基于卷积神经网络的乳腺癌病理图像分类  

Classification of Breast Cancer Pathological Images Based on Convolutional Neural Network

作　　者：蔡豪杰 王林[1] 王义兵 侍鹏 

机构地区：[1]盐城工学院电气工程学院,江苏 盐城

出　　处：《建模与仿真》2023年第5期4320-4331,共12页Modeling and Simulation

摘　　要：乳腺癌已经超过肺癌,成为世界第一大癌症。因此,乳腺癌的诊断就显得十分重要。为了提高对乳腺癌病理图像分类的准确率,提出了一种基于卷积神经网络的诊断方法。这种方法的出现,能做到快速对乳腺癌病理图像进行良恶性分类。一般来说,乳腺癌的病理图像结构十分复杂,为了增强网络的特征提取的能力,在卷积神经网络的基础上引进随机函数链神经网络和CA注意力机制。因为乳腺癌数据集太少,使用数据增强去扩充数据集。分别进行横向实验与消融实验,实验结果表明,优化后的卷积神经网络能有效提高分类的准确率。Breast cancer has already overtaken lung cancer as the No. 1 cancer in the world. Therefore, the di-agnosis of breast cancer is of great importance. To improve the accuracy of classifying pathological images of breast cancer, a diagnostic method based on convolutional neural network (CNN) is pro-posed. This method makes a quick and automatic benign and malignant diagnosis for breast cancer pathology images. In general, the structure of pathological images of breast cancer is very complex. In order to enhance the capability of feature extraction, Random Vector Functional Link Neural Network (RVFLNN) and Coordinate Attention (CA) are introduced based on CNN. Because there are too few breast cancer datasets, data enhancement is used to augment the datasets. Ablation ex-periments and horizontal experiments were conducted. The experimental results show that the op-timized CNN can improve the accuracy of classification effectively.

关 键 词：卷积神经网络 函数链神经网络 注意力机制 病理图像 乳腺癌 特征提取 数据集 数据增强 

分 类 号：R73[医药卫生—肿瘤]