Wavelet-Based Multi-Class Seizure Type Classification System

TL;DR

This paper introduces a novel wavelet-based method for automatic multi-class seizure type classification from EEG signals, achieving state-of-the-art accuracy and setting new benchmarks on a large dataset.

Contribution

It proposes a new feature extraction and classification approach using Dual-tree Complex Wavelet Transform for seizure type identification from EEG data.

Findings

Achieved 99.1% weighted F1-score for seizure-wise classification.

Achieved 74.7% weighted F1-score for patient-wise classification.

Set new benchmark results on the TUH EEG Seizure Corpus dataset.

Abstract

Epilepsy is one of the most common brain diseases that affect more than 1\% of the world's population. It is characterized by recurrent seizures, which come in different types and are treated differently. Electroencephalography (EEG) is commonly used in medical services to diagnose seizures and their types. The accurate identification of seizures helps to provide optimal treatment and accurate information to the patient. However, the manual diagnostic procedures of epileptic seizures are laborious and highly-specialized. Moreover, EEG manual evaluation is a process known to have a low inter-rater agreement among experts. This paper presents a novel automatic technique that involves extraction of specific features from EEG signals using Dual-tree Complex Wavelet Transform (DTCWT) and classifying them. We evaluated the proposed technique on TUH EEG Seizure Corpus (TUSZ) ver.1.5.2 dataset and compared the performance with existing state-of-the-art techniques using overall F1-score due to class imbalance seizure types. Our proposed technique achieved the best results of weighted F1-score of 99.1\% and 74.7\% for seizure-wise and patient-wise classification respectively, thereby setting new benchmark results for this dataset.