A Single-Channel EEG Approach for Sleep Stage-Independent Automatic Detection of REM Sleep Behavior Disorder

A Single-Channel EEG Approach for Sleep Stage-Independent Automatic Detection of REM Sleep Behavior Disorder | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search A Single-Channel EEG Approach for Sleep Stage-Independent Automatic Detection of REM Sleep Behavior Disorder View ORCID Profile Gabriele Salvatore Giarrusso , View ORCID Profile Irene Rechichi , View ORCID Profile Alessandro Cicolin , View ORCID Profile Gabriella Olmo doi: https://doi.org/10.1101/2025.05.28.25328491 Gabriele Salvatore Giarrusso 1 Department of Control and Computer Engineering , Politecnico di Torino, 10129, Turin, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Gabriele Salvatore Giarrusso For correspondence: gabriele.giarrusso{at}polito.it Irene Rechichi 1 Department of Control and Computer Engineering , Politecnico di Torino, 10129, Turin, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Irene Rechichi Alessandro Cicolin 2 Department of Neuroscience “Rita Levi Montalcini”, University of Turin , 10126 Turin, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Alessandro Cicolin Gabriella Olmo 1 Department of Control and Computer Engineering , Politecnico di Torino, 10129, Turin, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Gabriella Olmo Abstract Full Text Info/History Metrics Data/Code Preview PDF Abstract Rapid Eye Movement (REM) Sleep Behavior Disorder (RBD) is a parasomnia characterized by the loss of physiological muscle atonia during REM sleep, often manifesting through dream-enacting behavior. Idiopathic RBD is largely considered a prodromal stage of neurodegenerative diseases, with a conversion rate to overt α -synucleinopathies of up to 96% after 14 years. Currently, the diagnostic procedure relies on time-consuming and labor-intensive inspection of polysomnography (PSG). This study proposes a Machine Learning (ML), stage-agnostic framework for the automatic detection of RBD subjects through unstaged, single-channel EEG sleep data from 58 subjects (32 RBD). The best model achieved 86.21% accuracy, 90.6% sensitivity, and 87.9% F-1 score, demonstrating strong predictive power. This study is the first to explore whole-night EEG data for RBD detection, paving the way for scalable, lightweight clinical decision support systems for early neurodegenerative screening and risk assessment. Clinical relevance This study presents a lightweight, clinical decision support tool to enhance RBD detection and support early interventions in neurodegenerative diseases. I. INTRODUCTION Rapid Eye Movement (REM) Sleep Behavior Disorder (RBD) is a parasomnia characterized by the loss of physiological muscle atonia during REM sleep, which manifests through dream-enacting, oftentimes accompanied by violent behavior. Idiopathic RBD is considered a prodromal neurodegenerative stage, with a conversion rate to overt α -synucleinopathy of up to 96% after a 14-year follow-up [ 1 ]. Hence, this makes it a powerful tool as an early marker of neurodegenerative disorders such as Parkinson’s Disease (PD), Dementia with Lewy Bodies, and multiple system atrophy [ 2 ]. Criteria for the diagnosis of RBD require the lack of atonia, and the evidence of dream-enacting behavior [ 3 ], assessed through video-polysomnography (PSG) [ 4 ]. However, the process of scoring PSG, both for sleep staging and for subsequent diagnostic procedures, is manual and based on visual assessments by experts, resulting in a time-consuming, and labor-intensive procedure. In the literature, various Machine Learning (ML) approaches have been proposed to support the diagnostic process of RBD [ 5 ], offering promising solutions through the automated analysis of PSG-derived signals – such as electroen-cephalography (EEG), electromyography (EMG), and electrooculography (EOG) [ 6 ], [ 7 ], [ 8 ], or inertial data [ 9 ]. Some advancements have been made by tackling Deep Learning (DL) approaches [ 10 ]. Recent studies specifically focused on RBD detection from EEG signals, particularly in the REM and Slow Wave Sleep (SWS) stages [ 11 ], [ 12 ], reporting encouraging results. However, a limitation persists: these approaches heavily rely on sleep staging, as they bank on hand-engineered features tailored to characterize the SWS and REM stages. To overcome this potential limitation, this study proposes a ML-based, stage-agnostic framework for the automatic