AbstractPurpose Facet syndrome is a condition that may cause 15–45 % of chronic lower back pain. It is commonly diagnosedand treated using facet joint injections. This needle techniquedemands high accuracy, and ultrasound (US) is a potentiallyuseful modality to guide the needle. US-guided injections,however, require physicians to interpret 2-D sonographicimages while simultaneously manipulating an US probe andneedle. Therefore, US-guidance for facet joint injectionsneeds advanced training methodologies that will equip physi-cians with the requisite skills.Methods We used Perk Tutor—an augmented reality train-ing system for US-guided needle insertions—in a configu-ration for percutaneous procedures of the lumbar spine. Ina pilot study of 26 pre-medical undergraduate students, weevaluated the efficacy of Perk Tutor training compared to tra-ditional training.Results The Perk Tutor Trained group, which had accessto Perk Tutor during training, had a mean success rate of61.5 %, while the Control group, which received traditionaltraining, had a mean success rate of 38.5 % (p = 0.031).No significant differences in procedure times or needle path
E. Moult and M. Welch were supported by the NSERC USRAprogram. G. Fichtinger was supported as a Cancer Care OntarioResearch Chair. T. Ungi was supported as an Ontario Ministry ofResearch and Innovation Postdoctoral Fellow.
E. MoultDepartment of Electrical and Computer Engineering,Queen’s University, Kingston, ON, Canada
T. Ungi (B) · M. Welch · J. Lu · G. FichtingerSchool of Computing, Queen’s University, Kingston, ON, Canadae-mail: [email protected]; [email protected]
R. C. McGrawSchool of Medicine, Queen’s University, Kingston, ON, Canada
lengths were observed between the two groups.Conclusions The results of this pilot study suggest that PerkTutor provides an improved training environment for US-guided facet joint injections on a synthetic model.
The clinical motivation for this work is diagnosis and treat-ment of lower back pain emanating from the zygapophyseal,or facet, joints, which are synovial joints linking the articularprocesses of adjacent vertebrae. Injury and degeneration offacet joints is estimated to be the underlying cause of 15–45 %of chronic lower back pain [2]. As discussed in [1], diagnosisof lower back pain through traditional physical examinationsand imaging studies is unreliable. Filling this void, facet jointinjections, which involve the injection of anesthetic into thejoint, have become a commonly used diagnostic tool. Theunderpinning rationale of the procedure is that delivery ofanesthetic to a pain-causing joint will result in a temporarycessation of pain, while delivery of anesthetic to a healthyjoint will leave pain levels unchanged [12]. Thus, the proce-dure allows physicians to determine whether or not a patient’spain is caused by damage to a particular facet joint.
Accurate needle tip placement is of utmost importanceto the success of a facet joint injection. As such, computedtomography (CT) and fluoroscopy imaging are often used toguide needle insertion; however, more recently ultrasound(US) has also been shown to be a promising modality forguidance [6,7]. US is a particularly attractive alternative toCT and fluoroscopy because it is affordable, time efficient,
123
832 Int J CARS (2013) 8:831–836
and obviates radiation risks. There is, however, significantdifficulty associated with interpreting sonographic imageswhile simultaneously handling the US probe and needle; thisdifficulty results in a long learning curve as discussed in [3].The adoption of US as a guidance modality for facet jointinjections is further hindered by the fact that many physiciansare unfamiliar interpreting sonographic images of the spine.Taken together, these factors motivate the need for trainingmethods that will equip physicians with the necessary skillsfor US-guided facet joint injections. Addressing this need, wesuggest that augmented reality (AR) technology can be lever-aged to provide visualization options that will greatly facil-itate the process of learning US-guided facet join injections.
Prior work using AR for US-guided facet joint injections
Clarke et al. [4] and Moore et al. [11] described an AR systemthat allows US slices and a needle model to be overlaid ontop of a CT spine model. Although the system was not usedfor training, it was shown to be a useful guidance tool forphysicians during facet joint injections.
