided sur-gery system has been introduced, as shownin Figure 7-3.

4. Intraoperative Tomography: CT/MRI.Use of this technology requires high-speedimaging and data transfer. The requirementsare highly procedure-dependent. Conven-tional picture archiving and communica-tions servers (PACS) are generally not fastenough for intraoperative use. As a result, touse intraoperative imaging effectively, anew standard for local transfer needs to bedeveloped to circumvent the slowness ofcurrent PACS standards.

204 Workshop Report

REGISTRATION

In addition to examining particular changes andimprovements of current imaging technologies, ourWorking Group also examined technical require-ments for registration of the spine. As spinal ver-tebrae move relative to each other during the courseof surgery, frequent spatial and temporal registra-tions must be performed by the image guidancesystem. Three types of registration are described:

1. Transcutaneous Registration.Percutane-ous procedures such as biopsies and injec-tions might benefit from a method of trans-cutaneous registration, one that may betracked and/or uses three-dimensional ultra-sound. The image guidance system mustthen support the image processing needed toextract necessary information for registra-tion assessment and tracking.

2. 2D-to-3D Registration. Intraoperative ul-trasound, laparoscopy, and intraoperativefluoroscopy are all two-dimensional imag-ing modalities. These 2D images should bereformatted to be related to 3D data sets,such as computed tomography. Three-di-mensional position information can be ob-tained from two views and should be sup-ported by the system.

3. Temporal Registration. The frequency ofrepeat registration and allowable time forregistration is procedure-dependent. Tem-poral registration time is most critical insurgical procedures for tumor and defor-mity, but should be a basic component ofany spinal guidance system.

INTRAOPERATIVE DATA INTEGRATION

Preoperative planning can be crucial to the successof the surgery, especially in cases of deformity. Theintegration of a preoperative plan with an intraop-erative reality using some of the data streams indi-cated above can speed the surgery and allow forbetter agreement between desired configuration andaccomplished tasks.

Acquisition and assessment of intraoperativedata and integration of these data into surgicalplanning and procedures are important technicalrequirements for image-guided surgery of thespine. Guidance systems should be viewed as in-formation appliances. If standards are developedwhich allow measurement devices in the operatingroom to talk to and interact with the guidancesystem, this intraoperative data can be used tooptimize the intervention. These data may be pre-sented visually, as shown in the sample 3D visual-ization of a brain tumor in Figure 7-4. These datacan include:

1. Thermal data for radio frequency ablationand cryoablation.

2. Neurophysiological data such as evoked po-tentials.

3. Mechanical data measure elements which canbe used to modify biomechanical models withpatient-specific information. In addition, suchprocesses can allow for the quantification ofthe rigidity of spinal instrumentation.

Figure 7-3: FluoroNav™ fluoroscopy-based image-guided surgery system (Courtesy of Medtronic—SurgicalNavigation Technologies).

Figure 7-4: 3D tumor visualization (Courtesy of RichardBucholz, MD, St. Louis University).

Workshop Report 205

4. Pathological data identifying the tumor typeand distribution that may impact the natureof the procedure being performed.

FEEDBACK OF INFORMATION TO THE

INTERVENTIONALIST

The capability for providing intraoperative data tothe interventionalist is a critical technical require-ment of surgical imaging systems. Design issuesthat were raised in this regard by our WorkingGroup were:

1. Feedback of data during surgical procedureson the spine is necessary. Surgical guidancesystems are based on the concept of displayof position. In spinal surgery, perhaps morethan any other type of surgery, the feedbackof the position of connected structures is ofcritical importance.

2. The mode of data display should be flexibleand available as needed. As the systemsevolve into information appliances, the han-dling of data flowing into the system and theselective display of that information are vi-tal intraoperative processes. System designshould allow for the flexible display of in-formation and control over data sources andoperative effectors.

RESEARCH PRIORITIES

Several priorities for intensive research were iden-tified within this group:

● High-speed networks (and image transferstandards for effective use of these networks),should be developed and placed immediatelyin research centers to facilitate the use ofimaging in the operating room.

● Research into the development of user-con-figurable graphical user interfaces should besupported, and the ergonomics of system-sur-geon interaction should be carefully pursued.

● Registration techniques must be simplified,and enabled using low-cost techniques such asfluoroscopy or ultrasound.

● Intraoperative navigational systems should bedeveloped with an open interface to facilitatetheir transformation into information appli-ances capable of acquiring and displaying in-

formation from diverse data sources, includ-ing imaging sources.

● Intraoperative inter-vertebral motion and theeffect of this motion upon registration accu-racy is poorly understood, and is critical forcreating effective image-guided systems forsurgery of the spine. An animal model for thetesting of spinal image guidance systemsshould be developed to study this motion.