detection of RBD subjects employing single-channel EEG sleep data. This solution not only promotes lightweight and minimally-invasive sleep studies but also seeks to simplify the diagnostic process by enhancing its robustness and scalability. The rest of the paper is structured as follows: Section II provides a description of the employed datasets, in terms of subjects and data. Section III illustrates the analysis pipeline, including signal pre-processing, feature extraction, and the ML framework. Section IV reports the experimental results and their interpretation; finally, Section V discusses the clinical implications of this work, addresses the limitations and highlights directions for future developments. II. MATERIALS A. Subjects The study employed PSG recordings from 58 subjects – 26 Healthy Controls (HC) and 32 RBD subjects – from two sources: a public and a private datasets. The former is the CAP Sleep Database [ 13 ], publicly available on PhysioNet [ 14 ], consisting of PSG recordings of 16 HC (9 males, aged 32 ± 5 years) and 22 RBD (19 males, aged 70 ± 6 years), widely employed in the literature [ 6 ], [ 7 ]. The private database includes PSG data from 10 RBD subjects (8 males, aged 62 ± 6 years) and 10 HC (6 males, aged 37 ± 16 years). It was recorded at the Regional Center for Sleep Medicine (Molinette University Hospital, Turin) [ 7 ]; data collection was performed in concordance with the Declaration of Helsinki and approved by the local ethical committee (No. 00384/2020). The mean length across all recordings was 6.91 ± 1.19 hours. B. Data All PSG recordings were manually scored by sleep experts according to the American Academy of Sleep Medicine (AASM) scoring criteria [ 15 ], which classify sleep as REM sleep, and Non REM (NREM) sleep (further divided in N1, N2, SWS, from light to deep). As previously introduced, and in the perspective of developing lightweight screening tools for supporting sleep studies, this work relies on a single EEG channel and adopts only essential pre-processing steps. In more detail, only recordings from the central EEG channel (C3-A2, or C4-A1) were considered. Then, EEG recordings were exploited and merged, so as to obtain two different feature sets to be employed in the ML framework. Specifically, for the first feature set, the whole NREM sleep (stages SWS, N2, N1) was considered as a uniform segment. Then, for the second feature set, a comprehensive approach was tackled by taking into account the whole sleep, regardless of the stage (REM, SWS, N2, N1). This procedure is further detailed in Section III-A. III. METHODS This section describes the procedures adopted in the analysis, carried out through custom-written code in MATLAB® vR2023b and Python v3.11. A. Signal Pre-Processing The signals underwent minimal pre-processing, to ensure applicability in natural environments and wearable scenarios. The EEG recordings were band-pass filtered (0.01–40 Hz) through a cascade of two zero-phase, anti-causal Butterworth filters – low-pass and high-pass. Then, the EEG segments were processed so as to follow two different analysis pipelines ( Figure 1 ). Specifically, as described earlier, first only NREM segments were considered (SWS, N2, N1), then, all sleep was taken as a whole, stage-independent segment. Finally, features were extracted separately on the two EEG segments, as described in the following section. Download figure Open in new tab Fig. 1. Pipeline chart. B. Feature Extraction A comprehensive, multi-domain extraction process was carried out, considering key attributes found in topic-related studies [ 16 ], [ 17 ], [ 18 ], [ 11 ], [ 8 ], along with the division of EEG features into Time Domain (TD), Frequency Domain (FD), Time-Frequency Domain (TFD), and Nonlinear Domain (NND). This guaranteed the inclusion of morphological and spectral characteristics of EEG, together with details regarding its non-stationary and non-linear nature, expression of the underlying intricate brain mechanisms from which it derives. Table I summarizes the employed features; they were extracted in 30-seconds epochs, to match the AASM sleep scoring criteria. To ensure EEG signal stationarity, 2-seconds windows were employed for the FD, TFD, and NND features [ 12 ]. Several statistics, including Mean, Standard Deviation, 75 th percentile, and Main Mode (MM) were then employed to summarize the obtained values, as in [ 12 ], and employed as distinct features ( Figure 2 ). This