TongueCaps: A Model for the Multiclassification of Tongue Color

Abstract

AbstractTongue colors can reflect physiological information in the human body and assist Chinese medicine doctors in completing syndrome differentiation and treatment, so the classification of human tongue color is of great significance. However, for the three more common tongue colors, light red, red and deep red, their color tone is all red, in addition, because the color rendering of the tongue image is affected by the color temperature of the ambient light, camera equipment and color rendering equipment, it is difficult to classify the color of tongue. In our work, we propose TongueCaps, a model based on the combination of residual block and capsule, to realize end-to-end computer-aided tongue color classification. This paper carries out classification experiments on three categories of tongue color with RGB images, light red tongue (382 cases), red tongue (312 cases), and deep red tongue (104 cases). Furthermore, we tested TongueCaps in comparison with VGG16, ResNet18. The model effect was evaluated by accuracy, sensitivity, specificity. Compared with the results, TongueCaps showed the best performance.KeywordsTongue colorResidual blockCapsuleArtificial intelligence