Seeing What Matters: Lesion-Aware High-Resolution Patch Discovery and Fusion for Chest X-ray Report Generation

2026-07-07

AI Summary

Existing chest X-ray report generation methods are constrained by the fixed token budget of visual encoders, forcing the use of low-resolution inputs that often omit critical lesion details. This work proposes LePaX, a framework that emulates radiologists’ diagnostic workflow of “localize first, scrutinize later.” LePaX introduces Learnable Spatial Resolution Allocation (LSRA) and Global-Regional Fusion (GRF) mechanisms to achieve efficient high-resolution perception without increasing token count. Coupled with a spatially aligned resolution back-projection strategy, LePaX significantly outperforms current approaches across multiple chest X-ray benchmarks, achieving superior clinical accuracy and language generation quality while using less than one-tenth the visual tokens required by naive high-resolution patching.

Abstract

Despite rapid advances in chest X-ray (CXR) foundation models, most radiology report generation (RRG) systems still rely on heavily downsampled inputs (e.g., 256x256) due to the fixed visual token budgets of pretrained vision encoders, suppressing subtle yet clinically important cues present in native-resolution images. However, enabling high-resolution (high-res) perception remains challenging: naive tiling causes prohibitive token inflation, while global compression suppresses subtle lesions and degrades diagnostic fidelity. Inspired by radiologists' workflow, localizing suspicious regions before detailed high-res assessment. We propose Lesion-Aware High-Resolution Patch Discovery and Fusion for Chest X-ray Reporting (LePaX), the first RRG framework that enables efficient high-res CXR perception (up to 1920x1920) without increasing the vision-token count. LePaX formulates high-res perception as a constrained spatial resolution allocation problem under a fixed token budget and introduces two key components: Learnable Spatial Resolution Allocation (LSRA), which learns a spatial utility map that adaptively allocates limited high-res capacity to diagnostically relevant regions, enabling targeted extraction of high-res patches from native CXRs; and Global-Regional Fusion (GRF), which performs token-preserving region-to-global refinement by projecting high-resolution regional evidence back onto the global feature grid through spatially aligned resolution write-back, avoiding token inflation. Experiments on multiple CXR benchmarks demonstrate that LePaX consistently improves both clinical and linguistic metrics while enabling native-resolution CXR perception with over 10x fewer visual tokens than naive high-res tiling.

Problem

• chest X-ray report generation
• high-resolution perception
• lesion-aware
• visual token budget
• radiology report generation

Innovation

• high-resolution perception
• lesion-aware allocation
• spatial resolution allocation
• token-efficient fusion
• chest X-ray report generation