癫痫模型中痫样棘波自动检测的新算法及其应用

A new algorithm for automatically detecting epileptiform spikes and its application in epilepsy models

癫痫的典型特征是神经元群体产生异常同步发放,在记录的神经电信号中呈现为痫样棘波。为了定量分析癫痫发生时的痫样棘波,本文设计了一种基于窗口的检测法用于自动检测大鼠海马CA1区急性癫痫模型的棘波信号,即群峰电位(PS),并计算其特征参数。实验结果表明,在钾离子通道拮抗剂4-氨基吡啶(4-AP)和γ-氨基丁酸A型受体拮抗剂印防己毒素(PTX)诱导的癫痫模型中,该算法可直接从原始宽频带记录信号中正确识别PS波。两种模型中的PS检出率分别为94.2%±1.6%(n=11)和95.9%±1.9%(n=12),且误检率分别为3.5%±2.3%(n=11)和4.8%±2.3%(n=12),远小于普通阈值法的误检率。比较4-AP和PTX模型的PS波特征,结果显示:4-AP诱导的PS波具有较宽的波形,发放较分散,发放间隔主要分布于100~700 ms范围内。而PTX诱导的PS则呈现爆发式发放,发放率较高,发放间隔主要分布于2~20 ms范围内,使得每秒PS幅值之和显著大于4-AP模型。因此,PTX模型的神经元群体同步发放活动比4-AP模型要强烈。总之,该棘波检测新算法可以正确识别和分析痫样棘波,为癫痫发生机制的研究和癫痫治疗新方法的评估提供了一种有用的数据分析工具。

Epilepsy is characterized by abnormally synchronized firing of neuronal populations, which is presented as epileptiform spikes in neural electrical signal recordings. In order to investigate the epileptiform spikes quantitatively, we designed a new window-based algorithm to automatically detect population spikes （PS） in acute epilepsy models in rat hippocampus CA 1 region, and to calculate characteristic parameters of PS. Results show that the algorithm could recognize PS waveforms directly in wideband recording signals in epilepsy models induced by 4- aminopyridine （4-AP）, a potassium channel blocker, or by picrotoxin （PTX）, an antagonist of y-aminobutyric acid A-type receptor. The PS detection ratios of the two epilepsy models were 94.2%±1.6% （n= 11） and 95.9%±1.9% （n= 12）, respectively. The false positive ratios were 3.5%±2.3% （n= 11） and 4.8%±2.3% （n= 12）, which were significantly lower than those of the conventional threshold method. Comparisons of the PS patterns between the 4-AP model and the PTX model showed that the PS of the 4-AP model had wider waveforms and fired more dispersedly with intervals mainly in the range of 100-700 ms. The PS of the PTX model fired as Burst with a higher firing rate and with intervals mainly in the range of 2-20 ms, resulting in a larger sum of spike amplitudes per second than the 4-AP model. Thus, the synchronous firing of neuronal populations in the PTX model was more intense than that in the 4-AP model. In conclusion, the new algorithm of PS detection can correctly detect and analyze epileptiform population spikes. It provides a useful tool of data analysis for investigating the underlying mechanism of seizure generation and for evaluating new therapeutics of epilepsy.