A partnership between precision health-care technology company PROTXX and Edmonton-based AltaML has resulted in a new benchmark in concussion diagnostics, the results of which were recently published in the journal Sensors.
PROTXX, which has offices in Menlo Park and Calgary, provided the proprietary technology — a phybrata (or physiological vibration acceleration) sensor that sits on the bone behind a person's ear. AltaML's machine learning analytics interpreted the data to provide a diagnostic solution, explained Alex Hope, data science manager for AltaML.
The sensor measures microscopic motions in the head, which are captured and interpreted in a variety of ways to make a diagnosis. The companies found that ML algorithms enabled the sensor to make a more precise diagnosis than previous approaches to ML-based concussion diagnostics.
"There's a major burden within the health-care system around getting professionals to actually diagnose concussions," Hope told Taproot. "They're also really hard to diagnose, because sometimes they're not fully clinical."
The remote sensor can help address some of those challenges from a resource perspective and may also catch other issues occurring inside someone's brain that aren't detectible through behavioural tests.
Other health-care scenarios where machine learning could be used include diagnostics in medical imaging, drug discovery, care prediction, and disease identification. Celia Wanderley, chief customer officer and head of services for AltaML, said the sensor technology could also be valuable in diagnosing neurophysiological conditions such as multiple sclerosis and Parkinson's disease.
"The information associated with our health has been rapidly expanding for years, and we are reaching the point of each individual being able to generate petabytes of health-related information," Wanderley said. (A petabyte is a million gigabytes.) The available data ranges from genetics to activity levels tracked through fitness monitors.
"Increasingly it is a challenge to maintain a holistic picture of a person's health," Wanderley said. "Using this information in an effective manner is a daily struggle for clinicians, and machine learning/artificial intelligence are often put forward as a solution to this problem."