SECTION 8: WORKSHOP SUMMARYPRESENTATION

By Michael W. Vannier, M.D.†University of Iowa

Contents

● Planning objective and process● Significance of the problem● Background● Needs and opportunities● Strategy● Detailed recommendations● Recommendations specific to spinal surgery● Summary

† Editor’s note: Dr. Vannier served as the Workshop rapporteur and was tasked with the job of summarizing theWorkshop on the last day. This chapter is his summary talk. His presentation described the conference objective and process,the significance of the problem, some background, needs, opportunities, and overall recommendations.

Figure 8-1. Workshop planning process. The planningprocess includes needs assessments, design and applicationof an intervention, and evaluation of the results, whichrefines and adapts future needs to the changing environ-ment.

206 Workshop Report

PLANNING OBJECTIVE AND PROCESS

The objective of the conference is to determine thetechnical requirements for image-guided proce-dures in the spinal column, spinal cord, andparaspinal region. The planning process (Figure8-1) begins with a needs assessment, which in-cluded a pre-Workshop questionnaire and the for-mation of Working Groups.

This process is intended to design, develop,deploy, and evaluate new systems that will be use-ful in spinal interventions, which will need to bevalidated through outcomes assessment. The ulti-mate goal, of course, is improved clinical outcomesfor spinal disorders, especially low-back pain.

The Workshop is organized into six WorkingGroups, each of which has a specific area of con-centration (Figure 8-2). Inplanning and simulation(Group 1), the purpose is to choose the best strategyfrom the various interventions possible, as well asto optimize the intervention chosen.Intraoperativeimaging and endoscopy(Group 2) deals with col-lecting data;registration and segmentation(Group3) with preparing the data; andmodeling(Group 4)with organizing knowledge in the context which isultimately more useful.Instrumentation, tooling,and robotics(Group 5) assist in the interventionitself, while systems architecture and user inter-faces(Group 6) is concerned with the integration ofthese components.

Group Purpose1. Planning and

simulationChoose best alternative

(optimization)2. Imaging Collect data3. Registration and

segmentationPrepare data

4. Modeling Organize knowledge5. Instruments and

robotsIntervene

6. System architecture Integrate

Figure 8-2. Working Groups and Purpose.

There are several general questions regardingimage-guided spinal interventions that should beanswered in this report. First of all, whom are wetrying to serve? What do we want to do? Why dowe want to do it, and why is it important? How willthis be accomplished? How will we know if we’vegot it right or not?

SIGNIFICANCE OF THE PROBLEM

To underscore the importance of the topic ad-dressed here, consider the following summary of anoverview of low-back pain by one of the nation’s

leading experts in outcomes analysis of interven-tions related to low-back pain:

Up to 80 percent of all adults will eventuallyexperience back pain. Its possible causes are mul-tifarious and mysterious. Why some people expe-rience it is as hard to understand as why manyothers don’t. Fortunately, treatment options are im-proving, and they usually involve neither surgerynor bed rest.3

However, there are cases when surgery isindicated. According to Deyo,3 surgical consulta-tion with CT or MR imaging is indicated for pa-tients with persistent or progressive neurologicaldeficits or persistent sciatica with nerve root ten-sion signs. Also, acute radiculopathy with bilateralneurologic deficits, saddle anesthesia, or urinarysymptoms is suggestive of cord compression orcauda equina syndrome and requires urgent surgi-cal referral.8

Based on the medical literature and a needsassessment, a problem statement for Image-GuidedSpinal Interventions can be formulated as follows:

Low-back pain is a prevalent and poten-tially correctable source of disability. Surgi-cal treatment, when indicated, produces vari-able outcomes that may be improved by less-invasive, image-guided procedures. Reducedoverall disability, lower cost of treatment,fewer complications, and less variability inoutcomes may be realized by using image-guided technology.

Based on the problem statement, we developedspecific goals for image-guided spine procedures. It isessential that treatment be individualized, that thetechniques employed have the promise to optimizeinterventions, that variability be reduced, and that theoverall efficiency be improved.

BACKGROUND

The image-guided spine surgery process is com-plex. As shown in Figure 8-3, components of theprocess include preoperative imaging, data prepa-ration including modeling and segmentation, sim-ulation (if applicable), then registering sources ofdata and applying these to the intervention on thepatient. The interventions are monitored, corrected,or extended, according to the results of intraoper-ative imaging.

