Warby (2014): Sleep-spindle detection: crowdsourcing and evaluating performance of experts

Sleep-spindle detection: crowdsourcing and evaluating performance of experts, non-experts and automated methods

PDF: Warby_NM_xx_1-8_suppl-info_2014_spindles-crowdsourcing

Warby 2014

Simon C Warby1, Sabrina L Wendt1,2, Peter Welinder3, Emil G S Munk1,2, Oscar Carrillo1, Helge B D Sorensen4, Poul Jennum2, Paul E Peppard5, Pietro Perona3 & Emmanuel Mignot1

Sleep spindles are discrete, intermittent patterns of brain activity observed in human electroencephalographic data. Increasingly, these oscillations are of biological and clinical interest because of their role in development, learning and neurological disorders. We used an Internet interface to crowdsource spindle identification by human experts and non-experts, and we compared their performance with that of automated detection algorithms in data from middle- to older-aged subjects from the general population. We also refined methods for forming group consensus and evaluating the performance of event detectors in physiological data such as electroencephalographic recordings from polysomnography. Compared to the expert group consensus gold standard, the highest performance was by individual experts and the non-expert group consensus, followed by automated spindle detectors. This analysis showed that crowdsourcing the scoring of sleep data is an efficient method to collect large data sets, even for difficult tasks such as spindle identification. Further refinements to spindle detection algorithms are needed for middle- to older-aged subjects.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s