Less Is More? Selective Visual Attention to High-Importance Regions for Multimodal Radiology Summarization
This work addresses the challenge of improving multimodal models for radiology report summarization, which is incremental as it builds on existing methods by introducing selective visual attention.
The paper tackled the problem of automated radiology report summarization by challenging assumptions that more visual input is always better, showing that selectively focusing on pathology-relevant visual patches instead of full images yields substantially better performance, achieving state-of-the-art results with 29.25% BLEU-4 and 69.83% ROUGE-L scores.
Automated radiology report summarization aims to distill verbose findings into concise clinical impressions, but existing multimodal models often struggle with visual noise and fail to meaningfully improve over strong text-only baselines in the FINDINGS $\to$ IMPRESSION transformation. We challenge two prevailing assumptions: (1) that more visual input is always better, and (2) that multimodal models add limited value when findings already contain rich image-derived detail. Through controlled ablations on MIMIC-CXR benchmark, we show that selectively focusing on pathology-relevant visual patches rather than full images yields substantially better performance. We introduce ViTAS, Visual-Text Attention Summarizer, a multi-stage pipeline that combines ensemble-guided MedSAM2 lung segmentation, bidirectional cross-attention for multi-view fusion, Shapley-guided adaptive patch clustering, and hierarchical visual tokenization feeding a ViT. ViTAS achieves SOTA results with 29.25% BLEU-4 and 69.83% ROUGE-L, improved factual alignment in qualitative analysis, and the highest expert-rated human evaluation scores. Our findings demonstrate that less but more relevant visual input is not only sufficient but superior for multimodal radiology summarization.