AI Model Detects Respiratory Diseases Through Cough Sounds, Enhancing Global Healthcare

A recent report highlighted by Google Research reveals that artificial intelligence is being harnessed to detect respiratory diseases such as tuberculosis (TB) and chronic obstructive pulmonary disease (COPD) by analysing cough sounds. The AI model, named Health Acoustic Representations (HeAR), was developed by Google and launched in March.

HeAR's Capabilities

HeAR is trained on 300 million audio samples, including roughly 100 million cough sounds. This extensive dataset allows the model to identify patterns in health-related sounds, offering a new frontier in medical audio analysis. According to Shravya Shetty, Google Research’s Director of Engineering, HeAR consistently outperforms other models across various tasks, showing its ability to generalise across different audio inputs.

Application in Tuberculosis Detection

Salcit Technologies, a respiratory healthcare company based in India, has integrated HeAR into its Swaasa app. The app analyses cough sounds to detect TB, a treatable yet often undiagnosed disease due to limited access to healthcare. By leveraging AI, Swaasa aims to improve TB diagnosis and make healthcare more accessible and affordable worldwide.

Zhi Zhen Qin, a digital health specialist with the Stop TB Partnership, praised the potential of AI-powered tools like HeAR, stating that they could revolutionise TB screening and detection by providing low-impact, accessible solutions to those in need.

AI's Role in Early Disease Detection

While AI-generated deepfakes dominate much of the discussion around artificial intelligence, the technology’s promise in early disease detection continues to drive research. In June, the AI model EMethylNET, developed by researchers at the University of Cambridge, showed potential in early cancer detection, diagnosis, and treatment planning.

Dr. Gabriel Zada, a neurosurgery specialist at Keck Medicine of USC, emphasised the importance of AI in determining tumour types during surgery, which can significantly impact treatment decisions. Similarly, New York-based medtech company Ezra uses AI to scan vital body areas for early cancer detection, improving patient outcomes dramatically.

Dr. Daniel Sodickson, Ezra's medical advisor, noted that early detection increases five-year survival rates for cancer from 20% to 80%, potentially saving billions of lives.