Whom do we want to serve? There are many

Workshop Report 207

constituencies that are affected by spine abnormal-ities and the interventions to treat them. Each hasdifferent needs and motivations. These constituen-cies include the patients and their families, theclinicians who treat spinal disorders, as well asemployers and insurance companies who are af-fected by the productivity loss and financial conse-quences of persons who experience low-back pain.Patients want pain relief, and for chronic pain theywant long-term relief.

These needs lead us to a vision for what wewould like to achieve; namely, that low-back paintreatment with image guidance improves outcomeswhile reducing overall variability, costs, and com-plications. This yields benefits for patients, physi-cians, employers, and the public at large.

Low-back pain is a major concern in spinalinterventions because it is so common, but thescope of abnormality in the spine is vast and hasbearing in many different clinical areas.

Our shared vision is:

Low-back pain treatment with image guid-ance improves outcomes while reducingoverall variability, cost, and complications.Benefits accrue to patients, payors, employ-ers, and the public at large.

The scope of disease and abnormality in thespine is vast and includes trauma, deformity, de-generation, neoplasm, and other disorders. Imagesare often available, especially preoperatively, in 2Dand 3D for spinal abnormalities, but the images

themselves do not depict function. In particular,pain and disability are not shown on the images.While the anatomy is well delineated on the im-ages, particularly the bony geometry, pain and dis-ability are usually evaluated subjectively.

There are many imaging modalities available,and they have overlapping and unique capabilities,which will continue to evolve. There are severalpossible interventions for many conditions, includ-ing surgical alternatives, and many of the interven-tions are not well standardized.

Surgical procedures for image-guided spinalsurgery (IGSS) are either anatomic, ablative, oraugmentative. The anatomic procedures correct the“cause”; ablative procedures destroy pain path-ways; and augmentative procedures modulate paintransmission. IGSS, in general, could be useful inthe spine and has already proven its value in se-lected procedures such as pedicle screw insertion.However, these systems must be made more real-time and interactive. Since a completely integratedsystem is likely to be much more useful than apartial implementation, the benefits of these sys-tems may not be fully realized until an integratedsystem is available.

It is important to emphasize that all spinesurgery is image-guided, whether there is a director indirect use of images in the operating room.Preoperative imaging is clearly the standard ofcare, and typically includes plain radiographs sup-plemented by myelography, CT, and MRI. Costconstraints discourage using multiple modalities,and so usually a single modality such as a CT orMRI scan is employed.

With regards to low-back pain, the structuralabnormalities in the spinal images themselves andthose foundin vivo are not equivalent. Both struc-tural abnormalities and low-back pain are preva-lent, but their correlation is not high. The predictivevalue of imaging in low-back pain is also imper-fect, but for different reasons. It is not clear thatinterventions that improve appearance in imagingwill reduce or eliminate symptoms; thus the imagesare not that valuable as a predictor of outcomes.Patient selection to receive a given treatment andoutcomes measurement across populations is anunderstudied area.

An article inSpine1 reported on the diagnosticaccuracy of MRI, work perception, and the psycho-social factors in identifying symptomatic disk her-niations. This prospective study involved patients(study group) with symptomatic disk herniationsand asymptomatic volunteers (control group)matched for age, sex, and work-related risk factors.

Figure 8-3: Image-guided spinal surgery process dia-gram.

208 Workshop Report

The researchers found that for a risk factor-matched group of asymptomatic individuals, discherniation had a substantially higher prevalence(76%) than previously reported in an unmatchedgroup. They concluded that MR images in individ-uals with minor disc herniations (i.e., protrusion,contained disks) are not a causal explanation ofpain because many asymptomatic subjects (63%)had comparable morphologic findings. Thus, in thisexample, imaging is limited in its diagnostic pre-dictive value for spinal abnormalities.

For evaluating IGSS system performance, thereis almost no quantitative information available. Thislack of measurements makes comparison of experi-ence from various groups difficult. It is not clear thatthere are any standardized ways of treating the samedisease between groups. Again, we see that a highdegree of variability and uncertainty exists. The out-comes are not well defined; non-technical, societalfactors strongly influence the results and may be moresignificant than surgical factors.

On the other hand, we have a tremendousamount of operational knowledge, and a good deal oftechnology is already available. We have detailedknowledge of anatomy, including an atlas in elec-tronic form. We have expertise in biomechanics, ma-terial properties, and kinematics and dynamics. Thereare many imaging modalities available, including in-traoperative systems. Surgical instruments, appli-ances, and prostheses are all well developed.

Image-guided spinal surgery (IGSS) proce-dures are classified as: decompression (largest vol-ume of cases), stabilization (high volume of proce-dures), and deformity correction (highest risk ofundesirable outcomes).

