Covid-detection through AI: Phase-1 of study in Mumbai’s NESCO facility ends; voice samples of 2,000 patients collected

The NESCO jumbo Covid-19 centre at Goregaon has completed the first phase of screening suspected Covid-19 patients through their voice samples by using artificial intelligence (AI). The centre has collected 4,000 voice samples from 2,000 suspected Covid-19 patients for the pan-India study and a preliminary report on the research is expected to be released next month.

The study commenced in August with an Israel-based company. The report is likely to establish the effectiveness of the first-of-its kind diagnostic tool to detect Covid-19.

“We have collected 4,000 samples from 2,000 patients. Each patient has given two voice samples. The first sample was given when the patients were suspected of having Covid-19 during their admission to the centre, while the second sample was given after their recovery,” said Dr Neelam Andrade, dean of Nair Dental Hospital, who is in-charge of the NESCO facility.

The samples have been sent to a firm Vocalis Health for analysis.

“Along with the voice samples, we have also conducted Covid-19 tests on suspected patients. This will help experts to identify if the voice analysis technique can actually detect infected patients,” said Dr Andrade.

According to the experts, the study is based on the hypothesis that the novel coronavirus infection may affect the voice of individuals, given the interdependence between the respiratory and speech systems in the body. When symptoms of Covid-19 manifest, an individual starts developing breathing problems that affect the amount of air exhaled. This exhaled air interacts with inflamed muscles on its journey to produce voice or speech. These interactions impact voice modulations — measurable qualities that form the basis of their biomarkers. Depending on its variations, an individual can be detected with Covid-19.

An individual will be diagnosed based on vocal biomarkers (VB). For instance, if a person’s VB is below the standard point — 0.5, they will be considered negative for the virus. But if anyone records VB above the standard point, the patient will be suspected to be infected with Covid-19 and an RT-PCR test will be conducted for confirmation.

“With lockdown restrictions being lifted, this diagnosis can prove to be helpful as it provides reports in just 30seconds. The preliminary report will be available next month. The final reports, which will come in January, will help in deciding the effectiveness of the technique,” said Suresh Kakani, additional commissioner, BMC.

In the pan-India study, 10,000 samples will be collected all across the country. “If it shows promising results, we have plans to use the programme in hospitals as well. It is non-invasive and can deliver results quicker than other tests,” said Kakani.