(Monday, 7/30/2018) 9:30 AM - 10:00 AM
Room: Exhibit Hall | Forum 5
Purpose: True knowledge of the radiation dose received by cervical cancer patients is necessary to correctly assess the accuracy of the treatment and to predict associated toxicities. Currently, due to the lack of a method to obtain the total dose from external-beam radiotherapy (EBRT) and brachytherapy (BT), assumptions need to be made. We aim to develop an automated clinically implementable algorithm to perform deformable image registration (DIR) and obtain the cumulative doses from EBRT and BT.
Methods: An automated workflow has been developed to perform DIR of EBRT and BT images using a commercial DIR software with in-house Matlab extensions. The intensity-based algorithm was biased by masking different structures (bladder, rectum, sigmoid, uterus, HRCTV, applicator region, and bones) to achieve a combined structure-based and intensity-based algorithm. We used EBRT and BT images as reference and compared the DIR results. When BT was the reference, a virtual applicator region (AR) was created in the EBRT CT. Dice similarity coefficient (DSC) between reference and deformed contours and cumulative rectum and bladder D2cc for four patients are used to evaluate the method.
Results: The average DSC for bladder and rectum are 0.94 and 0.89, respectively, using EBRT as reference and 0.92 and 0.91 with BT as the reference. There are not significant differences in the accumulated D2cc regarding the image used as reference. The computational time is around 4min.
Conclusion: The automated EBRT-BT DIR method presented here yields high agreement of reference and deformed contours. Since using EBRT and BT images as reference provides similar results, we recommend using the EBRT as reference as it does not require to create an artificial AR contour. This tool provides accurate knowledge of the radiation received in a fast way and takes us a step closer to real personalized medicine in radiotherapy for cervical cancer.
Funding Support, Disclosures, and Conflict of Interest: Work funded by the CTRMI at UC San Diego
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