To accomplish IGSS, there are four majortasks: diagnosis, planning, intervention, and evalu-ation. Diagnosis includes detection and character-ization as well as outcomes prediction and progno-sis estimation. Planning includes a first-guessapproximation, simulation (and optimization),treatment selection, and often involves image seg-mentation and labeling. Intervention requires reg-istration, localization and orientation, and intrapro-cedural navigation with and without real-timeupdates. Finally, evaluation is done to assess im-mediate and subsequent late outcomes.

NEEDS AND OPPORTUNITIES

Image-guided spinal surgery (IGSS) is performedto satisfy unmet needs of spine surgery. However,there are many alternatives and we must identifythe best of several feasible alternatives. Spine ab-normalities involve highly prevalent disease(s)

with multiple presentations and etiologies. Thereare major costs to society when less effective andefficient treatment is used. Many treatment optionsare available, but individualization (selection andoutcomes) is less predictable. Variability is high inIGSS.

Variability in IGSS refers to patient, disease,procedure, device, and operator characteristics.Each of these entities contributes to a perceivedneed for individualization on a case-by-case basis.This is a fundamental observation that applies toIGSS.

The barriers to wider use of IGSS are the ab-sence of proven technology and economic value.From the technology standpoint, there is no clearevidence that IGSS works in a majority of cases. Thistechnology is rapidly evolving and integrated systemsare not widely available. Multicenter IGSS trials areseldom reported. From the economic and publichealth policy perspective, the principal barriers toIGSS are its added cost, which is not specificallyreimbursed in many cases. Since the methods areexperimental in many cases, the proof of benefit isabsent for most applications. There is a general lackof randomized, controlled, multicenter clinical trialsof IGSS methods and technology.

We seek to improve outcomes, both immediateand short-term, through pain reduction and by rapidreturn to work. In the long term, we seek freedomfrom chronic as well as acute pain in these patients,with lower overall disability. Restoration and mainte-nance of structural integrity (for destructive patho-logic processes, especially metastases) are importantin some cases. From the economic and public healthpolicy standpoint, treatment outcomes should be pre-dictable, while IGSS would ideally minimize compli-cations (improve safety) and lower costs (to payor,government, employer, etc.).

In general, we aspire to reduce the variability inoutcomes, reduce total costs, and assure the best pos-sible results with the fewest complications in individ-ual cases. We observe that most variability is due to afew sources and surgeons are susceptible to informa-tion overload when too much information is presentedthrough a suboptimal user interface.

From the surgeon’s perspective, we shouldoffer newer IGSS interventions such as interstitialheating, cryotherapy, and accessories such as theMammotome™ (vacuum biopsy/removal). Thesetechnologies promise better utility, convenience,and efficiency with fewer limitations (errors). IGSScan provide more certainty, thereby increasing thesurgeon’s confidence, which is consistent with fun-damental surgical precepts. IGSS promises to fa-

Workshop Report 209

cilitate accommodation of individual differences byunblinding the operator. IGSS may avoid compli-cations and complete the procedure as planned inthe largest number of cases.

STRATEGY

We defined technical requirements for further de-velopment of image-guided spinal surgery in thefollowing six categories:

1. Planning and simulation2. Guidance and localization3. Monitoring and control4. Instruments and systems5. Evaluation6. Training and career development

Our recommendations for each of these categoriesare given in the next section, followed by recom-mendations that are specific to spinal surgery.

DETAILED RECOMMENDATIONS

Planning and Simulation

● Better definition of tumor and other surgicaltarget margins or boundaries utilizing variousmedical imaging techniques is needed (corre-lating with spatially registered histology toestimate the capabilities of the various imag-ing modalities in defining the boundaries forvarious anatomical regions).

● Development of real-time image-processingtechniques, particularly rapid methods ofmodel creation, three-dimensional rendering,and accurate segmentation of anatomic tissuesfor various imaging modalities.

● Research in surgical planning and simulation,particularly trajectory planning for needleplacement, the basic surgical application oftrajectory planning today.

● Improvement, via more complex, automatedtechnologies, of current registration or imagefusion methods for different medical imagingmodalities, especially video-based and laser-scanning techniques with prospectively cre-ated models.

Guidance and Localization

● Development of flexible and untethered sen-sors to provide anatomical fiducial marks orinformation on the position of needles, cathe-ters, and surgical instruments for tracking ofinstruments or for fusing patient and imagecoordinate systems.

● Development of computational systems andalgorithms to enable “instantaneous” recon-struction, reformation, and display of the im-age data so as to enable real-time following ofa physician’s actions during a procedure (e.g.,advancing a catheter or needle).

