Google has introduced a revolutionary AI system capable of diagnosing tuberculosis and other respiratory diseases by analyzing the sounds of coughs and breathing. This innovative model, developed using 300 million audio samples that include coughs, sneezes, and breathing patterns, aims to detect diseases like TB through subtle acoustic signals.
In collaboration with Salcit Technologies, an Indian AI startup specializing in respiratory healthcare, Google plans to integrate this technology into smartphones. This will make it accessible to high-risk populations in areas with limited healthcare facilities.
“HeAR AI model can detect subtle differences in cough patterns, which can enhance diagnostic accuracy and speed for identifying tuberculosis,” said Shravya Shetty, Director and Engineering Lead at Google Health, during a recent media roundtable. “By analyzing patients’ coughs, we could triage patients earlier in their healthcare journey, identifying those at higher risk and determining who needs follow-up testing to confirm tuberculosis.”
Once tuberculosis is initially diagnosed, it can be effectively treated. The HeAR AI technology operates by analyzing two-second audio snippets of coughing, sneezing, breathing, and sniffling to establish a baseline of what healthy respiratory sounds like. The AI model then searches for anomalies in new audio samples that might indicate a health issue.
The system compares the recorded sounds against a database of known coughs to identify signs of the disease. This allows patients to decide whether to seek further medical consultation. The collaboration between Google and Salcit Technologies aims to improve the accuracy and effectiveness of early respiratory illness detection.
Statistics reveal that tuberculosis caused approximately 1.3 million deaths globally in 2022, with India accounting for nearly 25% of these deaths annually.
AI models are also proving useful in screening for other conditions, such as cancer. Similar technologies are being employed to detect early signs of breast cancer, myopia, and heart disease. Radiologists are utilizing AI tools to expedite medical imaging analyses. For instance, researchers at the University of Louisville developed an AI system last year that, through MRI scans of toddlers, predicts autism with 98.5% accuracy.