Ungi et al. [14] outlined a navigation system for US-guided facet joint injections using tracked US snapshots. Intheir set-up, the positions and orientations of the US trans-ducer and needle were electromagnetically tracked, allowingthese two tools to be displayed in a virtual 3-D scene. Thesystem facilitated navigation by allowing users to acquireUS snapshots of the facet joints, mark needle entry and tar-get locations on these snapshots and then use these landmarksto guide the insertion. The system was presented as a navi-gation tool, rather than a training system, and was shown tosignificantly improve needle placement.
Contribution
We configured Perk Tutor for US-guided facet joint injectiontraining. Perk Tutor is an open-source training platform forUS-guided needle insertions, with a strong focus on being re-configurable, allowing it to be used as a trainer for multipleprocedures [15]. Our Perk Tutor configuration is designedto encourage learning by allowing trainees to visualize thepositions of tools (needle and US probe) with respect to thelive ultrasound sequence and the vertebral column. To inves-tigate the effectiveness of Perk Tutor as a training tool forUS-guided facet joint injections, we performed a pilot studyon 26 pre-medical undergraduate students.
Materials and methods
A basic overview of the components of Perk Tutor is shown inFig. 1. In the sections below, we outline the key components
Fig. 1 Schematic showing the basic hardware and software compo-nents of Perk Tutor
of the system as they exist in our Perk Tutor configuration.The pilot study’s experimental protocol is also outlined.
Phantom
Perk Tutor makes use of a synthetic phantom of the vertebralcolumn and surrounding tissues. Using a synthetic phantomallows users to learn the procedure without putting patients atrisk; synthetic phantoms also have advantages compared tocadavers and animal models in that they lower cost and avoidethical concerns. The phantom used in Perk Tutor is com-prised of a plastic model of the vertebral column suspended inan opaque tissue-like polyvinyl chloride-based plastisol gel(Fig. 2b, c); the properties of the plastisol gel were adjustedso as to give the needle insertion a feeling similar to realtissue. The gel is covered with a rubber sheet imitating theskin layer. The model of the vertebral column was created byfirst segmenting the L2–L5 vertebrae of a human-patient CTand then printing this segmentation using a rapid prototyp-ing machine. The vertebral column is physically attached toa base and attached to the base is a vertical post to which anelectromagnetic (EM) reference sensor is fixed. This refer-ence sensor is used for tracking purposes as discussed in thesubsequent section. Additionally, the phantom has six fidu-cial divots allowing for coordinate registration, a process thatis discussed later.
Hardware
Perk Tutor enables visualization of the tool positions andorientations with respect to the US sequence and vertebral
123
Int J CARS (2013) 8:831–836 833
Fig. 2 a A section of the vertebral column model with arrows point-ing to the gaps between the articular processes, where the facet jointswould be situated. These gaps are the insertion targets. b The vertebralcolumn model in its bucket before being immersed in gel. c The Perk
Tutor phantom, consisting of the vertebral column model immersed ingel and covered with a rubber skin; here, the rubber skin is rolled backto reveal the gel. Also visible is the reference sensor fixed in to its postin the top right corner
column model; to allow for simultaneous tool tracking andimage acquisition, the SonixTouchTM US system was used inconjunction with the SonixGPSTM extension. The SonixGPSconsists of a DriveBay EM tracker (by AscensionTM) and amoveable arm, which holds the EM transmitter. In all, threetools are tracked by Perk Tutor: the US probe (UltrasonixL14-5/38 linear transducer), the needle, and a reference sen-sor. The US probe’s EM sensor is embedded in the probeitself, the needle’s EM sensor is contained within the stylet,and the reference sensor is secured to the post on the phan-tom’s body (Fig. 2c). The reference sensor is a necessary partof the tracking configuration as it provides the common coor-dinate system in which the other two tools—the US probeand needle—are tracked.
