Automatic REM sleep detection associated with idiopathic rem sleep Behavior Disorder

TY - GEN

T1 - Automatic REM sleep detection associated with idiopathic rem sleep Behavior Disorder

AU - Kempfner, J

AU - Sørensen, Gertrud Laura

AU - Sorensen, H B D

AU - Jennum, P

PY - 2011

Y1 - 2011

N2 - Rapid eye movement sleep Behavior Disorder (RBD) is a strong early marker of later development of Parkinsonism. Currently there are no objective methods to identify and discriminate abnormal from normal motor activity during REM sleep. Therefore, a REM sleep detection without the use of chin electromyography (EMG) is useful. This is addressed by analyzing the classification performance when implementing two automatic REM sleep detectors. The first detector uses the electroencephalography (EEG), electrooculography (EOG) and EMG to detect REM sleep, while the second detector only uses the EEG and EOG. Method: Ten normal controls and ten age matched patients diagnosed with RBD were enrolled. All subjects underwent one polysomnographic (PSG) recording, which was manual scored according to the new sleep-scoring standard from the American Academy of Sleep Medicine. Based on the manual scoring, an automatic computerized REM detection algorithm has been implemented, using wavelet packet combined with artificial neural network. Results: When using the EEG, EOG and EMG modalities, it was possible to correctly classify REM sleep with an average Area Under Curve (AUC) equal to 0.900.03 for normal subjects and AUC 0.810.05 for RBD subjects. The performance difference between the two groups was significant (p 0.01). No significant drop (p 0.05) in performance was observed when only using the EEG and EOG in neither of the groups. Conclusion: The overall result indicates that the EMG does not play an important role when classifying REM sleep.

AB - Rapid eye movement sleep Behavior Disorder (RBD) is a strong early marker of later development of Parkinsonism. Currently there are no objective methods to identify and discriminate abnormal from normal motor activity during REM sleep. Therefore, a REM sleep detection without the use of chin electromyography (EMG) is useful. This is addressed by analyzing the classification performance when implementing two automatic REM sleep detectors. The first detector uses the electroencephalography (EEG), electrooculography (EOG) and EMG to detect REM sleep, while the second detector only uses the EEG and EOG. Method: Ten normal controls and ten age matched patients diagnosed with RBD were enrolled. All subjects underwent one polysomnographic (PSG) recording, which was manual scored according to the new sleep-scoring standard from the American Academy of Sleep Medicine. Based on the manual scoring, an automatic computerized REM detection algorithm has been implemented, using wavelet packet combined with artificial neural network. Results: When using the EEG, EOG and EMG modalities, it was possible to correctly classify REM sleep with an average Area Under Curve (AUC) equal to 0.900.03 for normal subjects and AUC 0.810.05 for RBD subjects. The performance difference between the two groups was significant (p 0.01). No significant drop (p 0.05) in performance was observed when only using the EEG and EOG in neither of the groups. Conclusion: The overall result indicates that the EMG does not play an important role when classifying REM sleep.

U2 - 10.1109/iembs.2011.6091498

DO - 10.1109/iembs.2011.6091498

M3 - Conference article

SN - 0589-1019

VL - 2011

SP - 6063

EP - 6066

JO - Annual International Conference of the IEEE Engineering in Medicine and Biology - Proceedings

JF - Annual International Conference of the IEEE Engineering in Medicine and Biology - Proceedings

ER -