1

Image Registration:The Challenge for QA Centers

Kenneth Ulin, Ph.D.Marcia M. Urie, Ph.D.

Quality Assurance Review Center

Types of 3D Imaging:CTMRIPET

Imaging Studies Used for:

Staging

Protocol Eligibility

Target Volume Definition

Adaptive Radiation Therapy

Treatment Delivery

Assessment of Response

Outcome Analysis

Copyright ©Radiological Society of North America, 2004

Kapoor , V. et al. Radiographics 2004;24:523-543

QA for Multi-Institutional StudiesRequiring Registration of Different Imaging Sets

Two Approaches to QA:

Require a credentialing exercise to ensure that an institution has the tools and expertise to adequately perform the image registration required by the protocol.

And/or

Verification of the image registration for individual patients on protocol.

QARC: Image Fusio n Benchmark(Developed For COG Low Grade Glioma Protocol)

MR –CT for Target Delineation

2

“fuse” DICOMCT & MR

scans

QARC FUSION Bench mark

Display on CT

QARC FUSION Benchmark

Outline target on2 MR slices

Report the centerof the target andthe center of thisrod on the mostinferior CT slice.

QARC FUSION Bench mark

Software System Type of matching Software System Type of matching

BrainSc an automatic Phili ps Pinnacle automatic

manual & auto manual & auto

match points manual

Corvus match points Pinnacle/Syntegra automatic

manual & auto

Varian Eclipse automatic Plato match points

manual & auto PLUNC manual

match points

CMS Focal automatic Radionics XKnife match points

manual & auto

In-House manual Oncentra MasterPlan automatic

QARC FUSION Benchmark

Results from 11 software systems

3

Acceptability crit eria

(from first 17 submissions): 3 mm

QARC FUSION Bench mark

True target center: average of all submissions.

The uncertainty (one standard deviation of the mean): 0.2 mm

Average error in the position of the center of the target: 1.4 mm

One standard deviation in the displacement of the target center from thetrue target center: 1.7 mm

QARC FUSION Benchmark

Erro r Distribution

0

2

4

6

8

10

12

14

0.0 - 0.5 0.6 - 1.0 1.1 - 1.5 1.6 - 2.0 2.1 - 2.5 2.6 - 3.0 3.1 - 3.5

Error Range (mm)

No

.in

each

ran

ge

QARC FUSION Bench mark

Registration Method # benchmarks Average Error (mm)

Manual 11 1.1

Automatic 27 1.6

Manual & Automatic 8 1.3

Match Points 5 1.4

Overall 51 1.4

51 submissions from 45 institutions

QARC FUSION Benchmark

Results:

~80% approved on first submission

15 by dosimetrists

35 by physicists

1 by radiation oncologist

4

QARC FUSION Benchmark

Conclusion: Studies requiring registrationof MR and CT for target delineation needto add 2-3 mm to PTVs and PRVs toaccount for registration uncertainties.

Credentialing of institutions for use of image registration in a clinical trial can be best accomplished by:

20%

20%

20%

20%

20%

10

A. A questionnaire that asks detailed questions about the institution’s software and procedures.

B. Distribution of two scan sets of a specially designed phantom, for which the correct transformation matrix is known.

C. Re-registering the image sets for the first patient enrolled on thestudy.

D. Distribution of two scan sets of an actual patient, for which the correct transformation matrix is approximately known.

E. Requiring the involvement of radiation oncologist, physicist, and dosimetrist in the credentialing exercise.

Explanation:A. A questionnaire does not test ability to perform the image

fusion.B. A phantom test is too easy.C. Re-registering the image sets for the first patient enrolled

on the study is labor intensive, and the correct answer is not known.

E. Involvement of radiation oncologist, physicist, and dosimetrist in the process is desirable, but is not a credentialing test and does not usually happen in practice.

Answer: (D) Credentialing of institutions for use of image registration in a clinical trial can be best accomplished by distribution of two scan sets of an actual patient, for which the correct transformation matrix is approximately known.

PET & PET/CTPET/CT

Kostakoglu, L. et al. Radiographics 2004;24:1411-1431

Copyright ©Radiological Society f North America, 2004

DICOMcoordinatesregistrationbetweenPET and CTscans

5

Robinson et al.: Copper fiducial markers for CT and PETMedical Physics, Vol . 31, No. 9, September 2004

PET CTPET CT

Fiducial markers : mark alignment locations for PET and CT study

Lavely et al.: Validation of CT-PET image regis trationMedical Phys ics, Vol. 31, No. 5,2004

Yellow4.0–6.0 mm

Green2.0–4.0 mm

Magenta8.0 mm

PET CTPET CT Using fiducial markers

Copyright ©Radiological Society of North America, 2004

Kostakoglu, L. et al. Radiogr aphics 2004;24:1411-1431

PETPET & CTCT

Clinical trials (and QA centers) allowing subjective expertise ofradiation oncologists for correlation of PET with CT planning scans.

SUV calculation:

DICOM standard does not include all required informationfor SUV calculations,

so recalculation by QA centers is difficult.

Quantitative PETSUV calculations (Standardized Uptake Values):

great interest in clinical trials to assess

staging

response

outcome

6

Image Guided Adaptive Therapy

Tomotherapy

MV CT

kV CT

Cyberknife

ultrasound

Challenge for QA Centers ofImage Guided Adaptive Therapy

Planning CT MV CT

Planning CT kV CTon boardimaging

Challenge for QA Centers ofImage Guided Adaptive Therapy

Challenge for QA Centers ofImage Guided Adaptive Therapy

Planning CT DRRon-board plain images

7

Challenge for QA Centers ofImage Guided Adaptive Therapy

CT UltrasoundCost-benefit of individual patient review must be weighed relative tothe time & labor to:

gather

transmit

organize

archive

review

Challenge for QA Centers ofImage Guided Adaptive Therapy

Currently QA centers are concentrating on verifying theaccuracy of the entire process for each institution.

ATC (Advanced Technology Consortium) isdeveloping credentialing criteria

Challenge for QA Centers ofImage Guided Adaptive Therapy

We wish to acknowledge the companies that have providedmany of the images shown in this presentation.

QA for Multi-Institutional StudiesRequiring Registration of Different

Imaging Sets

Challenge

verifying institutions’ systems and capabilities

(credentialing)

verifying individual protocol patient’s treatment

8

ReferencesReferences

�� Phantom validation of coPhantom validation of co--registration of PET and registration of PET and CT for imageCT for image--guided radiotherapy, W. C. guided radiotherapy, W. C. LavelyLavelyet al, Med. Phys. 31(5):1083et al, Med. Phys. 31(5):1083--1092, 2004.1092, 2004.

�� Image registration and data fusion in radiation Image registration and data fusion in radiation therapy, M. L. Kessler, BJR 79:S99therapy, M. L. Kessler, BJR 79:S99--S108, 2006.S108, 2006.

�� Results of a MultiResults of a Multi--Institutional Benchmark Test Institutional Benchmark Test for Cranial CT/MR Image Registration, K. for Cranial CT/MR Image Registration, K. UlinUlinand M. and M. UrieUrie, Med. Phys. 33(6): 2048, 2006., Med. Phys. 33(6): 2048, 2006.