Evaluation of Kampo disease states using facial images

Abstract

In the traditional Japanese (Kampo) medicine, diagnosis is based on the observation of facial color and expression. Disease states that are diagnosable from facial characteristics include blood stagnation, blood deficiency, and yin deficiency. However, these diagnoses are subjective and require lots of experience. In this paper, we aimed to quantify facial diagnosis by creating a system to evaluate the severity of disease states using feature values obtained from facial images. A Kampo physician evaluated the facial images and rated them from 1 to 5 according to severity of disease states. We verified the accuracy of this system using the mean squared error calculated from the difference between the physician evaluation scores and the system estimates. The mean squared error was close to zero, indicating that the system has high accuracy. The selection of feature values using this system corresponded with those facial regions used by the Kampo physician in diagnosing the disease states.