Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CLIP Based Region-Aware Feature Fusion for Automated BBPS Scoring in Colonoscopy Images

Yujia Fu, Zhiyu Dong, Tianwen Qian, Chenye Zheng, Danian Ji, Linhai Zhuo

TL;DR

The paper tackles the subjectivity of BBPS scoring in colonoscopy by introducing a high-quality HDFD dataset and a CLIP-based region-aware feature fusion framework. The method combines a global CLIP visual encoder with adapters and a localized fecal feature branch using anchors and textual priors, fused via gating to predict BBPS scores without explicit segmentation. Extensive experiments on HDFD and the public NERTHU dataset demonstrate state-of-the-art performance and robust ablations confirm the value of adapters and the fecal-feature module. This work advances automated bowel cleanliness assessment and supports potential clinical deployment in computer-aided colonoscopy analysis.

Abstract

Accurate assessment of bowel cleanliness is essential for effective colonoscopy procedures. The Boston Bowel Preparation Scale (BBPS) offers a standardized scoring system but suffers from subjectivity and inter-observer variability when performed manually. In this paper, to support robust training and evaluation, we construct a high-quality colonoscopy dataset comprising 2,240 images from 517 subjects, annotated with expert-agreed BBPS scores. We propose a novel automated BBPS scoring framework that leverages the CLIP model with adapter-based transfer learning and a dedicated fecal-feature extraction branch. Our method fuses global visual features with stool-related textual priors to improve the accuracy of bowel cleanliness evaluation without requiring explicit segmentation. Extensive experiments on both our dataset and the public NERTHU dataset demonstrate the superiority of our approach over existing baselines, highlighting its potential for clinical deployment in computer-aided colonoscopy analysis.

CLIP Based Region-Aware Feature Fusion for Automated BBPS Scoring in Colonoscopy Images

TL;DR

The paper tackles the subjectivity of BBPS scoring in colonoscopy by introducing a high-quality HDFD dataset and a CLIP-based region-aware feature fusion framework. The method combines a global CLIP visual encoder with adapters and a localized fecal feature branch using anchors and textual priors, fused via gating to predict BBPS scores without explicit segmentation. Extensive experiments on HDFD and the public NERTHU dataset demonstrate state-of-the-art performance and robust ablations confirm the value of adapters and the fecal-feature module. This work advances automated bowel cleanliness assessment and supports potential clinical deployment in computer-aided colonoscopy analysis.

Abstract

Accurate assessment of bowel cleanliness is essential for effective colonoscopy procedures. The Boston Bowel Preparation Scale (BBPS) offers a standardized scoring system but suffers from subjectivity and inter-observer variability when performed manually. In this paper, to support robust training and evaluation, we construct a high-quality colonoscopy dataset comprising 2,240 images from 517 subjects, annotated with expert-agreed BBPS scores. We propose a novel automated BBPS scoring framework that leverages the CLIP model with adapter-based transfer learning and a dedicated fecal-feature extraction branch. Our method fuses global visual features with stool-related textual priors to improve the accuracy of bowel cleanliness evaluation without requiring explicit segmentation. Extensive experiments on both our dataset and the public NERTHU dataset demonstrate the superiority of our approach over existing baselines, highlighting its potential for clinical deployment in computer-aided colonoscopy analysis.
Paper Structure (17 sections, 4 equations, 3 figures, 5 tables)

This paper contains 17 sections, 4 equations, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. BBPS Scoring and Clinical Significance
  4. Transfer Learning for Medical Image Analysis
  5. HDFD Dataset Construction
  6. Construction Process
  7. Evaluation & Comparison
  8. Framework
  9. Preliminaries
  10. Method Overview
  11. Framework
  12. Training & Inference
  13. Experiments
  14. Experimental Setup
  15. Main Results
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Illustration of BBPS scoring system and our annotation process. Top: Examples of BBPS scores 0--3. Bottom: Expert consensus filtering criteria to assure quality.
  • Figure 2: t-SNE visualization of feature distributions. Different colors indicate different BBPS scores. (a) Our HDFD dataset; (b) Public NERTHU dataset.
  • Figure 3: The architecture of proposed region-aware feature fusion framework. Red denotes trainable network components, while blue represents frozen parameters.