基于移动端的非侵入式贫血检测系统研究与设计  

Research and Design of Non-Invasive Anemia Detection System Based on Mobile Terminals

作　　者：张盛 胡国清[1] 陈佳 付西敏 赵芮 王莛昱 刘丹 ZHANG Sheng;HU Guoqing;CHEN Jia;FU Ximin;ZHAO Rui;WANG Tingyu;LIU Dan(School of Mechanical and Automotive Engineering,South China University of Technology,Guangzhou 510000,China;School of Chinese Medicine,Guangdong Pharmaceutical University,Guangzhou 510000,China;Guangdong Huanyu Smart Medical Technology Co.,Ltd.,Guangzhou 510000,China;School of Humanities and Management,Guangdong Medical University,Dongguan 523000,China;School of Electronics and Control Engineering,Chang'an University,Xi'an 710000,China;School of Mathematics and Computer Science,Hebei Normal University for Nationalities,Chengde 067000,China)

机构地区：[1]华南理工大学机械与汽车工程学院,广州510000 [2]广东药科大学中医学院,广州510000 [3]广东寰宇智慧医疗科技有限公司,广州510000 [4]广东医科大学人文与管理学院,广东东莞523000 [5]长安大学电子与控制学院,西安710000 [6]河北民族师范学院数学与计算机科学学院,河北承德067000

出　　处：《计算机测量与控制》2024年第11期63-71,共9页Computer Measurement &Control

基　　金：国家自然科学基金(51105213);。

摘　　要：针对全球范围内的贫血健康问题,以及传统侵入式贫血检测会带来的不适,设计了一种基于结膜图像的非侵入式贫血检测系统,结合了图像处理算法和移动端开发技术,实现方便快捷的非侵入式贫血检测;通过数据集的构建与数据清洗、去光点、数据增强等预处理方法,在分类算法ResNet34基础上,改进残差块以提高网络对关键信息的学习能力,使用卷积层组的整体跳跃连接,便于深层信息和浅层信息的融合;开发了一款非侵入式贫血检测微信小程序,适用于日常的贫血检测;实验结果表明改进后的结膜贫血分类方法同原始基线网络和其他经典分类网络相比,分类效果更好,分类准确率、精确度、召回率和F_(1)分数分别达到了0.918、0.940、0.888和0.913,经实际应用移动端软件达到了完全识别的效果;表明该系统具有较强的分类准确性和可用性,也为非侵入式贫血检测提供了有益的参考。To address the global issue of anemia and the discomfort associated with traditional invasive anemia testing,this paper designs a non-invasive anemia detection system based on conjunctival images.This system integrates image processing algorithms and mobile development technologies to achieve convenient and rapid non-invasive anemia detection.Through the construction of datasets,data cleaning,light spot removal,data enhancement,and other preprocessing methods.On the basis of the classification algorithm ResNet34,the system improves the residual block to enhance the network's ability to learn key information.The overall skip connection of convolutional layer groups is used to achieve the integration of deep and shallow information.A non-invasive anemia detection WeChat mini-program is developed,and suitable for daily anemia testing.Experimental results show that the improved conjunctival anemia classification method is superrior to the original baseline network and other classical classification networks,achieving the classification accuracy,precision,recall,and F_(1)score of 0.918,0.940,0.888,and 0.913,respectively.The mobile application software achieves complete recognition in practical applications.This indicates that the system has strong classification accuracy and usability,providing a valuable reference for non-invasive anemia detection.

关 键 词：贫血检测 微信小程序开发 图像分类 健康监护 未病先防 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]