Spatio-temporal learning and explaining for dynamic functional connectivity analysis: Application to depression.

BACKGROUND: Functional connectivity has been shown to fluctuate over time. The present study aimed to identifying major depressive disorders (MDD) with dynamic functional connectivity (dFC) from resting-state fMRI data, which would be helpful to produce tools of early depression diagnosis and enhance our understanding of depressive etiology. METHODS: The resting-state fMRI data of 178 subjects were collected, including 89 MDD and 89 healthy controls. We propose a spatio-temporal learning and explaining framework for dFC analysis. A yet effective spatio-temporal model is developed to classifying MDD from healthy controls with dFCs. The model is a stacking neural network model, which learns network structure information by a multi-layer perceptron based spatial encoder, and learns time-varying patterns by a Transformer based temporal encoder. We propose to explain the spatio-temporal model with a two-stage explanation method of importance feature extracting and disorder-relevant pattern exploring. The layer-wise relevance propagation (LRP) method is introduced to extract the most relevant input features in the model, and the attention mechanism with LRP is applied to extract the important time steps of dFCs. The disorder-relevant functional connections, brain regions, and brain states in the model are further explored and identified. RESULTS: We achieved the best classification performance in identifying MDD from healthy controls with dFC data. The top important functional connectivity, brain regions, and dynamic states closely related to MDD have been identified. LIMITATIONS: The data preprocessing may affect the classification performance of the model, and this study needs further validation in a larger patient population. CONCLUSIONS: The experimental results demonstrate that the proposed spatio-temporal model could effectively classify MDD, and uncover structural and temporal patterns of dFCs in depression.

Reduced myelin content in bipolar disorder: A study of inhomogeneous magnetization transfer.

BACKGROUND: Previous neuroimaging and pathological studies have found myelin-related abnormalities in bipolar disorder (BD), which prompted the use of magnetic resonance (MR) imaging technology sensitive to neuropathological changes to explore its neuropathological basis. We holistically investigated alterations in myelin within BD patients by inhomogeneous magnetization transfer (ihMT), which is sensitive and specific to myelin content. METHODS: Thirty-one BD and 42 healthy controls (HC) were involved. Four MR metrics, i.e., ihMT ratio (ihMTR), pseudo-quantitative ihMT (qihMT), magnetization transfer ratio and pseudo-quantitative magnetization transfer (qMT), were compared between groups using analysis methods based on whole-brain voxel-level and white matter regions of interest (ROI), respectively. RESULTS: The voxel-wise analysis showed significantly inter-group differences of ihMTR and qihMT in the corpus callosum. The ROI-wise analysis showed that ihMTR, qihMT, and qMT values in BD group were significantly lower than that in HC group in the genu and body of corpus callosum, left anterior limb of the internal capsule, left anterior corona radiate, and bilateral cingulum (p < 0.001). And the qihMT in genu of corpus callosum and right cingulum were negatively correlated with depressive symptoms in BD group. LIMITATIONS: This study is based on cross-sectional data and the sample size is limited. CONCLUSION: These findings suggest the reduced myelin content of anterior midline structure in the bipolar patients, which might be a critical pathophysiological feature of BD.

Connectome architecture modulates the gray matter atrophy in major depression disorder patients with diverse suicidal ideations.

