Objective To promote the objectification and intelligence of clinical TCM tongue diagnosis in ulcerative colitis(UC).Methods Daosheng DS01-B tongue and face diagnosis information collection system was used to prospectively collect tongue image pictures of patients with ulcerative colitis damp-heat in the large intestine syndrome(DCSR)and spleen deficiency dampness retention syndrome(PXSY),with totaling 1096 images.After UC tongue image segmentation,preprocessing,and data augmentation,a data set of UC tongue images was formed.Based on ResNet50,a UC tongue image classification model was constructed using feature fusion methods and attention modules.Results The UC tongue image classification model constructed had better classification performance,with an average F1 value of 85.09%,an AUC value of 0.83 for PXSY,and an AUC value of 0.81 for DCSR,both of which were higher than the VGG11 and ResNet50 models.Conclusion The constructed UC tongue image classification model can effectively identify DCSR and PXSY,providing a new approach to improve the accuracy and objectivity of UC differentiation and assisting in the intelligence of TCM tongue diagnosis.

Objective To promote the objectification and intelligence of clinical TCM tongue diagnosis in ulcerative colitis(UC).Methods Daosheng DS01-B tongue and face diagnosis information collection system was used to prospectively collect tongue image pictures of patients with ulcerative colitis damp-heat in the large intestine syndrome(DCSR)and spleen deficiency dampness retention syndrome(PXSY),with totaling 1096 images.After UC tongue image segmentation,preprocessing,and data augmentation,a data set of UC tongue images was formed.Based on ResNet50,a UC tongue image classification model was constructed using feature fusion methods and attention modules.Results The UC tongue image classification model constructed had better classification performance,with an average F1 value of 85.09%,an AUC value of 0.83 for PXSY,and an AUC value of 0.81 for DCSR,both of which were higher than the VGG11 and ResNet50 models.Conclusion The constructed UC tongue image classification model can effectively identify DCSR and PXSY,providing a new approach to improve the accuracy and objectivity of UC differentiation and assisting in the intelligence of TCM tongue diagnosis.