Monitoring and Control

● Definition of the temporal resolution requiredfor various image-guided therapeutic proce-dures, taking into consideration the physicalcharacteristics of the specific imaging modal-ities and the dynamic properties of the moni-tored procedures, specifically for multislicevolumetric monitoring.

● For MRI, development of new pulse se-quences designed specifically for therapeuticapplications rather than diagnostic applica-tions. A particularly important need is thedevelopment of highly temperature-sensitivepulse sequences to enable monitoring of “heatsurgery.”

● Investigations to correlate the factors affectingenergy deposition or abstraction (e.g., pulseduration, pulse energy, and power spectrum)with histological and physiological changes inthe tissue and resulting image changes. Thepurpose of this correlation is to determinemechanisms of thermal damage and the bio-physical changes that take place during vari-ous thermal surgical procedures such as inter-stitial laser therapy, cryoablation, and high-intensity focused ultrasound treatment. Suchinvestigations need to be undertaken for var-ious anatomic regions and medical conditionsfor which such therapy might be appropriate.

● Investigation of the range of medical condi-tions amenable to treatment with minimallyinvasive techniques that are made possible byexpanded capabilities for visualization duringa procedure via the various medical imagingmodalities.

Instruments and Systems

● Although prototypical MRI systems havebeen created that provide direct and easy ac-cess to the patient, more research and devel-opment is required to further optimize thegeometric configuration of these systems.Similar requirements are appropriate for theother imaging modalities, particularly CT.

● For MRI-guided biopsy and therapy, magnet-compatible needles and other equipment using

210 Workshop Report

materials that do not cause image distortionsin a magnetic field need to be identified anddeveloped. Accessible and easy-to-use guid-ance systems are required to perform localiza-tion or biopsy of lesions detected by MRIalone.

● Development of high-performance 2D detec-tor arrays for CT and other X-ray imagingmodalities are needed, as are less-expensive2D transducer arrays for ultrasound. Appro-priate means for acquiring, reconstructing,and displaying the data are also required.

● Improved methods of inexpensively shieldingthe magnetic field to enable inexpensive ret-rofitting of existing MRI systems for use incurrent operating rooms need to be developed.

● There is a need for integrating imaging meth-ods with therapeutic procedures, includingfeedback systems between data display de-vices and image information, computer-as-sisted image-controlled surgical tools, roboticarms, and instruments.

● Creation and development of new instrumentsand tools to accomplish new tasks enabled bythe availability of image-guided therapy, es-pecially specialized surgical tools such asMRI-guided therapy.

Evaluation

● Developin vitro and in vivo models for eval-uation and measurement.

● Define relevant outcomes/effects standardsfor human applications of image guidancesystems.

● Inaugurate translational clinical trial mecha-nisms and support for biomedical imagingsciences and engineering.

● Develop and foster adoption of clear regula-tory guidelines.

Training and Career Development

● Develop multidisciplinary curricula focusingon “invention” and creativity, aimed at dis-cerning and overcoming roadblocks betweendisciplines.

● Train professionals in medical physics, ap-plied mathematics, computer science, bio-medical imaging science, and biomedical en-gineering to develop basic methods, and linktheir training with translational clinical re-search programs.

● Develop multidisciplinary training sites, and

include corporate partners and a mix of NIH,NSF, and industrial support for the implemen-tation of such programs.

RECOMMENDATIONS SPECIFIC TO SPINAL

SURGERY

Research should focus on:

● Biomechanical evaluation of structural stabil-ity, load capacity, and movement

● Bone-implant and disk-nerve interaction● Imaging in the presence of metal● Separation of scar from tumor and normal

tissue (tissue characterization)● Spine-specific atlases and instruments● Systems, techniques, and equipment designed

(in part) and validated by spine surgeons forspine surgeons

SUMMARY

In summary, recommendations for image-guidedspine surgery are made that encompass treatmentselection and optimization; real-time 3D imaging;integration of imaging and therapy for seamless,flexible systems that monitor progress on-line; andoutcomes evaluation of translational researchthrough standards, metrics, regulatory issues, andreimbursement perspectives. In addition, there arerecommendations for multidisciplinary training andthe formation of academic/industrial/governmentconsortia to work towards realizing these neededdevelopments in image-guided spine surgery.

SECTION 9: SPECIAL SESSIONSThree special sessions were presented during theWorkshop: a keynote talk titled “The OperatingRoom of the Future,” and panels on “OutcomesMeasurement for Spine Interventions,” and “Ther-apy Teams of the Future.” These sessions are sum-marized here by Barbara Hum (first session) andAudrey Kinsella (sessions 2 and 3). For more de-tails, see the full report on the Web as noted inSection 1.

