What is the clinical evidence from this study?

Study design: Other. Population: Epilepsy. Intervention: Long Short-Term Memory (LSTM) classifier vs. Support Vector Machine (SVM) classifier. Primary outcome: Classification accuracy for pre-ictal vs ictal EEG signals.

What does this research mean for the field?

A double-layered Long Short-Term Memory (LSTM) classifier using Hurst Exponent and ARMA features detects pre-ictal versus ictal epilepsy seizures in neonatal EEG signals with 99.17% accuracy, outperforming previously used SVM classifiers. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

January 1, 2019Open Access

Detection of Epilepsy Seizures in Neo-Natal EEG Using LSTM Architecture

Key Result

A double-layered Long Short-Term Memory (LSTM) classifier using Hurst Exponent and ARMA features detected pre-ictal versus ictal epilepsy seizures in EEG signals with 99.17% accuracy.

Structured PICO

Does a double-layered LSTM classifier improve the detection accuracy of epileptic seizures from EEG signals compared to traditional classifiers?

Population

Publicly available EEG time series dataset from University of Bonn, Germany, containing 5 segments (A-E) of 100 recordings each. Segments A and B from healthy volunteers, C and D (seizure-free) and E (seizure activity) from 5 patients with epilepsy.

Intervention

Long Short-Term Memory (LSTM) classifier (double-layered memory units) using Hurst Exponent and Auto-regressive Moving Average (ARMA) features extracted from EEG signals.

Comparator

Support Vector Machine (SVM) classifier and single-layered LSTM.

Outcome

Classification accuracy of EEG signals into pre-ictal, inter-ictal, and ictal states.

A double-layered LSTM architecture using Hurst Exponent and ARMA features provides high accuracy (up to 98%) for detecting epileptic seizures from EEG signals.

Main Result

Absolute Event Rate: 99.17% vs 97.27%

Limitations

Binary classes are rarely available in real-life problems
LSTM architecture requires hyperparameter tuning and standardization to avoid biasing

Abstract

Epilepsy is the most unpredictable and recurrent disease among neurological diseases. Early detection of epileptic seizures can play a critical role in providing timely treatment to patients especially when a patient is in a remote area. This paper uses deep learning framework to detect epilepsy in the Electroencephalography (EEG) signal. The dataset used is publicly available and has a recording of three kinds of EEG signals: pre-ictal, inter-ictal (seizure-free epileptic) and ictal (epileptic with seizure). The proposed Long Short-Term Memory (LSTM) classifier classifies these three kinds of signals with up to 95% accuracy. For binary classification such as detection of inter-ictal or ictal only, its accuracy increases to 98%. The EEG signal is modelled as wide sense non-stationary random signal. Hurst Exponent and Auto-regressive Moving Average (ARMA) features are extracted from each signal. In this work, two different configurations of LSTM architecture: single-layered memory units and double-layered memory units are also modelled. After standardising the features, double-layered LSTM approach gives the highest accuracy in comparison to previously used Support Vector Machine (SVM) classifier and proved to be computationally efficient at Graphics Processing Unit (GPU).

Ask AI

Helpful

Bookmark

View Full Paper