Enhancing Prediction for Depressive Symptoms in Subthreshold Depression by Personalized Parcellation
Poster No:
1793
Submission Type:
Abstract Submission
Authors:
Liangfang Li1, Junji Ma1, Yue Gu1, Xitian Chen1, Jiehui Qian1, Ying Lin1, Zhengjia Dai1
Institutions:
1Sun Yat-sen University, Guangzhou, Guangdong
First Author:
Co-Author(s):
Introduction:
Subthreshold depression (SD) is characterized by clinically relevant depressive symptomatology, yet not meet the diagnosis criteria of major depressive disorder (MDD) (Cuijpers & Smit, 2004). Since MDD can be better considered as a continuum and SD is the early stage of MDD, exploring the neuropathology of SD is warranted for understanding the dynamic course of depression-related brain changes from mild to major depression (Hwang et al., 2015), which aids clinical intervention and treatment. However, few studies have studied the SD population and previous works mainly use group-level atlas to assess functional organization and detect case-control difference (Hwang et al., 2015; Zhu et al., 2019), which may ignore the individualized features and miss vital brain-behavior associations that are critical for understanding disease processes (Michon et al., 2022; Zhao et al., 2021).
Methods:
We collected the structural MRI (sMRI) and Resting-state fMRI (R-fMRI) data, depression and anxiety scores from 65 SD participants with BDI-II score >13 (Beck et al., 1996). Depressive and anxiety symptoms were measured by Beck Depression Inventory–II (21 items) and Beck Anxiety Inventory scale (21 items). The R-fMRI data was pre-processed using the fMRIprep pipeline (Esteban et al., 2019), including tissue segmentation, surface reconstruction, unstable volumes removal, slice-timing, head motion correction, regressing out nuisance variables, and filtering. After aligning to the structural images, the R-fMRI data was smoothed and projected to fsaverage6 surface space. First, the multi-session hierarchical Bayesian model (Kong et al., 2021) was used to perform individualized parcellation procedure, during which we initialized with the 400-region group-level atlas (Schaefer et al., 2018) and then iteratively refined individualized region boundary. Second, correlations between items and individualized functional connectivity (FC) matrices were calculated to get 42 FC-symptom correlation matrices, which were then clustered into symptom domains using Ward's hierarchical clustering analysis (Murtagh & Legendre, 2014). Third, to identify FCs that were reliably related to a given symptom domain, we ran permutation tests by randomly averaging the equal number of FC-symptom matrices 10000 times. Connection with p < 0.0001 was kept as a feature, and the correlation between FC and domain score was set as weight. Last, we used the weighted sum of FC values as the predictive symptom score and assessed the correlation between predicted and actual symptom scores. Same analyses directly using group-level atlas were repeated as a baseline situation.
Results:
We identified three symptom domains, all of which can be estimated by related individualized FCs (depressive domain: r = 0.784, p < 0.001; cognitive and somatic domain: r = 0.630, p < 0.001; anxiety domain: r = 0.755, p < 0.001). The contributing individualized FCs to symptom estimation were mainly between-network connections involving the DMN, vATN, and MOT networks (Yeo et al., 2011). Meanwhile, the FC created using group-level atlas also significantly related to these symptom domains (depressive domain: r = 0.744, p < 0.001; cognitive and somatic domain: r = 0.610, p < 0.001; anxiety domain: r = 0.704, p < 0.001). Predictive group-level FCs were also dominated by between-network connections, many of which originate from vATN and primary networks including VIS and MOT. Notably, statistical comparisons of correlation values showed that prediction performance of individualized FC model outperformed that of group-level FC model (zs > 1.668, ps < 0.048).
·Fig. 1 Predictive individualized FC for estimating three symptom domains.
·Fig. 2 Contributing group-level FC for estimating three symptom domains.
Conclusions:
We identified three different symptom domains for the SD population, and individualized FC estimated the symptom scores with higher accuracy than the group-level FC. These findings highlight the necessity of considering individual variability in brain functional organization, which can facilitate the detection of more effective neural biomarkers for SD and MDD.
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 2
Modeling and Analysis Methods:
fMRI Connectivity and Network Modeling 1
Novel Imaging Acquisition Methods:
BOLD fMRI
Keywords:
Affective Disorders
Anxiety
FUNCTIONAL MRI
Machine Learning
Psychiatric
Psychiatric Disorders
1|2Indicates the priority used for review
Provide references using author date format
② Cuijpers, P. (2004), 'Subthreshold depression as a risk indicator for major depressive disorder: A systematic review of prospective studies', Acta Psychiatrica Scandinavica, vol. 109, no. 5, pp. 325–331.
③ Esteban, O. (2019), 'fMRIPrep: A robust preprocessing pipeline for functional MRI', Nature Methods, vol. 16, no. 1, pp. 111-116.
④ Hwang, J. W. (2015), 'Subthreshold depression is associated with impaired resting-state functional connectivity of the cognitive control network', Translational Psychiatry, vol. 5, no. 11, pp. e683-e683.
⑤ Kong, R. (2021), 'Individual-Specific Areal-Level Parcellations Improve Functional Connectivity Prediction of Behavior', Cerebral Cortex, vol. 31, no. 10, pp. 4477–4500.
⑥ Michon, K. J. (2022), 'Person-specific and precision neuroimaging: Current methods and future directions', NeuroImage, vol. 263, pp. 119589.
⑦ Schaefer, A. (2018), 'Local-Global Parcellation of the Human Cerebral Cortex from Intrinsic Functional Connectivity MRI', Cerebral Cortex, vol. 28, no. 9, pp. 3095–3114.
⑧ Yeo, B. T. T. (2011), 'The organization of the human cerebral cortex estimated by intrinsic functional connectivity', Journal of Neurophysiology, vol. 106, no. 3, pp. 1125–1165.
⑨ Zhao, Y. (2023), 'Individualized functional connectome identified replicable biomarkers for dysphoric symptoms in first-episode medication-naïve patients with major depressive disorder', Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, vol. 8, no. 1, pp. 42-51.
⑩ Zhu, Y. (2019), 'Connectome-based biomarkers predict subclinical depression and identify abnormal brain connections with the lateral habenula and thalamus', Frontiers in psychiatry, vol. 10, pp. 371.