procedure, resulting in a (58, 900) feature matrix, was repeated for all patients, eventually obtaining two feature sets: View this table: View inline View popup Download powerpoint TABLE I. F eatures employed in the study , according to their domain . ⋄ : adapted from cited study . Download figure Open in new tab Fig. 2. (A) Each EEG signal is divided into 30-seconds (TD) or 2-seconds (FD, TFD, NND) epochs, based on the specific domain D. (B) Features F (TD: 32, FD: 165, TFD: 20, NND: 8, total: 225) are extracted for each epoch, and then (C) summarized through four statistics, namely Mean, Standard Deviation, 75 th percentile, and Main Mode, resulting in 900 features. The process is repeated for each Subject S (1,…, 58), (D) , resulting in a feature matrix (58, 900). Set NREM : features extracted from ( SWS + N 2 + N 1) sleep Set ALL : features extracted from ( REM + SWS + N 2 + N 1) sleep C. Statistical Analysis, Feature Normalisation, and Feature Selection Statistical tests were performed on the extracted features to test their relevance for the task, through the open-source software Jamovi [ 21 ]. First, a Shapiro-Wilk normality test was carried out; then, independent samples testing was conducted, through the Student’s t-test and Mann-Whitney U test for normally and non-normally distributed features, respectively. Z-score normalization was performed on the analyzed attributes. Feature selection was conducted in both feature sets through the Multiple Threshold Spatially Uniform ReliefF (MultiSURF) feature selection method [ 22 ]. It is a ReliefF-based algorithm [ 23 ], which proved to be effective in highdimensional biomedical datasets, possibly with non-linear relations among features, as the ones explored in this study. Features were ranked according to their importance score, and the top-10 features were employed for the classification task. For the sake of clarity, feature selection was performed in all training folds to limit the risk of data leakage. D. Classification Framework Supervised ML models were employed to carry out the automatic detection of RBD subjects. Namely, a Support Vector Machine (SVM), a K-Nearest Neighbors (KNN), a Random Forest (RF), an eXtreme Gradient Boosting (XGBoost), and a Logistic Regression (LR) classifier were adopted for both feature sets. Given the relatively small sample size (58 total subjects, 38 RBD), a Leave-One-Subject-Out (LOSO) Cross Validation (CV) approach was adopted to ensure proper generalization of the models. In this procedure, model training is repeated N times ( N : total number of subjects); at each iteration, the classifier is trained on ( N − 1) subjects, and tested on the held-out subject. Hyperparameters were optimized on training data only, through a k -fold ( k =5) combined with a Grid Search approach, to further enhance the robustness of the models. Table II provides a summary of the employed classifiers and the optimized parameters, along with their search range. View this table: View inline View popup Download powerpoint TABLE II. S ummary of the employed classifiers and the searched hyperparameters , ( parameter and range ). IV. RESULTS A. Feature Inspection As mentioned in Section III-B, a total of 225 features were extracted divided according to the domain. Specifically: 32 for TD, 165 for FD, 20 for TFD, and for in NDD ( Figure 2 ); for each feature, a total of four statistics were computed and employed as predictors, resulting in the (58, 900) final feature set (58 subjects, cf. III-B). The independent samples testing revealed a substantially large number of features exhibiting high-to-very-high statistical significance ( p : < 0.001 − 0.05). In an attempt to limit the computational load required in the classification step, a preliminary data reduction step was carried out by keeping only features with p < 0.05 (Mann-Whitney U). This resulted in Set NREM set to have 361 features and 347 for Set ALL . Feature selection was repeated on all training folds; as the MultiSURF approach ranks features according to their importance, the resulting top-10 features in each fold were used for classification. Table III reports the top-10 most recurrent features across all training folds, in terms of Frequency of Occurrence (FO) (%). As appreciable, PKF FREM , representing the peak frequency in the FREM band (2–8 Hz), proved to be a meaningful predictor for detecting RBD, being selected in 98% and 100% of folds in the Set NREM (PKF