A comparative study of methods for automatic detection of rapid eye movement abnormal muscular activity in narcolepsy

Alexander Neergaard Olesen; Matteo Cesari; Julie Anja Engelhard Christensen; Helge Bjarup Dissing Sorensen; Emmanuel Mignot; Poul Jennum

doi:10.1016/j.sleep.2017.11.1141

A comparative study of methods for automatic detection of rapid eye movement abnormal muscular activity in narcolepsy

Alexander Neergaard Olesen, Matteo Cesari, Julie Anja Engelhard Christensen, Helge Bjarup Dissing Sorensen, Emmanuel Mignot, Poul Jennum

Department of Clinical Medicine

4 Citations (Scopus)

Abstract

Objective: To evaluate rapid eye movement (REM) muscular activity in narcolepsy by applying five algorithms to electromyogram (EMG) recordings, and to investigate its value for narcolepsy diagnosis. Patients/methods: A modified version of phasic EMG metric (mPEM), muscle activity index (MAI), REM atonia index (RAI), supra-threshold REM EMG activity metric (STREAM), and Frandsen method (FR) were calculated from polysomnography recordings of 20 healthy controls, 18 clinic controls (subjects suspected with narcolepsy but finally diagnosed without any sleep abnormality), 16 narcolepsy type one without REM sleep behavior disorder (RBD), nine narcolepsy type one with RBD, and 18 narcolepsy type two. Diagnostic value of metrics in differentiating between groups was quantified by area under the receiver operating characteristic curve (AUC). Correlations among the metrics and cerebrospinal fluid hypocretin-1 (CSF-hcrt-1) values were calculated using linear models. Results: All metrics excluding STREAM found significantly higher muscular activity in narcolepsy one cases versus controls (p < 0.05). Moreover, RAI showed high sensitivity in the detection of RBD. The mPEM achieved the highest AUC in differentiating healthy controls from narcoleptic subjects. The RAI best differentiated between narcolepsy 1 and 2. Lower CSF-hcrt-1 values correlated with high muscular activity quantified by mPEM, sMAI, lMAI, PEM and FR (p < 0.05). Conclusions: This automatic analysis showed higher number of muscle activations in narcolepsy 1 compared to controls. This finding might play a supportive role in diagnosing narcolepsy and in discriminating narcolepsy subtypes. Moreover, the negative correlation between CSF-hcrt-1 level and REM muscular activity supported a role for hypocretin in the control of motor tone during REM sleep.

Original language	English
Journal	Sleep Medicine
Volume	44
Pages (from-to)	97-105
ISSN	1389-9457
DOIs	https://doi.org/10.1016/j.sleep.2017.11.1141
Publication status	Published - Apr 2018

Access to Document

10.1016/j.sleep.2017.11.1141

Cite this

@article{0ad13b0f2a684794a928fcc28190178b,

title = "A comparative study of methods for automatic detection of rapid eye movement abnormal muscular activity in narcolepsy",

abstract = "Objective: To evaluate rapid eye movement (REM) muscular activity in narcolepsy by applying five algorithms to electromyogram (EMG) recordings, and to investigate its value for narcolepsy diagnosis. Patients/methods: A modified version of phasic EMG metric (mPEM), muscle activity index (MAI), REM atonia index (RAI), supra-threshold REM EMG activity metric (STREAM), and Frandsen method (FR) were calculated from polysomnography recordings of 20 healthy controls, 18 clinic controls (subjects suspected with narcolepsy but finally diagnosed without any sleep abnormality), 16 narcolepsy type one without REM sleep behavior disorder (RBD), nine narcolepsy type one with RBD, and 18 narcolepsy type two. Diagnostic value of metrics in differentiating between groups was quantified by area under the receiver operating characteristic curve (AUC). Correlations among the metrics and cerebrospinal fluid hypocretin-1 (CSF-hcrt-1) values were calculated using linear models. Results: All metrics excluding STREAM found significantly higher muscular activity in narcolepsy one cases versus controls (p < 0.05). Moreover, RAI showed high sensitivity in the detection of RBD. The mPEM achieved the highest AUC in differentiating healthy controls from narcoleptic subjects. The RAI best differentiated between narcolepsy 1 and 2. Lower CSF-hcrt-1 values correlated with high muscular activity quantified by mPEM, sMAI, lMAI, PEM and FR (p < 0.05). Conclusions: This automatic analysis showed higher number of muscle activations in narcolepsy 1 compared to controls. This finding might play a supportive role in diagnosing narcolepsy and in discriminating narcolepsy subtypes. Moreover, the negative correlation between CSF-hcrt-1 level and REM muscular activity supported a role for hypocretin in the control of motor tone during REM sleep.",

author = "Olesen, {Alexander Neergaard} and Matteo Cesari and Christensen, {Julie Anja Engelhard} and Sorensen, {Helge Bjarup Dissing} and Emmanuel Mignot and Poul Jennum",

year = "2018",

month = apr,

doi = "10.1016/j.sleep.2017.11.1141",

language = "English",

volume = "44",

pages = "97--105",

journal = "Sleep Medicine",

issn = "1389-9457",

publisher = "Elsevier",

}

TY - JOUR

T1 - A comparative study of methods for automatic detection of rapid eye movement abnormal muscular activity in narcolepsy

