Evidence-Based Evaluation of Skin Flash for Prone Breast Treatments by Using Edge Detection On Daily Exit Dosimetry Images
W Gu*, S Yoon, E Hubley, A Haertter, G Freedman, N Taunk, J Zou, B Teo, C Kennedy, L Dong, T Li, University of Pennsylvania, Philadelphia, PA
Presentations
TH-IePD-TRACK 5-2 (Thursday, 7/29/2021) 3:00 PM - 3:30 PM [Eastern Time (GMT-4)]
Purpose: In breast radiotherapy, skin flash is important to account for setup uncertainties and intrafraction motion. On Halcyon linear accelerator, exit dosimetry images are automatically acquired at each fraction, providing a method to assess the breast position with respect to the radiation field borders. We developed a tool to analyze the interfractional skin flash and breast anatomy variation using breast edge detection technique based on the exit dosimetry images.
Methods: 288 daily exit dosimetry images were acquired for 9 patients undergoing 16-fraction prone breast treatments, with two opposed fields. The field border and breast contour map were acquired for every portal image using Canny algorithm. Algorithm parameters were derived automatically from a gradient-based initial estimate, to eliminate manual tuning. Noise and couch edges were removed combining local information and pixel connectivity. The distance between field border and apex of breast edge was calculated as the flash amount per fraction, and compared with the value from plans. The breast shape and position variation were evaluated by the range of breast contour change in beam’s-eye-view (BEV) at three positions, upper, middle and lower part of the breast.
Results: The mean and standard deviation of difference between fractional flash and planned flash is -1.8±3.8mm (negative sign represents actual breast closer to field border than planned). The variation of breast contour during treatment course has a mean and std of 9.5±2.8mm.
Conclusion: An edge detection tool was developed for breast radiotherapy based on Halcyon exit dosimetry images, which is used to investigate skin flash and breast position change with treatment progression. Breast contour in BEV for this patient cohort varies up to 17mm through the treatment course. This tool could be useful to automate daily breast contour detection, and to facilitate ongoing physics chart check by alerting user in case of insufficient skin flash
ePosters
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email