Resp-Agent: An Agent-Based System for Multimodal Respiratory Sound Generation and Disease Diagnosis

Researchers introduced Resp-Agent, an AI system that uses multimodal deep learning to generate respiratory sounds and diagnose diseases. The system addresses data scarcity and representation gaps in medical AI through an autonomous agent-based approach and includes a new benchmark dataset of 229k recordings.

• →Resp-Agent uses an Active Adversarial Curriculum Agent to actively identify diagnostic weaknesses and schedule targeted synthesis.
• →The system combines clinical text with audio tokens to capture both long-range clinical context and millisecond-level acoustic details.
• →A new benchmark dataset Resp-229k containing 229k respiratory recordings paired with clinical narratives was released.
• →The system outperforms existing approaches in diagnostic accuracy under data scarcity conditions.
• →The research addresses fundamental challenges in medical AI including information loss and limited training data availability.