MIT OpenCourseWare http://ocw.mit.edu

HST.583 Functional Magnetic Resonance Imaging: Data Acquisition and AnalysisFall 2006

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.

Cerebral MR Perfusion Imaging

2006

A. G. Sorensen MD

MGH-HST Center for Biomarkers in Imaging

A. A. Martinos Center

Massachusetts General Hospital

Harvard Medical School & Massachusetts Institute of Technology

Division of Health Sciences and Technology

HST.583: Functional Magnetic Resonance Imaging: Data Acquisition and Analysis, Fall 2006 Harvard-MIT Division of Health Sciences and Technology Course Director: Dr. Randy Gollub

Why Perfusion With MRI?

• A number of neurological illnesses can be traced to abnormal blood flow–Ischemic Stroke–Cerebral Neoplasia–Many, many others, from Alzheimer’s to MS

• MRI is heavily used already• It works!

Tracer Measurement over time

Dynamic DSC Perfusion DataSeconds

MRI image removed due to copyright restrictions.See Sorensen, A. G., and P. Reimer. Cerebral MR Perfusion Imaging.Stuttgart, Germany: Thieme Medical Publishers, 2000.

DWI / PWI mismatch vs. Final Infarct

Four MRI images removed due to copyright restrictions.See Figure 2 (D, E, G, H) in Sorensen A. G., et al. Radiology 210 (1999): 519-527.

Time = BrainBetter outcomes with earlier treatmentMeta-analysis of 6 trials, 2775 patients

The ATLANTIS, ECASS, and NINDS rt-PA Study Group Investigators. "Association of outcome with early stroke treatment: pooled analysis of ATLANTIS, ECASS, and NINDS rt-PA stroke trials." Lancet 363 no. 9411 (2004): 768-774.

Courtesy Elsevier, Inc., http://www.sciencedirect.com. Used with permission.

Diffusion and Perfusion in Stroke

• DWI is now widely used• PWI is in increasing use

– No approved contrast agent– Most use some sort of timing map (e.g., time to

peak)– ASL continues its testing / development

• Desmoteplase or other innovations suggest MRI use may become more widespread

Initial core ?= DWI LesionOr maybe CT edema?

Follow-up Infarct(can be variable)

Initial PWI Lesion

The so-called ‘ischemic penumbra’ --now accepted (too widely?)

MGH / AGS

MTT ≠ Tissue at Risk

N= 90; see Schaefer et al AJNR 24 (2003): 436

• DWI is core, but PWI not penumbra when PET is used. Heiss et al Stroke 2004 35(11 Suppl 1)2671

Courtesy of Dr. Pamela Schaefer. Used with permission.

Problems (and Potential Solutions)• Monitoring Treatment?• Closing the gap between imaging and clinical outcome

– Location matters!• MTT as currently calculated may be overestimating

due to delay– Choice of AIF? (delay, dispersion)– New CBF approaches: Local AIF, circular deconvolution

• Other markers besides perfusion– Microvascular status– Vasomotion

• Imaging Time: still too much time in the scanner

Monitoring Treatment with Diffusion / Perfusion MRI

• A number of clinical trials have DWI/PWI now as endpoints

• One new treatment: hyperoxia for ischemic stroke

• Evidence that treatment with 100% oxygen at room pressure widens the therapeutic window

Example: NBO attenuates DWI abnormalities72-yr woman with right MCA stroke, treated with NBO

Baseline (13 hrs after onset)Large DWI lesionLarger MTT lesionRMCA occlusion

During NBO (after 4 hrs)DWI lesion smaller (arrows)Stable MTTStable RMCA occlusion

Post-treatment (after 24 hrs)DWI lesion growing MTT smallerStable RMCA occlusion

Singhal et al MGH

Images removed due to copyright restrictions.See figure 2 in Singhal, A. B., et al. "A Pilot Study of NormobaricOxygen Therapy in Acute Ischemic Stroke." Stroke 36 (2005): 797.

NBO: mostly

transient effects!

