基于MobileNet V2-ELM的肝硬化识别

Identification of Liver Cirrhosis Based on MobileNet V2-ELM

针对传统卷积神经网络的结构参数量大,识别率较低等问题,提出利用一种轻量级模型结构MobileNet V2与超限学习机(ELM)相结合的方法对肝硬化进行识别。首先,采用迁移学习的方法,在ImageNet数据集上进行预训练后得到权重和参数,避免数据产生过拟合现象。为提高识别肝硬化准确率,将模型的全连接层特征以向量形式输出,送入ELM进行分类,替代原有的softmax分类器。实验结果表明,该方法识别率高于当前的仅使用深度学习或者机器学习等方法且运算速率较高。

Aiming at the problems of large volume parameters and low recognition rate of traditional convolutional neural networks,a lightweight model structure,MobileNet V2 and over-limit learning machine(ELM),was proposed to identify cirrhosis.First,using the migration learning method,the pre-training on the ImageNet dataset yields weights and parameters to avoid over-fitting of the data.In order to improve the accuracy of identifying liver cirrhosis,the full connection layer features of the model are output in vector form and sent to the ELM for classification to replace the original softmax classifier.The experimental results show that the recognition rate of this method is higher than the current method of using only deep learning or machine learning and the operation rate is high.