Overlapping Volume Histogram (OVH) as a Tool for OAR Delineation Outliers Detection and DVH Prediction
T Teo1*, M Gopalakrishnan2, D Cutright3, I Das4, (1) Northwestern Memorial Hospital, Chicago, IL, (2) Northwestern Memorial Hospital, Chicago, IL, (3) University of Chicago Medicine, Chicago, IL, (4) Northwestern Memorial Hospital, Northwestern University Feinberg School of Medicine, Chicago, IL
Presentations
PO-GePV-T-237 (Sunday, 7/25/2021) [Eastern Time (GMT-4)]
Purpose: n contrast to conventional usage of OVH to determine the lowest feasible dose to an organ-at-risks (OAR), this study examines the feasibility of using OVH in detecting outliers incurred in the delineation of OARs in treatment plans, and determining the optimal OVH (amongst a large cohort of patient OVH) in predicting dosimetric DVH properties.
Methods: An OVH characterizing the spatial relationship between the planning-target-volume (PTV) and an OAR was implemented by iteratively expanding the PTV, and determining the volume of overlap between the OAR and the expanded PTV. The OVH was determined for a cohort of 71 head-and-neck cancer cases. In evaluating the use of OVH as a tool for detecting outliners in OAR delineation, the OVH was applied to the entire data set and mean and median OVH from the OARs to PTV-1, -2, and -3 were evaluated along with the 1-standard-deviation of the OVH of all patient plans. Properties derived from the mean and median OVH profile were evaluated for accuracy in DVH prediction.
Results: A smaller spread in the OVH curves, and a closer spacing between the mean and median OVH curves are observed for the serial organs (e.g. spinal cord) vs the parallel OARS (e.g. right parotids). For parallel organs, the OVH is confounded by the laterality of the OAR with respect to the PTV. The median OVH, which has a steeper gradient and lies close to the center of the OVH distribution, is relatively insensitive to outliers. Larger differences in the predicted DVH was observed between using mean and median OVH features for parallel organs.
Conclusion: New insights on the sensitivity and feasibility of using OVH as a tool in detecting outliers of target contours is presented. Prediction of DVH for serial OAR is sensitive to features extracted from the mean and median OVH.
ePosters
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email