Improved spindle detection through intuitive pre-processing of electroencephalogram

Detection of K-complexes and sleep spindles (DETOKS) using sparse optimization

“…The bandpass filter used is also excited by transients present in the input EEG. Much of the effort of these algorithms is to either pre-process (Jaleel et al, 2014), or post-process (Motamedi-Fakhr et al, 2014 the bandpass filtered data, so as to distinguish spindles from transients.…”

Learning machines and sleeping brains: Automatic sleep stage classification using decision-tree multi-class support vector machines

“…It is possible that higher performances could be achieved by exploring the discriminative power of further sleep specific neuronal phenomena: Quantifying the presence of K-complex waves (Colrain, 2005;Loomis et al, 1938), sleep spindles (Andrillon et al, 2011;Contreras and Steriade, 1996), bursts of high-frequency gamma oscillations (Ayoub et al, 2012;Dalal et al, 2010;Le Van Quyen et al, 2010;Valderrama et al, 2012;Worrell et al, 2012), monofractal and multifractal properties of the human sleep EEG (Weiss et al, 2009(Weiss et al, , 2011Zorick and Mandelkern, 2013) and including them in the proposed DSVM method could potentially lead to an even better classification. The detection of some of these phenomena might be enhanced by recent methodological developments (Ahmed et al, 2009;Babadi et al, 2012;Chaibi et al, 2012Chaibi et al, , 2013Chaibi et al, , 2014Jaleel et al, 2014;Nonclercq et al, 2013;O'Reilly and Nielsen, 2014a,b;Warby et al, 2014;Worrell et al, 2012). Furthermore, features such as cross-frequency interactions, longrange coupling among distant electrodes and long-range temporal correlations may also provide efficient novel markers for distinct sleep stages.…”

Multichannel Sleep Spindle Detection using Sparse Low-Rank Optimization

“…For the experiments that follow, the value of λ 2 is varied in the range (25,35) for the DREAMS database and in the range (45,48) for the MASS database. The value of the threshold c is varied in the range (0.5, 3) for both the databases.…”

“…Filtering based approaches vary from basic methods, which utilize a bandpass filter with constant or adaptive thresholds, to advanced methods that use time-frequency information along with bandpass filtering. Most of the filtering based methods involve pre-processing of the desired channel of the EEG (usually a central channel) for arti-fact removal [35]. One of the first automated detectors to be proposed used a bandpass filter in conjunction with an amplitude threshold [56].…”