● The Operating Room of the Future.Accordingto Dr. Don Long, this operating room requires thecoordination of numerous functions, many ofwhich are now available but are not as readilyaccessible as they should be. The new designshould be modular, where centrally located, high-end imaging modalities such as computed tomog-raphy (CT) and magnetic resonance imaging (MRI)will be at hand for intraoperative use. Other com-ponents of the room include:

Workshop Report 211

● an easy-to-maneuver patient transport system;● improved optics;● robotic assistance for hand support and instru-

ment control; and● enhanced virtual reality capabilities.

All of these features are much needed to increaseprecision and enhance the outcomes of surgicalprocedures performed on the spine.

● Outcomes Measurement for Spine Interven-tions. Both clinical and technical measurement ofpatient health outcomes were scrutinized in thissession led by Drs. Richard North and DanielClauw. Measures for spinal interventions havechanged over time, they note. Currently, clinicaloutcome reportage of spinal interventions may relyon patients’ assessments of “success”, and techni-cal outcome reportage on clinicians’ views.Achieving consensus among these two groups maybe an elusive goal.

Gathering adequate measurement data aboutthe procedures and patients’ response is, however,the key requirement needed for improving the qual-ity of spinal interventions. Drs. North and Clauwreviewed the generic clinical outcomes measure-ment tools that are currently in use, such as theMcGill questionnaire, and the shortcomings of eachas they apply to patients who have undergone spi-nal interventions. They recommend the followingprocedures for outcomes researchers:

● Use more than one outcome measurement toolto obtain a more complete picture of results

● Understand the differences in outcomes mea-surements (e.g., generic or body-part-specifictools) and the varied results that can be ex-pected from the different tools.

The presenters also reviewed technical outcomemeasurement procedures, noting that analysis hasbecome a good deal more complex than simplyasking the question: Did the device work? Expec-tations of “successful” interventions are higher, anddiffer from the strictly technical assessments ofpositive outcomes used in the past.

It is a challenge, the speakers concluded, tokeep current about what is or can be correctlymeasured and what data are needed to render acomplete picture of patient health outcomes from,say, a spinal intervention. Nevertheless, gatheringadequate outcomes measurement data about theprocedures and patients’ response is the key re-quirement needed not only for documenting what

was accomplished, but for improving the quality ofspinal and all clinical interventions.

● Therapy Teams of the Future. A panel ofengineers, scientists, and physicians discussed thistopic. The panelists were Drs. Heinz Lemke, JamesAnderson, Martha Gray, Richard Bucholz, Ron Ki-kinis, and Thomas Whitesides.

Creating these teams, it is thought, can en-courage multidisciplinary collaboration betweenengineers, scientists, and physicians. Developmentof these teams will be a critical step toward meetingthe challenges of exponential growth of new andsometimes complex technologies that are now be-ing experienced in health sciences and technologydevelopment.

Multidisciplinary education is necessary, thepanel members said, if programs are to preparestudents to meet the challenges posed by thisgrowth. The panelists shared information abouttheir institutions’ programs in clinical and engi-neering problem-solving tasks shared by their stu-dents. Less independent and more interactivemodes are critical in multidisciplinary programs,they noted.

The training process was described by the pan-elists in some detail, with a note that multidisciplinaryteam training is not akin to “cross training.” Engineersare not being trained to become spine surgeons, forinstance. Rather, clinicians can be taught the princi-ples of engineering, engineers the principles of med-icine. Modular curricula development is also a part ofthe program planning and delivery, so that differentlevels of sophistication can meet learners’ variedneeds on an as-needed basis.

The most important goal of these therapyteam training programs, as suggested by the partic-ipants in this session, is to broaden students’ expo-sure to multidisciplinary problems. This means fa-cilitating

● constant interaction among clinicians, scien-tists, and engineers;

● exchanges between institutions and from in-stitutions to industry (one speaker noted: “Thebest way to transfer technology is to transferpeople.”);

● group problem-solving venues.

Each of these means of exposure can, it issuggested, help each learner appreciate how realis-tic his or her expectations should or can be abouttechnologies and clinical applications; and howthey can best contribute to the multidisciplinary,problem-solving team effort.

212 Workshop Report

APPENDIX A: WORKSHOP PROGRAMThe full program can be found on the Web as notedin Section 1.

Day 1 (Sunday)

0830–1200 Overview & vision0840–0930 Clinical state-of-the-art: Dietrich

Gronemeyer, M.D.1010–1030 Spine surgery in the 21st century:

Don Long, M.D.1040–1100 Coupling information to action:

Russell Taylor, Ph.D.1140–1155 Advanced technology direction

for U.S. Army Medical R&D:Conrad Clyburn, Greg Mogel,M.D.