Images removed due to copyright restrictions.See figure 1a,c in Singhal, A. B., et al. "A Pilot Study of NormobaricOxygen Therapy in Acute Ischemic Stroke." Stroke 36 (2005): 797.

Problems (and Potential Solutions)• Monitoring Treatment?• Closing the gap between imaging and clinical outcome

– Location matters!• MTT as currently calculated may be overestimating due

to delay– Choice of AIF? (delay, dispersion)– New CBF approaches: Local AIF, circular deconvolution

• Other markers besides perfusion– Microvascular status– Vasomotion

• Imaging Time: still too much time in the scanner

• RANTTAS (Stroke 1999 30:293-298):

• ASPECTS (Lancet 2000 355:1670): r=0.56

Lesion Size Alone Correlates Moderately

Courtesy of Lippincott Williams & Wilkins. Used with permission.

Outcome measure

BI (n=150)

GOS (n=170)

NIHSS (n=131)

Mortality (n=191)

0.43 (0.28-0.58)

0.53 (0.40-0.65)

0.54 (0.41-0.68)

0.31 (0.18-0.45) 0.32 (0.16-0.49)

0.56 (0.40-0.71)

0.46 (0.29-0.64)

0.59 (0.47-0.71)

All infarcts (95% CI) Visible infarcts only (95% CI)

Figure by MIT OpenCourseWare.

Why doesn’t imaging better predict / improve outcome?

• Specificity of techniques– Not all lesions are equal

• The Real Estate factor:–Location–Location–Location

This is in part because location matters…

Day 8 NIHSS = 11Volume ~ 2 cm3

Day 8 NIHSS = 2Volume ~ 2 cm3

Men

ezes

/ A

y /

Mar

tinos

-M

GH

-HS

TAtlas-based approach markedly improves imaging-based predictions of outcome

(n=46, p<0.01)

Figure by MIT OpenCourseWare.

0

5

10

15

20

25

30

0

5

10

15

20

25

30

0 5 10 15 20 25 0 5 10 15 20 25Measured NIHSS score

Volu

me

scor

es (p

redi

ctio

ns)

Hyb

rid A

tlas s

core

s (pr

edic

tions

)

Measured NIHSS score

R=0.53 R=0.82

Problems (and Potential Solutions)• Monitoring Treatment?• Closing the gap between imaging and clinical outcome

– Location matters!• MTT as currently calculated may be overestimating

due to delay– Choice of AIF? (delay, dispersion)– New CBF approaches: Local AIF, circular deconvolution

• Other markers besides perfusion– Microvascular status– Vasomotion

• Imaging Time: still too much time in the scanner

Global versus local AIF

Lorenz / Benner / Sorensen - Martinos MGH + HST

Lorenz, C., et al. "Automated Perfusion-Weighted MRI Using Localized Arterial Input Functions.“Journal of Magnetic Resonance Imaging 24, no. 5 (2006): 1133-1139.Copyright © 2006 Wiley-Liss Inc. Reprinted with permission of John Wiley & Sons., Inc.

Global Ipsi Global Contral Local

Follow-up

Estimation of CBF vs. AIFs

Lorenz / Benner / Sorensen - Martinos MGH + HST

Estimation of CBF vs. AIFs

Lorenz, C., et al. "Automated Perfusion-Weighted MRI Using Localized Arterial Input Functions.“Journal of Magnetic Resonance Imaging 24, no. 5 (2006): 1133-1139.Copyright © 2006 Wiley-Liss Inc. Reprinted with permission of John Wiley & Sons., Inc.

Lorenz, C., et al. "Automated Perfusion-Weighted MRI Using Localized Arterial Input Functions.“Journal of Magnetic Resonance Imaging 24, no. 5 (2006): 1133-1139.Copyright © 2006 Wiley-Liss Inc. Reprinted with permission of John Wiley & Sons., Inc.

