B Anderson*, K Moore, C Bojechko, University of California San Diego, San Diego, CA

• 
  B Anderson

WE-A-202-3 (Wednesday, 7/13/2022) 7:30 AM - 8:30 AM [Eastern Time (GMT-4)]

Room 202

Purpose: To create a deep-learning model to predict in-vivo electronic portal imaging device (EPID) transit images for IMRT treatments. This model can be used to predict in-vivo images to identify machine and patient related errors that occur during beam delivery and are undetectable with current QA approaches.

Methods: A generative adversarial network (GAN) was trained to generate in-vivo EPID images using information from the on-treatment cone-beam CTs (CBCTs), and a pretreatment EPID image with no patient in the beam. We acquired 656 images from 37 patients previously treated on the Varian Halcyon. Treatment sites included abdomen, lung, skull, and extremities. CBCTs were collected immediately before treatment, giving an accurate representation of the anatomy. A 3-channel input image was used, consisting of the pretreatment EPID image, a ray tracing projection through the CBCT to the EPID panel and a projection to isocenter. Model training:validation set ratios were 510:146 images from 29:8 different patients. Prediction accuracy was assessed by comparing model-predicted and measured in-vivo EPID images with a 3%/3mm gamma pass rate.

Results: The gamma pass rate for the patients in the validation set was 90.1% on average, and 102/146 (70%) images had a gamma pass rate larger than 90%. The CBCT field-of-view influences the performance of the model, with larger field-of-view margins around the treatment field producing higher pass rates. The prediction time for a single image takes less than one second.

Conclusion: This model can predict in-vivo EPID images with an average gamma pass rate of 90%. The model can be further improved extending the field-of-view of the CBCT in the training and validation dataset. Image predictions from this model can be used to detect in-vivo treatment errors and changes in patient anatomy adding an additional layer of patient specific quality assurance.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email