文献类型：期刊文献

中文题名：CRANet:舌象图像分割与分类网络

英文题名：CRANet:Tongue Image Segmentation and Classification Network

作者：彭辰[1];李文举[2];钱志勤[1];徐蕾[3];罗琪[1];余依婕[1]

机构：[1]华东理工大学机械与动力工程学院,上海200237;[2]郑州市仁济医院南院内二科,郑州450004;[3]上海市嘉定区安亭医院后勤保障部,上海201800

年份：2025

卷号：51

期号：4

起止页码：564

中文期刊名：华东理工大学学报(自然科学版)

外文期刊名：Journal of East China University of Science and Technology

收录：;北大核心:【北大核心2023】；

基金：上海市“科技创新行动计划”自然科学基金面上项目(23ZR1416200)。

语种：中文

中文关键词：卷积神经网络;注意力机制;图像分割;图像分类;中医舌诊

外文关键词：convolutional neural network;attention mechanism;image segmentation;image classification;traditional chinese medicine tongue diagnosis

摘要：中医诊断中,舌诊是一种重要的诊断方法。然而,舌象分析的准确性和效率受主观因素影响较大。本文提出了基于卷积神经网络与注意力机制的舌体分割与分类模型(CRANet),并设计了卷积残差模块(CR Block)。该模型能够自动对舌体进行分割与特征识别,并据此对舌象进行分类,提高舌诊的客观性和准确性。在分割任务中,本文模型的平均准确率为99.43%,交并比(IoU)为97.50%,Dice系数为98.73%;在分类任务中,平均准确率为88.91%,精确率为86.37%,召回率为85.32%,F1分数为85.84%;并验证了基于卷积神经网络和注意力机制的舌象图像分割与分类方法在中医舌诊中的应用潜力。
In Traditional Chinese Medicine(TCM)diagnosis,tongue diagnosis is an important method.However,the accuracy and efficiency of tongue analysis are greatly influenced by subjective factors.This paper proposes a tongue segmentation and classification model based on convolutional neural networks and attention mechanism(CRANet),and designs a Convolutional Residual Block(CR Block).The model can automatically segment the tongue and recognize features,thus classifying tongue images and improving the objectivity and accuracy of tongue diagnosis.The proposed model achieved an average accuracy of 99.43%,an Intersection over Union(IoU)of 97.50%,and a Dice coefficient of 98.73%in the segmentation task.In the classification task,it achieved an average accuracy of 88.91%,a precision of 86.37%,a recall of 85.32%,and an F1 score of 85.84%.This verifies the application potential of the tongue image segmentation and classification method based on Convolutional Neural Networks and Attention Mechanism in TCM tongue diagnosis.