Coughing can be a sign of health issues, but what if AI could tell exactly what’s wrong just by listening to your cough? Google Research recently shared a report showing how a team in India is using an AI model to identify diseases like tuberculosis (TB) and chronic obstructive pulmonary disease (COPD) from the sound of a cough.
Google launched its Health Acoustic Representations (HeAR) model in March. This model was trained on 300 million audio samples, with around 100 million focused on cough sounds. It’s designed to pick up patterns in health-related sounds, such as how someone coughs or speaks.
The AI can also analyze tone, pitch, and speed in speech to detect issues like dementia. According to Google Research Director Shravya Shetty, HeAR performs better than other models across different tasks and microphones.
In India, Salcit Technologies uses HeAR in their Swaasa app to listen to coughs and detect tuberculosis. This is crucial since many TB cases go undiagnosed due to limited healthcare access. AI tools like HeAR could help make diagnosis easier, more affordable, and accessible worldwide.
While AI discussions often focus on deepfakes, advancements like this show how AI can be a powerful tool for early detection of serious diseases.
Source: decrypt
