Automatic detection of sleep transients and its applications in sleep spindle enhancement

Patti, C 2018, Automatic detection of sleep transients and its applications in sleep spindle enhancement, Doctor of Philosophy (PhD), Engineering, RMIT University.


Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Patti.pdf Thesis application/pdf 25.94MB
Title Automatic detection of sleep transients and its applications in sleep spindle enhancement
Author(s) Patti, C
Year 2018
Abstract Sleep spindles and K-complexes are transient events that are observed in electroencephalography (EEG) signals of the mammalian brain. In human sleep, spindles and K-complexes have been positively correlated with significant functions such as memory consolidation and intelligence. This dissertation addresses two issues related to sleep spindles and K-complexes, i.e., one is the use of automatic detection algorithms to detect sleep spindles and K-complexes, and the other is the exploration of auditory stimulation to enhance sleep spindle count during sleep.

The numbers of sleep spindles and K-complexes during sleep vary significantly between different subjects. This implies that automatic detection algorithms need to be developed with sensitivity to such differences. In this dissertation, multiple detection methods are presented. The primary methods presented here use the expectation maximisation (EM) clustering technique in order to be sensitive to inter-subject differences. It was found that clustering certain features using the EM algorithm produced results that were able to capture the aforementioned inter-subject differences. The results for sleep spindle detection were evaluated on two public databases and a private database, with a total of 27 subjects. While simultaneous addressing the problem of inter-subject differences, comparison with existing methods in literature showed that the performance of the EM based clustering technique was on par with what had been reported in the existing literature. The results for K-complex detection were evaluated on one public database with 6 subjects. Compared with existing methods in literature, the K-complex detection algorithm presented in this work showed poorer performance but the algorithm was effective in capturing inter-subject differences. Due to the lack of availability of multiple public K-complex databases, a conclusive finding could not be reached for the use of clustering algorithms to detect the K-complex.

This dissertation also presents results observed and analysed using two different methods to enhance sleep spindles during 90-minute naps. One method used auditory stimulation synchronised to a post spindle refractory period to enhance sleep spindle count. Sleep spindle detection algorithms developed using clustering techniques in the first part of the dissertation were used to synchronise auditory stimulation to a post spindle refractory period. The other method to enhance sleep spindles was to use sensorimotor rhythm (SMR) neurofeedback (NF) prior to a 90-minute nap. Three groups consisting of 2 subjects each were used in this study. Group 1 received auditory stimulation only for 5 sessions, and Group 2 received auditory stimulation and NF for 10 sessions, while Group 3 received NF only for 10 sessions. The results showed that both Group 1 and Group 2 showed an increase in sleep spindles compared with the baseline, while Groups 3 showed negative performance compared with the baseline. Results showed that only group 1 showed an improvement in memory consolidation compared with the baseline, while Groups 2 and 3 showed negative performance compared with the baseline. These results indicate that auditory stimulation synchronised to a post spindle refractory period is effective in enhancing sleep spindles and memory consolidation, while SMR NF showed an inverse relationship in enhancing sleep spindles and memory consolidation.

Overall, the research presented in this dissertation made novel contributions with the application of clustering techniques to sleep transient detection and auditory stimulation synchronised to a post spindle refractory period.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Engineering
Subjects Biomechanical Engineering
Neurosciences not elsewhere classified
Computer Perception, Memory and Attention
Keyword(s) Sleep
Sleep Spindle
K-complex
Auditory Stimulation during Sleep
Automatic detection
Multivariate Gaussian Mixture Model
Expectation Maximisation
Versions
Version Filter Type
Access Statistics: 32 Abstract Views, 45 File Downloads  -  Detailed Statistics
Created: Tue, 27 Nov 2018, 15:40:46 EST by Anna Koh
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us