Authors :
Presenting Author: Hsin Tung, MD – Neurological Institute, Taichung Veterans General Hospital, Taichung, Taiwan
syu-Jyun Peng, PhD – Professional Master Program in Artificial Intelligence in Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan; Yen-Chuan Chang, MD – College of Medicine, Taipei Medical University.
Rationale: Electroencephalography (EEG) records the neuronal activity, and the EEG background represents the brain function. We quantify the EEG into the values of power and connectivity and try to explore their relationship with the performances in different cognitive domains.
Methods: We retrospectively enrolled the subjects who had received both EEG exam and the neuropsychiatric test since June 2022 to December 2022. Cognitive Abilities Screening Instrument (CASI) was used, composing of nine aspects. We excluded subjects who had severe psychiatric disorders and the total CASI scores less than three.
EEG was performed in awake status with eye-closing for at least ten minutes. The electrodes were placed based on the international 10-20 system. Ninety-second period of signals without epileptiform discharges and electrode/muscle artifacts were collected. Independent component analysis was applied to preprocess and remove blinking artifacts. EEG signals were decomposed into four frequency bands: delta, theta, alpha, and beta, and then partitioned using 4-second interval with 50% overlapping. We calculated the power of four frequency bands of each epoch individually. The power of 19 electrodes were averaged, and then their partial correlation with the nine aspects of cognitive function was calculated by adjusting age and educational years.
The network topology was quantified into the parameters of the graph theory. Four parameters, global efficiency, betweenness centralis, clustering coefficient, and average nodal strength were calculated. We used proportional threshold from 0.1 to 0.9, with each 0.05 increasement to select the edges for analysis. Then, the correlation between the parameters and the scores of the nine cognitive aspects were studied.
Results:
Totally, there were 80 subjects were enrolled and then analyzed, including 30 male and 50 female. The age ranged from 28 to 88, and the educational year was from 0 to 19. In the average power aspect, we found long-term memory was negatively correlated with delta and theta band power (
p = < 0.001, 0.004), and positively correlated with beta band (
p = 0.001). Mental manipulation scores had negative correlation with delta band (
p = 0.007). The alpha band power did not show the relationship with the cognitive functions.
In the network topology aspect, the average nodal strength of delta band was negatively related to the scores of the short-term memory and the total scores. Higher average nodal strength and the clustering coefficient of alpha band are associated with better drawing scores.
Conclusions:
Our study disclosed that neuronal activation and their activity coherence are related to different cognitive domains. Lower-frequency band, like delta oscillation, was associated with cortical deafferentation and suggestive of inactive thalamo-cortical inputs. Thus, the subjects had poor long-term memory and mental manipulation. The alpha band coherence was related to the function of visual construction, which suggests well-organized alpha oscillation characterizes better cognitive processes.
Funding: None