1. Stromer, Daniel 15.03.2016 Pattern Recognition Lab (CS 5) Data Completeness Estimation for 3D C-Arm Scans with Rotated Detector to Enlarge the Lateral Field-of-View

2. 2 Outline ●Introduction ●Materials and Methods ● Scan Modes and Field-of-View Enlargement ● C-arm CT 3-D Reconstruction ●Results ●Discussion and Outlook

3. 3 Introduction ●3-D imaging uses Short- and Large Volume Scan ●Patient may be still too large to be completely covered  Find method to enlarge the detector‘s field-of-view in lateral direction

4. 4 Materials and Methods

5. 5 C-arm CT geometry source detector z Cone Beam Geometry

6. 6 Short Scan x x detector covered object ●200 degree circular scan (180 degrees + fan angle) source detector covered object

7. 7 Large Volume Scan source shifted detector covered object source shifted detector ●360 degree circular scan with shifted detector x x

8. 8 Helical Large Volume Scan ●Similar to Large Volume Scan ●Helical orbit with pitch p ●Increased axial coverage

9. 9 Field-of-View Enlargement u v u v u v u v Diamond Scan Short Scan Large Volume Scan

10. 10 Data Completeness Estimation

11. 11 Short Scan coverage Short Scan Radon sphere coverage and forward projection of the coverage volume Diamond Short Scan Radon sphere co- verage and forward projection of the coverage volume

12. 12 Large Volume Scan (LVS) coverage Large Volume Scan Radon sphere coverage and forward projection of the coverage volume Diamond LVS Scan Radon sphere coverage and forward projection of the coverage volume

13. 13 Helix coverage Forward projected volume coverage for Helix Diamond Large Volume Scan

14. 14 Coverage Results ●Coverage increased by 25.3 % for the Short Scan and about 27.3 % for the Large Volume Scans ●Using a helical trajectory compensates the axial loss Scan Mode Radon sphere coverage Short Scan2613276625.3 Large Volume Scan47560513027.3 Helix LVS47560513027.3

15. 15 C-arm CT 3-D Reconstruction

16. 16 Basics: Iterative Reconstruction ●Inverse problem of Radon transform is reformulated as system of equations: ●Minimization of objective function (SSD):

17. 17 SART Reconstruction ●Compute orthogonal projections of the current approximation to the hyperplanes ●Compute the centroid and use it for the next iteration

18. 18 Total Variation Based Reconstruction

19. Results

20. 20 Configuration

21. 21 Reconstruction Results Scan Mode Short Scan0.21340.1370 Diamond Short Scan0.20260.1242 Standard LVS0.06080.0549 Diamond LVS0.03220.0217 Helical LVS0.06410.0488 Helical Diamond LVS0.02290.0151  eTV lowers the RMSE compared to SART by 29.1 %

22. 22 Large Volume Scan Diamond Large Volume Scan VHP eTV Reconstructions Central Slices (z=15 of 30)

23. 23 Diamond Large Volume Scan Helix Diamond Large Volume Scan VHP eTV Reconstructions Outmost Slices (z=30 of 30)

24. 24 Discussion and Outlook ●Coverage gain with presented method:  Short Scan 25 %,LVS: 27 % Helix: axial gain ●Reconstruction:  eTV lowers RMSE compared to SART by 29 % But: ●Only coverage > 0.9 was considered ●Different detector settings may fit better ●Other eTV parameters may increase quality

25. 25 Thank you very much for your attention