AI can detect tuberculosis early by listening to your cough

The same underlying technology that powers hugely popular generative AI models, such as those from major tech companies like OpenAI, is now being used to scan for early signs of lung disease. Google, one of the leaders in new AI models, is working with a healthcare startup that analyzes massive data sets of coughs and sneezes to detect signs of tuberculosis or other respiratory illnesses before they worsen. It’s one of many ways rapidly evolving technology is rapidly changing early disease detection across the healthcare sector. However, what happens once the initial diagnosis is made still requires substantial human clinical expertise.

Google’s massive database of coughs and stuffy noses

Earlier this year, Google released details about what they called a new self-supervised deep learning model for healthcare Representation of health acoustics (To belong). The model was trained on approximately 300 million two-second audio clips of people coughing, sneezing, breathing and sniffling. This varied set of audio clips was reportedly collected from non-copyrighted public data from around the world. For a sense of scale, the cough model alone was trained on 100 million different cough sounds. All of this data should, in theory, show patterns of what a healthy respiratory system sounds like. The trained AI model can then use that knowledge to look for abnormalities in a new patient audio clip that could indicate a potential health risk.

More recently Google announced in a blog post, it had partnered with an India-based respiratory healthcare startup called Salcit Technologies to apply these findings in the real world to look for early signs of tuberculosis. Bloomberg reported this week about the collaboration. Salcit has its own product called Swaasa, which allows users to record an audio file of them coughing using their mobile device’s microphone. An AI model then compares that audio to a database of cough sounds to look for indicators of the fatal but treatable disease. From there, patients can decide whether they want to see a doctor for further treatment. By merging their proprietary model with HeAR, the two companies expect to increase the product’s effectiveness and accuracy for early detection of respiratory diseases. 1.3 million people worldwide reportedly died of tuberculosis in 2022. India responsible for almost 25% of these deaths annually.

The predictive properties of AI help healthcare professionals detect different types of diseases faster. Research has already shown that these models can be effective screening for possible cancerous tumors that might otherwise go unnoticed. Similar models are also used to detect early signs of breast cancer, look for early signs of breast cancer, myopiaAnd heart disease. Radiologists are already using GenAI tools accelerate the pace of medical imaging analysis. AI’s impact on diagnoses could even extend beyond chronic conditions that are typically noticed later in life. Last year, researchers at the University of Louisville discovered created an AI system they say they can parse MRI scans of toddlers to predict with 98.5% accuracy whether or not they will be clinically diagnosed with autism.