Automated Cervical Os Segmentation for Camera-Guided, Speculum-Free Screening

2025-09-12

AI Summary

Real-time automatic segmentation of the cervical os in speculum-free cervical cancer screening remains challenging due to limited visual cues and anatomical variability in vaginal endoscopic video. Method: We propose a lightweight segmentation framework based on an endoscopy-video-pretrained vision transformer (EndoViT/DPT), fine-tuned on speculum-free vaginal endoscopy images. Five encoder-decoder architectures were systematically compared using ten-fold cross-validation and external validation on phantom data, with performance evaluated via IoU, Dice coefficient, detection rate, and localization error. Contribution/Results: EndoViT/DPT achieves the first robust, real-time cervical os segmentation in this setting—attaining a Dice score of 0.50±0.31, detection rate of 87%±33%, and inference speed of 21.5 FPS—significantly outperforming existing methods. The approach enables reliable, visualization-guided sampling by non-expert operators across low- and high-resource settings.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Calibration & Uncertainty Quantification
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Security and Privacy: Browser and end-point security
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Cervical cancer is highly preventable, yet persistent barriers to screening limit progress toward elimination goals. Speculum-free devices that integrate imaging and sampling could improve access, particularly in low-resource settings, but require reliable visual guidance. This study evaluates deep learning methods for real-time segmentation of the cervical os in transvaginal endoscopic images. Five encoder-decoder architectures were compared using 913 frames from 200 cases in the IARC Cervical Image Dataset, annotated by gynaecologists. Performance was assessed using IoU, DICE, detection rate, and distance metrics with ten-fold cross-validation. EndoViT/DPT, a vision transformer pre-trained on surgical video, achieved the highest DICE (0.50 pm 0.31) and detection rate (0.87 pm 0.33), outperforming CNN-based approaches. External validation with phantom data demonstrated robust segmentation under variable conditions at 21.5 FPS, supporting real-time feasibility. These results establish a foundation for integrating automated os recognition into speculum-free cervical screening devices to support non-expert use in both high- and low-resource contexts.

Problem

• Automated cervical os segmentation for speculum-free screening
• Real-time visual guidance in transvaginal endoscopic images
• Integration into non-expert cervical screening devices

Innovation

• Deep learning for cervical os segmentation
• Vision transformer outperforms CNN methods
• Real-time segmentation supports non-expert screening