Page 59 - MEGIN Book Of Abstracts - 2023

Medical University, Nanjing 210029, China; Department of investigation provides a better understanding of the
Psychiatry, The Affiliated Brain Hospital of Nanjing Medical neuroelectrophysiological process underlying cognitive
University, Nanjing 210029, China; Nanjing Brain Hospital, impairments in patients with UD and BD.
Medical School, Nanjing University, Nanjing 210093, China;
School of Biological Sciences and Medical Engineering, Keywords: Bipolar depression, Cognitive impairment,
Southeast University, Nanjing 210096, China. Electronic MEG, Magnetoencephalography, Resting-state, Unipolar
address: [email protected]; School of Biological Sciences depression
and Medical Engineering, Southeast University, Nanjing
210096, China; Child Development and Learning Science, Key Journal of affective disorders (2023), Vol. 321 (36181913)
Laboratory of Ministry of Education, Nanjing 210096, China. (0 citations)
Electronic address: [email protected]

BACKGROUND Unipolar depression (UD) and bipolar MEG-based Classification and Grad-CAM
depression (BD) showed convergent and divergent Visualization for Major Depressive and Bipolar
cognitive impairments. Neural oscillations are linked Disorders with Semi-CNN (2022)
to the foundational cognitive processes. We aimed to
investigate the underpinning spectral neuronal power Huang, Chun-Chih; Low, Intan; Kao, Chia-Hsiang; Yu,
patterns by magnetoencephalography (MEG), which Chuan-Yu; Su, Tung-Ping; Hsieh, Jen-Chuen; Chen,
combinates high spatial and temporal resolution. We Yong-Sheng; Chen, Li-Fen
hypothesized that patients with UD and BD exhibit
common and distinct patterns, which may contribute ABSTRACT Major depressive disorder (MDD) and bi-
to their cognitive impairments. polar disorder (BD) are two major mood disorders with
partly overlapped symptoms but different treatments.
METHODS Group cognitive tests were performed. Eyes However, their misdiagnosis and mistreatment are
closed resting-state MEG data were collected from 61 common based on the DSM-V criteria, lacking objective
UD, 55 BD, and 52 healthy controls (HC). Nonparamet- and quantitative indicators. This study aimed to de-
ric cluster-based permutation tests were performed velop a novel approach that accurately classifies MDD
to deal with the multiple comparison problem on and BD based on their resting-state magnetoencepha-
channel-frequency MEG data. Correlation analysis of lography (MEG) signals during euthymic phases. A re-
cognitive dysfunction scores and MEG oscillation were visited 3D CNN model, Semi-CNN, that could automati-
conducted by Spearman or partial correlation analysis. cally detect brainwave patterns in spatial, temporal,
and frequency domains was implemented to classify
RESULTS Wisconsin Card Sorting Test showed similar wavelet-transformed MEG signals of normal controls
cognitive impairment in patients with UD and BD. and MDD and BD patients. The model achieved a test
Moreover, patients with BD exhibited extensive cogni- accuracy of 96.05% and an average of 95.71% accuracy
tive deficits in verbal executive functions and visuospa- for 5-fold cross-validation. Furthermore, saliency maps
tial processing. Compare to HC, both patients with UD of the model were estimated using Grad-CAM++ to vi-
and BD showed increased frontal-central beta power sualize the proposed classification model and highlight
while high gamma power was decreased in UD groups disease-specific brain regions and frequencies. Clinical
during the resting-state. The significant correlations Relevance - Our model provides a stable pipeline that
between cognitive function and average beta power accurately classifies MDD, BD, and healthy individuals
were observed. based on resting-state MEG signals during the euthy-
mic phases, opening the potential for quantitative and
CONCLUSIONS Patients with BD had more cognitive
impairments on different dimensions than those with
UD, involving disrupted beta power modulations. Our

 ontents  Index 38
C