Abstract:
Accurately measuring and quantifying the severity of psychiatric disorders is difficult as there are often no objective measures. Therefore, diagnosis relies on interviews and self-administered questionnaires which can be ambiguous and difficult to standardise. Researchers have studied quantitative measures (e.g., neuroimaging) to assist in the assessment of patients with psychiatric disorders. In our works, disease pertinent images (e.g., dysphoric images for patients with depression) are presented on a computer screen and the individuals’ gaze positions are measured using an eye tracker. Differences in which the patients view the images (i.e., visual scanning behaviour) was used to provide quantitative measures of their psychiatric state. Two lines of work are presented in this talk: 1) novel eye-tracking paradigms to characterise patients with eating disorders, depression and Alzheimer’s disease, and 2) novel analytic methods that use deep neural networks to learn spatial-temporal interactions of individuals’ visual scanning behaviour. The neural networks were successful in differentiating patients with bipolar and unipolar disorder and study apathy in Alzheimer’s disease.