Akademik Veri Yönetim Sistemi
Neural Computing and Applications, cilt.38, sa.6, 2026 (Scopus)
Restless legs syndrome (RLS) is one of the most common sleep movement disorders diagnosed by detailed clinical and polysomnographic (PSG) evaluation. This study was conducted to facilitate the PSG diagnosis of RLS and to increase the physician’s ability to interpret PSG analysis. The aim of this article is to develop a new diagnostic model that will assist PSG in detecting RLS from electroencephalogram (EEG) signals during sleep staging, without examining electromyography (EMG) recordings during PSG analysis. EEG signals of 60 volunteers with known RLS (n = 32) and nonRLS (n = 28) (Age; 38.2 ± 5.31 years and Body Mass Index; 34.1 ± 3.47) were analyzed by PSG measurement. These data were obtained directly from our own sleep laboratory records and do not contain any public datasets. Additionally, all EEG channels were filtered between 0.5–38 Hz and processed using a standardized methodology. A novel algorithm was developed for the EEG-motion model by training the EEG signals using machine learning methods to help determine whether a person has RLS. EEG-motion was compared with the classical method ('PSG-EMG-classical’; RLS diagnoses made by a doctor analyzing EMG recordings from PSG). A strong positive correlation was found between the results of the classical method and the novel model, and this correlation was statistically significant (p < 0.001). Confusion matrix [accuracy (99%), sensitivity (98%), precision (100%), F1-Score (99%)] and ROC analysis (AUC: 99%) were performed to evaluate the EEG-motion model (p = 0.00). These results demonstrate that the proposed EEG-only approach can reliably detect RLS without the need for EMG recordings, representing a significant methodological contribution by simplifying PSG analysis and reducing electrode usage. EEG-motion can be integrated into PSG measurements in sleep laboratories. The novel model will increase the interpretation power of physicians and researchers during the analysis by providing the input of the patient’s periodic leg movement information from the EEG recording during sleep staging, which is the first stage of long PSG analysis.