1200–1330: Lunch and Working Group meeting 1:Define the current status in image-guided proce-dures of the spine for your Working Group.

1415–1530 Diseases/procedures1415–1430 Interventions: John Mathis, M.D.1430–1445 Trauma: John Kostuik, M.D.1445–1500 Tumors: Elizabeth Bullitt, M.D.1500–1515 Deformity: David Polly, M.D.1500–1530 Degenerative disease: Richard

North, M.D.1600–1700 Working Group presentations by

technical leaders

1800–1900: Dinner and Working Group meeting 2:Clinical requirements. Using clinical areas identi-fied in questionnaire, discuss how image guidancemight be applied.

Day 2 (Monday)

0830–0930 Working Group presentations byclinical leaders

0930–0940 Deformable modeling: ChristosDavatzikos, Ph.D.

0940–0950 Display technology: HeinzLemke, Ph.D.

0950–1000 Accuracy issues: Neil Glossop,Ph.D.

1030–1040 Intraoperative CT: Frank Feigen-baum, M.D.

1040–1050 Open MRI for spine procedures:Eric Woodard, M.D.

1100–1145 Special session on outcomes anal-ysis: Daniel Clauw, M.D.; RichardNorth, M.D.

1200–1330: Lunch and Working Group meeting 3:Technical requirements. Based on the clinical re-quirements developed in meeting 2, define the tech-nical requirements for these applications and brain-storm potential solutions.

1415–1530 Therapy Teams of the Future (spe-cial session)

Panel chair: Heinz Lemke, Ph.D.Speakers/panel members: Martha Gray,Ph.D.; James Anderson, Ph.D.; Richard Bu-cholz, M.D.; Ron Kikinis, M.D.; Tom White-sides, M.D.

1830–2030 Group dinner: sponsor presenta-tions

NSF: Sohi Rastegar, Ph.D.NIH/NCI: Larry Clarke, Ph.D.Picker International & DePuy MotechAcroMed: Lou Arata, Ph.D.

Day 3 (Tuesday)

0830–1100 Working Groups present reports1100–1200 Summary speaker and discussion:

Michael Vannier, M.D.1200–1300 Group lunch1300–1500 Working Group leaders outline re-

ports1500 Depart

APPENDIX B: WORKSHOP PARTICIPANTSAnderson, James Ph.D. Johns Hopkins Medical InstitutionsArata, Lou Ph.D. Picker International & DePuy Motech AcroMedBarrick, Fred M.D. Inova Fairfax HospitalBascle, Benedicte Ph.D. Siemens Corporate ResearchBlezek, Dan Ph.D. Mayo ClinicBrazaitis, Michael M.D. Walter Reed Army Medical CenterBzostek, Andrew M.S. Johns Hopkins University

