TY - JOUR T1 - Multichannel Sleep Spindle Detection using Sparse Low-Rank Optimization JF - bioRxiv DO - 10.1101/104414 SP - 104414 AU - Ankit Parekh AU - Ivan W. Selesnick AU - Ricardo S. Osorio AU - Andrew W. Varga AU - David M. Rapoport AU - Indu Ayappa Y1 - 2017/01/01 UR - http://biorxiv.org/content/early/2017/04/10/104414.abstract N2 - Background We consider the detection of sleep spindles simultaneously across the frontal, central and occipital channels of sleep EEG in a single run.New Method We propose a multichannel sleep spindle detection method utilizing a multichannel transient separation algorithm based on a sparse optimization framework. The proposed transient separation algorithm decomposes the multichannel EEG into the sum of an oscillatory and a transient component. Consecutive overlapping blocks of the multichannel oscillatory component are assumed to be of low-rank whereas the transient component is assumed to be piecewise constant with a zero baseline. We estimate both the components by minimizing a convex objective function using an iterative algorithm. The multichannel oscillatory component is used in conjunction with the Teager operator for detecting sleep spindles.Results and comparison with other methods The performance of the proposed method is evaluated using an online single channel EEG database and compared with 7 state-of-the-art automated detectors. The by-event F1 scores for the proposed spindle detection method averaged 0.67 ± 0.03. The average false discovery rate for the proposed method was 31.3 ± 0.04%. For an overnight 6 to 8 channel EEG signal, the proposed algorithm takes on an average 2 minutes to detect sleep spindles.Conclusions Comparable F1 scores and fast run times make the proposed spindle detector a valuable tool in answering the open question of studying the dynamics of sleep spindles and tracking their propagation overnight across the scalp in sleep EEG. ER -