Loading...

Detection of High Frequency Oscillations from ECoG Recordings in Epileptic Patients

Gharebaghi Asl, Fatemeh | 2022

185 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55460 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Hajipour, Sepideh; Sinaei, Farnaz
  7. Abstract:
  8. The processing of brain signals, including the electrocorticogram (ECoG) signal, is widely used in the investigation of neurological diseases. Conventionally, the ECoG signal has frequency components up to the range of 80 Hz. Studies have proven that in some conditions, such as epilepsy, the brain signal includes frequency components higher than 80 Hz, which are called high-frequency oscillations (HFO). Therefore, HFOs are recognized as a biomarker for epilepsy. The aim of this thesis is to review the previous methods of detecting HFOs and to present new methods with greater efficiency in the direction of diagnosis or treatment of epileptic patients. For this purpose, we used the ECoG data of the Fedele group. Next, the previous methods for detecting these oscillations were reviewed and new methods were presented, among which the best result was obtained by presenting a method based on graph learning of time samples and classification using the DGCNN network. According to this method, the time data structure is a graph structure, which is different in HFO and non-HFO intervals. That is, by considering the sequence of time samples as the vertices of a graph and training the adjacency matrix of the resulting graph using time data, different graphs are obtained for HFO and non-HFO intervals. In addition to the adjacency matrix, other features such as RMS, STE, LL and Teager distinguish the intervals. Therefore, we consider these features as features of graph vertices that play a role in increasing classification accuracy. We used the DGCNN network to classify the time-trained graphs with vertices labels (extracted features). Finally, the DGCNN network achieved 90.7% sensitivity and 93.3% specificity
  9. Keywords:
  10. High Frequency Oscillations (HFO) ; Epilepsy ; Deep Convolutional Neural Networks ; Electrocorticogram (ECoG)Signal ; Deep Graph Convolutional Neural Networks (DGCNN) ; Brain Waves

 Digital Object List

 Bookmark

No TOC