背部痧象特征要素数显划分方法  

Digital display division method of characteristic elements of back Sha images

作　　者：李斌[1] 胡广芹[1] 李霄[1] 张新峰 LI Bin;HU Guangqin;LI Xiao;ZHANG Xinfeng(Faculty of Environmental and Life Sciences,Beijing University of Technology,Beijing 100124;Faculty of Information Science,Beijing University of Technology,Beijing 100124)

机构地区：[1]北京工业大学环境与生命学部,北京100124 [2]北京工业大学信息学部,北京100124

出　　处：《北京生物医学工程》2024年第1期23-28,共6页Beijing Biomedical Engineering

基　　金：北京市中医药科技发展基金(QN-2020-08、QYSF-2020-06);国家重点研发计划(2018YFC1707705)资助。

摘　　要：目的通过对痧象图片数据的研究,提出一种数显的客观方法来对背部痧象特征要素进行准确辨识,以避免主观认知不同带来的偏差,帮助人们快速辨识痧象的相关特征,为中医医师的诊断提供一定的客观依据,减少主观化成分。方法首先,采用深度语义分割算法模型对采集的背部痧象图片进行分割,以获取只包含痧象的部分区域,去除图片中其他无关部分,尽可能减少对后续实验所带来的影响。然后针对分割得到的图像,结合中医五脏六腑在背部的映射区,采用关键区域检测的方法对痧象颜色特征进行3种区间的识别;根据中医主观上对于痧象形状的划分,采用像素统计方法将痧象形状划分为片状和点状两种。最后,通过准确度、精确度、召回率、F1值对客观识别的结果进行数据上的评价。结果本文提出的客观数显方法在痧象颜色特征划分上的准确率达到了80.56%,在痧象形状特征的划分上准确率达到了89.60%,能够较为准确地对痧象特征要素进行划分,具有一定的可行性。结论本文提出的数显方法能够对痧象的颜色和形状特征进行较为准确的辨识,可以在很大程度上避免主观认知所造成的辨识不同问题,在辅助医师诊断以及用于中医痧象特征的识别教学方面具有较大的应用前景,可为正确了解痧象特征信息提供帮助。Objective Through the study of the image data of the Sha,an objective method of digital display is proposed to accurately identify the characteristic elements of the Sha on the back,in order to avoid the deviation caused by different subjective cognition,help people quickly identify the relevant features of the Sha,provide a certain objective basis for the diagnosis of traditional Chinese medicine doctors,and reduce the subjective components.Methods First,the deep semantic segmentation algorithm model was used to segment the collected Sha images on the back.The purpose was to obtain only part of the area containing Sha and remove other irrelevant parts in the images,so as to minimize the impact on the subsequent experiments.Then,according to the segmented images,combined with the mapping area of the five viscera in the back of traditional Chinese medicine,the color features of the Sha were identified in three intervals by the method of key region detection.According to the traditional Chinese medicine subjective division of the shape of the Sha,the pixel statistics method was used to divide the shape of the Sha into flake and spot.Finally,the accuracy,precision,recall rate and F1 value were used to evaluate the objective identification results.Results The objective digital display method proposed in this paper achieved an accuracy rate of 80.56%in the color feature and 89.60%in the shape feature of the Sha images,which could accurately classify the feature elements of the Sha images and had certain feasibility.Conclusions The digital display method proposed in this paper can identify the color and shape features of Sha more accurately,and can avoid the different identification problems caused by subjective cognition to a large extent.It has a great application prospect in assisting doctors in diagnosis and teaching of traditional Chinese medicine image features recognition,and provides help for correctly understanding the feature information of Sha.

关 键 词：中医痧象 信息化 机器学习 特征分类 图像处理 

分 类 号：R318[医药卫生—生物医学工程]