摘要
为了改善在宫颈细胞的分类工作中,出现的将异常的病变细胞与正常细胞判断混淆的误诊问题,提出了一种细胞生物学特征-卷积神经网络联合分类方法。首先,使用ResNet分类网络提取出特征向量,然后再将其与手动提取的DNA指数、细胞核/浆比特征一起输入到全连接层,并使用基于MSE损失值的逻辑回归分类,对宫颈细胞进行分类识别。使用5折交叉验证法在Heer数据集上的实验结果表明,这种将卷积神经网络与细胞生物学特征相结合的联合分类方法相较于ResNet卷积神经网络,分类结果的整体准确率提高4%,达到了95%;同时优化MSE损失函数的方法在准确率达到瓶颈的情况下,能够将严重错分率由2.10%降为0.248%,且保持了细胞的整体识别准确率。提出的方法进行计算机辅助检测,能够提升宫颈细胞分类工作准确率、降低误诊率。
In order to improve the misdiagnosis problem that confuses abnormal diseased cells with normal cells in the classification of cervical cells, a cell biology feature-convolutional neural network joint classification method was used to investigate that in this paper. First, use the ResNet classification network to extract the feature vector, and then input it into the fully connected layer together with the manually extracted DNA index and the nuclear/plasma ratio feature, and use the logistic regression classification based on the MSE loss value to classify the cervical cells recognition. The experimental results on the Heer data set using the 5-fold cross-validation method show that the combined classification method that combines the convolutional neural network with the biological characteristics of cells is compared with the ResNet convolutional neural network. The overall classification results The accuracy rate is increased by 4% to 95%;at the same time, the method of optimizing the MSE loss function can reduce the serious misclassification rate from 2.10% to 0.248% when the accuracy rate reaches the bottleneck, and the accuracy rate of identifying abnormal diseased cells can reach 97%. It can be seen that using the method proposed in this article for computer-assisted detection can improve the accuracy of cervical cell classification, and reduce the rate of misdiagnosis.
作者
郑旎杉
曾立波
ZHENG Nishan;ZENG Libo(Electronic Information School,Wuhan University,Wuhan 430072,China)
出处
《激光杂志》
CAS
北大核心
2021年第12期212-218,共7页
Laser Journal
基金
国家科技支撑计划(No.2011BAF02B00)。
关键词
宫颈细胞
细胞分类
细胞生物学特征
卷积神经网络
cervical cells
cell classification
cell biological characteristics
convolutional neural network
