Diagnosing asymptomatic covid-19 using only cough recordings

Article written By Laurence Goldberg

Researchers at Massachusetts Institute of Technology (MIT) have developed an artificial intelligence (AI) model that can distinguish between asymptomatic people with covid-19 and healthy individuals. 

One of the problems with containing SARS-CoV2 infection is the number of people who turn out to have had the disease without exhibiting symptoms who could, unwittingly, have passed it to others. A quick, cheap, non-invasive screening test could be useful for identifying such people at an early stage. Apparently there are differences between the coughs of asymptomatic covid patients and those of healthy individuals but these differences are not discernible to the human ear. 

Building on technology that had recently been developed for the diagnosis of Alzheimer’s disease the researchers developed an AI model that they ‘trained’ on thousands of forced cough recordings that people voluntarily submitted through web browsers and devices such as mobile phones and laptops. The model focuses on subtle changes in four biomarkers – vocal cord strength, sentiment, lung and respiratory performance and muscular degradation – and identifies a characteristic pattern of changes for covid-19.

The final model achieves 97.1% discrimination accuracy on subjects diagnosed with an official test.

The important implication of this is that the model could be built into a mobile device app for regular (e.g. daily) screening. A user could log in daily, cough into their phone, and instantly get information on whether they might be infected and therefore should confirm with a formal test.

“The effective implementation of this group diagnostic tool could diminish the spread of the pandemic if everyone uses it before going to a classroom, a factory, or a restaurant,” says co-author Brian Subirana, a research scientist in MIT’s Auto-ID Laboratory. This work was reported in the IEEE Open Journal of Engineering in Medicine and Biology (September 2020) . An audio description of the work is also available. This is helpful because it illustrates the different coughs and clearly shows how difficult it is for the human ear to distinguish between them.