Automatic artifact detection for whole-night polysomnographic sleep recordings.
Dorothée
Coppieters ‘t Wallant1*,
Sarah
L.
Chellappa1,
Giulia
Gaggioni1,
Mathieu
Jaspar1,
Christelle
Meyer1,
Vincenzo
Muto1,
Gilles
Vandewalle1,
Pierre
Maquet1, 2, 3 and
Christophe
Phillips1, 4
-
1
University of Liège, Cyclotron Research Centre, Belgium
-
2
Centre Hospitalier Universitaire de Liège, Neurology, Belgium
-
3
Walloon Excellence in Lifesciences and Biotechnology (WELBIO), Belgium
-
4
University of Liège, Department of Electrical Engineering and Computer Science, Belgium
Objective: Detecting of bad channels and artifacts for whole-night polysomnographic recordings is very time consuming and tedious. We therefore developed an automatic procedure to automatize this job.
Method: The method uses temporal and spectral parameters estimated from the data, and thresholds that are systematically adapted to the recording analyzed. First, bad channel detection proceeds per individual channel per scoring window (of 20 or 30 seconds), leaving usable episodes available for further processing. Afterwards artifacts are searched for over one-second epochs and across all channels: ‘slow undulation’ linked to breathing and sweating, ‘rapid transient activity’, and ‘movements and arousals’. Automatic artifact detection is compared to manual detection, using four criteria: inter-rater reliability (S), sensitivity (Se), overlap of detected artifacts (C) and False Discovery Rate (FDR).
Result: 43 recordings (from 23 different subjects) were rated by one expert and our automatic method. By considering the manual detection as the gold standard, we obtained: S=93.8%, Se=78.2%, C=71.9% and FDR=51.7%. Inter-expert variability was tested by comparing the automatic detection with that of six different experts on four sleep recordings (7 scoring for each recording) from a single subject: Sauto=97.7% vs Sexperts=98.4%, Seauto=86.2% vs Seexperts=72.7%, Cauto=73.2% vs Cexperts=90.3% and FDRauto=54.7% vs FDRexperts=27.5%. Post hoc checks showed that automatic detection usually found artifacts that had been overlooked by the experts.
Conclusion: Tests performed over different datasets show that our automatic method is robust and reproducible, as well as more reliable than different experts between them. Moreover it works (on a standard PC) much faster than manual detection
Acknowledgements
FRIA-FNRS-WBI-FMRE-BBSRC-FEDER-RADIOMED-ARC-ULg.
Keywords:
Polysomnography,
Electroencephalography,
artifacts,
bad channel,
Sleep,
automatic detection
Conference:
Belgian Brain Council 2014
MODULATING THE BRAIN: FACTS, FICTION, FUTURE, Ghent, Belgium, 4 Oct - 4 Oct, 2014.
Presentation Type:
Poster Presentation
Topic:
Clinical Neuroscience
Citation:
Coppieters ‘t Wallant
D,
Chellappa
SL,
Gaggioni
G,
Jaspar
M,
Meyer
C,
Muto
V,
Vandewalle
G,
Maquet
P and
Phillips
C
(2014). Automatic artifact detection for whole-night polysomnographic sleep recordings..
Conference Abstract:
Belgian Brain Council 2014
MODULATING THE BRAIN: FACTS, FICTION, FUTURE.
doi: 10.3389/conf.fnhum.2014.214.00076
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
08 Jul 2014;
Published Online:
02 Sep 2014.
*
Correspondence:
Ms. Dorothée Coppieters ‘t Wallant, University of Liège, Cyclotron Research Centre, Liège, Belgium, d.coppieters@ulg.ac.be