Software
The open-source software package PLUS, which is run on theUS machine, is used by Perk Tutor to acquire US image dataand track the US probe and needle [8]. 3D Slicer, anotheropen-source software package, is run on a personal com-puter (PC) and is used to visualize the virtual models of thespine phantom and tracked tools [5]. PLUS and 3D Slicerare connected through the OpenIGTLink network protocol[13]. This connection allows PLUS to send 3D Slicer datadescribing the positions and orientations of the needle andthe US probe with respect to the reference sensor. PLUS alsosends 3D Slicer a stream of US images. Using the VolumeRes-liceDriver 3D Slicer module, this arrangement allows the vir-tual tracked tool models—the US Probe and needle—and US
stream to be overlaid on top of the virtual vertebral columnmodel (Fig. 3a). In addition to the VolumeResliceDriver mod-ule, two other custom written 3D Slicer modules are used:the Transform Recorder module, which allows tool trajec-tories to be recorded, and the PerkEvaluator module, whichallows recorded tool trajectories to be analyzed. The specificfunctionalities that the PerkEvaluator module provides arediscussed in the “Data analysis method” section of this paper.
System registration
To show the US images and virtual needle relative to the vir-tual model of the vertebral column, the virtual model must beregistered to the physical model. The registration proceeds asfollows: (1) The (tracked) needle tip is touched to a sequencefiducial divots on the physical model, and at each divot, the3-D position is recorded to form the physical fiducial set,(2) the corresponding sequence of fiducial divots is selectedon the virtual model to form the virtual fiducial set, and (3)the two fiducial sets—physical and virtual—are rigidly reg-istered, using the paired-point matching technique, to obtainthe transformation that relates the physical and virtual mod-els. It should be noted that during this process the referencesensor is secured in its slot on the physical model of thevertebral column (Fig. 2c).
Participants
Twenty-six pre-medical undergraduate students, with noprior needle insertion experience, participated in the study.
123
834 Int J CARS (2013) 8:831–836
Fig. 3 a A snapshot of Perk Tutor’s 3D Slicer scene with an enlargedcopy of the US image shown in the upper right corner. The needle andUS plane are shown overlaid on top of the virtual model of the vertebral
column. b A trainee being instructed how to perform US-guided facetjoint injections using Perk Tutor
Informed consent was obtained from all subjects, and ethicsapproval for the study was granted by the Queen’s UniversityHealth Sciences and Affiliated Teaching Hospitals ResearchEthics Board.
Experimental protocol
The subjects were randomized into a Control group anda Perk Tutor Trained group. Both the Control group andthe Perk Tutor Trained group received a 10 min introduc-tory lesson detailing the relevant anatomy, explaining theprocedure, and introducing US image interpretation andneedle handling techniques. During this introductory les-son, subjects were also shown a successful US-guided facetjoint injection on the Perk Tutor phantom. After the les-son, both the Control group and the Perk Tutor Trainedgroup had a 10 min practice session in which they wereable to practice US-guided facet joint injections on thephantom; questions were permitted during the practice ses-sion. As no anesthetic was used in the sessions, subjectswere only asked to position the needle as if they weregoing to perform an injection. During the practice session,the Control group only had access to the US machine,while the Perk Tutor Trained group had access to the USmachine and Perk Tutor. After the practice session, bothgroups were asked to perform US-guided facet joint injec-tions on the L3–L4 right, L3–L4 left, L4–L5 right andL4–L5 left facet joints of the Perk Tutor phantom. Eachevaluation began with the subject holding the tools abovethe phantom and ended when subjects indicated that theybelieved the needle was correctly positioned for injectionof the anesthetic. No feedback was given during testing
and neither group had access to Perk Tutor. The nee-dle trajectories were tracked during testing for subsequentanalysis.
Data analysis method
The needle trajectories from each trial were analyzed by aphysician who was blinded as to which group the given tra-jectory belonged. An insertion was classified as a success ifthe needle tip was positioned in—or right on the edge of—the gap between the articular processes (Fig. 2a). The successrates of the two groups were compared using the chi-squaretest.
In addition to trials being classified as a success or a fail-ure, the following four metrics were derived from the needletrajectories using the PerkEvaluator module.
1. Total time: time from the beginning of the trial until thesubject indicates that the needle is correctly positionedfor injection of anesthetic.
2. Total path: total distance that the needle travels insideand outside of the phantom body.
3. Time inside: time that the needle tip spends within thephantom body.
4. Path inside: distance that the needle tip travels within thephantom body.
Using the D’Agostino–Pearson normality test, the met-ric data were determined to be not normally distributed. Assuch, the Mann–Whitney U test was used to compare the twogroups, with the alternative hypothesis being that the medianmetric values of the two groups differed.
