OAK

Classification of EEG Induced by Various Thermal Stimuli using Deep Convolutional Neural Network

Metadata Downloads
Abstract
With the growing emphasis on quantifying pain in the field of biomedical engineering, there has been a surge in studies aiming to develop pain-related biomarkers by analyzing bio-signals such as electroencephalogram (EEG). While recent EEG studies have investigated cold and hot pain, they typically focused on one stimulus condition at a time and required relatively long stimulation periods to gather EEG data. However, pain recognition in real-world scenarios occurs rapidly, and various internal and external factors can influence an individual's perception of pain.
In this thesis research, EEG signals were collected during the experience of Thermal Grill Illusion (TGI; 20-40 ℃, 23-37 ℃, and 26-34 ℃), as well as during Cool (20 ℃, 23 ℃, and 26 ℃) and Warm (34 ℃, 37 ℃, and 40 ℃) stimuli. The recorded EEG signals spanned a duration of 7 s, and classification was performed using deep convolutional neural network-based classifiers.
Among the three convolutional neural network (CNN) models used for classification (EEGNet, ShallowConvNet, and DeepConvNet), ShallowConvNet exhibited the best performance in accurately classifying the six normal stimuli (20 ℃, 23 ℃, 26 ℃, 34 ℃, 37 ℃, and 40 ℃). Therefore, ShallowConvNet was selected as the classifier for TGI, Cool, and Warm stimuli classification. The classification accuracies for the three-class classification of TGI, Cool, and Warm stimuli were similar, with an average accuracy of approximately 47 %. The classification results were further analyzed using a confusion matrix, revealing a bias in predicted labels towards the strongest and weakest stimuli.
To identify feature differences, DeepLIFT was employed, which allowed the observation of somatosensory cortex activity during the classification of Cool and Warm stimuli. For the classification of TGI stimuli, the central area (Cz, CP2) demonstrated a high contribution, particularly for the 20-40 ℃ and 23-37 ℃ TGI stimuli. This finding suggests the involvement of the primary somatosensory cortex (S1) in perceiving and modulating thermal stimuli, indicating that the TGI illusion arises due to S1 activation, resulting in the perception of the stimulus as hotter than the corresponding warm stimulus.
Author(s)
Jihoon Baek
Issued Date
2023
Type
Thesis
URI
https://scholar.gist.ac.kr/handle/local/19003
공개 및 라이선스
  • 공개 구분공개
파일 목록
  • 관련 파일이 존재하지 않습니다.

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.