Automatic Pain Recognition Systems in Dementia: Are We There Yet?

New methodological and technical advances, especially in machine learning, have paved the way for the development of automatic pain recognition systems. These systems hold the potential for continuous pain monitoring even in individuals with advanced stages of dementia who are not able to communicate pain verbally. In the last decade, several studies have been published that showed promising results, especially for the automatic detection of facial expressions of pain. However, most often these systems are trained with video material of young faces in ideal lightning conditions and with individuals often showing strong facial expressions, which limits the generalizability of the findings. Thus, in order to use these systems to validly assess clinical pain states in individuals with dementia, more research is still required. In this presentation, novel approaches will be presented that are especially tailored for the assessment of clinical pain in older individuals with dementia. These approaches range from automatic facial expression analyses to multi-model assessment of various pain behaviors of individuals with dementia in a living lab environment. Short video clips will be presented to better illustrate these new approaches.