Semi-automated detection of polysomnographic REM sleep without atonia (RSWA) in REM sleep behavioral disorder

Iva Milerska, Vaclav Kremen, Vaclav Gerla, Erik K. St Louis, Lenka Lhotska

Research output: Contribution to journalArticlepeer-review


We aimed at evaluating semi-automatic detection and quantification of polysomnographic REM sleep without atonia (RSWA). As basic requirements, we defined lower time demand, the possibility of comparison of several evaluations and ease of examination for neurologists. We focused on well-known primary processing of surface electromyographic signals and selected recordings that were free of technical artifacts that could compromise automated signal detection. Thus we created a comprehensive method consisting of several modules (data preprocessing, signal filtration, envelopes creation, detection of ECG QRS complexes, iterative RSWA detection, detection evaluation and interactive visualization). The original dataset consisted of 7 individual polysomnography (PSG) recordings of individual human adult subjects with REM sleep behavior disorder (RBD). RSWA detection was performed with three different methods for envelope creation (envelope by moving average filter, envelope by Savitzky–Golay filtration and peaks interpolation). Best RSWA detection was achieved using the envelope by moving average filter (average precision 64.24 ± 12.34% and recall 91.63 ± 10.07%). The lowest precision was 42.86% with 100% recall.

Original languageEnglish (US)
Pages (from-to)243-252
Number of pages10
JournalBiomedical Signal Processing and Control
StatePublished - May 2019


  • ECG QRS detection
  • Iterative method
  • Linear envelope
  • Polysomnography
  • REM
  • REM sleep behavior disorder
  • REM sleep without atonia
  • Savitzky–Golay filter
  • Semi-automatic detector

ASJC Scopus subject areas

  • Signal Processing
  • Health Informatics


Dive into the research topics of 'Semi-automated detection of polysomnographic REM sleep without atonia (RSWA) in REM sleep behavioral disorder'. Together they form a unique fingerprint.

Cite this