Now, an IISc project to 'hear' Covid-19 symptoms

The Project Coswara team from the Indian Institute of Science (IISc) is building a diagnostic tool for Covid-19 based on respiratory, cough and speech sounds. It requires users to provide a recording of breathing sounds, cough sounds, vowel sounds and count for five to seven minutes.

So far, the team has collected more than 1,100 voice samples. The Indian Council of Medical Research (ICMR) had in May reviewed their protocol and asked them to come back after sample collection for validation of their diagnostic tool.

Follow live updates on the coronavirus here

The collected data is analysed using signal processing and machine learning techniques. The sharpness and frequency of the voice help them distinguish between Covid-19 and non-Covid-19 patients. The team aims to release the diagnosis tool as a web or mobile application that provides a score to indicate the probability of infection.

The researchers are looking for volunteers — both healthy and Covid-19-positive — to provide voice samples through the web tool https://coswara.iisc.ac.in and reach a target pool of more than 2,000 samples. The data from healthy individuals will help the team set the right threshold.

Dr Sriram Ganapathy, assistant professor, IISc, who is leading the project told DH that production of sounds is compromised in Covid-19 patients. “We look for distinctive patterns in cough. Regular patients’ cough is harmonic in nature,” he said, adding that in Covid-19 patients, the cough has a lot of vibrations. “We analyse signals with computer algorithms and create a model that identifies these vibrations to detect Covid-19,” he added.

“A paper published by the University of Oklahoma, University of Michigan and a facility in Ukraine showed accuracy of 95%,” he said.