Truace
Subscribe

Deep Learning-Based Detection of Sessile Serrated Lesions in Colonoscopy Images: Optimization of a YOLOv3 Framework With Mucus Cap Feature Enhancement for Sessile Serrated Lesion Detection

Background and aim Sessile serrated lesions (SSLs) are difficult to detect during colonoscopy via standard white-light imaging (WLI) because they are flat, have indistinct borders, and are often hidden by mucus caps. Although narrowband imaging (NBI) can help, its effectiveness is inconsistent. This study evaluated the feasibility of an artificial intelligence (AI)-assisted approach incorporating mucus cap features and both WLI and NBI for SSL detection and diagnosis. Methods This study developed an AI model for…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-1754· cite & verify
Deep Learning-Based Detection of Sessile Serrated Lesions in Colonoscopy Images: Optimization of a YOLOv3 Framework With Mucus Cap Feature Enhancement for Sessile Serrated Lesion Detection
"Endoscopy" by Yuya Tamai is licensed under CC BY 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/2.0/.

In brief

Researchers built an AI model to find sessile serrated lesions, which are flat, indistinct and often covered by mucus, in colonoscopy images. They trained it on 1980 images collected between 2023 and 2025 and added mucus cap features to a standard detection framework.

The results suggest that targeting mucus caps can make AI detection of these easily missed lesions more reliable, but the evidence comes from images rather than live procedures. The authors call it an incremental step and say prospective, multicenter studies are needed before clinical deployment.

Main points

  1. Image-level accuracy was 98.7%, with 100.0% sensitivity and 97.4% specificity.
  2. Detection rates were 95.9% for lesions without a mucus cap and 98.4% for lesions with one.
  3. For morphology-based typing, the white-light model reached 85.8% accuracy (AUC 0.89).
  4. The narrowband imaging model performed better at typing, with 91.6% accuracy (AUC 0.94).

The gain

Adding mucus cap screening to a YOLOv3-based model raised lesion-level detection of sessile serrated lesions from 77.4% to 95.0% in the study's colonoscopy images.

The rundown

Sessile serrated lesions are hard to detect with standard white-light imaging because they are flat, have indistinct borders and are often hidden by mucus caps. Narrowband imaging can help, but the study describes its effectiveness as inconsistent.

The team evaluated an AI approach that uses both white-light and narrowband imaging and builds mucus cap features into the detection framework. Performance was measured by detection rate, accuracy and AUC.

The authors describe the work as real-world clinical evidence that supports AI models for these lesions and helps improve generalizability across clinical settings, while noting it is not a novel framework.

Sources

  1. Peer-reviewedJournal of Gastroenterology and Hepatology2026-10-07

The debate