Initial response time measurement in eye movement for dementia screening test

This paper proposes a method for detecting eye movement by using a generic RGB camera such as the one in a laptop computer or a smartphone. This method is designed for diagnosis of brain dysfunction such as dementia. For such a diagnostic purpose, precise detection of the eye movement is required, e.g., ± 1 video frame. However, noisy iris localization makes it difficult. The proposed method detects the moments when the iris begins and finishes moving by binary classification of eye movement velocity. This classification is achieved by discriminant analysis for optimization of these moments. We developed two diagnostic systems using the aforementioned function. The first one is designed just for diagnosis in which a subject gazes at a target point that appears on a white screen. The second one is merged with a generic web browser in which the gaze of a subject is measured while he/she reads texts displayed on the browser. Experimental results with these two systems demonstrated that the proposed method can detect the moments when the iris begins and finishes moving more accurately than required for a dementia screening test.