AU - Olesen, Alexander Neergaard

AU - Cesari, Matteo

AU - Christensen, Julie Anja Engelhard

AU - Sorensen, Helge Bjarup Dissing

AU - Mignot, Emmanuel

AU - Jennum, Poul

PY - 2018/4

Y1 - 2018/4

N2 - Objective: To evaluate rapid eye movement (REM) muscular activity in narcolepsy by applying five algorithms to electromyogram (EMG) recordings, and to investigate its value for narcolepsy diagnosis. Patients/methods: A modified version of phasic EMG metric (mPEM), muscle activity index (MAI), REM atonia index (RAI), supra-threshold REM EMG activity metric (STREAM), and Frandsen method (FR) were calculated from polysomnography recordings of 20 healthy controls, 18 clinic controls (subjects suspected with narcolepsy but finally diagnosed without any sleep abnormality), 16 narcolepsy type one without REM sleep behavior disorder (RBD), nine narcolepsy type one with RBD, and 18 narcolepsy type two. Diagnostic value of metrics in differentiating between groups was quantified by area under the receiver operating characteristic curve (AUC). Correlations among the metrics and cerebrospinal fluid hypocretin-1 (CSF-hcrt-1) values were calculated using linear models. Results: All metrics excluding STREAM found significantly higher muscular activity in narcolepsy one cases versus controls (p < 0.05). Moreover, RAI showed high sensitivity in the detection of RBD. The mPEM achieved the highest AUC in differentiating healthy controls from narcoleptic subjects. The RAI best differentiated between narcolepsy 1 and 2. Lower CSF-hcrt-1 values correlated with high muscular activity quantified by mPEM, sMAI, lMAI, PEM and FR (p < 0.05). Conclusions: This automatic analysis showed higher number of muscle activations in narcolepsy 1 compared to controls. This finding might play a supportive role in diagnosing narcolepsy and in discriminating narcolepsy subtypes. Moreover, the negative correlation between CSF-hcrt-1 level and REM muscular activity supported a role for hypocretin in the control of motor tone during REM sleep.

AB - Objective: To evaluate rapid eye movement (REM) muscular activity in narcolepsy by applying five algorithms to electromyogram (EMG) recordings, and to investigate its value for narcolepsy diagnosis. Patients/methods: A modified version of phasic EMG metric (mPEM), muscle activity index (MAI), REM atonia index (RAI), supra-threshold REM EMG activity metric (STREAM), and Frandsen method (FR) were calculated from polysomnography recordings of 20 healthy controls, 18 clinic controls (subjects suspected with narcolepsy but finally diagnosed without any sleep abnormality), 16 narcolepsy type one without REM sleep behavior disorder (RBD), nine narcolepsy type one with RBD, and 18 narcolepsy type two. Diagnostic value of metrics in differentiating between groups was quantified by area under the receiver operating characteristic curve (AUC). Correlations among the metrics and cerebrospinal fluid hypocretin-1 (CSF-hcrt-1) values were calculated using linear models. Results: All metrics excluding STREAM found significantly higher muscular activity in narcolepsy one cases versus controls (p < 0.05). Moreover, RAI showed high sensitivity in the detection of RBD. The mPEM achieved the highest AUC in differentiating healthy controls from narcoleptic subjects. The RAI best differentiated between narcolepsy 1 and 2. Lower CSF-hcrt-1 values correlated with high muscular activity quantified by mPEM, sMAI, lMAI, PEM and FR (p < 0.05). Conclusions: This automatic analysis showed higher number of muscle activations in narcolepsy 1 compared to controls. This finding might play a supportive role in diagnosing narcolepsy and in discriminating narcolepsy subtypes. Moreover, the negative correlation between CSF-hcrt-1 level and REM muscular activity supported a role for hypocretin in the control of motor tone during REM sleep.

U2 - 10.1016/j.sleep.2017.11.1141

DO - 10.1016/j.sleep.2017.11.1141

M3 - Journal article

C2 - 29530376

SN - 1389-9457

VL - 44

SP - 97

EP - 105

JO - Sleep Medicine

JF - Sleep Medicine

ER -

A comparative study of methods for automatic detection of rapid eye movement abnormal muscular activity in narcolepsy

Abstract

Access to Document

Fingerprint

Cite this