FREM,Mean ) and Set ALL sets (PKF FREM,Std ), respectively. The most recurrent feature in Set NREM was SEFd(95-50) θ,Mean (100% of folds), which appeared significantly predictive also in Set ALL (95% of folds). Considering only feature importance, in the Set NREM set, the most informative feature was SSp β,Std (STD of the spectral skewness in β -band), ranked first in 79% of folds. This parameter is also featured among the most recurrent features, being selected overall in 98% of all training folds. Regarding Set ALL , the feature PKF FREM,Std , selected in the totality of training folds, emerged also as the most informative for the classification, being ranked top position in 69% of folds. These results suggest that the β EEG band holds great predictive power for detecting RBD through sleep EEG, as also supported by [ 24 ]. Similar considerations can be done for FREM (2–8 Hz), in line with [ 18 ], and the θ band. Figure 3 and Figure 4 depict the distribution (obtained through Kernel Density Estimation) in the two classes of the most recurrent features for Set NREM and Set ALL , respectively. As appreciable, these plots further confirm the discriminative power of the selected parameters. View this table: View inline View popup Download powerpoint TABLE III. T op 10 most recurrent F eatures in the two training sets , with frequency of occurrence (FO) and statistical significance *** : p < 0.001. < 0.001 *** 53% Download figure Open in new tab Fig. 3. Distribution of the top 5 most recurrent features in Set NMREM , ordered following descending FO values. Download figure Open in new tab Fig. 4. Distribution of the top 5 most recurrent features in Set ALL , ordered following descending FO values. B. RBD Detection Performance The explored classifiers exhibited moderate-to-high classification accuracies in both feature sets, achieving Macro Averaged Accuracy (MAA) values of 78.28 ± 0.02 % and 81.03 ± 0.03 % for Set NREM and Set ALL , respectively, across all models. These results not only indicate robust classification performance but also suggest the high predictive power of the selected parameters, as already revealed in the statistical analysis. Notably, these results become significant considering the minimal signal processing applied to the EEG, suggesting the practicability of the proposed approach. The Area Under the Curve (AUC) values, across the models, of 0.89 ± 0.02 for Set ALL and 0.86 ± 0.03 for Set NREM further demonstrate the strong discrimination ability for the two classes (HC and RBD), which remains consistent through the two feature sets. For the sake of brevity, Table IV reports the LOSO-CV classification performance of the best models in the two feature sets, in terms of Accuracy, Sensitivity, Precision, F1-score, and AUC. In Set NREM a LR (C=50, max iterations: 15, solver: Newton-CG) emerged as the best model, with 81% Accuracy, 84.4% Sensitivity, and an AUC of 0.91. As appreciable, higher performance metrics were attained in Set ALL ; the best model was an XGBoost (eta: 0.1, gamma: 0.0001, max depth: 3, number of trees: 25), showing an 86.2% Accuracy, 90.6% Sensitivity, and 0.91 AUC, suggesting high predictive power of the selected features. A moderately high value of F1-score was also obtained (87.9%), indicative of good model robustness. These performances are in line with previous works employing the same set of subjects [ 7 ], [ 12 ], which achieved a MAA of 86.21%, and AUC of 0.92 ( Figure 6 ). Although seemingly lower than those obtained through stage-specific approaches ( Figure 5 ), the performance of this study are encouraging, particularly taken into account that continuous sleep data exhibit very diverse characteristics. As shown in Figure 5 , Set NREM yields lower Accuracy, Sensitivity, and Specificity compared to SWS alone, but exhibits higher Precision and AUC. This suggests that Set NREM is less prone to false positives while maintaining a good detection trade-off across various thresholds. Likely, the presence of lighter and more variable sleep stages (N1, N2) could possibly introduce confounding elements compared to the more stable SWS. Nevertheless, a direct comparison with previous works is challenging, as the analyses in [ 7 ], [ 12 ] relied on hand-engineered features from specific and distinct sleep stages – i.e., SWS and REM. To the best of the Authors’ knowledge, this study is the first to consider the whole EEG sleep segments without stage-dependent feature