Can Your Cough Reveal COVID‑19? AI Model Detects Asymptomatic Infections via Smartphone

Although the current coronavirus is less destructive than the 2002 SARS outbreak, its danger lies in a prolonged asymptomatic incubation period, allowing infected individuals to unknowingly spread the virus; strong preventive measures, especially testing and isolating asymptomatic carriers, are essential to control large‑scale outbreaks.

While nucleic‑acid testing remains the primary detection method, a recent MIT News release announced a novel approach that requires only a cough sound recorded on a smartphone, with an artificial‑intelligence model determining whether the user is infected with COVID‑19.

The model works by having patients submit cough recordings via a browser, phone, or computer; a neural‑network trained on tens of thousands of cough and speech samples distinguishes asymptomatic individuals from healthy ones. When presented with new recordings, the system correctly identified 98.5% of confirmed COVID‑19 coughs, including 100% of asymptomatic coughs (recordings from people without symptoms but who tested positive).

The research team is now integrating the model into a user‑friendly application. If the FDA approves widespread use, the app could become a free, convenient, non‑invasive testing tool, allowing users to cough into their phone daily and receive immediate feedback on their infection risk.

The study was published in the IEEE Journal of Medical and Biological Engineering .