DWT DECOMPOSITION OF EEG SIGNAL AND SOURCE LOCALIZATION USING ICA-ELORETA TO ESTIMATE ACTIVE BRODMANN AREA AT 3D BRAIN CORTEX MODEL FOR BASIC HAND MOTOR ACTIVITY

• Main Author: Masitoh
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/56832
• Institution: Institut Teknologi Bandung

Electroencephalogram (EEG) is an electrophysiological measurement instrument to monitor electrical activity in the brain by measuring the electrical potential difference on scalp. The electrical activity recorded by the EEG electrodes is a superposition of various sources of electrical activity in the cerebral cortex. This condition causes the electrical potential difference recorded by the electrodes on the scalp does not represent the location of the brain activity source. Therefore, to estimate the source of brain electrical activity, a method known as EEG source localization is needed. If this is implemented in the medical rehabilitation process, especially in handling cases of motor disability after stroke, this will help doctors to visualize active brain regions of patients. Thus, feedback from patients can be obtained as information about the improvement of patient's abilities during the rehabilitation process. In this study, EEG source localization was performed by estimating the current density distribution in the brain. Independent Component Analysis (ICA) and exact-Low Resolution Brain Electromagnetic Tomography (eLORETA) were used as EEG source localization method. ICA is a signal decomposition algorithm to obtain independent components. The independent component was then localized using eLORETA which transforms the measured electrical potential difference on the scalp into current density distribution across the entire brain volume by considering a three-skeletal spherical model consisting of the skin, skull and cerebral cortex. By integrating the 3D brain cortex model of the MNI 152 template, information about the estimated active brain regions when the subject performs certain activities can be identified. EEG data recording was performed on healthy male subjects using 21 electrodes with 10-20 placement system. Subject was asked to perform basic hand motor activity i.e gripping without the load and with the load for each the right and left hand. The recorded EEG signal was then segmented to separate the baseline condition and movement condition. The denoising process using a Bandpass Filter (BPF) 5-15 Hz and 6-level Discrete Wavelet Transform (DWT) decomposition was carried out to obtain Mu waves at the frequency of 8-13 Hz which is related to motor activity in the sensorimotor cortex i.e Brodmann Area (BA) 4 and 6. ICA was then used to obtain independent components whose quality of the independence was assessed by the stability index. The EEG signal from the DWT results shows higher average stability index than BPF one. The independent components obtained using ICA were plotted onto scalp topographic map to provide visualization of the calculation of Mu signal power on the scalp. An independent component is selected based on neuroscience theory related to Event-Related Desynchronization (ERD) and contralateral phenomena in the motor cortex. The selected ICA component was then localized using eLORETA by defining the position of the 21 electrodes used in this study. The results of eLORETA localization showed that the central area (Brodmann Area 6) on the left hemisphere is more active when the subject makes gripping movement in the right hand, and vice versa. This confirmed the contralateral phenomenon which says the left hemisphere is associated to right-side body activity, and vice versa. From the results of statistical calculations for Brodmann Area 4 and 6, there was a decrease in the average current density when gripping activity compared to baseline condition, in line with the decrease in Mu signal power calculated at electrodes C3 and C4, which are closest to the motor cortex. This confirmed the occurrence of Event-Related Desynchronization (ERD) of Mu signal in the sensorimotor cortex when there is motor activity. Whereas for the comparison of the gripping activity without the load and with the load, the average power and current density of Mu signal when gripping activity with the load is smaller than the activity without load. This means greater desynchronization occurs in the gripping activity with the load. ?