extraction, making the analysis more generalizable and clinically practical. Additionally, as displayed by the confusion matrix in Figure 7 , the relatively low number of False Negatives (FN) in Set ALL further supports the feasibility of the proposed approach, suggesting its suitability to clinical screening applications, that seek to minimize the number of FN. View this table: View inline View popup Download powerpoint TABLE IV. RBD classification performance (%) of the best algorithms Download figure Open in new tab Fig. 5. Performance comparison with topic-related studies. SWS, REM, and REM+SWS represent the averaged results obtained using various REM and SWS data combinations in [ 11 ], [ 12 ]. NREM and REM+NREM represent the novelty of the current work. Download figure Open in new tab Fig. 6. Performance comparison with best results in previous works. For the sake of clarity: (REM+SWS) represents the averaged results from [ 11 ], [ 12 ], (REM+NREM) represents the novelty of the current work. Download figure Open in new tab Fig. 7. Confusion matrix for the best model in Set ALL (XGBoost). V. CONCLUSION AND FUTURE WORK RBD is largely considered a prodrome to α -synucleinopathies, making it an important actor in neurodegenerative diseases. Currently, its diagnosis requires protracted manual labor, significantly impacting diagnostic times, and possibly leading to interrater variability. Although various ML have been proposed to aid the diagnostic process, they still rely on sleep staging (either manual or automated), inevitably requiring full PSG assessments, with significant patient discomfort. In the perspective of developing minimally-invasive, scalable sleep studies, this paper proposes a ML-based framework for the automatic detection of RBD subjects, through the analysis of single-channel EEG and stage-agnostic features. The preliminary investigation results obtained in Set NREM , when compared with the previous works from the same research group [ 11 ], [ 12 ], suggest that sleep stages N1 and N2 have a minimal contribution in the binary classification task, fostering the feasibility and practicability of a stage-agnostic approach. Indeed, the performance metrics in Set ALL (86.21% Accuracy, and Sensitivity above 90%), although slightly lower compared to stage-specific approaches on the same datasets [ 11 ], [ 12 ], suggest the feasibility of efficiently detecting RBD through unstaged EEG. Indeed, a Cohen’s κ of 72% was obtained with the best classification algorithm (XGBoost) on Set ALL , indicative of substantial agreement between the developed framework and manual assessment. Hence, these findings support the development of clinical decision support systems for RBD without requiring a full-PSG evaluation and conventional sleep scoring, to be used in large-scale screening protocols for neurodegenerative risk assessment. Although with promising results, this study is not without limitations. First, this framework should be tested on a larger number of subjects, possibly well-stratified for sex and age, in order to properly assess its scalability and applicability. Second, due to computational load restrictions, the Grid Search for hyperparameters optimization was conducted on a limited combination of parameters ranges. Future implementations should test more extensive parameters ranges with possibly beneficial effects on the classification performance. Moving forward, the best-performing model derived from Set ALL (cf. Section IV-B) and the top-10 most recurrent features (illustrated in Table III ), will be employed on unseen data to assess the proposed framework’s robustness and to externally validate its predictive capabilities. Finally, further tests with EEG recorded through wearable technology should be tackled, to assess the robustness and reliability of the proposed systems in natural environments. Overall, this study demonstrates the feasibility of employing unstaged EEG for the automatic detection of RBD, supporting advancements in lightweight, scalable clinical decision support systems for early identification of neurodegenerative states. Data Availability All data produced in the present study are available upon reasonable request to the authors ACKNOWLEDGMENT The drafting and publication of this work took place for G.G. in the context of the project PNRR-NGEU which has received funding from the MUR-DM 117/2023 (Italy), and for I.R. of the project PNRR-NODES which has