摘要
针对计算机辅助乳腺疾病诊断方法准确率低、耗时长等问题,提出一种基于改进的卷积神经网络(CNN)的乳腺疾病诊断方法.该方法从以下3个方面做了改进:(1)设计双通道卷积神经网络来解决单通道特征提取不充分的问题;(2)采用Dropout技术有效地防止过拟合现象;(3)采用支持向量机(SVM)代替传统的Softmax分类器以减少运算量,提高运算速度.测试结果表明:所提出的分类模型平均准确率高达92.31%,平均训练时间为968s,充分验证了该方法的有效性.
Aiming at the problem of low accuracy and long time consuming in computer-aided breast disease diagnosis,a new breast disease diagnosis method based on improved Convolutional Neural Network(CNN)is proposed.This method has been improved from the following three aspects:first,the design of a double channel convolutional neural network to solve the problem of inadequate single channel feature extraction;secondly,using Dropout technology to effectively prevent overfitting;finally,the support vector machine(Support Vector Machine,SVM softmax)to replace the traditional classifier in order to reduce computation to improve the speed of operation.After testing,the average accuracy of the proposed classification model is up to 92.31% and the average training time is 968 s,which fully validates the effectiveness of this method.
作者
赵京霞
钱育蓉
张猛
杜娇
ZHAO Jing-xia;QIAN Yu-rong;ZHANG Mcng;DU Jiao(School of Software,Xinjiang Univewity,Urumqi 830008,China)
出处
《东北师大学报（自然科学版）》
CAS
北大核心
2019年第2期65-70,共6页
Journal of Northeast Normal University(Natural Science Edition)
基金
国家自然科学基金资助项目(61562086)
新疆维吾尔自治区教育厅项目(XJEDU2016S035)
关键词
计算机辅助诊断
卷积神经网络
双通道
医学图像分类
computer aided diagnosis
convolutional neural network
double channel
medical image classification
