Putting assessment in the hands of patients
AI-based Automatic Assessment of Motor Deficits in PD
People with Parkinson’s disease might only see their doctors a few times a year, which poses a challenge for tracking the course of this complicated ailment, whose symptoms can vary from day to day or week to week.
If a clinical appointment falls on a “good” day with few problematic symptoms, the clinician could easily assume all is going well; conversely, if patients show up on a “bad” day, the conclusion may be that their condition is worsening. In fact, neither assessment may be correct, but there has been no way to confirm that — until now.
Dr. Babak Taati, an engineer with KITE, the research arm of the Toronto Rehabilitation Institute (TRI), has been studying this diagnostic challenge over the past few years. He works in the field of artificial intelligence (AI), a technology that has progressed to the point of following details of an individual’s symptoms based on physical movements shown on a video.
“The responsiveness of the computer vision machine learning techniques have been on par and in some cases even a little bit better than a clinician’s in detecting clinically important changes,” he says.
Taati’s work on how AI analyzes the body’s behaviour came to the attention of Yana Yunusova, a TRI speech researcher who had been exploring this technology’s ability to monitor the oro-facial (mouth and jaw) and speech communication health of stroke patients. The pair are collaborating on a single system to combine body and face tracking into an assessment tool that could track the course of Parkinson’s disease in an individual’s daily life.
“It would empower the patient if they could participate in their health care, if they could do something at home every day for five minutes.”
“It would empower the patient if they could participate in their health care, if they could do something at home every day for five minutes,” explains Taati.
He and Yunusova have developed a project to create an AI-based software package that could operate on hand-held electronic devices, such as smart phones and tablets. This app would use the camera on board these devices to record users as they go through a series of movements, such as pronouncing key words or tapping out a sequence. These are tests a clinician might conduct during office visits, but this convenient app will allow people to collect the same information more frequently.
“This data will help the neurologist get a much more complete picture of the patient’s condition and how it is changing,” says Taati. The information could also be anonymized and shared with other clinicians or researchers to help move the entire field forward.
“It’s a good tool for assessing the efficacy of new treatment options, be they pharmaceutical or exercises or surgery.”