Person Detection System to Help Patients Identify Visual Hallucinations

Abstract

As of 2021, Japan’s population is aging, and the number of patients with dementia continues to increase. Visual hallucinations occur as a core symptom in patients with dementia with Lewy bodies (DLB). Visual hallucinations involve perceiving forms or entities that do not actually exist and tend to cause patients to panic. In this study, we developed a system to help patients identify visual hallucinations to support their well-being. The proposed system uses the YOLO object-detection algorithm, an eye tracker, and an LED to present feedback. When the examinee’s line of sight remains within an area detected as a person in a field of view image for more than a specified gazing time, the LED attached to the eye tracker turns on to provide feedback. This allows users to identify a perceived form as a potential visual hallucination if no feedback is provided. We conducted two types of experiments with three examinees who acted as patients, and the results showed that they exhibited normal behavior in 99.2% of interactions.