The irony is inescapable: a research project on technology that demands a gruelling form of human labour.
For as much as 50 hours a week over a period of several months, a team at the University of Regina has been staring at video screens, manually coding different non-verbal pain expressions—coding a mind-boggling 50-thousand frames of video.
Every single analysis is then cross-checked by a colleague to ensure accuracy. The whole process was made all the more complicated when the software they were originally using for coding proved to be inadequate to the task and they had to purchase a more expensive version.
“It’s a massive undertaking,” says the co-leader of the project, Thomas Hadjistavropoulos, the Research Chair in Aging and Health at the university.
But it was the only effective way of gathering the raw data for an AGE-WELL-funded project that could revolutionize the assessment of pain in long term care facilities. The goal of PAIN-ASSESS is to adapt facial recognition technologies to alert staff when a person with dementia is experiencing pain.
It is a pressing need. Hadjistavropoulos points to studies that indicate that people suffering from severe Alzheimer’s are far less likely to receive an analgesic medication than older adults in general. They often cannot communicate what they are feeling and overworked staff do not have the time to conduct regular pain assessments on every single resident.
“Pain is under-assessed and under-treated in long term care facilities,” he says.
The results can be a cascade of damaging outcomes. People with dementia suffering from unrecognized pain can turn aggressive. Staff then sometimes prescribe psychotropic medications, when in fact what is really needed is pain relief. Psychotropic medications can in turn increase the risk of death.
The University of Regina team members staring at video screens for all those hours are looking at the reactions of people with dementia and cataloguing expressions that indicate they are in pain. It could be a lowered eyebrow or a wince—things that a staff person in a long term care facility might note, if only they had the time to be watching each resident every hour of the day.
The raw data provided by Hadjistavropoulos’s group is being shared with their collaborating team at the Toronto Rehabilitation Institute (TRI), which is headed by Dr. Babak Taati, a Scientist with TRI’s AI & Robotics Team.
Taati says advertisers already use facial recognition technology to recognize emotions—a useful tool when surveying test audiences in market research, but it is unclear whether the algorithms would work on older faces that might have wrinkles and that perhaps might express feelings differently due to the effects of dementia.
“This is expanding into a different area,” says Taati.
His team will use the data collected in Regina to develop new algorithms that will not only note the signs of pain, but attempt to capture the indicators even when using a lower-quality (and lower cost) camera. The goal would be to have multiple cameras throughout a care facility, not only in each room, but also in common areas to effectively detect any resident who is suffering.
“The interdisciplinary nature of the project and the focus on affordability bring up really interesting technical challenges,” he says.
Given the sensitivities of privacy, the cameras will never be storing video. Their purpose is solely to advise staff when a resident should be assessed for pain.
The project has a health psychologist leading a team in Regina, in collaboration with a computer engineer leading a team in Toronto. Dr. Ken Prkachin, a University of Northern British Columbia expert on non-verbal pain expressions, is also a key co-investigator. The PAIN-ASSESS team operates within a cluster of researchers at the University of Alberta and University of Toronto who work on related projects. In other words, PAIN-ASSESS is a typical AGE-WELL project, where experts in diverse fields are brought together to find synergies from their complementary skills.
If all goes well, they will have a prototype ready within 3 years and be ready to start testing it in two long term care facilities by year 4. Success would produce a virtuous circle: more people with dementia would get the pain medication they need when they need it, leading to a decrease in incidents of aggression and stress levels of hard working staff could be lowered.
Thomas Hadjistavropoulos has high hopes for the potential of PAIN-ASSESS:
“It could change everything.”