123
Int J CARS (2013) 8:831–836 835
Table 1 Needle metrics
Metric Perk Tutor Trained Control p
Total time (s) 66 ± 6 73 ± 8 0.70
Total path (mm) 1366 ± 185 1803 ± 290 0.93
Time inside (s) 296 ± 45 243 ± 28 0.97
Path inside (mm) 42 ± 16 25 ± 3 0.91
Results and discussion
The Perk Tutor Trained group (n = 13) had a mean successrate of 61.5 %, while the Control group (n = 13) had a meansuccess rate of 38.5 %. The success rate of the Perk TutorTrained group was significantly higher than the success rateof the Control group (p = 0.031). The tool trajectory metricsare shown in Table 1, expressed as mean ± standard error ofthe mean (SEM). Neither the needle path lengths nor theinsertion times showed a statistically significant differencebetween the two groups.
We believe that the primary reason why Perk Tutor is ben-eficial to trainees is that its 3-D needle-US-spine overlay(Fig. 3a) helps trainees to build connections between the 2-DUS images and the 3-D anatomy of the spine. In particular,we suspect that the overlay assists trainees in learning to con-struct 3-D mental models of the spine from 2-D US images.Evidence of this was observed during the training and test-ing sessions. Specifically, we observed that the Perk TutorTrained group was more competent at identifying whether ornot the needle tip was correctly positioned in the facet joint,a task that requires determining the 3-D needle tip positionwith respect to spine anatomy. In contrast, we noticed thatthe Control group was less likely to identify incorrectly posi-tioned needles. Furthermore, the types of errors that the Con-trol group made betrayed an inability to correctly relate 3-Dneedle position to 2-D US images. The most notable of sucherrors occurred when trainees would insert the needle too far,passing the tip in one side of the facet joint and out the other.This mistake suggests that the trainee is able to correctly con-strain the needle position with respect to two dimensions, butis unable to correctly position the needle with respect to thethird dimension—in this case, the depth of insertion.
This study is not without its limitations, though. Withrespect to the subject pool, it should be noted that Perk Tutoris ultimately aimed at training medical students and not pre-medical undergraduate students—who were the subjects ofthis study. While this is indeed true, we believe that the resultsof the study are still valid because Perk Tutor was designedfor trainees who have little to no experience with US-guidedinsertions, a criteria that pre-medical undergraduate studentsfit well. The limited number of subjects in the study is anotherdrawback of the study; however, while we agree that addi-tional testing is required to further strengthen results, we
believe that the number of subjects was sufficient to showthe beneficial effects of Perk Tutor training. To address theabove concerns, we are currently in the process of organizinga larger scale experiment with medical student subjects.
In addition to the above considerations, it could also beargued that the improved performance of the Perk TutorTrained group on the synthetic phantom may not necessar-ily translate into improved performance in the clinical set-ting. Indeed, this is an ever present question associated withsimulation-based training. While it is out of the scope of thispaper to refute or confirm the efficacy of simulation-basedtraining, achievement of competence requires practice, andfor reasons of patient safety and learner comfort, early stagepractice is best done in a simulated environment [9,10]. Thisfact has resulted in simulation-based training being increas-ingly relied upon in medical education programs and is initself motivation to develop improved methodologies—suchas Perk Tutor—for simulation-based training. One precau-tionary measure that we will adopt in future studies is touse separate models for training and testing. This will bedone with the aim of preventing users from “memorizing”the model; however, we expect that if anything this will onlyskew the results in favor of the Perk Tutor Trained groupfurther.
Conclusions
In this work, we present a Perk Tutor configuration designedto provide augmented reality training for US-guided facetjoint injections. A pilot study on 26 pre-medical under-graduate students was performed to evaluate the effective-ness of the training system. The group that had access toPerk Tutor performed significantly better than the groupwho did not have access to Perk Tutor, suggesting that PerkTutor’s enriched teaching environment positively influencedtrainees’ learning. In particular, we suspect that Perk Tutorimproves trainees’ ability to visualize 3-D needle position,enabling them to identify, and correct for, erroneous inser-tions.
Acknowledgments The authors wish to thank Jennifer Andrea andKim Garrison for their kind assistance in setting up the Perk Tutorsystem.
