AI Models Powered by Google’s HeAR Technology Can Detect Early Signs of Diseases Like Tuberculosis Through Cough Sounds

Who doesn’t get annoyed when one is suffering from some form of cough? The irritation to pain to the sounds, we always have a tough time when we are coughing because of some sickness. However, have you ever thought that the sounds generated from our speech and breath can give medical specialists quite a lot of information about our health?

Bioacoustics sounds give out subtle clues, which medical professionals widely use to screen, diagnose, monitor, and manage various health conditions like Tuberculosis (TB) or even chronic obstructive pulmonary disease (COPD).

This is the age of advancement through AI. Researchers at Google understand the potential that sound holds when it comes to giving out health signals. Today, when there are high-quality microphones connected to smartphones, accessibility to such sounds is easy for researchers. Thus, researchers at Google have been venturing into the possibility of AI gathering health information from sonic data.

Sounds our bodies make — like coughs — contain valuable health information. Learn how AI models built using HeAR, a bioacoustic foundation model trained on 100 million cough sounds, can help flag early signs of diseases like tuberculosis ↓ https://t.co/Mrqgspa0vm

— Google (@Google) August 22, 2024

This year, Google presented the Health Acoustic Representations (HeAR). For the unversed, HeAR is a bioacoustics foundation model, with the help of which researchers can create models capable of listening to human sounds and deciphering signs of disease.

HeAR has received training on 300 million samples of diverse and de-identified datasets. Among these, about 100 million cough sounds alone were used to specifically train the cough model.

HeAR has managed to achieve high performance with less training data. According to reports by Google, models trained with HeAR have managed to achieve high performance with less training data. On the other hand, it stands above other models on a wide spectrum of tasks and when it comes to generalizing across microphones.

Researchers can now access HeAR to develop custom bioacoustics models with less data, setup, and computation.

Swaasa®, a product by Salcit Technologies (Indian based respiratory healthcare company) analyses cough sounds and assesses lung health with the help of AI. In the latest developments, it is now exploring the capabilities of HeAR to expand the horizon of its bioacoustics AI models. At its onset, with the help of HeAR, Swaasa® is trying to enhance its early detection of TB based on cough sounds.