Loading...

Diagnosis of Depressive Disorder using Classification of Graphs Obtained from Electroencephalogram Signals

Moradi, Amir | 2022

62 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56535 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Hajipour, Sepideh
  7. Abstract:
  8. Depression is a type of mental disorder that is characterized by the continuous occurrence of bad moods in the affected person. Studies by the World Health Organization (WHO) show that depression is the second disease that threatens human life, and eight hundred thousand people die due to suicide every year. In order to reduce the damage caused by depression, it is necessary to have an accurate method for diagnosing depression and its rapid treatment, in which electroencephalogram (EEG) signals are considered as one of the best methods for diagnosing depression. Until now, various researches have been conducted to diagnose depression using electroencephalogram signals, most of which were based on feature extraction and classification methods. Today, in classification problems, the use of methods based on graphic processing has received a lot of attention. In this research, with the help of features extracted from the used data, we formed graphs of brain communication and obtained classification results in different frequency bands by using methods based on graph features (graphlet) and kernel graph. In the obtained results, the best classification related to the semi-rotational kernel was the PLV feature in theta band, the accuracy of which reached 73%, and also the graphlet-based classification results were about 74%. In the end, by using the multi-kernel method and the combination of different kernels, we saw an improvement in the classification results
  9. Keywords:
  10. Electroencephalogram Signals Classification ; Electroencphalogram Signal ; Depressive Disorder ; Graph Classification ; Brain Connectivity ; Kernel Methods ; Graphlet-Based Method

 Digital Object List

 Bookmark

No TOC