Improving seizure detection performance reporting: analysing the duration needed for a detection

The impact of signal normalization on seizure detection using line length features

Accurate automated seizure detection remains a desirable but elusive target for many neural monitoring systems. While much attention has been given to the different feature extractions that can be used to highlight seizure activity in the EEG, very little formal attention has been given to the normalization that these features are routinely paired with. This normalization is essential in patient-independent algorithms to correct for broad-level differences in the EEG amplitude between people, and in patient-dependent algorithms to correct for amplitude variations over time. It is crucial, however, that the normalization used does not have a detrimental effect on the seizure detection process. This paper presents the first formal investigation into the impact of signal normalization techniques on seizure discrimination performance when using the line length feature to emphasize seizure activity. Comparing five normalization methods, based upon the mean, median, standard deviation, signal peak and signal range, we demonstrate differences in seizure detection accuracy (assessed as the area under a sensitivity-specificity ROC curve) of up to 52 %. This is despite the same analysis feature being used in all cases. Further, changes in performance of up to 22 % are present depending on whether the normalization is applied to the raw EEG itself or directly to the line length feature. Our results highlight the median decaying memory as the best current approach for providing normalization when using line length features, and they quantify the under-appreciated challenge of providing signal normalization that does not impair seizure detection algorithm performance.

Discriminating between best performing features for seizure detection and data selection

Seizure detection algorithms have been developed to solve specific problems, such as seizure onset detection, occurrence detection, termination detection and data selection. It is thus inherent that each type of seizure detection algorithm would detect a different EEG characteristic (feature). However most feature comparison studies do not specify the seizure detection problem for which their respective features have been evaluated. This paper shows that the best features/algorithm bases are not the same for all types of algorithms but depend on the type of seizure detection algorithm wanted. To demonstrate this, 65 features previously evaluated for online seizure data selection are re-evaluated here for seizure occurrence detection, using performance metrics pertinent to each seizure detection type whilst keeping the testing methodology the same. The results show that the best performing features/algorithm bases for data selection and occurrence detection algorithms are different and that it is more challenging to achieve high detection accuracy for the former seizure detection type. This paper also provides a comprehensive evaluation of the performance of 65 features for seizure occurrence detection to aid future researchers in choosing the best performing feature(s) to improve seizure detection accuracy.