Detail of the student project

Topic:Automatic sleep EEG patterns detection
Department:Katedra kybernetiky
Supervisor:Mgr. Elizaveta Saifutdinova , Ing. Václav Gerla Ph.D.
Announce as:BP,PMI
Description:An electroencephalogram (EEG) is a recorded electrical activity in the brain. Presence of special EEG patterns are very helpful in brain activity evaluation and as consequence in diagnosing of neurological disorders. Such as in sleep medicine sleep spindles and k-complexes mark a second sleep stage and decreased sleep spindles density can be connected with some cognitive issues. In the reality a trained clinician evaluates these signals manually. Sometimes EEG signal continue for a few hours and evaluation became a very tedious work and of course it leads to mistakes.
Automatic EEG pattern detection helps to evaluate EEG signals using objective criteria which, in common, is not strictly defined. Machine learning techniques are able to compute the criteria using training examples or without them. Standard steps for such method are follows: preprocessing, segmentation, feature extraction, feature selection and classification (clustering).
We have implemented simple automatic EEG pattern detection algorithm for sleep spindles in Matlab. Your task will be to gain an understanding of how it works and improve the algorithm (probably it would be one of the steps of the algorithm) in such a way that showed the best values of the metrics. The task helps to comprehend basic principles of automatic detection methods in signals and using machine learning technique in particular.
Bibliography:Adeli, Hojjat: Automated EEG-Based Diagnosis of Neurological Disorders: Inventing the Futute of Neurology. CRC Press. 2010
Gerla, V.: Automated Analysis of Long-Term EEG Signals. PhD Thesis. CVUT Prague, 2012
Sanei, S.; Chambers, J.A.: EEG Signal Processing. Wiley, 2007
Realization form:kod v matlabu
Responsible person: Petr Pošík