摘要
确定肺癌类型对于病人后续治疗方案的选择至关重要,但肺癌诊断必须由专业病理医师在显微镜下观察活体组织切片确定,诊断过程耗时长,且病理医师之间很难取得较好的一致诊断。随着病理全切片扫描设备的普及,病理图像可在计算机上保存、观察、分析,使通过现代数字图像处理技术辅助诊断或提供决策支持成为可能。对于亿级像素的全切片病理图像(WSI),通过在图像中提取小块、训练分类网络,并根据验证结果调整网络参数,可得到较好的分类准确率。测试时,聚合全图中小块分类结果得到最终类别。使用基于块的分类方法,卷积神经网络模型在WSI分类任务中取得了较好的效果,有望通过现代深度学习方法对肺癌提供辅助诊断及决策支持。
The diagnosis of lung cancer must be determined by a professional pathologist after looking at the biopsy under a microscope,which is time consuming.Besides,determining the subtype of lung cancer is crucial for the follow-up treatment plan,but it is difficult for pathologists to obtain a decent consistent diagnosis.With the popularity of virtual microscopy devices,it is convenient that pathological images can be stored,observed,and analyzed on computers,which make it possible to assist in diagnosis or provide decision support through modern computerized image processing techniques.For the very large GigaPixel pathological whole-slide images(WSI),patches were extracted and used to train the convolutional neural network model in this paper.After the parameters of network were adjusted based on the verification results,decent classification accuracy were ultimately obtained.During test phase,the classification results of the patches extracted from a whole image were aggregated to get the final result.By the patched-based classification method,the convolutional neural network model has achieved good results in the classification of pathological WSIs.It is hoped that modern deep learning methods can be used to assist in the diagnosis of lung cancer and provide decision support.
作者
宁静艳
俞晨
程年
刘芃
NING Jing-yan;YU Chen;CHENG Nian;LIU Peng(Automation School,Southeast University,Nanjing 210096,China;Department of Integrated TCM & Western Medicine,Jiangsu Cancer Hospital,Nanjing 210000,China)
出处
《软件导刊》
2019年第2期141-144,共4页
Software Guide
关键词
卷积神经网络
肺癌
病理图像分类
计算机辅助诊断
convolutional neural network
lung cancer
pathological image classification
computer-aided diagnosis
