摘要
不同癫痫发作类型呈现不同脑网络变化特征。为在保证模型性能的前提下提取有效的特征量,提出一种基于二维脑网络特征选择的癫痫发作类型分类算法。选择脑电信号20个通道中的19个通道构建出20个不同的相干性脑网络,分别提取6个网络特征,计算各网络和通道的贡献度。提取信号的时域和频域特征,应用随机森林计算特征个体贡献度,根据通道和特征个体贡献度进行二维迭代选择,对经选择后的特征集进行模型训练与测试。实验结果表明,该实验所用的癫痫发作类型分类方法在测试数据集上取得了较好的分类效果。
Different seizure types present different characteristics of brain network changes.In order to extract the effective feature quantity under the premise of ensuring the performance of the model,a seizure type classification algorithm based on two-dimensional brain network feature selection is proposed.Firstly,19 of the 20 channels of EEG signals are selected to construct 20 different coherent brain networks,and 6 network features are extracted to calculate the contribution of each network and channel.At the same time,the time domain and frequency domain features of the signal are extracted,the random forest is applied to calculate the individual contribution degree of the features,and then the two-dimensional iterative selection is carried out according to the individual contribution degree of the channel and the individual contribution of the features.Finally,the selected feature set is trained and tested.The experimental results show that the seizure type classification method used in this experiment achieves a good classification effect on the test dataset.
作者
吴端坡
励杰
应娜
WU Duanpo;LI Jie;YING Na(School of Communication Engineering,Hangzhou Dianzi University,Hangzhou 310018,China)
出处
《实验室研究与探索》
CAS
北大核心
2023年第12期73-78,共6页
Research and Exploration In Laboratory
基金
国家重点研发计划项目(2021YFE0100100)
浙江省自然科学基金联合基金项目(LBY21H090001)
杭州电子科技大学省属高校基本科研业务费项目(GK239909299001-401)
杭州电子科技大学高等教育教学改革研究实验技术专项项目(SYYB202308)。
关键词
相干性脑网络
通道贡献度
特征贡献度
二维特征选择
coherence brain network
channel contribution
feature contribution
two-dimensional feature selection
