Decraene spie-09

702 views

Published on

Presentation by M. De Craene at SPIE Medical Imaging 2009

Published in: Technology, Health & Medicine
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
702
On SlideShare
0
From Embeds
0
Number of Embeds
194
Actions
Shares
0
Downloads
5
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Decraene spie-09

  1. 1. Non-stationary Diffeomorphic Registration: Application toEndo-Vascular TreatmentMonitoring<br />M. De Craene2,1,O. Camara1,2, B.H. Bijnens1,2,3, A.F. Frangi1,2,3<br />Center for Computational Imaging and Simulation Technologies in Biomedicine (CISTIB), Barcelona Spain.<br />1. Information and Communication Technologies Department, Universitat Pompeu Fabra, Barcelona, Spain<br />2. Networking Center on Biomedical Research - Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN<br />3. Catalan Institution for Research and Advanced Studies (ICREA).<br />
  2. 2. Context.Aneurysm Recurrence after Coiling<br />Causes<br />Compaction of the coil mass<br />Aneurysm growth<br />Related factors<br />Packing [Johnston,Kai]<br />Ratio between the volume of inserted coil and the aneurysm volume<br />Shown to be a strong predictor of aneurysm recurrence<br />Others [Cottier]<br />Size<br />Treatment during the acute phase<br />Rupture status<br />2<br />Example of “coil compaction” : DSA image [Steinman]<br />Cottier et al. Neuroradiology45, pp. 818–824, 2003.<br />Johnston et al. Stroke 39(1), pp. 120–125, 2008.<br />Kai et al. Neurosurgery 56, pp. 785–791, 2005.<br />Steinman et al. American Journal of Neuroradiology24, pp. 559–566, 2003.<br />
  3. 3. Image-based quantification of aneurysm recurrence and evolution over time<br />Objectives<br />Visualize several time points in a common frame of coordinates<br />Compute coil and aneurysm volume curves over time<br />Local deformation maps<br />3<br />t0 pre<br />t1 post<br />t2 pre<br />t1 pre<br />t0 post<br />t3 post<br />t4 pre<br />t4 post<br />t5 pre<br />
  4. 4. Characterize evolution using non-rigid registration<br />Local deformation maps<br />Where is the aneurysm growing?<br />4<br />
  5. 5. Challenges<br />Accuracy for detecting small volume changes and retreat if necessary<br />Flexibility for detecting small and large volume changes<br />Depends on time follow-up interval, aneurysm location, …<br />Invertibilityof the non-rigid mapping to ensure correctness of the volume change estimate<br />5<br />Patient 1<br />Patient 2<br />Patient 3<br />
  6. 6. Non-rigid registration and diffeomorphisms<br />Popular pairwise diffeomorphic registration schemes<br />Mainly optimize a dense velocity field <br /><ul><li>Higher computational cost, no implicit regularization as offered by FFD (except [Rueckert])
  7. 7. Simple optimization scheme based on first derivatives (except [Hernandez])
  8. 8. )</li></ul>6<br />Beg et al. Int. J. Comput. Vis. 61 (2), pp. 139–157, 2005.<br />Hernandez et al. MMBIA’07 , 2007.<br />Rueckert et al. MICCAI’06, LNCS 4191, pp. 702–709, 2006.<br />Vercauteren et al. MICCAI’07, LNCS 4792, pp. 319–326, 2007.<br />
  9. 9. LDFFD diffeomorphic non-rigid registration<br />Transformation = Concatenation of FFD transformations<br />Strong coupling between phases: the first transformation influences all subsequent time steps<br />Mutual information metric: ITK, Mattes´ implementation<br />LBFGS optimizer: ITK<br />7<br />v(x;t0)<br />v(x;t1)<br />v(x;t2)<br />v(x;t3)<br />u(x;t2)<br />time<br />For k=1:M (number of time steps)<br />
