基于深度卷积神经网络的脑电图异常检测

为解决EEG自动检测的错误率非常高的问题，提出了一种基于深层卷积神经网络(CNN)对脑电图进行异常检测的方法:首先，对多个异构数据源按标准进行重构和预处理，生成了有118 716个样本的训练集和有12 022个样本的测试集；然后，构建有快捷连接的深层CNN模型，以自动化学习ECG特征并进行分类识别; 接着，将模型在训练集上进行试验与调参，保存了性能最好的模型参数; 最后，在测试集上进行预测.预测结果显示该模型达到了94.33%的分类准确率.通过所提方法对脑电信号进行处理与分析，能够自动提取EEG特征并进行异常识别，从而达到快速检测与辅助诊疗的目的.

The Detection of Anomaly in Electroencephalogram with Deep Convolutional Neural Networks

In order to solve the problem of high error rate in EEG automatic detection, a method of EEG anomaly detection based on deep convolution neural network (CNN) is proposed. Firstly, multiple heterogeneous data sources are reconstructed and preprocessed according to the standard, and a training set with 118 716 samples and a test set with 12 022 samples are generated. Secondly, a deep CNN model with fast connection is constructed. Then, the model is tested and adjusted on the training set, and the best model parameters are saved. Finally, the model is predicted on the test set. The prediction results show that the model achieves 94.33% classification accuracy. EEG features can be automatically extracted and abnormal recognition can be carried out through the processing and analysis of EEG signals with the proposed method, so as to achieve the purpose of rapid detection and auxiliary diagnosis.