Loading...

Analysis of Epileptic Rats' EEG and Detection and Prediction of Epileptic Seizures

Niknazar, Mohammad | 2010

891 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 40218 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Vosoughi Vahdat, Bijan; Shamsollahi, Mohammad Bagher
  7. Abstract:
  8. Epilepsy is one of the most significant neurological disorders that about one percent of people suffer from it. Epilepsy can only be controlled, and so far no cure for it has been provided. Despite the many advances in the treatment of diseases, for a quarter of patients there is no medical treatment solution for controlling epileptic seizures. In the studies of medical groups on the epilepsy, one approach is employment of some models for each type of epilepsy. These types may be created in the animals to allow studying of the mechanism of epilepsy and also finding drugs of treatment or controlling seizures for each type of epilepsy. There is a type of epilepsy that is called absence epilepsy and is usually seen in children. Animal model for this type of epilepsy is PTZ model. This model is also applicable to Myoclonic epilepsy in human. In this project the EEG signals of induced epileptic rats by PTZ model, were analyzed in interictal, preictal, ictal and post ictal periods. Dataset was acquired at Pasteur Institute of Iran under specific protocol. The goal of this project is detection and prediction of epileptic seizures in the induced epileptic rats by PTZ model. In this work in addition to these data for the epileptic seizure detection, a dataset provided by Epilepsy Center of Freiburg, and for the epilepsy-related EEG classification, a dataset provided by Epilepsy Center at the University of Bonn, are used. First, important methods and features, used for processing of single channel epileptic EEGs for detecting and predicting of epileptic seizures and classifying epilepsy-related EEGs, and results of other’s works have been explained. Then results of implementing the methods and features on human and rat datasets used in the project have been presented. Using statistical analysis of different features in ictal period and before and after it, it has been shown that statistical characteristics of some features have significant differences in these periods and therefore they can be used for epileptic seizure detection. The best result for human dataset was achieved by proposed dissimilarity index which led to Pvalue < 0.001. For rat dataset the best result was achieved by fuzzy similarity index which led to Pvalue < 0.001. Also for the first time, the value of average diagonal length in the recurrence plots was successfully applied for epileptic seizure detection in rats. In addition to above mentioned seizure detection approach, another approach was used. In the second approach, by calculating a feature in consecutive windows and drawing resulted index according to time and comparing this index with a threshold, the moment of seizure onset was estimated. For the first time, nonlinear energy and coastline features were used in this approach and acceptable results were obtained. The best result was achieved by coastline feature which led to mean of 2 seconds delay in its true detections. We also predicted epileptic seizures by calculating a feature in consecutive windows and drawing resulted index according to time and comparing this index with a threshold and achieved good results. Using proposed dissimilarity index in Delta sub-band, epileptic seizures in rats were predicted with mean of 299.5 seconds. For classifying epilepsy-related EEGs two new methods were presented. One method is based on recurrence quantification analysis and led to accuracy of 98.67% which provided the best result for the dataset.
  9. Keywords:
  10. Electroencphalogram ; Epileptic Seizure Prediction ; Epilepsy ; Epileptic Rats ; Pentylenetetrazole (PTZ)Model ; Epileptic Seizure Detection ; Electroencephalogram Signals Classification

 Digital Object List

  • محتواي پايان نامه
  •   view

 Bookmark

No TOC