Automatic sleep spindles identification and classification with multitapers and convolution

Automatic sleep spindles identification and classification with multitapers and convolution

Sleep spindles are isolated transient surges of oscillatory neural activity present during sleep stages 2 and 3 in the nonrapid eye movement (NREM). They can indicate the mechanisms of memory consolidation and plasticity in the brain. Spindles can be identified across cortical areas and classified as either slow or fast. There are spindle transients across different frequencies and power, yet most of their functions remain a mystery. Using several electroencephalograms (EEG) databases, this study presents a new method, called the “spindles across multiple channels” (SAMC) method, for identifying and categorizing sleep spindles in EEGs during NREM sleep. The SAMC method uses a multitapers and convolution (MT&C) approach to extract the spectral estimation of different frequencies present in sleep EEGs and graphically identify spindles across multiple channels. The characteristics of spindles, such as duration, power, and event areas, are also extracted by the SAMC method. Comparison with other state-of-the-art spindle identification methods demonstrated the superiority of the proposed method with an agreement rate, average positive predictive value, and sensitivity of over 90% for spindle classification across the three databases used in this paper. The computing cost was found to be, on average, 0.004 seconds per epoch. The proposed method can potentially improve the understanding of the behaviour of spindles across the scalp and accurately identify and categorise sleep spindles.

This article is part of a UniSQ Thesis by publication. See Related Output.