Abstract
Colorectal cancer mostly evolves from colon polyps, and deep learning-based analysis is limited by scarce pixel-level annotations and significant intestinal mucosal texture heterogeneity. Besides, ImageNet pre-trained backbones suffer severe domain shift with endoscopic images, failing to capture subtle lesion boundaries. To address these bottlenecks, we fuse two public endoscopic datasets (LDPolypVideo & HyperKvasir) to build ColoTexture-Hybrid, a large-scale unlabeled intestinal endoscopic dataset containing 27,124 clinical static colonoscopy images, designed specifically for domain self-supervised pre-training of gastrointestinal vision models.
Keywords
- Colorectal Cancer Screening and Detection
- Colonoscopy
- Colorectal cancer
- Domain (mathematical analysis)
- Texture (cosmology)
- Fuse (electrical)
- Cancer
- Open access
Source
This is a preprint: it has not been peer reviewed yet, and its findings may change.
Published in OSF Preprints (OSF Preprints). The abstract is shown as published and remains the work of its authors; read the full article at the source.
Prognia lists this article for reference and education. It is not medical advice. Always read the original and follow current guidelines and local protocols before acting on it.
