Researchers are making AI-led efforts to establish voiceprints and other biomarkers for a host of pathologies
Little over a decade now, researchers have been using AI and machine-learning to spot biomarkers for, among other diseases, dementia, heart disease and depression. (Representative image)
For some time now, researchers have been trying to see if diseases can be linked to voiceprints—subtle variation with normal speaking that could give away an underlying physiological/psychological condition. For instance, from analysing the speech of normal individuals and those with the chronic obstructive pulmonary disorder (COPD), it is possible to map variations in the frequency of breathing to detect shortness of breath. So, an undiagnosed case can be mapped against data from hundreds of thousands of such speech samples to indicate, if not accurately diagnose, COPD. A report in Nature speaks of companies and research groups working on speech-based detection of Covid-19 (at least four companies across the globe are working on this). Some others are even working on mapping how a Covid-19 cough differs from other coughs.
Little over a decade now, researchers have been using AI and machine-learning to spot biomarkers for, among other diseases, dementia, heart disease and depression. Automated vocal analysis—through an intelligent home management system, say—is also under consideration. The complexity of human speech, variations and possibly underlying neuro, pulmonary and cardiac pathologies form a gateway to the use of AI to detect disease through picking up even the slightest aberrations. This has been used to detect Parkinson’s with 99% accuracy—UK researchers used AI and voice data to map 10 of 132 acoustic features in the saying of “ahh” that are associated with the disease. Similarly, correlation forms the basis of AI-led diagnosis of Alzheimer’s. While AI-focused diagnosis, including from facial-scanning, are an exciting new field, researchers/industry and policy must contend with concerns over privacy and the ethical ramifications of incorrect determination.