Google AI will recognize Lung Health Issues and Tuberculosis from Cough Sounds

Google and Salcit Technologies, an Indian respiratory healthcare company, have announced their partnership. An artificial intelligence tool called Swaasa was created by Salcit to evaluate lung function and analyse cough noises. As a result of this partnership, Swaasa will improve the precision of its tuberculosis detection models by utilising Google's HeAR (Health Acoustic Representations), a bioacoustic AI model that was made available in March.

HeAR, a foundational AI model, has been trained on 300 million pieces of audio data from diverse and de-identified sources. Specifically, the cough model within HeAR was trained on approximately 100 million cough sounds. According to Shravya Shetty, Director and Engineering Lead at Google Health, HeAR is capable of detecting subtle differences in cough patterns, significantly improving diagnostic accuracy and speeding up the detection process for tuberculosis.

HeAR is not limited to detecting a single disease; it is intended as a general acoustic model that can be fine-tuned for various health-related sounds and use cases. The model’s goal is to power screening tools that can assist physicians and health workers in identifying respiratory diseases using just the microphones on smartphones.

Researchers interested in using HeAR for further developments can request access to the HeAR API, which will be available via Google Cloud. Google hopes this will pave the way for advanced diagnostic tools and monitoring solutions for tuberculosis, lung diseases, and more, ultimately improving global health outcomes.

Google’s commitment to AI-driven healthcare extends beyond respiratory diseases. In March, the company partnered with Apollo Radiology International (ARI) in India to deploy AI models for early detection of tuberculosis, lung cancer, and breast cancer. Over the next decade, ARI plans to provide three million free AI-powered screenings for these diseases.

Additionally, Google is collaborating with Taiwan’s Chang Gung Memorial Hospital (CGMH) to develop AI models that can detect early signs of breast cancer through ultrasound images. This approach aims to make breast cancer screening more accessible, particularly in low-resource environments.