Lesion-Aware Visual-Language Fusion for Automated Image Captioning of Ulcerative Colitis Endoscopic Examinations

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Escamilla, Alexis Ivan Lopez, Ochoa, Gilberto, Al, Sharib
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909768231157760
author Escamilla, Alexis Ivan Lopez
Ochoa, Gilberto
Al, Sharib
author_facet Escamilla, Alexis Ivan Lopez
Ochoa, Gilberto
Al, Sharib
contents We present a lesion-aware image captioning framework for ulcerative colitis (UC). The model integrates ResNet embeddings, Grad-CAM heatmaps, and CBAM-enhanced attention with a T5 decoder. Clinical metadata (MES score 0-3, vascular pattern, bleeding, erythema, friability, ulceration) is injected as natural-language prompts to guide caption generation. The system produces structured, interpretable descriptions aligned with clinical practice and provides MES classification and lesion tags. Compared with baselines, our approach improves caption quality and MES classification accuracy, supporting reliable endoscopic reporting.
format Preprint
id arxiv_https___arxiv_org_abs_2509_03011
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Lesion-Aware Visual-Language Fusion for Automated Image Captioning of Ulcerative Colitis Endoscopic Examinations
Escamilla, Alexis Ivan Lopez
Ochoa, Gilberto
Al, Sharib
Computer Vision and Pattern Recognition
Artificial Intelligence
We present a lesion-aware image captioning framework for ulcerative colitis (UC). The model integrates ResNet embeddings, Grad-CAM heatmaps, and CBAM-enhanced attention with a T5 decoder. Clinical metadata (MES score 0-3, vascular pattern, bleeding, erythema, friability, ulceration) is injected as natural-language prompts to guide caption generation. The system produces structured, interpretable descriptions aligned with clinical practice and provides MES classification and lesion tags. Compared with baselines, our approach improves caption quality and MES classification accuracy, supporting reliable endoscopic reporting.
title Lesion-Aware Visual-Language Fusion for Automated Image Captioning of Ulcerative Colitis Endoscopic Examinations
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2509.03011