2018 年 17 巻 4 号 p. 381-393
In this study, we tried to extract and formalize medical experts' professional skills of schizophrenia symptom evaluation that make use of patients' facial information. In our experiment, eight medical experts were asked to evaluate 147 videos depicting 16 patients' faces using a rating scale representing the states of disease progression. The results of the experiment show that, with regard to the most severe state patients, different clusters are characterized by different patterns of these facial features. The results also show that patients' states of disease get worse with the sequential accumulation of several variables among these feature: a downward head and eyes, a loose mouth, and stiff facial expressions. These suggest that facial muscles control is related to symptom evaluation of schizophrenia. It demonstrates the possibility of constructing an automatic system for schizophrenia symptom evaluation using combinations of the six facial features of patients.