计算机工程与应用 ›› 2020, Vol. 56 ›› Issue (8): 136-142.DOI: 10.3778/j.issn.1002-8331.1901-0015

• 模式识别与人工智能 • 上一篇    下一篇

脑磁图脑功能连接网络癫痫棘波识别方法研究

张航宇,李彬,尹春丽,刘凯,王玉平,张军鹏   

  1. 1.四川大学 电气信息学院 自动化系,成都 610065
    2.首都医科大学 宣武医院 神经内科,北京 100053
  • 出版日期:2020-04-15 发布日期:2020-04-14

Study on Recognition Method of Epileptic Spike in Brain Functional Connectivity Network of Magnetoencephalogram

ZHANG Hangyu, LI Bin, YIN Chunli, LIU Kai, WANG Yuping, ZHANG Junpeng   

  1. 1.Department of Automation, College of Electrical Engineering and Information Technology, Sichuan University, Chengdu 610065, China
    2.Department of Neurology, Xuanwu Hospital, Capital Medical University, Beijing 100053, China
  • Online:2020-04-15 Published:2020-04-14

摘要:

脑磁图(MEG)现在被广泛用于临床检查及很多领域的医学研究中,基于静息态的脑磁图脑网络分析能用于研究大脑生理或病理机制。脑磁图分析对癫痫疾病的诊断具有重要的参考价值。对癫痫脑磁信号的自动分类可以及时对患者的情况作出判断,在临床上有很重要的意义。现有文献中对癫痫脑电信号的自动分类方法的研究已比较充分,但对癫痫脑磁信号的研究比较薄弱。提出了一种基于脑功能连接网络的全频段机器学习癫痫脑磁棘波信号自动判别方法,对四种分类器进行了综合判别对比,选择了效果最优的分类器,判别准确率可达到93.8%。因此,该方法在脑磁图癫痫棘波的自动识别与标记方面有较好的应用前景。

关键词: 静息态脑磁图, 脑功能网络, 机器学习, 特征提取

Abstract:

Magnetoencephalography(MEG) is now widely used in clinical examination and medical research in many fields. Resting-state-based MEG brain network analysis can be used to study the physiological or pathological mechanism of the brain. Magnetoencephalogram analysis has important reference value for the diagnosis of epilepsy. The automatic classification of epileptic magnetoencephalic signals can make timely judgments about the patient’s condition, which is of great clinical significance. The existing literature on the automatic classification of epileptic EEG signals has been fully studied, but the study of epileptic MEG signals is relatively weak. In this paper, an automatic discrimination method of epileptic magnetoencephalic spike wave signals based on full-band machine learning based on brain functional connectivity network is proposed. Four classifiers are synthetically discriminated and compared, and the classifier with the best effect is selected. The discriminant accuracy can reach 93.8%. Therefore, this method has promising application prospects in automatic recognition and marking of epileptic spikes in MEG.

Key words: resting state Magnetoencephalography(MEG), brain function network, machine learning, feature extraction