Room: Exhibit Hall
Purpose: To evaluate the practicability of deformable image registration(DIR) in pancreatic cancer(PC) radiotherapy for stomach dose accumulation, by assessing DIR accuracy with balloon phantom then performing clinical dose accumulation with DIR and rigid registration(RR).
Methods: We designed a simple balloon phantom using piled condoms hanged in a water tank, with markers on the surface. Perform CT scan each time when injecting contrast medium into the balloon to simulate the deformation. All CTs were transferred to MIM Maestro6.6 to calculate dice similarity index(DSI) and perform DIR in pairs. Each marker was manually identified to check the surface point registration accuracy before and after DIR. Secondly we enrolled 5 patients underwent PC radiotherapy, each with off-line verification CT scan every 5 fractions. Stomach dose accumulation was accomplished with DIR and RR.
Results: Phantom study indicated that even with DIR, surface points registration still had intolerable error unless DSI was small enough. The averaged point distance before DIR was from 1.52 to 3.85cm. Significant reduction was observed after DIR. The minimum deviation was 0.44cm with DSI=0.89. The maximum was 1.95cm with DSI=0.25. The higher the DSI, the smaller the registration error. Points around polar zoon had relatively smaller error than those around equatorial zoon. DIR corrected the balloon volume deviation well, from as high as 221% to under 5%. Clinical data demonstrated that stomachâ€™s DSI varied from 0.609 to 0.774. With RR the difference between planned and accumulated dose could reach to 10.6%, 9.7%, and 9.9% for D1cc, D3cc, and D5cc respectively. Thereâ€™s no significant difference between DIR and RR based dose accumulation.
Conclusion: We suspect the accuracy of DIR point dose accumulation due to previous QA test. Utilizing DIR for stomach dose accumulation in PC radiotherapy seems inappropriate maybe unnecessary. Future works are needed with more delicate phantom and more clinical data.