(Sunday, 7/14/2019)
Room: ePoster Forums
Purpose: PTVs are not routinely defined when treating intact-breast cancer so tangential beams are manually defined to cover the whole breast. We investigate a fully-convolutional networks to autosegment the 50% iso-dose volume as a surrogate to PTV to guide tangential beams placement.
Methods: CT scans and treatment plans from 599 intact-breast patients treated with opposed tangential fields to the whole left breast were used in this study. Patients were split into training (n=359)/cross-validation (n=120)/test(n=120) sets. The 50% iso-dose volume masks were generated for each patient. A 3D region-of-interest (ROI) was defined about the left breast. CT scan and 50% iso-dose masks voxels within the ROI were extracted and used as input and ground-truth, respectively, for our model. A 3D fully-convolutional neural network was trained to autosegment the 50% iso-dose volume. This volume was then imported into TPS and used to guide tangent beam placement. The predicted and clinical beams are compared in terms of gantry angle, collimator angle and medial collimator size for a subset of test set (n=26).
Results: Average gantry angle difference between the clinical and predicted beams was 3.4 degrees (range:0-7), average collimator angle difference was 1.8 degrees (range:0-6), and average medial collimator size difference was 0.2 cm (range:0-0.9). Patient disease presentation varied largely: one patient had a breast implant, one patient had significant medial extent of disease, and one patient had disease which was located more laterally. For these three cases, differences between clinic and predicted values were relatively large, whereas for patients where disease was found more central to the breast the differences were found to be smaller.
Conclusion: Autosegmented 50% iso-dose volumes can be used for guidance to place tangential beams for left-sided breast cancer patients. The predicted beams’ angles/distances agreed well with clinical beams within a few degrees and millimeters for most cases.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
