Deep reinforcement learning framework for thoracic diseases classification via prior knowledge guidance

Weizhi Nie, Chen Zhang, Dan Song, Lina Zhao, Yunpeng Bai, Keliang Xie, Anan Liu

Abstract: Highlights•Novel multi-agent reinforcement learning framework for abnormality classification.•Guided by prior knowledge, our diagnostic agent boosts performance.•Transformer-based module addresses agent correlation in multi-info fusion.•Exploration-added model tackles classification, similarity, and efficiency.•Clinical evaluations confirm alignment with professional expectations.