Google Creates AI That Could Diagnose Diseases Just By Listening

© Andrea Piacquadio/Pexels

Healthcare is a sector where artificial intelligence has has already largely started to prove itself. The Foresight tool developed by the Institute of Psychiatry, Psychology & Neuroscience, the advances of the start-up Owkin or even the detection early Alzheimer's; examples abound in recent times.

This time, it's Google that's hitting hard. The Silicon Valley giant has just unveiled an AI capable of diagnosing diseases by simply listening to the sounds of the human body. A feat that combines bioacoustics – this discipline combining biology and acoustics – with the power of machine learning.

Millions of sounds to save lives

At the heart of this innovation is HeAR (Health Acoustic Representations), an AI model fed by no less than 300 million sound samples. Coughs, sniffles, sneezes, breathing… So many bodily noises gleaned from YouTube or recorded in hospitals, like the one in Zambia where patients came to be screened for tuberculosis.

Shravya Shetty, director of engineering research at Google in Mountain View, explains: “Our AI detects early signs of illness based on subtle differences in cough patterns. The goal is to enable even people with little training to detect respiratory illnesses.” A boon for regions where sophisticated diagnostic equipment like X-ray machines are in dire need. Sub-Saharan Africa, Southeast Asia or Latin America, areas of the globe that could greatly benefit from this technology.

Tuberculosis in the crosshairs

This system could well revolutionize the fight against tuberculosis, a real scourge that still claims nearly 4,500 lives every day, according to figures from the World Health Organization. In India, where the disease kills 250,000 people each year, this early detection would be a major asset in stopping its spread.

200% Deposit Bonus up to €3,000 180% First Deposit Bonus up to $20,000

This is precisely where Salcit Technologies, an Indian startup specializing in AI for respiratory health, comes in. By combining Google's model with its own AI called Swaasa (which means “breath” in Sanskrit), Salcit is refining the accuracy of tuberculosis diagnosis and lung health assessment.

A pocket stethoscope

The best part is that this AI model is compatible with any smartphone. In fact, the Swaasa mobile application allows users to record a simple 10-second cough sample. “It's the equivalent of taking a blood sample, but processed in the cloud instead of in a lab,” says Salcit co-founder Manmohan Jain. With 94% accuracy and a cost of just Rs 200 (around €2.40), compared to Rs 3,000 (€32.15) for a traditional spirometry test (a test that measures the lungs' ability to inhale and exhale air), the technology could potentially democratize the screening.

Already adopted by regulators and major players in Indian healthcare such as Apollo Hospitals and the Healing Fields Foundation, This innovation has already won the trust of some institutions.

However, the path to widespread adoption still seems a bit distant. Changing established clinical practices is no small feat, and the tool will have to prove itself to gain the trust of the medical profession. Not to mention other technical challenges, such as managing ambient noise during recordings or getting the application to work for populations unfamiliar with the technology. Of course, these models are not yet mature enough to consider large-scale commercialization, but they still provide a glimpse of a future where early diagnosis would be accessible to all.

• Google is developing an AI capable of detecting diseases by analyzing body sounds.
• The tool is mainly aimed at screening for tuberculosis in regions lacking advanced medical equipment.
• This technology could democratize early diagnosis by making it accessible and affordable on a large scale.

📍 To not miss any Presse-citron news, follow us on Google News and WhatsApp.

325.2k reviews

[ ]