Gray matter (GM) atrophy is well documented in patients with major depressive disorder (MDD), but its underlying mechanism remains unknown. This study aimed to examine the GM atrophy in MDD patients with diverse suicidal ideations (SIs) and to explore whether those alterations were driven by connections. GM volume was estimated in 163 patients with recurrent MDD (comprising 122 with SI [MDDSI] and 41 without SI [MDDNSI]) and 134 health controls (HCs). A two-sample t-test was used to identify GM volume abnormalities in MDD patients and their subgroups. Functional connectivity was computed between pairs of aberrant GM in both patients and HCs, which were further compared with the connectivity of random brain regions. A permutation test was performed to assess its significance. Propensity score matching (PSM) was further performed to validate the main results. Compared with HCs, the MDDNSI group exhibited GM atrophy in 24 regions, with the largest effect sizes found in the frontal and parietal lobes, while the MDDSI group exhibited more widespread GM atrophy involving 49 regions, with the largest effect sizes in the frontal lobe, parietal lobe, temporal lobe, and the limbic system. Furthermore, patients and HCs exhibited significantly increased functional connectivity between regions with GM atrophy compared with randomly selected regions (p < 0.05). PSM analysis presented similar results to the main analysis. MDD patients had diverse GM atrophy features according to their SI tendency. Moreover, connectome architecture modulates the GM atrophy in MDD patients, implying the possibility that connections drive these pathological changes.

Reduced coupling between global signal and cerebrospinal fluid inflow in patients with depressive disorder: A resting state functional MRI study.

BACKGROUND: Depressed patients often suffer from sleep disturbance, which has been recognized to be responsible for glymphatic dysfunction. The purpose of this study was to investigate the coupling strength of global blood­oxygen-level-dependent (gBOLD) signals and cerebrospinal fluid (CSF) inflow dynamics, which is a biomarker for glymphatic function, in depressed patients and to explore its potential relationship with sleep disturbance by using resting-state functional MRI. METHODS: A total of 138 depressed patients (112 females, age: 34.70 ± 13.11 years) and 84 healthy controls (29 females, age: 36.6 ± 11.75 years) participated in this study. The gBOLD-CSF coupling strength was calculated to evaluate glymphatic function. Sleep disturbance was evaluated using the insomnia items (item 4 for insomnia-early, item 5 for insomnia-middle, and item 6 for insomnia-late) of The 17-item Hamilton Depression Rating Scale for depressed patients, which was correlated with the gBOLD-CSF coupling strength. RESULTS: The depressed patients exhibited weaker gBOLD-CSF coupling relative to healthy controls (p = 0.022), possibly due to impairment of the glymphatic system. Moreover, the gBOLD-CSF coupling strength correlated with insomnia-middle (r = 0.097, p = 0.008) in depressed patients. Limitations This study is a cross-sectional study. CONCLUSION: Our findings shed light on the pathophysiology of depression, indicating that cerebral waste clearance system deficits are correlated with poor sleep quality in depressed patients.

Spatiotemporal discoordination of brain spontaneous activity in major depressive disorder.

BACKGROUND: Major depressive disorder (MDD) is a widespread mental health issue, impacting spatial and temporal aspects of brain activity. The neural mechanisms behind MDD remain unclear. To address this gap, we introduce a novel measure, spatiotemporal topology (SPT), capturing both the hierarchy and dynamic attributes of brain activity in depressive disorder patients. METHODS: We analyzed fMRI data from 285 MDD inpatients and 141 healthy controls (HC). SPT was assessed by coupling brain gradient measurement and time delay estimation. A nested machine learning process distinguished between MDD and HC using SPT. Person's correlation tested the link between SPT's and symptom severity, and another machine learning method predicted the gap between patients' chronological and brain age. RESULTS: SPT demonstrated significant differences between patients and healthy controls (F = 2.944, p < 0.001). Machine learning approaches revealed SPT's ability to discriminate between patients and healthy controls (Accuracy = 0.65, Sensitivity = 0.67, Specificity = 0.64). Moreover, SPT correlated with the severity of depression symptom (r = 0.32. pFDR = 0.045) and predicted the gap between patients' chronological age and brain age (r = 0.756, p < 0.001). LIMITATIONS: Evaluation of brain dynamics was constrained by MRI temporal resolution. CONCLUSIONS: Our study introduces SPT as a promising metric to characterize the spatiotemporal signature of brain function, providing insights into deviant brain activity associated with depressive disorders and advancing our understanding of their psychopathological mechanisms.