摘要
乳腺癌是女性中最常见的癌症类型,如果能在乳腺癌的早期确诊和治疗,可以显著提高患者生存率。组织病理学检查是癌症确诊的黄金标准,针对医生很难在乳腺癌病理组织切片上精确快速的描绘出患病区域,给出一种基于改进AlexNet和注意力机制的网络模型用于IDC的自动检测,首先将全视野数字病理切片按照坐标信息进行不重复切片,然后将小切片输入至网络模型中进行训练,预测和评估,最后将小切片的分类结果按照坐标信息还原到全视野数字病理切片上,通过对小切片的二分类间接完成IDC的检测任务。该方法最终取得了86.34%的准确率、78.20%的F1评分和84.12%的平衡准确率,具有一定的实用价值和科研意义。
Breast cancer is the most common type of cancer among women.If breast cancer can be diagnosed and treated early,it can significantly improve the survival rate of patients.Histopathologic examination is the gold standard for cancer diagnosis.It is difficult for doctors to accurately and quickly depict the disease area on pathological sections of breast cancer.This paper presents a network model based on improved AlexNet and attention mechanism for automatic detection of IDC.First,the whole field digital pathological patches are sliced according to coordinate information.Then it is input into the network model for training,prediction and evaluation.Finally,the classification results of patches are restored to the full field digital slices according to the coordinate information.Through the two classification of small patches,the detection task of IDC is completed indirectly.This method finally achieved 86.34%accuracy,78.20%F1 score and 84.12%balanced accuracy,which has certain practical value and scientific research significance.
作者
郭笑颜
王波
张剑飞
刘明
GUO Xiao-yan;WANG Bo;ZHANG Jian-fei;LIU Ming(College of Computer and Control Engineering,Qiqihar University,Heilongjiang Qiqihar 161006,China)
出处
《齐齐哈尔大学学报（自然科学版）》
2022年第3期32-36,41,共6页
Journal of Qiqihar University(Natural Science Edition)
基金
黑龙江省省属高校基本科研业务费科研项目(135409608)。
作者简介
郭笑颜(1996-),男,河南南阳人,在读硕士,主要从事深度学习和计算机辅助诊断研究,gxyhpu@163.com。
