Room: Room 202
Purpose: Previously we developed a prior contour based TV (PCTV) method to enhance edge sharpness using registered edge information from prior images. The edge enhancement in PCTV is affected by the registration accuracy, especially when deformation exists. This study aims to develop a Hybrid-PCTV method to account for registration errors in reconstruction to enhance its robustness and accuracy.
Methods: Planning-CT is deformably registered with on-board CBCT reconstructed with edge preserving TV (EPTV) method. Edges extracted from planning-CT are deformed based on the registration to generate on-board edges for PCTV. Reference CBCT is reconstructed from the simulated projections of the deformed planning-CT. Image similarity map is then calculated between reference and on-board CBCT image gradient to estimate registration accuracy. The hybrid-PCTV method uses PCTV edges at regions with high similarity and EPTV edges at regions with low similarity. The Hybrid-PCTV method was evaluated using both extended-cardiac-torso (XCAT) phantom, liver and head-neck patient data. In XCAT study, breathing amplitude change, tumor shrinkage and new tumor were simulated from CT to CBCT. In the patient study, projections were simulated from clinical CBCT for reconstruction. 36 hall-fan projections over 360Ëš, 45 hall-fan projections over 360Ëš and 34 full-fan projections over 200Ëš scan angle were used in the XCAT, liver and head-neck patient studies, respectively. Results were compared with both EPTV and PCTV methods.
Results: Compared to EPTV and PCTV, Hybrid-PCTV enhanced edges of bone and tumor in XCAT reconstruction, bony structures in head-neck CBCT and avoided the wrong edge enhancement in liver CBCT. In XCAT study, compared with ground truth, relative error were 1.94%, 0.62% and 0.41% and edge cross-correlation were 0.60, 0.72 and 0.74 for EPTV, PCTV and Hybrid-PCTV, respectively.
Conclusion: Hybrid-PCTV further improved the performance of PCTV by accounting for uncertainties in deformable registration and anatomy mismatch between prior and onboard images.
Funding Support, Disclosures, and Conflict of Interest: This study was supported by NIH grant R01 CA-184173.
Not Applicable / None Entered.