Boosting Phonocardiogram Classification Performance with Function Generated Data

Naoki Nonaka, Hiroshi Seki, Tomohiro Komatsu, Jun Seita

Keywords: Synthetic medical data, Phonocardiogram (PCG) analysis, Deep neural networks (DNNs), Cardiac disease classification

TL;DR: Pre-training with synthetic heart sound data and subsequent fine-tuning with real-world data improves cardiac abnormality detection accuracy by up to 17.1% compared to training from scratch, demonstrated across eight DNN architectures.

Track: Proceedings

Abstract:

Deep neural networks require large datasets, yet medical phonocardiogram (PCG) data are scarce due to privacy and disease rarity.
To address this challenge in PCG analysis, we present a function-generated PCG pipeline that synthesizes S1/S2 heart sounds with modulated noise to emulate aortic stenosis (AS), aortic regurgitation (AR), and mitral regurgitation (MR).
Across eight architectures, we compare real-only training, synthetic-only, and synthetic pretraining followed by real fine-tuning (Syn$\rightarrow$Real).
Syn$\rightarrow$Real consistently improves AUROC with average gains of $+15.3\%$ (AS), $+17.0\%$ (AR), $+17.1\%$ (MR) on BMD-HS, and $+7.1\%$, $+8.8\%$, $+6.1\%$ on a private cohort ($8,564$ recordings).
Furthermore, we show Syn$\rightarrow$Real is competitive with pretraining on out‑of‑domain real data, and combining it with multi‑stage real fine‑tuning yields the best overall performance, highlighting the complementary value of synthetic and real PCGs.
While synthetic‑only training generalizes poorly, pretraining on function‑generated PCGs consistently improves PCG classification over training from scratch, offering a practical path to mitigate data‑collection burdens and potentially reduce privacy and ethical exposure.

General Area: Models and Methods

Specific Subject Areas: Supervised Learning, Time Series

Data And Code Availability: Yes

Ethics Board Approval: No

Entered Conflicts: I confirm the above

Anonymity: I confirm the above

Submission Number: 40