论文部分内容阅读
Objective: An important issue in epileptology is the question whether informat ion extracted from the EEG of epilepsy patients can be used for the prediction o f seizures. Several studies have claimed evidence for the existence of a pre-se izure state that can be detected using different characterizing measures. In thi s paper, we evaluate the predictability of seizures by comparing the predictive performance of a variety of univariate and bivariate measures comprising both li near and non-linear approaches. Methods: We compared 30 measures in terms of th eir ability to distinguish between the interictal period and the pre-seizure pe riod. After completely analyzing continuous inctracranial multi-channel recordi ngs from five patients lasting over days, we used ROC curves to distinguish betw een the amplitude distributions of interictal and preictal time profiles calcula ted for the respective measures. We compared different evaluation schemes includ ing channelwise and seizurewise analysis plus constant and adaptive reference le vels. Particular emphasis was placed on statistical validity and significance.Re sults: Univariate measures showed statistically significant performance only in a channelwise, seizurewise analysis using an adaptive baseline. Preictal changes for these measures occurred 5-30 min before seizures. Bivariate measures exhib ited high performance values reaching statistical significance for a channelwise analysis using a constant baseline. Preictal changes were found at least 240 mi n before seizures. Linear measures were found to perform similar or better than nonlinear measures. Conclusions: Results provide statistically significant evide nce for the existence of a preictal state. Based on our findings, the most promi sing approach for prospectiveseizure anticipation could be a combination of biva riate and univariate measures. Significance: Many measures reported capable of s eizure prediction in earlier studies are found to be insignificant in performanc e,which underlines the need for statistical validation in this field.
Objective: An important issue in epileptology is the question whether informat ion extracted from the EEG of epilepsy patients can be used for the prediction of seizures. Several studies have called evidence for the existence of a pre-se izure state that can be detected using different characterizing measures. In thi s paper, we evaluate the predictability of seizures by comparing the predictive performance of a variety of univariate and bivariate measures, including both li near and non-linear approaches. Methods: We compared 30 measures in terms of th eir ability to distinguish between the interictal period and the pre-seizure pe riod. After completely analyzing continuous inctracranial multi-channel recordi ngs from five patients lasting over days, we used ROC curves to distinguish betwen the amplitude distributions of interictal and preictal time profiles calcula ted for the 各 measures. We compared different evaluation schemes including channelwise and seizurewise ana Repetitive constant and adaptive reference le vels. Re sults: Univariate measures showed arrested significant performance only in a channelwise, seizurewise analysis using an adaptive baseline. Preictal changes for these measures before seizures. Bivariate measures exhibiting high performance values reaching statistical significance for a channelwise analysis using a constant baseline. Preictal changes were found at least 240 mi n before seizures. Linear measures were found to performed similar or better than nonlinear measures. Provide significant significant evide nce for the existence of a preictal state. Based on our findings, the most promi sing approach for prospectiveseizure anticipation could be a combination of biva riate and univariate measures. Significance: Many measures reported capable of s eizure prediction in earlier studies are found to be insignificant in performanc e, which underlines the need for statistical validation in this field.