Pain is often poorly treated among individuals with severe dementia who have trouble communicating their discomfort. A Canadian research team is aiming to tackle this problem with the use of facial-recognition technology.
The team, co-led by Dr. Thomas Hadjistavropoulos at the University of Regina, has developed an image processing system to monitor and analyze the facial expressions of long-term-care residents. The computerized system then alerts nurses whenever it detects wincing, frowning and other expressions that suggest a resident is in pain.
Less than a decade ago, the idea of using an automated system to recognize pain in those with severe Alzheimer’s disease and other dementias “seemed like science fiction,” says Dr. Hadjistavropoulos, a professor of psychology and research chair in aging and health.
But now, he says, “I have come to firmly believe that the greatest solutions … in improving the quality of lives of these patients within our lifetime are far more likely to come from engineering than they are to come from the medical sciences.”
With no cure and only limited treatments available for those with Alzheimer’s disease, Dr. Hadjistavropoulos and his team are betting their work can, at least, help alleviate suffering. They plan to begin testing their system, supported by the federally-funded AGE-WELL network, at two long-term-care facilities in Regina later this year.
Depending on the study, the prevalence of pain among long-term-care residents is estimated to be as high as 80 per cent, Dr. Hadjistavropoulos says. Yet, this population often does not get adequate pain treatment, in large part, because those with advanced dementia may not effectively express their needs, and long-term-care facilities often lack the staffing resources to adequately detect their pain, he explains.
As a result, residents can become agitated and aggressive, leading staff to mistakenly attribute their behaviour to psychiatric disturbances. Residents are thus commonly treated with psychotropic medications, rather than the analgesic medications they need, Dr. Hadjistavropoulos says.
At Toronto’s Baycrest Health Sciences, Dr. Yael Goldberg, who is not involved in Dr. Hadjistavropoulos’s team, says the use of facial-recognition technology seems a novel approach that builds upon existing methods of detecting pain in patients. Currently, doctors and nurses typically rely on physical examinations and observational assessments to detect pain in older individuals who cannot communicate verbally. These may involve checklists or scoring systems that rate patients’ behaviours, such as breathing, vocalizations, body language, how readily they can be consoled, and facial expressions. She cautions, however, that some signs of pain, such as limping, cannot be detected through facial expressions alone.
Dr. Goldberg, a clinical psychologist and clinical neuropsychologist, says an estimated 60 per cent of all behavioural symptoms of dementia – from agitation to refusing to eat or shower – are actually due to unaddressed pain.
“It’s a real problem and it’s really something that needs to be considered when we think about what might be triggering or maintaining behaviours in people with dementia,” she says. “The first step should always be looking at is there pain there?”
The system that Dr. Hadjistavropoulos and his team have developed involves setting up a video camera in the rooms of long-term-care residents. To address privacy concerns, the camera does not record anything; the system merely processes the video images. A computer program then detects any expressions of pain on the residents’ faces, and triggers a light to turn on at the nursing station. It also sends a confidential e-mail to the nursing team, informing them of which patient requires further attention.
One of the main challenges for the research team has been to train its algorithms to read the faces of older people, says Dr. Babak Taati, who is also leading the project. There are plenty of commercially available facial recognition and facial-analysis algorithms that work well on the faces of young, healthy individuals. But these are not always able to distinguish wrinkles from frowns, for example. And they do not accurately interpret the expressions of adults who have cognitive and physical disabilities, says Dr. Taati, a scientist at KITE, the research arm of the Toronto Rehabilitation Institute.
To gather the raw data needed to train their algorithms, the researchers videotaped 102 older individuals, from various angles, half of whom were long-term-care residents, and reviewed the footage frame by frame to manually code the participants’ expressions of pain.
Mary Brachaniec of Moncton, N.B., says she believes an automated pain-detection system would not only help individuals with dementia, but would also provide peace of mind to their families.
Ms. Brachaniec’s parents, who died in 2018, both had Alzheimer’s disease. Although she says staff at her mother’s long-term-care facility provided the best care they could, there were times she would visit and find her in pain.
Her mother, who had fractured her pelvis owing to a fall, was not always able to express her discomfort to staff, but Ms. Brachaniec says she could tell when her mother was hurting through subtle signs, such as grimaces, moans or closing her eyes tight.
“Even with family there, even with staff doing their best, the system is constrained and pain ... can be missed, or it can be underestimated and under-treated," which can be scary for those with dementia and their loved ones, she says. “We don’t want people to suffer, right?”