Author: Steve B. Jiang, Chien-Yi Liao, Dan Nguyen, Daniel Yang, Hengrui Zhao 👨🔬
Affiliation: Medical Artificial Intelligence and Automation (MAIA) Lab & Department of Radiation Oncology, UT Southwestern Medical Center 🌍
Purpose:
Post-operative radiotherapy for prostate cancer requires precise contouring of the clinical target volume (CTV) to account for microscopic disease that is invisible in the image. However, the absence of the prostate gland and unclear anatomical landmarks make this process challenging. In addition, the contouring style is greatly influenced by the physician’s experience and preferences. This greatly complicates automatic segmentation algorithms that rely on consistent data but also typically need large amounts data from many physicians to develop. This study presents a novel auto-segmentation method incorporating individual physicians’ contouring styles to enhance accuracy and reduce manual workload.
Methods:
We developed CLIP-UNet, a deep learning-based model integrating a text encoder to encode physician-specific contouring styles into a latent vector. This information was combined with CT image features to guide segmentation. A dataset of 824 patients was divided into 699 training, 49 validation, and 76 testing data. The test set included recent cases from four physicians, while the training set represented data from seven physicians. CLIP-UNet was compared against a baseline UNet without physician-specific information and PSA-Net, a state-of-the-art (SOTA) model with physician-specific information.
Results:
CLIP-UNet achieved an average Dice score of 84.1%, outperforming the baseline UNet (82.0%) and the PSA-Net (83.1%). Incorporating physician-specific information improved segmentation consistency and reduced manual contouring effort, demonstrating its potential to enhance treatment planning for post-operative prostate cancer radiotherapy.
Conclusion:
By integrating physicians’ contouring styles, CLIP-UNet addresses inter-physician variability, improving CTV auto-segmentation precision. This approach personalizes radiotherapy planning according to physicians and improves treatment consistency.