Conflict of interest The authors declare that they have no conflict ofinterest.
References
1. Boswell M, Singh V, Staats PS, Hirsch JA (2003) Accuracy ofprecision diagnostic blocks in the diagnosis of chronic spinal painof facet or zygapophysial joint origin: a systematic review. PainPhys 6:449–456
123
836 Int J CARS (2013) 8:831–836
2. Boswell M, Colson J, Sehgal N, Dunbar E, Epter R (2007) A sys-tematic review of therapeutic facet joint interventions in chronicspinal pain. Pain Phys 10:229–253
3. Chen CP, Lew HL, Tsai WC, Hung YT, Hsu CC (2011) Ultrasound-guided injection techniques for the low back and hip joint. Am JPhys Med Rehabil 90:860–867
4. Clarke C, Moore J, Wedlake C, Lee D, Ganapathy S, Salbalbal M,Wilson T, Peters T, Bainbridge D (2010) Virtual reality imagingwith real-time ultrasound guidance for facet joint injection: a proofof concept. Anesth Analg 110:1461–1463
5. Fedorov A, Beichel R, Kalpathy-Cramer J, Finet J, Fillion-RobinJC, Pujol S, Bauer C, Jennings D, Fennessy F, Sonka M, Buatti J,Aylward S, Miller JV, Pieper S, Kikinis R, (2012) 3D slicer as animage computing platform for the quantitative imaging network.Magn Reson Imaging 30:1323–1341
6. Galiano K, Obwegeser A, Bodner G, Freund M, Maurer H, Kamel-ger F, Schatzer R, Ploner F (2005) Ultrasound guidance for facetjoint injections in the lumbar spine: a computed tomography-controlled feasibility study. Anesth Analg 101:579–583
7. Galiano K, Obwegeser AA, Walch C, Schatzer R, Ploner F, Gru-ber H (2007) Ultrasound-guided versus computed tomography-controlled facet joint injections in the lumbar spine: a prospectiverandomized clinical trial. Reg Anesth Pain Med 32:317–322
8. Lasso A, Heffter T, Pinter C, Ungi T, Fichtinger G (2012) Imple-mentation of the plus open-source toolkit for translational researchof ultrasound-guided intervention systems. In: MICCAI, pp 1–12
9. Ma IW, Brindle ME, Ronksley PE, Lorenzetti DL, Sauve RS, GhaliWA (2011) Use of simulation-based education to improve out-comes of central venous catheterization: a systematic review andmeta-analysis. Acad Med 86:1137–1147
10. Magee D, Zhu Y, Ratnalingam R, Gardner P, Kessel D (2007) Anaugmented reality simulator for ultrasound guided needle place-ment training. Med Biol Eng Comput 45:957–967
11. Moore J, Clarke C, Bainbridge D, Wedlake C, Wiles A, Pace D,Peters T (2009) Image guidance for spinal facet injections usingtracked ultrasound. In: MICCAI vol 523, pp 516–523
12. Sehgal N, Dunbar EE, Shah RV, Colson J (2007) Systematic reviewof diagnostic utility of facet (zygapophysial) joint injections inchronic spinal pain: an update. Pain Phys 10:213–228
13. Tokuda J, Fischer GS, Papademetris X, Yaniv Z, Ibanez L, ChengP, Liu H, Blevins J, Arata J, Golby AJ, Kapur T, Pieper S, BurdetteEC, Fichtinger G, Tempany CM, Hata N (2009) OpenIGTLink: anopen network protocol for image-guided therapy environment. IntJ Med Robot 5:423–434
14. Ungi T, Abolmaesumi P, Jalal R, Welch M, Ayukawa I, NagpalS, Lasso A, Jaeger M, Borschneck DP, Fichtinger G, Mousavi P(2012) Spinal needle navigation by tracked ultrasound snapshots.IEEE Trans Biomed Eng 59(10):2766–2772
15. Ungi T, Sargent D, Moult E, Lasso A, Pinter C, McGraw R,Fichtinger G (2012) Perk Tutor: an open-source training platformfor ultrasound-guided needle insertions. IEEE Trans Biomed Eng59(12):3475–3481
