Sleep Disorder Diagnosis Using EEG Signals and LSTM Deep Learning Method

Diagnosing brain disorders, particularly sleep disorders, has emerged as a critical area of research in both basic sciences and engineering. Sleep disorders encompass a range of conditions, including excessive or insufficient sleep, frequent awakenings, difficulty falling asleep, and challenges in reaching deep sleep stages. A key application of sleep disorder diagnosis lies in supporting the medical community through the development of advanced algorithms and models capable of accurately identifying these disorders based on individual brain signals, such as those captured by electroencephalography (EEG). These sophisticated approaches facilitate the creation of personalized treatment programs, paving the way for more targeted and effective interventions. Historically, statistical pattern recognition methods were regarded as the gold standard for analyzing brain signals and identifying sleep disorders. However, with the rapid evolution of deep learning, these traditional methods have been increasingly supplanted by more advanced deep learning techniques. To fully harness the potential of these emerging technologies, further research is essential to explore the application of deep learning in analyzing brain signals for sleep disorder identification. Such efforts can address existing knowledge gaps and expand our understanding in this field. In this study, a comprehensive public database was utilized, consisting of 197 full-night sleep recordings from participants aged 25 to 101 years, including both male and female subjects. Following rigorous preprocessing stages—such as noise elimination and signal refinement—critical features were extracted and analyzed. Using deep learning techniques, particularly Long Short-Term Memory (LSTM) neural networks, the researchers achieved a remarkable accuracy rate of 93.3% in distinguishing between healthy and diseased classes. By incorporating advanced fusion techniques, the classification accuracy was further improved to 95%. Notably, the computational efficiency of these deep learning networks makes them highly promising for clinical applications in sleep disorder diagnosis. Their relatively low processing time suggests that such algorithms could be invaluable in clinical settings, enabling rapid and precise detection of sleep disorders.