This groundbreaking research combines large language models (LLMs) with non-negative matrix factorization (NMF) to improve cardiorespiratory sound separation - critical for accurate medical diagnostics.

• First integration of LLMs with NMF for medical sound separation
• Employs LLMs in a dual role: providing detailed insights for disease prediction and optimizing the NMF algorithm
• Delivers enhanced separation of heart and lung sounds from mixed recordings
• Creates a feedback loop system that continuously improves separation quality

This innovation matters because clearer separation of heart and lung sounds enables more accurate diagnosis of cardiorespiratory conditions, potentially improving clinical outcomes through earlier and more precise detection of abnormalities.

Large Language Model-based Nonnegative Matrix Factorization For Cardiorespiratory Sound Separation