摘要
意识障碍(disorders of consciousness,DOC)是由中枢神经系统受损导致个体感知能力出现障碍的一种神经系统疾病。DOC的预后评估能够辅助医师及时调整治疗措施,促进DOC患者的意识恢复。传统的预后评估手段包括行为量表、核磁共振成像等,但这些检查手段易受干扰、价格昂贵、难以实时评估。相关研究表明:脑电图(electroencephalogram,EEG)与DOC患者的意识水平显著相关,这为DOC的预后提供了一种潜在的评估手段。基于此,文中以EEG为数据源,提出了一种新的基于EEG的DOC预后评估方法。首先,对给定的DOC-EEG信号,采用小波包变换提取不同频带的脑电信号;其次,计算不同频带脑电信号的功率谱,并绘制对应功率谱的庞加莱散点图;然后,对每一庞加莱散点图定义3种新的度量指标:最大半径、区域密度以及密度变异性;进而,结合3种度量指标,提出基于功率谱庞加莱散点图的脑电特征提取方法(power spectra-Poincare plot-based feature,PPBF);最后,结合所提特征与随机森林分类器,提出了一种基于PPBF的DOC预后状态自动评估方法。采用西安某医院神经内科临床采集19名DOC患者的脑电数据对所提方法进行了验证,所得平均评估准确率85.54%,表明所提特征能够有效完成DOC预后状态的自动识别。
Disorders of consciousness(DOC)is a kind of neurological disorders,which could lead to impairment of perception ability.The prognosis assessment of DOC is helpful for physicians to adjust treatment measures in time and promote the recovery of consciousness of DOC patients.Traditional prognostic methods including behavior scale and nuclear magnetic resonance imaging,etc.,are often vulnerable to interference,expensive,and it is impossible for physicians to evaluate them in real time.Studies have shown a significant correlation between electroencephalogram(EEG)and levels of consciousness in DOC patients,which provides a potential way to assess the prognostic state of DOC patients.Based on this,a new EEG-based prognostic assessment method for DOC is proposed.Firstly,the DOC-EEG signals in different frequency bands are extracted by wavelet packet transform.Secondly,the power spectrum of EEG signals in different frequency bands are calculated,and the corresponding Poincare plots are drawn.Furthermore,three new measures are defined for each Poincare scatter plot:maximum radius,regional density and density variability.Then,the power spectra-Poincare plot-based feature(PPBF)is proposed based on the above three metrics.Finally,the PPBF is fed into random forest to complete assessment of the prognostic states of DOC.Simulation results on EEG signals collected from 19 DOC patients in the department of neurology of a hospital in Xi′an are conducted to verify the feasibility and effectiveness of the proposed method.The average assessment accuracy of the proposed method is 85.54%,indicating the proposed features have a good performance in the automatic recognition of DOC patient′s prognostic status.
作者
程雅楠
李斯卉
宋江玲
张瑞
CHENG Yanan;LI Sihui;SONG Jiangling;ZHANG Rui(Medical Big Data Research Center,Northwest University,Xi′an 710127,China)
出处
《西北大学学报（自然科学版）》
CAS
CSCD
北大核心
2021年第4期558-566,共9页
Journal of Northwest University（Natural Science Edition）
基金
国家自然科学基金面上项目(12071369)
陕西省重点研发计划资助项目(2019ZDLSF02-09-02)
国家自然科学基金青年项目(62006189)。
关键词
意识障碍
预后评估
脑电信号
功率谱-庞加莱散点图
随机森林
disorder of consciousness(DOC)
prognostic assessment
electroencephalogram(EEG)
power spectra-Poincare plot
random forest
