Page 113 - MEGIN Book Of Abstracts - 2023

RESULTS MDD demonstrated poorer cognitive perfor- velop a novel approach that accurately classifies MDD
mance in six domains compared to HC. The difference and BD based on their resting-state magnetoencepha-
in community detection (functional integration mode) lography (MEG) signals during euthymic phases. A re-
in MDD was frequency-dependent. MDD showed visited 3D CNN model, Semi-CNN, that could automati-
significantly decreased community dynamics in all cally detect brainwave patterns in spatial, temporal,
frequency bands compared to HC. Specifically, differ- and frequency domains was implemented to classify
ences in the visual network (VN) and default mode wavelet-transformed MEG signals of normal controls
network (DMN) were detected in all frequency bands, and MDD and BD patients. The model achieved a test
differences in the cognitive control network (CCN) were accuracy of 96.05% and an average of 95.71% accuracy
detected in the alpha2 and beta frequency bands, and for 5-fold cross-validation. Furthermore, saliency maps
differences in the bilateral limbic network (BLN) were of the model were estimated using Grad-CAM++ to vi-
only detected in the beta frequency band. Moreover, sualize the proposed classification model and highlight
community dynamics in the alpha2 frequency band disease-specific brain regions and frequencies. Clinical
were positively correlated with verbal learning and Relevance - Our model provides a stable pipeline that
reasoning problem solving abilities in MDD. accurately classifies MDD, BD, and healthy individuals
based on resting-state MEG signals during the euthy-
CONCLUSIONS Our study found that decreasing in mic phases, opening the potential for quantitative and
the dynamics of overlapping sub-networks may differ accurate brain-based diagnosis for the highly misdiag-
by frequency bands. The aberrant dynamics of over- nosed MDD/BD patients.
lapping neural sub-networks revealed by frequency-
specific MEG signals may provide new information on Annual International Conference of the IEEE Engineering
the mechanism of cognitive impairments that result in Medicine and Biology Society. IEEE Engineering in
from MDD. Medicine and Biology Society. Annual International
Conference (2022), Vol. 2022 (36086021) (0 citations)
Keywords: Cognitive function, Dynamic functional con-
nectivity, Magnetoencephalography, Major depressive
disorder, Overlapping sub-network Research on the MEG of Depression Patients Based
on Multivariate Transfer Entropy (2022)
Journal of affective disorders (2023), Vol. 320 (36179776)
(0 citations) Zhang, Xinyu; Xie, Jicheng; Fan, Changyu; Wang, Jun

Smart Health Big Data Analysis and Location Services
MEG-based Classification and Grad-CAM Engineering Research Center of Jiangsu Province, Nanjing
Visualization for Major Depressive and Bipolar University of Posts and Telecommunications, Nanjing, China
Disorders with Semi-CNN (2022)
ABSTRACT The pathogenesis of depression is complex,
Huang, Chun-Chih; Low, Intan; Kao, Chia-Hsiang; Yu, and the current means of medical diagnosis is single.
Chuan-Yu; Su, Tung-Ping; Hsieh, Jen-Chuen; Chen, Patients with severe depression may even have great
Yong-Sheng; Chen, Li-Fen physical pain and suicidal tendencies. Magnetoen-
cephalography (MEG) has the characteristics of ultra-
ABSTRACT Major depressive disorder (MDD) and bi- high spatiotemporal resolution and safety. It is a good
polar disorder (BD) are two major mood disorders with medical means for the diagnosis of depression. In this
partly overlapped symptoms but different treatments. paper, multivariate transfer entropy algorithm is used
However, their misdiagnosis and mistreatment are to study MEG of depression. In this paper, the subjects
common based on the DSM-V criteria, lacking objective are divided into the same brain region and the multi-
and quantitative indicators. This study aimed to de- channel combination between different brain regions,

 ontents  Index 92
C