Developing a novel epileptic discharge localization algorithm for electroencephalogram infantile spasms during hypsarrhythmia.

Infantile spasms (ISS) is a devastating epileptic syndrome that affects children under the age of 1 year. The diagnosis of ISS is based on the semiology of the seizure and the electroencephalogram (EEG) background characterized by hypsarrhythmia (HYPS). However, even skilled electrophysiologists may interpret the EEG of children with ISS differently, and commercial software or existing epilepsy detection algorithms are not helpful. Since EEG is a key factor in the diagnosis of ISS, misinterpretation could result in serious consequences including inappropriate treatment. In this paper, we developed a novel algorithm to localize the relevant electrical abnormality known as epileptic discharges (or spikes) to provide a quantitative assessment of ISS in HYPS. The proposed algorithm extracts novel time-frequency features from the EEG signals and localizes the epileptic discharges associated with ISS in HYPS using a support vector machine classifier. We evaluated the proposed method on an EEG dataset with ISS subjects and obtained an average true positive and false negative of 98 and 7%, respectively, which was a significant improvement compared to the results obtained using the clinically available software. The proposed automated method provides a quantitative assessment of ISS in HYPS, which could significantly enhance our knowledge in therapy management of ISS.

Automatic localization of epileptic spikes in eegs of children with infantile spasms.

A novel methodology is proposed for identifying epileptiform discharges associated with individuals exhibiting Infantile Spasms (ISS) also known as West Syndrome, which is characterized by electroencephalogram (EEG) recordings exhibiting hypsarrythmia (HYPS). The approach to identify these discharges consists of three stages: first - construct the time-frequency domain (TFD) of the EEG recording using matching pursuit TFD (MP-TFD), second - decompose the TFD matrix into two submatrices (W, H) using non-negative matrix factorization (NMF), and third - use the decomposed spectral and temporal vectors to locate the epileptiform discharges, referred to as spikes, during intervals of HYPS. The method was applied to an EEG dataset of five individuals and the identification of spike locations was compared with those which were visually identified by the epileptologists and those obtained using commercially available clinical analysis software. The MP-TFD method resulted in average true positive and false negative percentages of 86% and 14%, respectively, which represents a significant improvement over the clinical software, which achieved average true positive and false negative percentages of 4% and 96%, respectively.