Problems (and Potential Solutions)• Monitoring Treatment? • Closing the gap between imaging and clinical outcome

– Location matters!• MTT as currently calculated may be overestimating

due to delay– Choice of AIF? (delay, dispersion)– New CBF approaches: Local AIF, circular deconvolution

• Other markers besides perfusion– Microvascular status– Vasomotion

• Imaging Time: still too much time in the scanner

MRI Measure of VasomotionShotShot--toto--shot variability of SI in MRI time series due to noise:shot variability of SI in MRI time series due to noise:

§§ NonNon--physiological (e.g., thermal, scanner)physiological (e.g., thermal, scanner)§§ Physiological (e.g., cardiac, respiratory, vasomotion)Physiological (e.g., cardiac, respiratory, vasomotion)

Time (Image #)

SI

Hz

FT

Wang / Martinos -MGH -HST

Time (Image #)

SI

Noise quantified via standard Noise quantified via standard deviation (SD) or variance deviation (SD) or variance mapsmaps

MRI Measure of Vasomotion

SD Map

No spatial distribution for non-physiological noise

®Regional differences in SD reflect physiological noiseRescaled

CSF: cardiac, respiration

Parenchyma: vasomotion

MethodsSubjects: 32 acute stroke patientsSubjects: 32 acute stroke patientsMRI (1.5 T):MRI (1.5 T): •• Acute Perfusion: <12 h, TR=1.5s, 1.7 mm resolutionAcute Perfusion: <12 h, TR=1.5s, 1.7 mm resolution

•• FollowFollow--up T2: >5 dup T2: >5 d

CBF and CBV maps calculated from PWICBF and CBV maps calculated from PWIPrePre--bolus PWI (~15 images) used to calculate SD mapsbolus PWI (~15 images) used to calculate SD maps

pre-bolus

Wang / Martinos -MGH -HST

SD Differences in Normal TissueSDSD--PWI differences seen in normal tissue regions PWI differences seen in normal tissue regions Normal white SD < normal gray SD (31 of 32 patients)Normal white SD < normal gray SD (31 of 32 patients)

DWI CBF CBV F/U T2

SD

SI

Image # Image #

Gray White

Wang / Martinos -MGH -HST

See: Wang, H. H., A. G. Sorensen, et al. J Magn Reson Imaging 27 no. 4 (2008): 866-71.

SD Differences in Ischemia

No clear boundary with which to draw outlinesNo clear boundary with which to draw outlines®® Applied CBF, CBV lesion outlines to SD mapsApplied CBF, CBV lesion outlines to SD maps

SD

SI

Image #

DWI CBF F/U T2

CBV

See: Wang, H. H., A. G. Sorensen, et al. J Magn Reson Imaging 27 no. 4 (2008): 866-71.

Problems (and Potential Solutions)• Monitoring Treatment? • Closing the gap between imaging and clinical outcome

– Location matters!• MTT as currently calculated may be overestimating

due to delay– Choice of AIF? (delay, dispersion)– New CBF approaches: Local AIF, circular deconvolution

• Other markers besides perfusion– Microvascular status– Vasomotion

• Imaging Time: still too much time in the scanner

A Key to Speeding Up Scans?

Photos removed due to copyright restrictions.See, for example, http://www.nmr.mgh.harvard.edu/~fhlin/reprints/abstracts/ismrm2005_96channel_array.pdf

Why Perfusion MRI In Cancer?

• Correlation of angiogenesis with tumor grade

• Perfusion measures capillary hemodynamics, which may in turn reflect angiogenesis

• Measure: CBV = cerebral blood volumeCBF = cerebral blood flow

The first in vivo images of tumor angiogenesis

American Journal of Roentgenology 1939 42:891-899

Ide AG, Baker NH, Warren, SL. Vascularization of the Brown Pearce rabbit epithelioma transplant as seen in the transparent ear chamber

Image removed due to copyright restrictions.

Tumor Perfusion MRI: Two basic types

• Dynamic Susceptibility Contrast MRI– Traditional PWI like in stroke– First pass of Gd, T2 or T2* imaging– CBV, CBF, plus others like MTT, permeability, etc.

