New Delhi: Salcit Technologies, an India-based respiratory healthcare firm, has collaborated with Google to harness AI’s power to detect Tuberculosis and other diseases from cough sounds.
In 2020, Salcit Technologies developed ‘Swaasa’ which uses AI to analyze cough sounds and assess lung health. Under the partnership, Swaasa will utilize Google’s bioacoustic foundational AI, HeAR (Health Acoustic Representations) to improve the accuracy of its tuberculosis AI models.
HeAR was launched in March 2024 to help researchers create models that can listen to human sounds and identify early signs of diseases. Shravya Shetty, director and engineering lead at Google Health spoke about the efficiency of HeAR. She explained that the AI model has been trained on 300 million pieces of audio data created from a wide range of datasets. The cough model in particular has been trained on around 100 million cough sounds. The model can detect subtle differences in cough sounds and help in the early detection of respiratory diseases. In addition, it can be fine-tuned for different health-related sounds and use cases.
Google has made the HeAR API available for researchers via a Google Cloud API. According to Shetty, the goal is to power various screening tools to help doctors and physicians detect respiratory diseases by simply using microphones on their smartphones.