received funding from the MUR-M4C2 1.5 funded by the European Union NextGenerationEU (Grant agreement no. ECS00000036). Footnotes firstname.lastname{at}polito.it alessandro.cicolin{at}unito.it R eferences [1]. ↵ A. Galbiati , L. Verga , E. Giora , M. Zucconi , and L. Ferini-Strambi , “ The risk of neurodegeneration in rem sleep behavior disorder: a systematic review and meta-analysis of longitudinal studies ,” Sleep medicine reviews , vol. 43 , pp. 37 – 46 , 2019 . OpenUrl CrossRef PubMed [2]. ↵ R. B. Postuma , A. Iranzo , M. Hu , B. Högl , B. F. Boeve , R. Manni , W. H. Oertel , I. Arnulf , L. Ferini-Strambi , M. Puligheddu , et al. , “ Risk and predictors of dementia and parkinsonism in idiopathic rem sleep behaviour disorder: a multicentre study ,” Brain , vol. 142 , no. 3 , pp. 744 – 759 , 2019 . OpenUrl CrossRef PubMed [3]. ↵ M. J. Sateia , “ International classification of sleep disorders ,” Chest , vol. 146 , no. 5 , pp. 1387 – 1394 , 2014 . OpenUrl CrossRef PubMed [4]. ↵ M. Cesari , A. Heidbreder , E. K. St. Louis , et al. , “ Video-polysomnography procedures for diagnosis of rapid eye movement sleep behavior disorder (rbd) and the identification of its prodromal stages: guidelines from the international rbd study group ,” Sleep , vol. 45 , no. 3 , p. zsab257 , 2022 . OpenUrl CrossRef PubMed [5]. ↵ M. Cesari and I. Rechichi , “ Automatic and machine learning methods for detection and characterization of rem sleep behavior disorder ,” in Handbook of AI and Data Sciences for Sleep Disorders , pp. 197 – 217 , Springer , 2024 . [6]. ↵ N. Cooray , F. Andreotti , et al. , “ Detection of rem sleep behaviour disor-der by automated polysomnography analysis ,” Clinical Neurophysiology , vol. 130 , no. 4 , pp. 505 – 514 , 2019 . OpenUrl CrossRef PubMed [7]. ↵ I. Rechichi , A. Iadarola , M. Zibetti , A. Cicolin , and G. Olmo , “ Assessing rem sleep behaviour disorder: From machine learning classification to the definition of a continuous dissociation index ,” International Journal of Environmental Research and Public Health , vol. 19 , no. 1 , p. 248 , 2021 . OpenUrl [8]. ↵ M. Cesari , A. Portscher , A. Stefani , R. Angerbauer , A. Ibrahim , E. Brandauer , S. Feuerstein , K. Egger , B. Högl , and A. Rodriguez-Sanchez , “ Machine learning predicts phenoconversion from polysomnography in isolated rem sleep behavior disorder ,” Brain Sciences , vol. 14 , no. 9 , p. 871 , 2024 . OpenUrl PubMed [9]. ↵ A. Brink-Kjaer , J. Winer , J. M. Zeitzer , H. B. Sorensen , P. Jennum , E. Mignot , and E. During , “ Fully automated detection of isolated rapid-eye-movement sleep behavior disorder using actigraphy ,” in 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC) , pp. 1 – 5 , IEEE , 2023 . [10]. ↵ K. M. Gunter , A. Brink-Kjær , E. Mignot , H. B. Sørensen , E. During , and P. Jennum , “ Svit: A spectral vision transformer for the detection of rem sleep behavior disorder ,” IEEE Journal of Biomedical and Health Informatics , 2023 . [11]. ↵ I. Rechichi , F. Amato , A. Cicolin , and G. Olmo , “ Single-channel eeg detection of rem sleep behaviour disorder: The influence of rem and slow wave sleep ,” Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) , vol. 13346 LNBI , pp. 381 – 394 , 2022 . OpenUrl [12]. ↵ G. S. Giarrusso , I. Rechichi , and G. Olmo , “ Eeg-based detection of rem sleep behaviour disorder: Towards a stage-agnostic approach ,” in International Work-Conference on Bioinformatics and Biomedical Engineering , pp. 263 – 276 , Springer , 2024 . [13]. ↵ M. G. Terzano , L. Parrino , A. Sherieri , R. Chervin , S. Chokroverty , C. Guilleminault , M. Hirshkowitz , M. Mahowald , H. Moldofsky , A. Rosa , et al. , “ Atlas, rules, and recording techniques for the scoring of cyclic alternating pattern (cap) in human sleep .,” Sleep medicine , vol. 2 , no. 6 , pp. 537 – 553 , 2001 . OpenUrl CrossRef PubMed Web of Science [14]. ↵ A. L. Goldberger , L. A. Amaral , L. Glass , J. M. Hausdorff , P. C. Ivanov , R. G. Mark , J. E. Mietus , G. B. Moody , C.-K. Peng , and H. E. Stanley , “ Physiobank, physiotoolkit, and physionet: components of a new research resource for complex physiologic signals ,” circulation , vol. 101 , no. 23 , pp. e215 – e220 , 2000 . OpenUrl Abstract / FREE Full Text [15]. ↵ R. B. Berry , R. Brooks , C. E. Gamaldo , S. M. Harding , C. Marcus , B. V. Vaughn , et al. , “ The aasm manual for the scoring of sleep and associated events ,” Rules, Terminology and Technical Specifications, Darien, Illinois, American Academy of Sleep Medicine , vol. 176 , no. 2012 , p. 7 , 2012 . OpenUrl [16]. ↵ K. Aboalayon , M. Faezipour , W. Almuhammadi , and S. Moslehpour , “ Sleep stage classification using EEG signal analysis: A comprehensive survey and new investigation ,” Entropy (Basel) , vol. 18 , p. 272 , Aug . 