Automatic Diagnosis of Tongue Using Mask-RCNN

Abstract

The purpose of this paper is to develop a support system for tongue diagnosis in Traditional Chinese Medicine using artificial intelligence image recognition technology. The tongue has a direct or indirect connection with many viscera, and the five parts (tip of tongue, middle tongue, lingual sides of left and right, root of tongue) correspond to the heart/lung, stomach/spleen, liver/bile, and kidney, respectively. In this paper, for a given image of the tongue, a deep learning method, Mask R-CNN, is applied firstly to automatically recognize the tongue and cut out only the tongue. Then, the tongue is divided into five parts. For each part of the tongue, Mask R-CNN is applied again to judge the health status of the related viscera. Finally, experiments and the results are shown to verify our proposed method.