| Summary: | 博士 === 國立成功大學 === 資訊工程學系 === 102 === Accurate automatic spike detection is highly beneficial to clinical assessment of epileptic electroencephalogram (EEG) data. In this thesis, a new two–stage approach is proposed for epileptic spike detection. First, the k-point nonlinear energy operator (k-NEO) is used to detect all possible spike candidates. Then, different kinds of features are extracted and applied to these candidates for spike classification. Moment descriptors are first applied as the features to describe the EEG candidate data and the empirical mode decomposed candidate data for spike classification. The statistical moments give promising classification results, however, the moment method does not include the shape information which is critical for epileptic spike classification. We subsequently propose a novel spike model-based method for spike classification. Although spikes with slow waves frequently occur in epileptic EEGs, they are not used in conventional spike detection. The newly proposed system accommodates both the single spike and spike with slow wave in the spike model. Using the AdaBoost classifier, the system outperforms the conventional spike model in both two- and three-class EEG classification problems. It not only achieves better accuracy in spike classification but provides new ability to differentiate between spikes and spikes with slow waves. Consequently, the proposed system has better capability for assisting clinical neurologists in routine EEG examinations and epileptic diagnosis.
|
|---|
