[Paper Review] Automatic Identification of Epileptic Seizures from EEG Signals using Sparse Representation-based Classification

Sobhan Sheykhivand, Tohid Yousefi Rezaii+3 more|arXiv (Cornell University)|Oct 18, 2019

EEG and Brain-Computer Interfaces29 references4 citations

TL;DR

This paper proposes a fully automated epileptic seizure detection system using sparse representation-based classification (SRC) with self-learned EEG dictionaries, eliminating the need for manual feature extraction. It achieves 100% sensitivity, specificity, and accuracy in 8 out of 9 scenarios and demonstrates robustness to 0 dB noise, outperforming state-of-the-art methods across diverse multi-class seizure detection configurations.

ABSTRACT

Identifying seizure activities in non-stationary electroencephalography (EEG) is a challenging task, since it is time-consuming, burdensome, and dependent on expensive human resources and subject to error and bias. A computerized seizure identification scheme can eradicate the above problems, assist clinicians and benefit epilepsy research. So far, several attempts were made to develop automatic systems to help neurophysiologists accurately identify epileptic seizures. In this research, a fully automated system is presented to automatically detect the various states of the epileptic seizure. The proposed method is based on sparse representation-based classification (SRC) theory and the proposed dictionary learning using electroencephalogram (EEG) signals. Furthermore, the proposed method does not require additional preprocessing and extraction of features which is common in the existing methods. The proposed method reached the sensitivity, specificity and accuracy of 100% in 8 out of 9 scenarios. It is also robust to the measurement noise of level as much as 0 dB. Compared to state-of-the-art algorithms and other common methods, the proposed method outperformed them in terms of sensitivity, specificity and accuracy. Moreover, it includes the most comprehensive scenarios for epileptic seizure detection, including different combinations of 2 to 5 class scenarios. The proposed automatic identification of epileptic seizures method can reduce the burden on medical professionals in analyzing large data through visual inspection as well as in deprived societies suffering from a shortage of functional magnetic resonance imaging (fMRI) equipment and specialized physician.

Motivation & Objective

• To develop a fully automated, computationally efficient system for epileptic seizure detection from non-stationary EEG signals.
• To eliminate reliance on time-consuming manual feature extraction and preprocessing steps common in existing methods.
• To enhance clinical workflow by reducing neurophysiologist workload in interpreting large-scale EEG data.
• To improve detection accuracy and robustness in diverse, real-world seizure classification scenarios, including 2–5 class configurations.
• To support epilepsy research and clinical diagnosis in resource-limited settings lacking fMRI or specialized personnel.

Proposed method

• The method employs sparse representation-based classification (SRC) using a self-learned dictionary directly from raw EEG signals, bypassing traditional preprocessing and feature engineering.
• The dictionary is learned from EEG data using a sparse coding framework that captures intrinsic signal patterns associated with different seizure states.
• Classification is performed by representing each test EEG segment as a sparse linear combination of atoms from the learned dictionary and assigning the class with the minimum representation error.
• The approach is invariant to signal amplitude variations and maintains performance under low signal-to-noise ratios, including up to 0 dB noise.
• The system evaluates multiple configurations, including 2-class, 3-class, 4-class, and 5-class seizure state detection, to ensure generalization.
• No external preprocessing or handcrafted features are used—only raw EEG time-series data is input directly into the SRC framework.

Experimental results

Research questions

• RQ1 Can a fully automated seizure detection system achieve high accuracy without relying on manual feature extraction or preprocessing in EEG signals?
• RQ2 How does the proposed SRC-based method with self-learned dictionaries perform across diverse multi-class seizure detection scenarios (2 to 5 classes)?
• RQ3 To what extent is the proposed method robust to measurement noise levels commonly found in clinical EEG recordings?
• RQ4 How does the performance of the proposed method compare to state-of-the-art algorithms in terms of sensitivity, specificity, and accuracy?
• RQ5 Can this system effectively reduce the burden on clinicians and support epilepsy diagnosis in low-resource settings?

Key findings

• The proposed method achieved 100% sensitivity, specificity, and accuracy in 8 out of 9 tested scenarios, demonstrating exceptional detection performance.
• The system maintained robust performance under 0 dB measurement noise, indicating strong resilience to signal degradation.
• The method outperformed existing state-of-the-art algorithms in terms of sensitivity, specificity, and accuracy across all evaluated configurations.
• The absence of preprocessing and feature extraction steps significantly reduces computational overhead and increases automation.
• The approach is scalable and generalizable across multiple seizure state classification tasks, including complex 5-class scenarios.
• The method is particularly beneficial for clinical and research applications in regions with limited access to fMRI or specialized neurologists.