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

Pengfei Zhang, Tianxin Xie, Minghao Yang, Li Liu · 2026

Read Resp-Agent: An Agent-Based System for Multimodal Respiratory Sound Generation and Disease Diagnosis by Pengfei Zhang, Tianxin Xie, Minghao Yang, Li Liu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-15909.txt.

Description

Zhang and colleagues built Resp-Agent, a self-run multimodal system for lung sound generation and disease diagnosis. Lung sound work by deep learning faces two blocks. Turning signals into spectrograms drops brief sound events and clinical context, and training data is scarce with stark class imbalance.

A central controller, Thinker-A2CA, spots weak diagnoses and queues targeted synthesis in a loop. A Diagnoser weaves clinical text with audio tokens through broad attention and sparse audio anchors, catching long-range context and millisecond transients. A Generator adapts a text-only language model to make hard cases, splitting disease content from sound style. Resp-229k, a set of 229k recordings with distilled clinical notes, backs the work. Tests beat prior methods across varied settings, lifting robustness under scarce and skewed data.

Catalog record

CREATORPengfei Zhang, Tianxin Xie, Minghao Yang, Li Liu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE69 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-15909

Browse the full library · About the Tower of Babel