MambaGen: Efficient visual representation learning for automatic radiology report generation

Xiaodi Hou, Xiaobo Li, Simiao Wang, Mingyu Lu, Hongfei Lin, Yi-Jia Zhang

Abstract: Highlights•Proposes MambaGen, the SSM-based automatic radiology report generation model for modeling long-range dependencies with linear computational complexity.•Designs a mamba-visual recalibration module, which utilizes a twostage training strategy to effectively capture the efficient visual representation of medical images.•Outperforms competitive baselines through extensive experiments on publicly available datasets.

External IDs:dblp:journals/eswa/HouLWLLZ26