In a groundbreaking advancement in artificial intelligence, Google is leveraging audio signals to detect the early symptoms of illness. Using 300 million audio samples—ranging from coughs to labored breathing—Google has trained its AI foundation model to recognize signs of diseases like tuberculosis. In collaboration with Salcit Technologies, an AI startup in India focused on respiratory healthcare, Google is working to integrate this technology into smartphones. This innovation has the potential to transform healthcare in high-risk, underserved communities by offering more accessible diagnostic tools.

Google has a history of digitizing human senses, and its venture into bioacoustics is an extension of that. Bioacoustics, which merges biology and acoustics, uses AI to interpret sounds made by humans and animals to gain medical insights. In this field, Google’s new AI model, HeAR (Health Acoustic Representations), stands at the forefront. This model utilizes sound signals to detect early signs of illness, offering a promising tool for early diagnosis.

Generative AI, the same technology that powers ChatGPT, is enhancing bioacoustics, giving it new capabilities to monitor health. The HeAR model is designed to be deployed on smartphones, enabling it to track and screen individuals in areas where diagnostic tools like X-rays are unavailable.

Tuberculosis (TB) remains a major health challenge, causing around 4,500 deaths and 30,000 new infections each day, according to the World Health Organization. Early detection is crucial, yet millions of cases go undiagnosed, particularly in regions with limited healthcare access. In India, TB claims nearly a quarter-million lives annually, highlighting the urgent need for effective screening.

To address this, Google’s AI model has been trained on a massive dataset of 300 million audio samples, including coughs and breathing sounds from global sources. These sounds were gathered from publicly available, non-copyrighted materials such as YouTube videos and TB screenings in Zambian hospitals. By analyzing subtle differences in cough patterns, the AI can identify early signs of TB, potentially saving lives through early intervention and treatment.

Google’s collaboration with Salcit Technologies is poised to improve TB diagnosis and overall lung health assessments. Salcit is merging Google’s HeAR model with its own machine learning system, Swaasa, which is named after the Sanskrit word for breath. This partnership is expected to significantly enhance the monitoring and management of respiratory conditions, especially in remote areas with limited healthcare infrastructure.

The use of AI to detect diseases through sound signals marks a significant leap forward in medical technology. As AI models like HeAR evolve, they may expand their diagnostic capabilities beyond tuberculosis, detecting a wide range of respiratory illnesses and even cardiovascular conditions through sound analysis.

This innovative approach could revolutionize healthcare delivery, making advanced diagnostics more accessible to underserved populations around the world.

By Impact Lab