The researchers developed an algorithm that uses artificial intelligence (AI) to recognise the emotional state of Parkinson's disease patients based on physiological data collected from a smartwatch.
The approach could ultimately help clinicians accurately track patients' emotions and guide treatment decisions, the researchers said, noting that their preliminary findings “warrant confirmation in larger samples” as “open questions remain that should be addressed in further research.”
Their research:Supervised learning for automatic emotion recognition in Parkinson's disease from smartwatch signals“teeth, Expert Systems with Applications.
Although Parkinson's disease is best recognized by its characteristic motor symptoms, this neurodegenerative disease is accompanied by a variety of non-motor symptoms. Neuropsychiatric symptoms and mood changes, including anxiety, depression, apathy, are common in Parkinson's disease patients.
However, many of these psychological symptoms are highly subjective and difficult to measure or monitor in clinical practice.
AI algorithm research not focused on Parkinson's emotions
One way to monitor a person's emotional state is to measure corresponding physiological changes, such as increased heart rate and body temperature. These changes are related to the functioning of the autonomic nervous system, which is involved in regulating various involuntary bodily functions and is prone to being disrupted in Parkinson's disease.
These automated emotion recognition based on autonomic changes is a well-studied approach in healthy people but has not been thoroughly evaluated in people with Parkinson's disease, so scientists wanted to understand how this approach could be applied to patients with neurodegenerative diseases.
Eleven people with mild to moderate Parkinson's disease and eight people without the neurological disease were asked to watch a series of video clips designed to evoke a range of positive and negative emotions.
Each subject wore a smartwatch that collected physiological data such as heart rate, body temperature and skin electrical conductivity, and participants were also asked to report their emotional state as they watched each video.
The scientists then explored an AI-based approach to predicting a person's emotional state based on data collected from a smartwatch, comparing several different types of algorithms.
The results showed that the AI could identify whether an individual was feeling positive or negative emotions (emotional valence) and how strongly they felt those emotions (arousal level) with over 90% accuracy in both people with and without Parkinson's disease.
Predict the type and intensity of emotions
One algorithm performed significantly better than the other. The most important features used for prediction were skin temperature and galvanic skin response (electrical activity of the skin), which are known to be important for emotion recognition in healthy people.
Although overall accuracy was high, the researchers said it was harder for the algorithm to predict emotional valence than it was to predict arousal.
Autonomic dysfunction and motor symptoms in Parkinson's disease may reduce or fluctuate autonomic responses to felt emotions, making it more difficult to recognize patterns in physiological responses, the researchers noted.
The researchers say this is one reason why it's important to use data from Parkinson's patients to teach the algorithm how to make predictions, and not just from the general population. “It is possible to recognize emotions in these people with a high degree of accuracy in a low-cost, widely accepted and usable way,” they wrote.
“This discovery shows that these systems can be used in current clinical care to [people with Parkinson’s disease]”This may allow for the identification of the quality of emotions (regardless of their intensity) and may also guide therapeutic decisions to improve the quality of life of both patients and caregivers,” the researchers wrote.
The scientists say further studies are needed to test and validate the approach in larger patient groups, and there is a need to develop algorithms that can recognise more complex emotional states, such as anger or depression.
The researchers said they plan to further analyze the potential clinical significance of the results observed in their paper in a future publication.
