Article Text

Download PDFPDF

P034 A novel approach to quantify sleep-related rhythmic movement disorder using automatic 3D analysis
  1. Marcus Gall1,
  2. Rachel van Sluijs2,
  3. Elizabeth Wilhelm2,
  4. Heinrich Garn1,
  5. Peter Achermann3 and
  6. Cathy Hill4
  1. 1Sensing and Vision Solutions, Austrian Institute of Technology GmbH, Austria
  2. 2ETH, Switzerland
  3. 3Zurich Center for inter-disciplinary Sleep Research, University of Zurich, Switzerland
  4. 4School of Clinical and Experimental Sciences, Faculty of Medicine, University of Southampton, UK


Background Unlike other episodic sleep disorders, there are no agreed severity indices for sleep-related rhythmic movement disorder (RMD). Rhythmic movements (RMs) can be characterized by polysomnography, but sensor placement may inhibit movements. Actigraphy and 2D video can be used in the home, but have limitations. Actigraphy does not differentiate RMs from other movements, while manual scoring of 2D video is laborious.

We developed a sensitive method to detect and quantify RMs using automatic 3D video analysis.

Method Children with RMD (n=6, 4 male) aged 5–14 years were studied for two nights in a sleep laboratory. A ceiling-mounted camera captured 3D depth images, while another recorded 2D video, from lights off until lights on. We developed algorithms to analyze the characteristics of RMs and built a classifier to distinguish rhythmic from non-rhythmic movements based on 3D video data alone. Data from 3D automated analysis were compared to manual 2D video annotations in 1.5s segments to assess algorithm performance [figure 1]. Novel indices were developed: the RM index, frequency index and duration index to better characterize RMD severity.

Abstract P034 Figure 1

RM time of night distribution plot. Each data point shows how many 1.5 s segements are classified as RMs per 30 minutes, combining data across subjects. Automatic 3D analysis and manual 2D annotations show high agreement

Result Automatic 3D analysis demonstrated high levels of agreement with the manual approach (Cohen’s Kappa >0.9; F1-score >0.9). We also demonstrated how RM assessment can be improved using plots of our novel indices for ease of visualization.

Conclusion 3D video technology is widely available and can be integrated into sleep laboratories. Our automatic 3D video analysis algorithm yields reliable quantitative measurement of RMs, reducing the burden of manual scoring. Furthermore, our novel RMD severity indices offer standardized measures of utility to clinical and research practice.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.