基于深度学习的肠道肿瘤图像识别方法研究  

Research on Image Recognition Method of Intestinal Tumor Based on Deep Learning

作　　者：杨波 张立娜[2] YANG Bo;ZHANG Li-na(School of information engineering,Changchun University of Finance and Economics,Changchun 130122,China;School of information technology,Jilin Agricultural University,Changchun 130118,China)

机构地区：[1]长春财经学院信息工程学院,吉林长春130122 [2]吉林农业大学信息技术学院,吉林长春130118

出　　处：《电脑与信息技术》2021年第5期23-25,36,共4页Computer and Information Technology

基　　金：吉林省教育厅“十三五”科学技术重点项目(项目编号:JJKH20201256KJ)。

摘　　要：肠道肿瘤诊断目前主要依靠医务人员对于医学图像的经验判断,随着患者不断增加,对于医院和医生的诊断压力也逐渐加大。采用一种自动判断肠道肿瘤的方式对于解决目前肠道肿瘤诊断困难非常必要。文章研究利用深度学习方法针对肠道肿瘤图像进行特征提取和识别,实验采集了1600个医学图像数据,按照7:3比例分配训练集和测试集,采用ResNet50模型,经过训练的网络准确率达到97.95%,在一定程度上为肠癌的诊断提供了辅助诊断信息,具有一定的实用价值。At present,the diagnosis of intestinal tumor mainly depends on the medical staff's experience in medical images.With the increasing number of patients,the pressure of diagnosis for hospitals and doctors is gradually increasing.It is very necessary to adopt an automatic way to judge intestinal tumor for solving the current difficulties in the diagnosis of intestinal tumor.In this paper,deep learning method is used for feature extraction and recognition of intestinal tumor images.1600 medical image data are collected in the experiment,and the training set and test set are allocated according to the 7:3 ratio.Using resnet50 model,the trained network accuracy reaches 97.95%,which provides auxiliary diagnosis information for the diagnosis of colorectal cancer to a certain extent,and has certain practical value.

关 键 词：肠道肿瘤 深度学习 卷积神经网络 ResNet50 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]