Once considered discrete disciplines, biometrics and Artificial Intelligence are now more or less synonymous. The biometric technologies of today are AI-driven, often based on machine learning frameworks that enable biometric systems to refine their capability of detecting patterns in biological subjects.
For the most part, such biometric technologies are being used to authenticate and identify individuals, whether it’s in the passenger screening process at an airport terminal or in the fingerprint-scanning home button of a smartphone. Naturally, this applies in the healthcare sector, too; for example, biometrics are increasingly being used to reliably identify patients. But this technology is also starting to offer intriguing applications in another aspect of healthcare – medical diagnosis.
Looking for Symptoms
A striking example of this kind of application emerged in 2018, when AI specialist DeepMind, the University College of London, and the UK’s Moorfields eye hospital collaborated on a computer vision system that could help to save human vision from eye disease. As with so many other AI-driven biometric systems, this one was designed to find patterns in biological material – in this case, patterns pertaining to eye disease that could be spotted in retina scans. Remarkably, the researchers found that their AI system could identity eye disease with an accuracy of 94 percent, and that it could detect more than 50 different kinds of eye diseases more effectively than doctors can.
Even before that, researchers at the Mayo Clinic, in partnership with an Israeli startup called Beyond Verbal, were developing an AI-driven system designed to use voice biometrics to detect signs of heart disease. And researchers at Curtin University were exploring how face biometrics could be used to detect signs of patients’ pain – a notoriously murky area of clinical assessment in the healthcare field – as early as 2015.
Beyond Physical Ailments
Perhaps most remarkably, AI-driven biometric technologies may be able to help diagnose psychological illness in addition to physiological illness. Sonde Health announced last November that it had secured a patent for a new biometric diagnostic tool based on voice biometrics. Sonde Health’s system is designed to analyze short clips of patients’ speech in order to look for anomalies that may point to health conditions in their early stages, including depression and dementia; and what’s more, it also uses voice biometrics to identify patients.
Sonde Health is not the first organization to apply voice biometrics technology to the detection of psychological illness. In 2015, for example, researchers at New York University’s Langone Medical Center explored the use of machine learning AI software designed to scan voice biometrics for patterns associated with PTSD and depression, among other psychological illnesses. And even in the early stages of that system’s development, the researchers were finding that their system could diagnose PTSD with an accuracy of 77 percent.
Calling Doctor Apple
Diagnostic applications of biometric technology are even finding their way beyond the clinic and into mainstream consumer devices. This trail is currently being blazed by Apple, whose Apple Watch Series 4 introduced a sophisticated ECG reading capability thanks to embedded electrodes and a special sensor in the smartwatch’s back crystal. Lots of wristband wearables are designed to track various fitness metrics, but Apple’s smartwatch was unique in being able to detect abnormal heart rhythms, including those that might indicate Atrial Fibrillation, or AFib – a condition that can lead to stroke.
This is an exciting development for health researchers. Last November, medical genetics firm Invitae announced a new clinical research platform, developed in partnership with the American Heart Association’s Scientific Sessions, that would use the Apple Watch to research genetic impacts on AFib.
This may just be the beginning. AI and biometric technologies are advancing rapidly, and are increasingly finding their way into healthcare and medical applications. And while these technologies aren’t currently expected to actually replace trained physicians and medical experts, they may help to deliver valuable tools that can improve medical diagnoses and health outcomes, and possibly even save lives.
February 13, 2020 – by Alex Perala