  10. 10. LDFFD diffeomorphic non-rigid registration<br />8<br />∆u(x;t2)<br />∆v(x;t0)<br />v(x;t1)<br />v(x;t2)<br />∆ metric  ∆ intensity  ∆ mapped coordinate  ∆ transformation parameter<br />Parametric Jacobian<br />Similar expression can be found in LDDMM registration [Beg] when computing variational derivative<br />Parametric Jacobian of mth transformation<br />Jacobian of all transformations posterior to m <br />Beg et al. Int. J. Comput. Vis. 61 (2), pp. 139–157, 2005.<br />
  11. 11. LDFFD diffeomorphic non-rigid registration<br />9<br />u(x;t2)<br />v(x;t0)<br />Multi-resolution scheme in the temporal dimension<br />Initiate algorithm with 2 time steps<br />In the event that any of these parameters reaches a given threshold (0.4 the spacing between control points, as proposed by [Rueckert])<br />Interrupt optimization<br />Restore last set of valid parameters<br />Break the problematic time steps using square root computation [Arsigny]<br />with<br />Told<br />Tnew<br />Tnew<br />Arsigny et al. MICCAI’ 06, LNCS 4190, pp. 924-931, 2006.<br />Rueckert et al. MICCAI’06, LNCS 4191, pp. 702–709, 2006.<br />
  12. 12. LDFFD at work<br />10<br />
  13. 13. Results: aneurysm volume changes measured by non-rigid registration<br />11<br />Rueckert et al. IEEE Transactions on Medical Imaging 18(8), pp. 712-721, 1999.<br />Rueckert et al. MICCAI’06, LNCS 4191, pp. 702–709, 2006.<br />
  14. 14. Results: patient 1, second time point<br />12<br />t1<br />t2<br />LDFFD<br />[Verc06]<br />[Rueck99]<br />[Rueck06]<br />Rueckert et al. IEEE Transactions on Medical Imaging 18(8), pp. 712-721, 1999.<br />Rueckert et al. MICCAI’06, LNCS 4191, pp. 702–709, 2006.<br />Vercauteren et al. MICCAI’07, LNCS 4792, pp. 319–326, 2007.<br />
  15. 15. Results: patient 1, second time point<br />13<br />t1<br />t2<br />LDFFD<br />[Verc06]<br />[Rueck99]<br />[Rueck06]<br />Rueckert et al. IEEE Transactions on Medical Imaging 18(8), pp. 712-721, 1999.<br />Rueckert et al. MICCAI’06, LNCS 4191, pp. 702–709, 2006.<br />Vercauteren et al. MICCAI’07, LNCS 4792, pp. 319–326, 2007.<br />
  16. 16. 14<br />Results: Jacobian distributions<br />FFD<br />LDFFD<br />Diff FFD<br />Diff. Demons<br />
  17. 17. Results: displacement fields<br />15<br />[Rueck99]<br />LDFFD<br />[Rueck06]<br /> [Verc06]<br />Rueckert et al. IEEE Transactions on Medical Imaging 18(8), pp. 712-721, 1999.<br />Rueckert et al. MICCAI’06, LNCS 4191, pp. 702–709, 2006.<br />Vercauteren et al. MICCAI’07, LNCS 4792, pp. 319–326, 2007.<br />
  18. 18. Conclusions<br />LDFFD: non-stationary non-rigid registration algorithm<br />Dynamically finds the optimal number of time steps <br />Transformation invertibility<br />Keep the dimension of the optimization problem reasonably low<br />Applicable to quantify post interventional volume changes over subsequent follow-ups<br />Future work, <br />Exploit Jacobian-based local growth maps <br />Comparison to other coil compaction predictors published in the literature<br />Extension to motion and deformation estimation from image sequences: FIMH 09, Nice.<br />16<br />
  19. 19. Acknowledgements <br />This research has been partially funded by the Industrial and Technological Development Centre (CDTI) under the CENIT Programme (CDTEAM Project) and the Integrated Project @neurIST(IST-2005-027703), which is cofinanced by the European Commission. <br />The authors wish to acknowledge ElioVivasfor the acquisition of the intra-cranial aneurysm imaging data using 3D rotational angiography at Hospital General de Catalunya, San Cugat del Valles, Spain.<br />17<br />

×