2016 . OpenUrl [17]. ↵ A. Khosla , P. Khandnor , and T. Chand , “ A comparative analysis of signal processing and classification methods for different applications based on eeg signals ,” Biocybernetics and Biomedical Engineering , vol. 40 , no. 2 , pp. 649 – 690 , 2020 . OpenUrl CrossRef [18]. ↵ I. Rechichi , M. Zibetti , L. Borzì , G. Olmo , and L. Lopiano , “ Single-channel eeg classification of sleep stages based on rem microstructure ,” Healthcare Technology Letters , vol. 8 , pp. 58 – 65 , 6 2021 . OpenUrl PubMed [19]. A. Khosla , P. Khandnor , and T. Chand , “ A comparative analysis of signal processing and classification methods for different applications based on eeg signals ,” Biocybernetics and Biomedical Engineering , vol. 40 , no. 2 , pp. 649 – 690 , 2020 . OpenUrl CrossRef [20]. K. A. I. Aboalayon et al. , “ Sleep stage classification using eeg signal analysis: a comprehensive survey and new investigation ,” Entropy , vol. 18 , no. 9 , p. 272 , 2016 . OpenUrl [21]. ↵ “Jamovi - open statistical software for the desktop and cloud.” https://www.jamovi.org . Accessed: 2022-12-19 . [22]. ↵ D. Granizo-Mackenzie and J. H. Moore , “ Multiple threshold spatially uniform relieff for the genetic analysis of complex human diseases ,” in Evolutionary Computation, Machine Learning and Data Mining in Bioinformatics: 11th European Conference, EvoBIO 2013, Vienna, Austria, April 3-5, 2013. Proceedings 11 , pp. 1 – 10 , Springer , 2013 . [23]. ↵ R. J. Urbanowicz , M. Meeker , W. La Cava , R. S. Olson , and J. H. Moore , “ Relief-based feature selection: Introduction and review ,” Journal of biomedical informatics , vol. 85 , pp. 189 – 203 , 2018 . OpenUrl CrossRef PubMed [24]. ↵ C. O’Reilly , I. Godin , J. Montplaisir , and T. Nielsen , “ REM sleep behaviour disorder is associated with lower fast and higher slow sleep spindle densities ,” Journal of Sleep Research , vol. 24 , pp. 593 – 601 , June 2015 . OpenUrl CrossRef PubMed View the discussion thread. Back to top Previous Next Posted May 28, 2025. Download PDF Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following A Single-Channel EEG Approach for Sleep Stage-Independent Automatic Detection of REM Sleep Behavior Disorder Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share A Single-Channel EEG Approach for Sleep Stage-Independent Automatic Detection of REM Sleep Behavior Disorder Gabriele Salvatore Giarrusso , Irene Rechichi , Alessandro Cicolin , Gabriella Olmo medRxiv 2025.05.28.25328491; doi: https://doi.org/10.1101/2025.05.28.25328491 Share This Article: Copy Citation Tools A Single-Channel EEG Approach for Sleep Stage-Independent Automatic Detection of REM Sleep Behavior Disorder Gabriele Salvatore Giarrusso , Irene Rechichi , Alessandro Cicolin , Gabriella Olmo medRxiv 2025.05.28.25328491; doi: https://doi.org/10.1101/2025.05.28.25328491 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Health Informatics Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (299) Cardiovascular Medicine (4425) Dentistry and Oral Medicine (443) Dermatology (382) Emergency Medicine (607) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1507) Epidemiology (15221) Forensic Medicine (30) Gastroenterology (1123) Genetic and Genomic Medicine (6588) Geriatric Medicine (667) Health Economics (997) Health Informatics (4524) Health Policy (1368) Health Systems and Quality Improvement (1612) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15910) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (145) Nephrology (667) Neurology (6588) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1143) Occupational and Environmental Health (956) Oncology (3331) Ophthalmology (970) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (663) Pediatrics (1690) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5440) Public and Global Health (9219) Radiology and Imaging (2195) Rehabilitation Medicine and Physical Therapy (1369) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (710) Sports Medicine (529) Surgery (710) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ffb2d46a88041e2',t:'MTc3OTQ0NjE1NQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();