IISc Project Coswara

IISc Project Coswara: Diagnostic Tool To ‘Hear’ Covid-19 Symptoms

A Covid-19 diagnostic tool based on cough, respiratory, and speech sounds is being developed by researchers in the Project Coswara team from the Indian Institute of Science (IISc). Users are required to provide a recording of vowel sounds, cough sounds, breathing sounds, and count for five to seven minutes for the tool to diagnose Covid-19.

More than 1,100 voice samples have been collected by the team so far. For the validation of their diagnostic tool, the Indian Council of Medical Research (ICMR) had asked them to come back after sample collection, after reviewing their protocol in May.

Signal processing and machine learning techniques are used to analyze the collected data. Distinguishing between Covid-19 and non-Covid-19 patients is done with the help of the sharpness and frequency of the voice. The research team intents to launch and release the diagnosis tool as a web or mobile application through which a score to indicate the probability of infection can be provided.

The team wants to reach a target pool of more than 2,000 samples and are looking for both healthy and Covid-19-positive volunteers to provide voice samples through the web tool. The team

will be able to set the right threshold with the help of the data from healthy individuals.

Assistant professor at IISc, and the lead of the project, Dr. Sriram Ganapathy said that in Covid-19 patients, the production of sounds is compromised. Regular patients’ cough is harmonic in nature, while the cough has a lot of vibrations in Covid-19 patients, thus, the team looks for distinctive patterns in cough. In order to detect Covid-19, a model that identifies these vibrations is created by analyzing signals with computer algorithms

He added saying that an accuracy of 95% was shown in a paper published by the University of Oklahoma, a facility in Ukraine, and the University of Michigan.

 

 

Source
IISc Project Coswara: Diagnostic Tool To ‘Hear’ Covid-19 Symptoms and detect the disease