PulseAugur
EN
LIVE 14:29:43

Colonoscopy polyp segmentation benchmarks flawed, audit finds · 2 sources tracked

By PulseAugur Editorial · [4 sources] ·

A recent audit of 27 colonoscopy polyp segmentation benchmark papers published between 2015 and 2026 reveals significant inconsistencies in evaluation methodologies. The audit highlights three key issues: the omission of the crucial Hausdorff distance metric in 25 papers, the use of incompatible train/test split protocols across studies on the same datasets, and the lack of statistical significance testing for performance claims in 26 papers. These problems persist even in papers published after the introduction of the Metrics Reloaded framework, suggesting a need for better adherence to standardized reporting. To address these shortcomings, a new Polyp Segmentation Reporting Checklist (PSRC) is proposed. AI

IMPACT Inconsistent evaluation metrics in AI benchmarks can obscure true model performance, hindering reliable progress in critical applications like medical image analysis.

RANK_REASON The cluster contains academic papers detailing research findings and proposing new methodologies.

Read on arXiv cs.CV →

AI-generated summary · Google Gemini · from 4 sources. How we write summaries →

Colonoscopy polyp segmentation benchmarks flawed, audit finds · 2 sources tracked

COVERAGE [4]

  1. Hugging Face Daily Papers TIER_1 English(EN) ·

    Metrics or Mirage? An Audit of Evaluation Inconsistencies in Colonoscopy Polyp Segmentation Benchmarks

    Progress in colonoscopy polyp segmentation is routinely reported through leaderboard comparisons on a small set of public benchmarks. We argue that this apparent progress is difficult to verify: a systematic audit of \textbf{27 papers} published between 2015 and 2026 reveals thre…

  2. arXiv cs.CV TIER_1 English(EN) · Aisha Urooj, Zain Ul Abdien, Neelu Madan ·

    Metrics or Mirage? An Audit of Evaluation Inconsistencies in Colonoscopy Polyp Segmentation Benchmarks

    arXiv:2607.08203v1 Announce Type: new Abstract: Progress in colonoscopy polyp segmentation is routinely reported through leaderboard comparisons on a small set of public benchmarks. We argue that this apparent progress is difficult to verify: a systematic audit of \textbf{27 pape…

  3. arXiv cs.CV TIER_1 English(EN) · Neelu Madan ·

    Metrics or Mirage? An Audit of Evaluation Inconsistencies in Colonoscopy Polyp Segmentation Benchmarks

    Progress in colonoscopy polyp segmentation is routinely reported through leaderboard comparisons on a small set of public benchmarks. We argue that this apparent progress is difficult to verify: a systematic audit of \textbf{27 papers} published between 2015 and 2026 reveals thre…

  4. arXiv cs.CV TIER_1 English(EN) · Jiachi Zhang, Zhuoyu Wu, Wenqi Fang ·

    RIGS-Refiner: Risk-Guided Recursive Refinement in Prediction Space for Colonoscopy Polyp Segmentation

    arXiv:2607.03058v1 Announce Type: new Abstract: Post-refinement can improve colonoscopy segmentation after host inference, but many designs still rely on extra correction heads or multi-stage pipelines with non-negligible parameter or computational cost. For polyp segmentation, h…

RELATED ENTITIES

RELATED TOPICS