What question did this study set out to answer?

The aim is to determine if self-supervised pretraining can enhance segmentation performance in prostate radiotherapy when labeled data is limited.

February 21, 2026

Learning anatomy from unlabelled CT volumes: A self‐supervised framework for improving prostate radiotherapy segmentation

Key Points

The aim is to determine if self-supervised pretraining can enhance segmentation performance in prostate radiotherapy when labeled data is limited.
Utilized 322 pelvic CT volumes, split into training (258) and testing (64) sets.
Pretrained a lightweight 2D U-Net encoder using a slice-prediction task without labels.
Fine-tuned the pretrained model on multi-class segmentation with full or reduced labeled datasets.
Compared performance with baselines trained from scratch using different input channels and assessed accuracy using mean distance agreement and Dice similarity coefficient.
Self-supervised pretraining led to lower mean distance agreement across pelvic structures.
Significant reductions in bladder MDA from 0.600 mm to 0.547 mm and femoral heads from 1.370 mm to 0.994 mm were observed.
Prostate segmentation improved from 1.281 mm to 1.183 mm MDA, showing enhanced accuracy with self-supervised approaches.
Overall, segmentation accuracy improved without needing manual labels for pretraining.

Abstract

Abstract Background Accurate structure contouring on computed tomography (CT) is critical for prostate cancer radiotherapy, but it remains labour intensive and prone to interobserver variability, particularly for small, low‐contrast organs such as the prostate, seminal vesicles (SV) and penile bulb (PB). Deep‐learning models can automate this task. However, they typically require large, fully labelled datasets that are often unavailable in clinical settings. Purpose This study evaluated whether self‐supervised (label‐free) slice‐prediction pretraining could enhance segmentation performance, especially in scenarios with limited annotated data. Methods We used 322 pelvic CT volumes (215 from UMMC, 107 from TCIA), split 80:20 into training and testing sets (258 training, 64 testing patients). A novel lightweight 2D U‐Net encoder was first pretrained on unlabelled data using a slice‐prediction task across axial, sagittal, and coronal planes. The pretrained model was then fine‐tuned for multi‐class segmentation using either the full dataset or a reduced subset of 60 labelled patients. Baselines trained from scratch with 1‐channel or 3‐channel input were included for comparison. Segmentation accuracy was assessed using mean distance agreement (MDA) and Dice similarity coefficient (DSC). Paired t ‐tests with Bonferroni correction were applied to assess statistical significance. Results Models with self‐supervised pretraining achieved consistently lower MDA across all major pelvic structures. Notable improvements included reductions in bladder MDA from 0.600 mm to 0.547 mm, femoral heads from 1.370 mm to 0.994 mm, PB from 1.470 mm to 1.283 mm, rectum from 0.792 mm to 0.669 mm, prostate from 1.281 mm to 1.183 mm, and SV from 1.175 mm to 0.893 mm. Conclusions Self‐supervised pretraining via slice prediction enables anatomically informed feature learning and improves segmentation robustness under limited data conditions. This strategy enhances accuracy without reliance on manual labels during pretraining and is compatible with computationally lightweight architectures, making it well‐suited in resource‐constrained clinical environments.

Perguntar à IA

Bookmark