1 / 22

Registration of functional PET and structural MR images

Registration of functional PET and structural MR images. PVEOut satellite meeting Budapest, June 11 th 2004 Peter Willendrup & Claus Svarer Neurobiology Research Unit Copenhagen. Registration needed PVEOut.

asha
Download Presentation

Registration of functional PET and structural MR images

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Registration of functional PET and structural MR images PVEOut satellite meeting Budapest, June 11th 2004 Peter Willendrup & Claus Svarer Neurobiology Research Unit Copenhagen

  2. Registration needed PVEOut • Structural MR and functional PET image has to be registered/aligned as the structural information is applied to each voxel in the functional image • As image are coming from same subject only a rigid 6 parameter transformation has to be estimated: • 3 translations (along X, Y and Z axis) • 3 rotations (around X, Y and Z axis) NRU, 2004

  3. What automatic methods are available? • West J, Fitzpatrick JM, Dawant BM, et al. • Headmounted fiducials serves as ``Gold standard'' coregistration between the modalities (MR/CT/PET). • Coregistration parameters are kept for reference, and fiducials are removed from the datasets and replaced by artificial noise. • Methods are tested ``blindly'' - no knowledge of the Gold standard answer. AIR SPM NRU, 2004

  4. Why are the automatic approaches not always a good idea? • These methods are very well suited for registration of images where: • There in the PET image is an equal uptake in all brain regions • There is no inhomogenity variation in the MR images • This is not the case for all receptor PET images, e.g. 5-HT2A altanserin PET images where there are very limited uptake in Cerebellum Limited uptake NRU, 2004

  5. What manual methods have been proposed? Many different approches exist in the litterature • Landmark based: "Graphics applied to medical image registration", G. Q. Maguire, Jr., M. E. Noz, H. Rusinek, et al., Comput Graph Appl, 1991, vol. 11, pp. 20-29. • Surface based: "Accurate three-dimensional registration of CT, PET, and/or MR images of the brain", C. A. Pelizzari, G. T. Y. Chen, D.R. Spelbring, R. R. Wechselbaum, and C-T. Chen, J Comput AssistTomogr, 1989, vol. 13, pp. 20-26. • Image overlay: "Quantitative Comparison of Automatic and Interactive Methods for MRI-SPECT Image Registration of the Brain Based on 3-Dimensional Calculation of Error ”, Pfluger T, Vollmar C et al.: J Nucl Med 2000; 41:1823-1829 • Voxel based: "MRI-PET registration with automated algorithm", R. P. Woods, J. C. Mazziotta, and S. R. Cherry, J Comput Assist Tomogr, 1993, vol. 17, pp. 536-546. NRU, 2004

  6. MARSMultiple Algorithms for Registration of Scans • Modular design • The problem of coregistration can be divided into subtasks • Data selection • Registration • Visualisation / Inspection • Parameter I/O • Reslicing / Re-Interpolation • All subtasks realised by ‘plugins’ - easy inclusion of alternative method • Different registration approaches benefit from shared code NRU, 2004

  7. Main program MARS • This is now included in pvelab NRU, 2004

  8. MARS • Subtask modules • Registration • Interface to Air 5.0 - Roger P. Woods • Interface to SPM 2 - J. Ashburner et. al. • IIO (Interactive Image Overlay) - NRU * • IPS (Interactive Point Selection) - NRU * • Visualisation • Inspect (NRU visualisation program) * • Asterisk-marked will be further explained NRU, 2004

  9. Registration 1: Interactive Image Overlay NRU, 2004

  10. Registration 1: Interactive Image Overlay Translation and rotation of overlay image and surface by keyboard commands NRU, 2004

  11. Registration 1: Interactive Image Overlay NRU, 2004

  12. Registration 2: Interactive Point Selection NRU, 2004

  13. Registration 2: Interactive Point Selection NRU, 2004

  14. Inspection of registration Overlay Side by side NRU, 2004

  15. Evaluation study: Setup • Images (5 subjects) • T1 weighted MR images (MPRAGE) • 18F-Altanserin 5HT-2A receptor images • Simulated PET images • Evaluation by 7 volunteers • 3 rounds of MR / Altanserin registration • 1 round of MR / Simulated PET registration • Registration order randomised • Max. one ‘round’ of registrations pr. day • Images also registered using SPM99 and Air 3.0 NRU, 2004

  16. Evaluation study: Simulated PET • Simulated PET datasets • Good: known registration parameters • Bad: “easy” for cost fct. Based methods NRU, 2004

  17. Evaluation study: Altanserin PET • Altanserin PET images • Bad: Lack of gold standard registration method • Good: Real world ‘limited uptake’ images NRU, 2004

  18. Evaluation study • Error measure - Euclidean distance between transformation endpoints • Evaluated for 1% evenly distributed brain voxels. • Mean and std. dev. calculated • Mean transformation realized by 6-parameter estimation to mean of transformed voxels MR PET NRU, 2004

  19. Evaluation study: Simulated PET NRU, 2004

  20. Evaluation study: Altanserin PET NRU, 2004

  21. Evaluation study: Result SPM Air Mean manual No Altanserin binding should be seen in Cerebellum, Rotation problem? Too little binding in Altanserin image, Translation problem? NRU, 2004

  22. Two manual co-registration methods and the interface to two automatic methods have been implemented and incorporated in the PVEOut SW package (pvelab). Four registration methods are included: Interface to SPM 2 (J. Ashburner et. al.) Interface to Air (R. Woods) IIO (NRU) IPS (NRU) For FDG/flow type images, SPM and Air are preferred, with reported errors in the range 2-3 mm. For neuroreceptor type images, with limited binding in areas of the brain, the manual methods can be used and possibly preferred. Measured errors: Registration: Conclusion Simulated images F18-Altanserin images NRU, 2004

More Related