Purpose: A potential application of the Varian Ethos platform’s online adaptive capabilities is prostate SBRT. Importantly, the suitability of Ethos iCBCT images for delineation of target volumes must be validated. In this work, we aim to determine if the Ethos iCBCT offers sufficient image quality and soft-tissue contrast to accurately segment the prostate and seminal vesicles for the purpose of online adaptive prostate SBRT.
Methods: Two radiation oncologists independently segmented the prostate and seminal vesicles on five daily CBCT images for a prostate SBRT patient previously treated on our institution’s Ethos. The Dice similarity coefficient (DSC) and mean distance to agreement (MDA) were computed for both structures as well as for their union. Segmentations from both observers were also compared to unedited, automatically generated contours from the Ethos planning system.
Results: The mean DSC between the two radiation oncologist contours were 0.61, 0.57, and 0.76 for the prostate, seminal vesicles, and the union of the prostate and seminal vesicles respectively. Corresponding average MDA values were 4.9 mm, and 3.3 mm, and 2.8 mm. Overall mean DSC values for the two radiation oncologist contours compared to corresponding Ethos automatically generated segmentations were 0.70 and 0.65 with MDA values of 3.9 mm and 3.8 mm.
Conclusion: Inter-observer variability was substantial compared to other published studies, though differences were mitigated by examining the union of the prostate and seminal vesicles. Further work with an expanded number of images and observers is warranted.
Prostate Therapy, Segmentation, Treatment Planning