End-2-End COVID-19 Detection from Breath & Cough Audio

TL;DR

This paper presents the first end-to-end deep learning approach for COVID-19 detection from breath and cough audio, demonstrating promising results and scalability potential for population-wide testing.

Contribution

It introduces a novel deep neural network model, CIdeR, for diagnosing COVID-19 from audio, and provides a publicly available dataset and evaluation framework.

Findings

Achieved ROC-AUC of 0.846 in COVID-19 detection

Proposed a scalable, cost-effective audio-based testing method

Released data splits and model details for reproducibility

Abstract

Our main contributions are as follows: (I) We demonstrate the first attempt to diagnose COVID-19 using end-to-end deep learning from a crowd-sourced dataset of audio samples, achieving ROC-AUC of 0.846; (II) Our model, the COVID-19 Identification ResNet, (CIdeR), has potential for rapid scalability, minimal cost and improving performance as more data becomes available. This could enable regular COVID-19 testing at apopulation scale; (III) We introduce a novel modelling strategy using a custom deep neural network to diagnose COVID-19 from a joint breath and cough representation; (IV) We release our four stratified folds for cross parameter optimisation and validation on a standard public corpus and details on the models for reproducibility and future reference.