AI-based prediction of agitation in dementia patients

With the increasing age of the global population comes a significant increase in the number of people with dementia. Although dementia is characterized primarily by a loss of cognitive functioning, there are also behavioral and psychological symptoms that are clinically relevant. Specifically, agitation is one of the most important behavioral problems seen in dementia patients. Agitation alone can lead to a significant decline in a patient’s quality of life and possible (over)use of medication, as well as increasing the burden of the caregiver, typically leading to early institutionalization. Measuring agitation more directly, objectively and continuously will allow us to better understand the mechanisms of agitation. This increased understanding will, in turn, enable better prevention of agitation and provide a platform to better monitor the effect of both pharmacological and non-pharmacological interventions. This research consists of developing a methodology to measure and eventually predict levels of agitation in dementia patients. Due to there not yet being a directly correlated biometric linked to agitation, the approach to be developed will combine multiple physiological and environmental measurements. The devices used for these measurements will be suitable for the target population, and will consist of both wearable and non-contact technology. The various contextualized physiological data inputs acquired will be used to develop a model with the goal of measuring and ultimately predicting levels of agitation.This prediction will be facilitated by machine-learning algorithms. Specifically, the measurement/prediction of agitation can be seen as a multi-input classification problem. To facilitate the development of these algorithms, multiple data collection trials will be organized. Based on the large database resulting from these trials, a personalized machine-learning model will be developed to classify levels of agitation.