{"slug": "lux-a-lesion-aware-graph-conditioned-visual-language-architecture-for-endoscopic", "title": "LUX: A Lesion-Aware Graph-Conditioned Visual - Language Architecture for Explainable Endoscopic Captioning", "summary": "Researchers introduced LUX (Lesion-aware Unified eXplainable captioning), a graph-conditioned vision-language architecture for explainable endoscopic image captioning in ulcerative colitis, which constructs a lesion-centric scene graph from Grad-CAM and CBAM activation maps and integrates graph embeddings into a T5 decoder's cross-attention layers. LUX outperformed baseline and state-of-the-art medical captioning models across BLEU, METEOR, ROUGE-L, and CIDEr metrics, with particularly strong gains in CIDEr, while reducing hallucinated clinical findings and improving lesion-level grounding.", "body_md": "arXiv:2608.23853v1 Announce Type: new\nAbstract: The interpretation of endoscopic imagery in ulcerative colitis is complex and subjective, with variability in human assessment and subtle mucosal inflammation. Although deep learning has advanced automated analysis, most vision-language models rely on global visual embeddings that overlook the localized and relational nature of pathological evidence, limiting clinical reliability and interpretability.\nWe introduce LUX (Lesion-aware Unified eXplainable captioning), a graph-conditioned vision-language architecture for explainable endoscopic image captioning. LUX constructs a lesion-centric scene graph from Grad-CAM and CBAM activation maps, representing pathological regions as nodes and encoding their spatial and clinical relationships. These graph embeddings are integrated into the cross-attention layers of a T5 decoder, enabling generated words to attend to specific lesion nodes rather than only to global image features. This provides direct alignment between linguistic content and pathological evidence, supporting token-level interpretability and relational reasoning.\nLUX outperforms strong baseline and state-of-the-art medical captioning models across BLEU, METEOR, ROUGE-L, and CIDEr, with particularly strong gains in CIDEr. It also reduces hallucinated clinical findings and improves lesion-level grounding through stronger correspondence between generated tokens and localized pathological regions.", "url": "https://wpnews.pro/news/lux-a-lesion-aware-graph-conditioned-visual-language-architecture-for-endoscopic", "canonical_source": "https://arxiv.org/abs/2608.23853", "published_at": "2026-08-26 04:00:00+00:00", "updated_at": "2026-08-26 04:14:12.076101+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "computer-vision", "natural-language-processing"], "entities": ["LUX", "T5", "Grad-CAM", "CBAM"], "alternates": {"html": "https://wpnews.pro/news/lux-a-lesion-aware-graph-conditioned-visual-language-architecture-for-endoscopic", "markdown": "https://wpnews.pro/news/lux-a-lesion-aware-graph-conditioned-visual-language-architecture-for-endoscopic.md", "text": "https://wpnews.pro/news/lux-a-lesion-aware-graph-conditioned-visual-language-architecture-for-endoscopic.txt", "jsonld": "https://wpnews.pro/news/lux-a-lesion-aware-graph-conditioned-visual-language-architecture-for-endoscopic.jsonld"}}