A computer system for the face stimulus assessment with application to the analysis of dementia

Abstract

In this paper, we have developed a computer system for evaluating elements in the standardized image of the human face in the face stimulus assessment (FSA) and applied it to the analysis of the FSA pictures by dementia patients. The system automatically and objectively evaluates several elements in the FSA by developing new computer algorithms and using existing computer systems for art assessments. The algorithm divides the FSA into several areas, finds the main color of each area, compares the main color with one of its surrounding areas, and then grades the elements. The reliabilities of gradings between the human raters, and between the human raters and the computer system were measured by quadratic weighted kappa for each element. The correlation analysis is applied to analyze the relation between the grade and degree of dementia, the factor analysis is then applied to discriminate the difference between groups of normal healthy people and dementia patients, and the regression analysis is applied to estimate the probability of dementia. The system proposed in this paper was verified using case studies and analysis of sample results. The system is expected to promote the usage of the FSA and ease the research of FSA validity.