文献类型：期刊文献

中文题名：基于深度学习和随机森林的齿痕舌分类研究

英文题名：Research on Classification of Dentate Tongue Based on Deep Learning and Random Forest

作者：颜建军[1];李东旭[1];郭睿[2];燕海霞[2];王忆勤[2]

机构：[1]华东理工大学机械与动力工程学院智能感知与诊断研究所,上海200237;[2]上海中医药大学四诊信息综合实验室,上海201203

年份：2022

卷号：40

期号：2

起止页码：19

中文期刊名：中华中医药学刊

外文期刊名：Chinese Archives of Traditional Chinese Medicine

收录：CSTPCD;;北大核心:【北大核心2020】；

基金：国家自然科学基金(81673880,82074332,81302913,81270050,30901897,81173199)。

语种：中文

中文关键词：舌诊客观化;齿痕舌;机器学习;深度学习;随机森林

外文关键词：objectification of tongue diagnosis;dentate tongue;machine learning;deep learning;random forest

摘要：目的齿痕舌分类识别是目前中医舌诊客观化的重要研究内容之一。目前齿痕舌的识别方法上存在效率低、准确率不高的问题。方法提出了一种基于深度学习和随机森林的齿痕舌分类方法。首先分别准备齿痕舌样本,并利用深度学习分割模型进行舌体分割;在相关中医专家的指导下对每张舌图像的齿痕进行标注;基于YoloV5深度学习算法训练齿痕检测模型;根据检测出的齿痕区域,提取相关特征构建齿痕舌特征向量;建立随机森林(random forest,RF)模型,实现齿痕舌分类。结果采用该方法得到的总体分类准确率达到93.7%,取得了较好的结果。结论利用该方法进行齿痕舌分类,达到了较为满意的效果,为齿痕舌的识别研究提供了一种新的思路,对舌诊客观化和现代化具有一定借鉴意义和实用价值。
Objective The classification and identification of dentate tongue is one of the most important research contents in the objectification of TCM tongue diagnosis.At present,there are problems of low efficiency and low accuracy in the identification of the dentate tongue images.Methods A dentate tongue classification method base on deep learning and random forest(RF)was proposed.Firstly,tongue images were prepared and segmented by deep learning segmentation model.Under the guidance of TCM experts,the tooth mark regions were annotated in each image.The deep learning algorithm YoloV5 was used to train the detection model of tooth marks.According to the detected tooth marks,the relevant features were extracted to construct the feature vector of the dentate tongues.The classification model of dentate tongue images were established based on random forest.Results The overall classification accuracy obtained by this method reached a good result with 93.7%.Conclusion It provides a new way of thinking for the research on the recognition of dentate tongue images,and has certain reference significance and practical value for the objectification and modernization of tongue diagnosis.