Workshop Report 213

APPENDIX B: WORKSHOP PARTICIPANTS (cont’d)Bucholz, Richard M.D. St. Louis UniversityBullitt, Elizabeth M.D. University of North CarolinaBurgess, James M.D. Inova Fairfax HospitalCaplan, Norman M.S. Johns Hopkins UniversityCarignan, Craig Ph.D. University of MarylandChao, Ed Ph.D. Johns Hopkins UniversityChoi, Jae Jeong M.S. Georgetown University Medical CenterClarke, Larry Ph.D. National Institutes of HealthClauw, Daniel M.D. Georgetown University Medical CenterCleary, Kevin Ph.D. Georgetown University Medical CenterClyburn, Conrad B.S. U.S. ArmyDavatzikos, Christos Ph.D. Johns Hopkins UniversityDeli, Martin B.S. Witten/Herdecke UniversityDevey, Gilbert B.S. Georgetown University Medical CenterDuerk, Jeff Ph.D. Case Western Reserve UniversityFreedman, Matthew M.D. Georgetown University Medical CenterGalloway, Robert Ph.D. Vanderbilt UniversityGlossop, Neil Ph.D. Traxtal TechnologiesGoldberg, Randy M.S. Johns Hopkins UniversityGraham, Sarah M.S. Johns Hopkins UniversityGray, Martha Ph.D. Massachusetts Institute of TechnologyGregerson, Gene M.S. Visualization Technology, Inc.Gronemeyer, Dietrich M.D. Witten/Herdecke UniversityHata, Nobuhiko Ph.D. Brigham and Women’s HospitalHerman, William B.S. Food and Drug AdministrationHiggins, Gerald Ph.D. Ciemed TechnologiesHum, Barbara M.D. Georgetown University Medical CenterKikinis, Ron M.D. Brigham and Women’s HospitalKim, Yongmin Ph.D. University of WashingtonKimia, Ben Ph.D. Brown UniversityKostuik, John M.D. Johns Hopkins Medical InstitutionsLangrana, Noshir Ph.D. Rutgers UniversityLathan, Corinna Ph.D. Catholic University of AmericaLemke, Heinz Ph.D. Technical University of BerlinLevy, Elliot M.D. Georgetown University Medical CenterLindisch, David R.T. Georgetown University Medical CenterLiu, Alan Ph.D. National Institutes of HealthLiu, Yanxi Ph.D. Carnegie Mellon UniversityLoser, Michael Ph.D. Siemens Medical EngineeringLoew, Murray Ph.D. George Washington UniversityLong, Don M.D. Johns Hopkins Medical InstitutionsMathis, John M.D. Lewis-Gale Medical CenterMogel, Greg M.D. U.S. ArmyMun, Seong K. Ph.D. Georgetown University Medical CenterMurphy, Mike Ph.D. Louisiana State UniversityNavab, Nassir Ph.D. Siemens Corporate ResearchNorth, Richard M.D. Johns Hopkins Medical InstitutionsPeshkin, Michael Ph.D. Northwestern UniversityPolly, David M.D. Walter ReedRajpal, Monish M.S. Johns Hopkins UniversityRampersaud, Y. Raja M.D. University of TorontoReinig, Karl Ph.D. University of ColoradoSieber, Ann R.N. Johns Hopkins Medical InstitutionsShahidi, Ramin Ph.D. Stanford UniversityShin, Yeong Gil Ph.D. Seoul National UniversityStaab, Ed M.D. National Institutes of HealthTaylor, Russell Ph.D. Johns Hopkins UniversityTendick, Frank Ph.D. University of California, San FranciscoThakor, Nitish Ph.D. Johns Hopkins UniversityTraynor, Laura B.S. University of Utah

214 Workshop Report

APPENDIX C: REPORTBIBLIOGRAPHYNote: A list of approximately 85 references forrelated reading as suggested by the workshop par-ticipants is available on the Web site.

1. Boos N, Rieder R, et al. The diagnostic accuracy ofmagnetic resonance imaging, work perception, andpsychosocial factors in identifying symptomatic discherniations. Spine 1995;20(24):2613–2625.

2. Bucholz RD. Advances in computer aided surgery. In:Lemke HU, Vannier MW, Inamura K, Farman AG,editors: Computer Assisted Radiology and Surgery —Proceedings of the 12th International Symposium andExhibition (CAR’98), June 1998, Tokyo, Japan. Ex-cerpta Medica International Congress Series 1165.Amsterdam: Elsevier, 1998. p 577–582.

3. Deyo RA. Low-back pain. Sci Am 1998 (August);279(2):48–53.

4. DiGioia A, Kanade T, Wells P, editors. Second Inter-national Workshop Robotics and Computer AssistedMedical Interventions. June 23–26, 1996, Bristol, En-gland. Also in: Simon DA, Morgan FM et al., editors.Excerpts from the Final Report for the Second Inter-national Workshop on Robotics and ComputerAssisted Medical Interventions, June 23-26, 1996,Bristol, England. Comput Aid Surg 1997;2(2):69–101.

5. Jolesz FA, Shtern F. The operating room of the future:Report of the National Cancer Institute Workshop,Imaging-Guided Stereotactic Tumor Diagnosis andTreatment. Invest Radiol 1992;27 (4):326–328.

6. Taylor RH, Bekey GA, editors. NSF Workshop onComputer-Assisted Surgery, 1993, Washington, DC.

7. Viergever MA. Image guidance of therapy. IEEETrans Med Imaging 1998;17(5):669–671.

8. Wipf JS, Deyo R. Low back pain. Med Clin North Am1995;79(2):231–246.

APPENDIX B: WORKSHOP PARTICIPANTS (cont’d)Vannier, Michael M.D. University of IowaWang, Joseph Ph.D. Catholic University of AmericaWatson, Vance M.D. Georgetown University Medical CenterWells, William Ph.D. Harvard Medical SchoolWhitesides, Thomas M.D. Emory UniversityWoodard, Eric M.D. Brigham and Women’s HospitalYoo, Terry Ph.D. National Library of MedicineYun, David Ph.D. University of HawaiiZeng, Jianchao Ph.D. Georgetown University Medical CenterZheng, Qinfen Ph.D. University of MarylandZinreich, S. James M.D. Johns Hopkins Medical Institutions

Workshop Report 215