Beyond Classification: A Cough Regression Benchmark for Respiratory Acoustic Foundation Models

2026-06-13

AI Summary

This study addresses the lack of systematic evaluation of foundation models in regressing continuous health indicators—such as age, BMI, and disease probability—from cough audio, despite their strong performance in cough classification. The authors establish the first multi-model, multi-target cough regression benchmark, evaluating five foundation models (OPERA-CT/GT/CE, HeAR, and M2D+Resp) combined with linear, small MLP, and full MLP regression heads across three datasets (CoughVID, Coswara, and CIDRZ) using subject-independent splits. Results show that MLP-small outperforms linear probes in 23 out of 30 tasks; HeAR achieves an age prediction MAE of 9.12 years on Coswara; near-optimal performance can be attained with as few as 50 samples; and large-scale diverse pretraining effectively transfers to small clinical settings, whereas the reverse leads to significant degradation, revealing critical trade-offs among pretraining strategy, regressor capacity, and data scale.

Abstract

Respiratory acoustic foundation models (FMs) excel at cough classification, yet their ability to predict continuous health quantities from cough audio remains largely unexplored, despite the clinical value of passive age, BMI, and disease probability estimation in settings where physical measurements are unavailable. We introduce the multi-model, multi-target cough regression benchmark evaluating five FMs (OPERA-CT, OPERA-CE, OPERA-GT, HeAR, M2D+Resp) across six targets on three datasets under subject-disjoint protocols, comparing linear, MLP-small, and full MLP regression heads. MLP-small beats the mean-predictor baseline on all tasks and linear probing in 23 of 30 model x task cases, with full MLP overfitting on small clinical data but recovering on larger sets, revealing a dataset size x head-capacity trade-off. HeAR leads within-dataset age regression on Coswara (9.12 yr MAE); its CIDRZ result is excluded from headline claims owing to possible HeAR-CIDRZ pretraining overlap. OPERA-GT is favored over OPERA-CT on age in all three datasets, with the CIDRZ margin within seed variance, extending a generative-pretraining advantage from breath to cough. HeAR and M2D+Resp reach near-full performance at N = 50 samples while OPERA models require N = 400. Cross-dataset transfer is strongly asymmetric as large diverse data generalises to small clinical populations (CoughVID to CIDRZ: -0.17 yr) but not vice versa (CIDRZ to Coswara: +2.43 yr, +26.6%).

Problem

• cough regression
• respiratory acoustic foundation models
• continuous health prediction
• age estimation
• cross-dataset generalization

Innovation

• cough regression
• respiratory acoustic foundation models
• continuous health estimation
• cross-dataset transfer
• sample efficiency