F26. Automated chin EMG analysis for quantification of REM sleep without atonia. (May 2018)

• Record Type: Journal Article
• Authors: Jeppesen, Jesper; Otto, Marit; Frederiksen, Yoon; Hansen, Allan; Fedorova, Tatyana; Knudsen, Karoline; Nahimi, Adjmal; Brooks, David; Borghammer, Per; Sommerauer, Michael
• Is Part Of: Clinical neurophysiology. Volume 129(2018)Supplement 1
• Page Start: e76
• Publication Date: 2018-05
• Subjects: Neurophysiology -- Periodicals; Electroencephalography -- Periodicals; Electromyography -- Periodicals; Neurology -- Periodicals; 612.8
• DOI: 10.1016/j.clinph.2018.04.189
• Languages: English
• ISSNs: 1388-2457

Abstract

Abstract : Introduction: Rapid eye movement (REM) sleep behavior disorder (RBD) is defined by dream enactment due to a failure of normal muscle atonia. Visual assessment of this muscle activity is time consuming and rater-dependent. A computer program with an automated scoring algorithm would thus be a valuable asset for sleep professionals scoring RBD. Methods: An electromyogram (EMG) computer algorithm for scoring 'tonic' (2× root mean square (RMS) of baseline EMG activity for > 15 s per 30 s epoch), 'phasic' (max amplitude > 4×RMS baseline (or 2× during concomitant 'tonic' activity), 3 s epochs) and 'any' ('tonic' and/or 'phasic') submental muscle activity during REM sleep was created using the criteria's of McCarter et al. [1]. For validation of the computer algorithm, visual scoring of EMG activity done by two blinded board-certified sleep professionals was compared to scoring of the computer algorithm in the first consecutive 10 Parkinson's disease (PD) patients with RBD (PD RBD+ ) and 10 PD patients without RBD (PD RBD− ) using regression analysis and Kappa statistics. Subsequently, 29 non-RBD subjects (14 PD RBD− and 15 healthy controls) and 23 RBD subjects (16 PD RBD+ and 7 idiopathic RBD subjects) were analyzed with the computer algorithm to test for discrimination power between RBD and non-RBD subjects. Results: The computer algorithm showed high congruency with human ratings on single epoch level (using Kappa statistics: 'tonic' activity κ = 0.74 & κ = 0.78, 'phasic' activity κ = 0.80 & κ = 0.76, 'any' activity κ = 0.83 & κ = 0.82), which was similar to the κ values between the two human raters ('tonic' activity κ = 0.64, 'phasic' activity κ = 0.79, 'any' activity κ = 0.82). In the discrimination analysis, all RBD subjects were correctly identified by excess of submental muscle activity, when artifacts were removed before analysis. At an individual patient level, cut-off values with full differentiation between non-RBD and RBD subjects were obtained for 'phasic' and 'any' activity (cut off values at > 10% of REM sleep; range of non-RBD subjects, 1.3–8.2% for 'phasic' and 'any'; range of RBD subjects, 13.2–64.5% for 'phasic' and 13.3–77.4% for 'any', respectively). Conclusion: Our proposed automated computer algorithm allows a fast quantitation of 'tonic', 'phasic', and 'any' EMG activity with good agreement to visual ratings and was able to detect and rate REM sleep without atonia. The computer algorithm seems a promising tool to be used in the clinical setting for supporting a diagnosis of RBD.