Evaluation Of Deformable Contour Propagation

Purpose: Deformable contour propagation, when performed accurately, can potentially reduce the workload associated with manual contouring and minimize inter‐observer variability. In this work, we have compared propagated contours to physician‐drawn ones for gynecologic brachytherapy (GYN) and head and neck (HN) external beam radiotherapy cases.Methods: Eleven HN and sixteen GYN cases were randomly and retrospectively chosen for this study. These cases included their respective pre‐treatment (pCT) and follow‐up CT (fCT) scans along with their respective physician‐drawn structures. The pCT volume was registered using a deformable algorithm to the fCT and the resulting deformation field was applied to the pCT's structure set to propagate it to the fCT, this was done using Velocity Medical Solutions' VelocityAI software. The propagated contours were compared to the physician drawn‐contours using the Dice similarity coefficient (DSC).Results: Eleven organs at risk (OAR) and the clinical target volume (CTV) were analyzed for the HN cases and two OARs, as well as the Miami applicator and the CTV for the GYN cases. The propagated contours for the brain, brainstem, spinal cord, esophagus, eyes,larynx, mandible, oral cavity, parotids and the CTV yielded mean DSCs ranging between 0.635–0.979 for the HN cases with a poor outcome for the sphincter muscle (mean DSC = 0.576). The mean DSCs ranged between 0.702–0.804 for the Miami applicator, bladder, rectum and CTV for the GYN cases Conclusions: This work has shown that deformable contour propagation is quite accurate and requires minimal modification. The HN cases slightly outperformed the GYN cases primarily due to the fact that the GYN scans are low in contrast and contain high deformation tissues such as the bladder and rectum.