• Dynamic Contrast Enhanced MRI– Newer– T1-based– Permeability

35 year old male with 3 years of intermittent headache and

intermittent speech difficulties

T2 FSE T1 post Gd

Dynamic Susceptibility Contrast MRI - using Gd as T2 agent

MRI image removed due to copyright restrictions.See Sorensen, A. G., and P. Reimer. Cerebral MR Perfusion Imaging.Stuttgart, Germany: Thieme Medical Publishers, 2000.

70% Oligodendroglioma, 30% Grade 3 astrocytoma

MRI image removed due to copyright restrictions.See Sorensen, A. G., and P. Reimer. Cerebral MR Perfusion Imaging.Stuttgart, Germany: Thieme Medical Publishers, 2000.

High Grade Glioma

Low Grade Glioma .

MRI image removed due to copyright restrictions.See Sorensen, A. G., and P. Reimer. Cerebral MR Perfusion Imaging.Stuttgart, Germany: Thieme Medical Publishers, 2000.

26 year old female with III/IV glioma

MRI image removed due to copyright restrictions.See Sorensen, A. G., and P. Reimer. Cerebral MR Perfusion Imaging.Stuttgart, Germany: Thieme Medical Publishers, 2000.

Pre XRT 6 weeks post XRT 5 months post XRT

rCBV and 18FDG over time

0.0

5.0

10.0

15.0

20.0

25.0

30.0

Sign

al

chan

ge

1.0

10.0Vessel Radius

(µm)

GE, Sim

SE, SimGE, Exp

SE, Exp

Another potential maker: Mean Vessel Radius SE EPI

0.2 mmol/kgGd-DTPA

FLASH (GE)Iron Oxide

Gradient Echo versus Spin Echo EPI: simultaneous acquisition approach

3.0 Tesla, TR 1.5s, TE for GE: 28, TE for SE: 105 0.2 mmol/l Gd-DTPA.

GE vs SE DSC

Pilocytic astrocytoma

4.4 : 1 tumor/white on

GE

vs 1.75 : 1tumor/white on

SE

GE CBV SE CBV

T1 p Gad Ratio Map

M Pescitides / Martinos - MGH -HST

DCE MRI

•Dynamic Contrast Enhanced MRI•Rapid imaging during small dose of Gd•T1 based, not T2•Attempts to map “permeability”•Analysis: analytic versus empiric (‘curvology’)•Popular in assessing drug therapy

Sample DCE Time Course

Sub-cutaneous fatCortexChoroid Plexus

Injection at about 1.5 times real speed

Images every ~5 seconds

Example Mapping Permeability (“Ktrans ”)

Ktrans: Day -5 Day -2 Day 1 Day 26(second baseline - note AutoAlign)

Day -2 Day 1 Day 26

Post Gd and Ktrans both show change; Ktrans will quantify

MGH-Martinos3T TimTrio

Vascular Normalization after Anti-angiogenic treatmentWinkler et al, Cancer Cell 6 no. 6 (2004): 553-563

Courtesy Elsevier, Inc., http://www.sciencedirect.com. Used with permission.

Mechanistic Insights from MRI?• Testing the “Vascular Normalization” hypothesis• Blocking VEGF should result in:

– Smaller vessels– Less permeability– Less edema– Decreased mass effect– Probably no survival benefit without combination therapy, though…

Summary

• Perfusion MRI: Active, and still going strong

• New developments to watch out for:• Vasomotion• Therapy monitoring (stroke, tumor)• Location

Special Thanks to:Ona Wu Timothy Reese Chris Wiggins Andreas PotthastThomas Benner Bill Copen Thierry Huisman Takashi YoshiuraSuzanne Komili Mette Wiegell Ruopeng Wang Megan SalhusRaymond Huang Tara Ullrich Dave Tuch Van Wedeen

and many, many more...

• Larry Wald, Anders Dale, Bruce Rosen, and many students, faculty, and staff of the HST Martinos Center

Acknowledgements