• +1 (416) 597-3422 ext.7800
  • kite@uhn.ca

NEW TECHNOLOGY AIMS TO PREDICT FALLS BEFORE THEY HAPPEN

Using cameras and artificial intelligence, researchers at Toronto Rehab and KITE believe they can prevent injuries caused by unintended falls

Memory loss isn’t the only aspect of dementia that’s difficult for people with the disease, and their loved ones, to deal with. According to the Alzheimer Society, those with dementia are four to five times more likely to experience falls than people without cognitive impairment. What’s more, people with dementia can have difficulty communicating how they feel, so developing health issues can go unnoticed.

However, scientists at KITE, Toronto Rehab’s research arm, think they have found a way to detect such issues in dementia patients before they become life-threatening problems. How? In part, by using relatively inexpensive, readily available technology, including video cameras.

Dr. Andrea Iaboni, a geriatric psychiatrist and clinician researcher, and Dr. Babak Taati, a scientist with KITE, have developed AMBIENT, a system that uses a single camera to detect changes in movement patterns. “We want to learn more about how someone is doing by watching them walk, or whether we can detect a critical level of unsteadiness in someone,” she explains, adding that for care providers, AMBIENT “is almost like having an extra set of eyes.”

Here’s how it works: A video camera is mounted at the end of a particular hallway in Toronto Rehab’s Specialized Dementia Unit, which is devoted to providing expert assistance to people with behavioural challenges related to the disease. When a patient enters the hallway, a tiny tag tucked inside their pant leg activates the camera, which records as they walk the length of the corridor, switching off again as they exit. Next, a computer program processes the resulting video, tracking the location of various joints.

With that information, the software, which uses artificial intelligence technology, measures multiple features of the patient’s gait, such as steps per second and smoothness of stride. By comparing the current measurements to those taken days and weeks earlier, the system is capable of picking up subtle differences that can herald a sudden decline in health. More impressive still, AMBIENT can detect changes in gait that are associated with a risk of falling.

A high-risk population

AMBIENT was developed chiefly for older adults with dementia, says Dr. Iaboni, since that group is particularly vulnerable to deteriorations in health. One of the consequences of dementia is a higher risk of falling, as the brain disease attacks the motor system, in addition to memory and thinking ability. This is an especially big problem in long-term care facilities, where roughly 70 per cent of residents have dementia.

$2 billion: Annual direct healthcare costs of falling among seniors.


Source: Patient Safety Institute

Multiple factors make nursing home residents with dementia more likely to experience serious repercussions from health issues. Not only are patients not always able to communicate that they’re feeling unwell, but older people don’t always present with fevers, or other symptoms, explains Dr. Iaboni. As well, staffing constraints make it impractical to carry out daily medical exams. Consequently, a case of pneumonia or a urinary tract infection may be missed until the individual is gravely ill.

One of the consequences of dementia is a higher risk of falling, as the brain disease attacks the motor system, in addition to memory and thinking ability.

Why movement patterns matter

So why might the way a person moves help identify such problems at an earlier stage? Imagine you have a bad case of flu. “You drag your feet,” notes Dr. Iaboni. “You’re less coordinated.”

A smart system capable of recognizing such changes could alert staff to the need for a comprehensive medical assessment, and thus facilitate earlier diagnosis and intervention, be it a change in medications, or increased supervision.

“Our first step was to identify which features of walking were most predictive of imminent risk of falling,” says Dr. Iaboni. “This is quite a unique approach.” That’s because existing methods of evaluating fall risk, which rely on risk factors such as a history of previous falls, can only project the chance of such an event occurring over the longer term.

Ultimately, the team discovered that two gait characteristics contributed markedly to a critical level of unsteadiness, including how smoothly people transfer their weight from one foot to the other. “One of the unique features of our device is that we did find it is able to detect this sort of critical unsteadiness,” says Dr. Iaboni.

Our first step was to identify which features of walking were most predictive of imminent risk of falling. This is quite a unique approach. Dr. Andrea Iaboni

Building on a breakthrough

The next step? Moving beyond the proof of concept – which employed only weeks of measurements per individual – to create a more finely honed instrument. “To allow us to identify when someone is becoming at risk of falling in a more longitudinal way,” Dr. Iaboni explains, “we’re collecting months’ and years’ worth of data at a nursing home that’s part of UHN called Lakeside.”

Plans are also underway to test whether the system is better at measuring gait characteristics than an experienced healthcare provider. “Even if it’s just a proxy for a well-trained physician,” says Dr. Iaboni, “there’s value in that, because there’s a problem with having enough skilled staff in nursing homes.” The system could even eventually serve as a tool to supply bedside staff with the training to spot critical unsteadiness.

In future, the research team hopes to improve AMBIENT even further, perhaps replacing the radio frequency tags with facial recognition, and increasing its accuracy at predicting falls. They can also foresee a day when the software is sufficiently simplified that it could easily be adopted around the globe. “You could use any video camera, and most nursing and retirement homes already have video surveillance in common areas,” says Dr. Iaboni. “This is a really great example of a project that’s taken advantage of the innovative and fast-moving areas of computer vision and pose-tracking for a very important clinical application which should improve the lives of people with dementia.”