Reproduce “Towards Polyp Counting In Full-Procedure Colonoscopy Videos” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Luca Parolari, Andrea Cherubini, Lamberto Ballan, Carlo Biffi
Researchers at the University of Padova and Cosmo Intelligent Medical Devices presented a method for precise polyp counting in full-procedure colonoscopy videos, achieving a 3.9-fold reduction in polyp fragmentation rate using unsupervised clustering algorithms. Their approach, evaluated on the open-access REAL-Colon dataset, improved polyp re-identification over prior threshold-based methods while maintaining a controlled false positive rate.
Key insights
- Unsupervised clustering algorithms, particularly Affinity Propagation, substantially reduce polyp fragmentation rates compared to previous threshold-based re-association methods while effectively controlling false positives.
- The observed performance gains from clustering are significant even with a moderately sized training dataset, indicating its efficiency for polyp re-identification.
- An open-access framework, including standardized data splits and released code, provides a crucial foundation for reproducible research and advancements in automated colonoscopy reporting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.