Inviting proposals to develop an AI tool for interpretation of Line Probe Assay (LPA) results.

Cough Against COVID

Identifying The Risk Of COVID-19 From Cough Sounds

Until we find a cure or develop a vaccine for COVID-19, testing continues to be the first line of defense in the global fight against the pandemic. However, ramping up testing capacity is expensive and time consuming, particularly in rural and remote areas. There is a need for a simple, non-invasive technology to triage patients who may be infected with the SARS-CoV-2. We have built a cough analysis technology that could help in identifying at-risk patients, before lab-based tests, even if they’re asymptomatic.

The Problem

Almost 1 million people have died as a result of contracting COVID-19, while over 30 million have contracted the disease. The solution, as suggested by healthcare experts around the world, is to expand testing so those who catch the infection can be isolated and treated effectively. But ramping up testing comes with costs. Many parts of the world lack adequate supplies. It is an expensive undertaking that requires additional infrastructure and trained medical professionals, both of which may not be available. To add to this, the gold standard Real-time polymerase chain reaction (RT-PCR) test has a long lead time. This leaves the world’s population at risk, especially those with comorbidities. The effective utilisation of resources becomes critical.

After screening, there is a need for a triaging layer to identify those who have a high probability of testing negative for Covid-19, and thus enable better allocation of available tests. We have built an open-access AI tool that uses cough sounds, symptoms and contextual information to detect risk of being COVID-19 infected. Our model demonstrates that solicited-cough sounds can be used as a biomarker for the Novel Coronavirus. This holds true for both symptomatic and asymptomatic people.

With approvals from Governments of Bihar and Odisha and Municipal Corporation of Greater Mumbai (MCGM) and in partnership with Norway India Partnership Initiative (NIPI), Doctors for you (DFY), Krishna Institute of Medical Sciences, Karad (KIMSDU) and AIIMS Jodhpur, we collected over 3,500 solicited-cough sounds from COVID centres in four states of India. These included both COVID-positive and negative subjects, confirmed by the RT-PCR test. To the best of our knowledge, this is currently the largest COVID-19 cough dataset with verified ground truth labels from RT-PCR test results. At the current performance and at an assumed disease prevalence of 5%, our model can potentially increase testing capacity by 43%.

The study is supported by USAID and Bill & Melinda Gates Foundation.

Impact through AI

In a triaging setup, Cough Against COVID works on a basic smartphone and can be used to assess risk of being infected, without the need for any additional hardware or device. It provides instantaneous results, and will enable healthcare systems to create a funnel through which testing resources can be dedicated to those with a higher probability of being infected with COVID-19.