基于深度学习的肺部肿瘤图像识别方法  

Method for Lung Tumor Image Recognition Based on Deep Learning

作　　者：高雷鸣 肖满生[1] 向华政[1] Gao Leiming;Xiao Mansheng;Xiang Huazheng(College of Computer,Hunan University of Technology,Zhuzhou 412000,China)

机构地区：[1]湖南工业大学计算机学院,湖南株洲412000

出　　处：《计算机测量与控制》2020年第10期160-164,共5页Computer Measurement &Control

基　　金：湖南省自然科学基金资助项目(2018JJ4068,2018JJ4078)。

摘　　要：鉴于浅层卷积神经网络难以获取图像深层特征、易过度拟合导致分类效率和精度低的问题,因此,设计一种肺部肿瘤图像的深度学习识别模型;在运用样本扩充和迁移学习的基础上,对AlexNet卷积神经网络进行改善和提升,在每层网络数据输入之前对数据进行归一预处理,同时使用线性整流函数(ReLU),实现对肺部肿瘤表达性特征的快速获取,输出端经由三层全连接层和softmax算法进行分类;实验结果表明,此方法在网络收敛速率和分类精度方面取得更优性能,比基于AlexNet卷积神经网络分类精度提高5.66%以上,且具备良好的健壮性。In view of the fact that shallow convolutional neural networks are difficult to obtain the deep features of the image and are easy to overfit,which leads to low classification efficiency and accuracy,a deep learning recognition model for lung tumor images is designed.Based on the use of data augmentation and transfer learning,and the improvement and promotion of the AlexNet convolutional neural network,the data is subjected to a normal preprocessing before the data input of each layer of the network,while applying a linear rectification function(ReLU).Realize fast acquisition of lung tumor expression characteristics,and the output end is classified through three fully connected layers and Softmax algorithm.The experimental outcome indicate that the proposed method achieves better performance in terms of network convergence speed and classification accuracy,which is 5.66%higher than that based on the AlexNet convolutional neural network,and it has good robustness.

关 键 词：样本扩充 迁移学习 深度学习 归一预处理 医学图像识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]