Neurologists used AI to design a better stroke recovery program
New York University neurologist Heidi Schambra describes the post-stroke brain as ‘eager to learn’: in the first weeks of recovery, the brain is ripe to reform its connection to muscle and regain abilities it once had natural. “There’s a window of opportunity where we can really engage the brain,” Schambra said. “There’s this heightened plasticity that the trait itself puts into action.”
For many recovering patients, rehabilitation involves working on some form of reduced arm mobility – the series of movements the limb strings together to brush their teeth or butter their toast for breakfast. As primed as the brain may be to engage in these movements, rehabilitation researchers and therapists still don’t know how many of them patients should try to do during rehabilitation sessions. That’s because we don’t yet have a practical, accurate method for counting and identifying motions made in any given session, Schambra said.
Without clarity on the amount of training a patient receives, it is difficult to understand how to adjust it. In a new study published this week in Digital Health PLOS, Schambra and other NYU colleagues describe a new approach they’ve developed to pinpoint this data, using AI to count and identify arm movements about 370 times faster than human observers. It may be able to help stroke patients get closer to the right amounts of movement needed to jump-start their recovery.
To pull it all together and kick off this approach, the researchers filmed 41 recovering patients with high-definition video cameras as the patients performed typical rehabilitation activities for arm mobility, such as pouring a glass of water, moving objects on a shelf and wash your face. using a washcloth—and fitted them with wearable sensors to record motion capture data. Trained human coders watched the video recordings and marked the start and end of movements, categorizing them into five classes: a “reach” for an object, a “reposition” to get closer to it, a “transport” for to carry an object somewhere, a “stabilize” to keep an object stationary, or an “idle” to stand ready near one.
The team then trained the AI algorithms on these thousands of movement examples and the data models representing each class – such as a “scope” or a “carry” – until the program was able to count and classify movements by itself. The resulting tool, which can process 6.4 hours of recorded activity in 1.4 hours versus the 513.6 hours needed for human coders, brings a combination of pragmatism and precision the field has never seen. before, Schambra said.
The most common way in the field to measure the amount of training a stroke patient receives during rehabilitation is to take time, i.e. the hours the patient spends in the course of a given session. While practical, Schambra said, it’s imprecise. There’s an incredible range of what patients could actually do at that time. “Let’s say you go to the gym for an hour,” she said. “You could train for 45 minutes with high intensity cardio. I’ll probably sit in the sauna and have a smoothie. We could still hit the gym for an hour each, but we do very different physical amounts between the two of us.
With the tool’s ability to measure training intensity both quickly and accurately, Schambra sees it as the ticket to finding out how many moves will speed recovery after a stroke. She has seen recent animal models that indicate early, aggressive training with hundreds of movement repetitions per session, 10 times more than patients are expected to do.
Schambra hopes to use the new approach in intensity response trials — similar to what you’d see in drug trials — in which researchers prescribe stroke patients (grouped by their level of reduced mobility) increasing amounts movements and study how patients recover based on prescribed amounts. “We had to step back and build this tool so we could do these studies in the future,” Schambra said.
As a clearer picture of the movements performed in rehabilitation emerges and researchers study how much movement stroke patients need to engage in – particularly during the crucial first weeks of peak neuroplasticity after a stroke – each patient may have a better chance of returning to the basic abilities that get us through life easier than we think.