Sleep Disorder Diagnosis Using EEG Signals and LSTM Deep Learning Method

TL;DR

This study demonstrates that LSTM deep learning models can accurately classify sleep disorders using EEG signals, achieving over 95% accuracy and offering a fast, reliable diagnostic tool.

Contribution

The paper introduces an effective LSTM-based approach for sleep disorder detection from EEG data, with improved accuracy through fusion techniques, advancing clinical diagnostic methods.

Findings

LSTM models achieved 93.3% accuracy in classifying sleep disorders.

Fusion techniques increased accuracy to 95%.

Models demonstrated computational efficiency suitable for clinical use.

Abstract

Diagnosing sleep disorders is an important focus in neuroscience and engineering, as these conditions involve issues such as insufficient sleep, frequent awakenings, and difficulty reaching deep sleep. Accurate detection based on brain signals, particularly electroencephalography (EEG), enables development of personalized treatments. While statistical pattern recognition was once the standard for analyzing EEG, deep learning has become the dominant approach. In this study, we analyzed a public database of 197 full night sleep recordings from participants aged 25-101 years. After preprocessing and feature extraction, Long Short-Term Memory (LSTM) neural networks achieved 93.3% accuracy in distinguishing healthy from disordered sleep, which improved to 95% with fusion techniques. The models were also computationally efficient, suggesting strong clinical potential for rapid and precise sleep disorder diagnosis.