Mumbai students develop AI-based voice tool to detect COVID-19, Rome university conducting pilot run
New Delhi, Apr 16 () A university in Rome is conducting a pilot run for a patented artificial intelligence-based tool developed by three biotechnology students and a professor from Mumbai, which they claim can test COVID-19 through voice-based diagnosis using a smartphone.
The tool is being tested by the University of Tor Vergata in Rome and has already been tested on 300 individuals, and has yielded 98 per cent accurate results.
According to the students and the professor from DY Patil Institute of Bio Technology and Bio Informatics, Mumbai, the tool is based on a voice-based diagnosis through an app.
The team includes bioinformatics students -- Rashmi Chakraborty, Priyanka Chauhan and Priya Garg.
"While several foreign universities are trying to launch a voice-based AI tool for COVID-19 detection, this Indian tool is fully functional and currently in use in Italy to successfully identify COVID-19 patients. The students have a full-fledged working software with a rich database of patients and healthy samples. This tool is being currently used by the University of Rome to detect COVID-19 patients with 98 per cent accuracy," professor Santosh Bothe, who supervised the project, told .
"As someone speaks to the microphone on the app, the tool breaks down the voice in multiple parameters such as frequency and noise distortion. These values are then compared to a normal person's values and the patented technique then determines if the patient is positive or not," he added.
A team at the Indian Institute of Science (IISc), Bangalore, is also working on a diagnostic tool based on analysis of cough and respiratory sounds.
According to Giovanni Saggio, professor at the Rome university's Engineering department, the audio-based disease diagnosis tool can find coronavirus from the timbre of the voice.
"Each human voice has 6,300 parameters, and only a few units, less than a dozen, specifically characterize individuals. The human ear, apart from colds, is not able to distinguish them, but artificial intelligence does. Each one of our internal organs is sort of a resonator, so if we have a problem with our lungs or our heart, this is reflected in our voice," Saggio told .
"The same person has one voice when they are healthy, and another if they have a pathology. Since coronavirus compromises lungs and airwaves, the voice is definitely affected. The current novel coronavirus cases could be picked this way," he added.
Student researcher Priya Garg believes this tool can be of great impact in doing the first level of screening to identify positives and only those who tested positive can go for the lab tests.
"This can reduce the current bottleneck on the medical infrastructure, help the government to identify hot spot regions in advance through location tracking integration and is possibly the best way to reach out to the remotest part of India by testing through a smartphone," she said.
"When the country's medical infrastructure and the whole medical community is under immense pressure, this might be a great solution to reach patients remotely, without the risk of exposure, both to the patient and the lab personnel. While we don't propose the tool to replace current testing methods, the simplicity of this test can be used to screen and identify the COVID-19 positive cases at a very early stage of inception and hence contain the spread further," Garg added.
Her co-developer Rashmi Chakraborty said, "This voice-based diagnosis will fetch zero cost testing to the patient and no wait time. Diagnosis of COVID-19 positive patients can be done in minutes and medical advice on the next steps can be sought through a video or audio call to the listed doctors."
COVID-19 has killed at least 1,31,319 people around the world, with Europe accounting for more than two-thirds of the deaths. Europe has reported 88,716 deaths, while the United States has the single highest toll at 26,950. The US is followed by Italy with 21,645 deaths, Spain with 18,579 and France with 17,167.