In this article, the author talks about the recent research that supports the use of contemporary machine learning (ML) and deep learning (DL) technologies to diagnose COVID-19 using aspects of coughing audio. The article is also backed up with scientific findings.

The severity of illness and accompanying mortality rates are growing at an alarming pace over the globe as a result of the COVID-19 pandemic caused by the coronavirus SARS-CoV-2. Non-invasive testing that is both reliable and quick has become critical for COVID-19 diagnosis and monitoring data.

While reverse transcriptase-polymerase chain reaction (RT-PCR) testing is the most often employed approach, coughing sounds have been shown to offer important COVID-19 signatures that may be used to assist quick, noninvasive, and reliable screening procedures. As a result, spectral and waveform fingerprints linked with disease-related coughing patterns are being evaluated as prognostic biomarkers. 

Study findings 

1. The app detects important characteristics in coughs using machine learning. The developer tested it on 741 patients in the United States and India, including 446 with COVID-19. Participants filled out questionnaires about any symptoms they were having and coughed them into a smartphone equipped with the app.
2. According to early data, the app was likewise deemed "excellent," with an area under the curve of 0.9 for its ability to discern between two diagnostic groups. To validate that its algorithms correctly identify COVID-19, they were compared to the Breathe Easy dataset, which contains data from over 1,000 patients with a variety of non-COVID-19 respiratory diseases.
3. Additionally, it said that the app demonstrated "constant performance" during two time periods, one during which Delta was the dominant variety and another during which Omicron was the dominating variant. The findings indicated that the app properly identified COVID-19 in 92% of persons infected. By contrast, lateral flow tests are capable of detecting around 72% of symptomatic patients and 58% of asymptomatic instances.

The company said that the findings of its clinical study would be published in a peer-reviewed publication "in the next weeks."

Why does it matter?

The WHO has cautioned that the worldwide pandemic is far from finished and that health systems must be prepared to deal with increasingly hazardous varieties.

Due to the high sensitivity and specificity of the smartphone-based test, fast antigen or polymerase chain reaction (PCR) testing would be needed less often to confirm COVID-19 in these contexts.

Application of machine learning and algorithms to illness diagnosis

Similar to how a doctor's stethoscope works, the instant screening test utilises cough sounds to determine the presence of illness. The app detects 'cough signatures' similar to those detected by a stethoscope and eliminates the need for human interpretation.

The machine learning method has resulted in the development of algorithms for diagnosing illness based on cough and respiratory sounds, using a library of signatures from previously diagnosed clinical conditions.

According to a study of 741 individuals (of whom 446 were positive), the innovative technology effectively-identified COVID-19 in 92 per cent of those with infection, 'exceeding the real-world measured sensitivity of fast antigen testing. The app will be able to ensure that any COVID-19 and COVID-19 screening test developed accurately identifies COVID-19 and not other respiratory diseases by utilising a large dataset of clinically labelled non-COVID-19 cough samples, including coughs from patients with non-COVID-19 lower respiratory tract infections.

Disclaimer- The views and opinions expressed in this article are those of the author's and do not necessarily reflect the official policy or position of M3 India.

The author is a practising super specialist from New Delhi.