Health identification of traditional Chinese medicine with tongue images

Abstract

Physical examination, which refers to various organs, depends on a lot of medical apparatus and instruments. The diagnostic result according to the examination reports is objective and reliable, but people cannot get the result instantly. As an important diagnostic method in traditional Chinese medicine (TCM), tongue diagnosis is straightforward, which depends on a sophisticated Chinese medicine doctor to examine the tongue of a patient. However, due to its qualitative, subjective and experience-based nature, traditional tongue diagnosis has a very limited application in clinical treatment. In this paper, we bridge the gap between tongue samples and their physical examination reports. With tongue images and diagnostic results according to the physical examination reports taken into account, the Health Identification System (HIS) we developed explores the relations between the tongue and health condition of a patient. Given a new tongue image, HIS automatically suggests a diagnostic result as tongue diagnosis does. Experimental results derived from a group of 1426 patients as well as other 747 volunteers show that SVM with linear kernel can obtain better performance with less computational cost than SVMs with other kernels. Ultimately, we embed SVM with linear kernel into HIS to learn a model for the tongue diagnosis.