Automatic detection of oculomotor disorders

International audienceAssessment of oculumotor functions is a key part of any eye examination and several studies have shown that it could also help in the diagnostics of various neurological conditions. Objective measurement of oculomotor performances using eye tracking is however little used in orthoptics or clinical practice, perhaps partly because current eye tracking tools provide raw or little processed data that are of limited value without deeper analysis. The aim of this study was to develop more sensitive tests and analysis tools to be able to identify automatically patients with oculomotor disorders. 40 subjects (25 normals and 15 orthoptics patients) participated to the study at Brest University Hospital. Subjects had to perform series of 3 different orthoptics exercises (fixation, saccades, small pursuit) during which their eyes’ movements were recorded with a Tobii Pro Spectrum eye tracker. The spatial and temporal statistics of the trajectories were analysed using conventional metrics and a recurrent neural network. The study shows that an eye tracker is a valuable tool to quantify oculomotor performances in orthoptics practice. The results of this study open up interesting possibilities for the use of the various algorithms in future work