Exploring Finetuned Audio-LLM on Heart Murmur Features

TL;DR

This paper examines finetuning an audio large language model (Qwen2-Audio) for phonocardiogram (PCG) analysis to predict 11 expert-labeled murmur features beyond binary healthy/unhealthy classifications. It introduces a SSAMBA-based PCG segmentation front-end to boost noise robustness and generalization, and integrates a LoRA-finetuned Qwen2-Audio encoder with a Whisper-based frontend for PCG reasoning. Key contributions include achieving state-of-the-art performance on the majority of features, effective handling of long-tail murmur features with limited data, and demonstrated zero-shot robustness across external PCG datasets. The work highlights the potential of audio LLMs as clinician-support tools to enhance cardiovascular disease diagnosis, while recognizing that expert medical oversight remains essential for clinical decision-making.

Abstract

Large language models (LLMs) for audio have excelled in recognizing and analyzing human speech, music, and environmental sounds. However, their potential for understanding other types of sounds, particularly biomedical sounds, remains largely underexplored despite significant scientific interest. In this study, we focus on diagnosing cardiovascular diseases using phonocardiograms, i.e., heart sounds. Most existing deep neural network (DNN) paradigms are restricted to heart murmur classification (healthy vs unhealthy) and do not predict other acoustic features of the murmur such as timing, grading, harshness, pitch, and quality, which are important in helping physicians diagnose the underlying heart conditions. We propose to finetune an audio LLM, Qwen2-Audio, on the PhysioNet CirCor DigiScope phonocardiogram (PCG) dataset and evaluate its performance in classifying 11 expert-labeled murmur features. Additionally, we aim to achieve more noise-robust and generalizable system by exploring a preprocessing segmentation algorithm using an audio representation model, SSAMBA. Our results indicate that the LLM-based model outperforms state-of-the-art methods in 8 of the 11 features and performs comparably in the remaining 3. Moreover, the LLM successfully classifies long-tail murmur features with limited training data, a task that all previous methods have failed to classify. These findings underscore the potential of audio LLMs as assistants to human cardiologists in enhancing heart disease diagnosis.

Figures (1)

• Figure 1: Proposed PCG system incorporating a finetuned audio LLM Qwen2-Audio model with a front-end segmentation network