Pelvic Nodal Auto-Segmentation Using a Deep Image to Image Network (DI2IN) Auto-Segmentation Algorithm: Comparing Male vs. Female Pelvis

Abstract

Deep Learning approaches have shown significant benefits compared to atlas-based methods in improving segmentation accuracy and efficiency in auto-segmentation algorithms. The AI-Rad Companion Organs RT pelvic nodal auto-segmentation feature was trained and developed for use in the male pelvis. There is no real-world data on its usability in male pelvis and whether it can be used for female pelvic nodal anatomy. This work represents the first multi-institutional study to describe and evaluate an AI algorithm for auto-segmentation of the pelvic nodal region in female patients based on a deep image-to-image network (DI2IN). The AIRC algorithm uses a two-step approach for segmentation. In the first step, the target organ region in the optimal input image is extracted using a trained Deep Reinforcement Learning network (DRL), which is then used as input to create the contours in the second step based on DI2IN. We retrospectively evaluated AIRC pelvic nodal auto-segmentation in both male and female patients treated at our network of institutions. The automated pelvic nodal contours generated by AIRC were evaluated by one board-certified radiation oncologist, specializing in prostate and gynecologic malignancies. A 4-point scale was used, where 4 is clinically usable and 1 requires re-contouring. Pelvic nodal regions included the right and left side of the common iliac, external iliac, internal iliac, obturator and midline presacral nodes. A chi-squared test was then used to compare the scores of male and female pelvic nodal cases. Fifty-two female and 51 male patients were included in the study, representing a total of 468 and 447 pelvic nodal regions, respectively. 96% (450 pelvic nodal contours) and 99% (443 pelvic nodal contours) required no or minor edits for female and male patients, respectively (p = 0.004). The right internal iliac was the only nodal group with a statistically significant difference between female (92% requiring no or minor edits) and male (100% requiring no or minimal edits) patients, p = 0.04. The percentage of patients requiring no, or minor edits was 87% (45 patients) and 92% (47 patients) for female and male patients, respectively (p = 0.36). AIRC pelvic nodal auto-segmentation performed very well in both male and female pelvic nodal regions, with the male pelvic nodal regions performing better especially in the right internal iliac nodal group. It is usable in female pelvic nodal regions, with 96% of contours requiring no or minor edits. As auto-segmentation becomes more widespread, it may be important to have equal representation from all genders in training and validation of auto-segmentation algorithms.