Google has developed an AI model named Health Acoustic Representations (HeAR) that is capable of detecting lung diseases, such as tuberculosis and chronic obstructive pulmonary disease, by analyzing cough sounds. HeAR is a bio-acoustic AI model that can derive critical health insights from acoustic patterns. HeAR is trained on 300 million audio samples and 100 million cough sounds. One of the standout features of HeAR is its ability to generalize across different microphones, demonstrating its superior ability to capture meaningful sound patterns compared to other AI models. The model’s robust and diverse training data allows it to perform well even with limited data. This is particularly important in the healthcare sector where data can often be scarce.
Salcit Technologies, an India-based respiratory healthcare company, that developed Swassa - an AI tool that analyzes cough sound to assess lung health, is already planning to leverage HeAR to enhance its AI-driven models for early TB detection. The company is planning to develop Swassa further with the help of HeAR to expand its screening capabilities across India. This collaboration highlights the potential of AI to make healthcare more accessible and affordable, particularly in regions where early diagnosis is crucial for effective treatment. Google is also offering HeAR API for researchers interested in exploring HeAR.
