Auto‐segmentation of the clinical target volume using a domain‐adversarial neural network in patients with gynaecological cancer undergoing postoperative vaginal brachytherapy

Abstract

PurposeFor postoperative vaginal brachytherapy (POVBT), the diversity of applicators complicates the creation of a generalized auto‐segmentation model, and creating models for each applicator seems difficult due to the large amount of data required. We construct an auto‐segmentation model of POVBT using small data via domain‐adversarial neural networks (DANNs).MethodsCT images were obtained postoperatively from 90 patients with gynaecological cancer who underwent vaginal brachytherapy, including 60 and 30 treated with applicators A and X, respectively. A basal model was devised using data from the patients treated with applicator A; next, a DANN model was constructed using these same 60 patients as well as 10 of those treated with applicator X through transfer learning techniques. The remaining 20 patients treated with applicator X comprised the validation set. The model's performance was assessed using objective metrics and manual clinical evaluation.ResultsThe DANN model outperformed the basal model on both objective metrics and subjective evaluation (p<0.05 for all). The median DSC and 95HD values were 0.97 and 3.68 mm in the DANN model versus 0.94 and 5.61 mm in the basal model, respectively. Multi‐centre subjective evaluation by three clinicians showed that 99%, 98%, and 81% of CT slices contoured by the DANN model were acceptable versus only 73%, 77%, and 57% of those contoured by the basal model. One clinician deemed the DANN model comparable to manual delineation.ConclusionDANNs provides a realistic approach for the wide application of automatic segmentation of POVBT and can potentially be used to construct auto‐segmentation models from small datasets.

A deep learning‐based 3D Prompt‐nnUnet model for automatic segmentation in brachytherapy of postoperative endometrial carcinoma

“…See Table 1, there are only two papers studied in postoperative CC or EC. Yan et al 16 validations, and 25 tests. These observed discrepancies in our study about postoperative EC might be attributed to two primary factors.…”

“…With respect to postoperative target volume delineation, the shape of the target volume is regular for vaginal stump irradiation after radical hysterectomy, exhibits minimal applicator artifact, and presents lower model training complexity compared to non-surgical patients. 16,17 Although these networks exhibit commendable performance, their utility for specific image segmentation tasks is frequently constrained. The necessity for task-specific design and configuration demands meticulous fine-tuning, as minor adjustments in hyperparameters can result in substantial performance disparities.…”