摘要
本文针对癫痫脑电图(EEG)信号中的发作检测问题,分析了癫痫患者EEG信号中的特异性特征,在传统EEG信号时频域基础上提出了改进的脑网络特征。本文对EEG信号进行分解,并重构了EEG信号,发现在重构信号上癫痫发作和癫痫未发作表现出较大差异。因此在重构EEG信号上通过皮尔逊系数(PCC)构建脑网络,并在该网络上提取脑网络特征,最后将这些特征输入Bi-LSTM-Attention混合网络检测癫痫发作。该网络可以筛选出对癫痫发作检测结果具有决定性因素的特征,捕捉EEG时间序列中最重要的信息。为了评估本文的方法,在公开的CHB-MIT数据集上进行实验,获得了96.20%的准确率、96.80%的特异性和95.31%的敏感性,实验结果表明该方法在癫痫发作检测这个任务上具有不错的性能。
This study focuses on seizure detection within electroencephalogram(EEG)signals for epilepsy patients,analyzing distinctive features in epileptic EEG signals and introducing improved brain network characteristics based on the traditional EEG signal time-frequency domain.The paper involves the decomposition and subsequent reconstruction of EEG signals,revealing substantial differences between seizure and non-seizure states in these reconstructed signals.Consequently,a brain network is constructed on the reconstructed EEG signals using the Pearson Correlation Coefficient(PCC),from which brain network features are extracted.These features are then fed into a Bi-LSTM-Attention hybrid network to detect epileptic seizures.This network is capable of identifying key features crucial for the detection of epileptic seizures and capturing the most significant information in the EEG time series.To assess the effectiveness of this method,experiments were conducted on the publicly available CHB-MIT dataset,yielding an accuracy of 96.20%,a specificity of 96.80%,and a sensitivity of 95.31%.The results indicate that this method is highly effective in the task of epilepsy seizure detection.
作者
龚帅奎
蒋路茸
范骐凯
GONG Shuaikui;JIANG Lurong;FAN Qikai(School of Information Science and Engineering,Zhejiang Sci-Tech University,Hangzhou 310018,China)
出处
《智能计算机与应用》
2024年第2期69-75,共7页
Intelligent Computer and Applications
基金
浙江省重点研发计划(2022C03136)
国家自然科学基金(61602417)。
