Brain Connectivity

As the U.S. National Institute of Health notes, the rich biomedical data can greatly improve our knowledge of human health if new analytical tools are developed, and their applications are broadly disseminated. A major challenge in analyzing the brain as a complex system is about developing parsimonious multivariate methods, and particularly choosing the most appropriate one among the existing methods with respect to the study variables in this multidisciplinary field. This study provides a review on the most important multivariate methods to aid in helping the most appropriate ones with respect to the desired variables for each study.

Data augmentation improves the accuracy of deep learning models when training data are scarce by synthesizing additional samples. This work addresses the lack of validated augmentation methods specific for synthesizing anatomically realistic four-dimensional (4D) (three-dimensional [3D] + time) images for neuroimaging, such as functional magnetic resonance imaging (fMRI), by proposing a new augmentation method.

The proposed method, Brain Library Enrichment through Nonlinear Deformation Synthesis (BLENDS), generates new nonlinear warp fields by combining intersubject coregistration maps, computed using symmetric normalization, through spatial blending. These new warp fields can be applied to existing 4D fMRI to create new augmented images. BLENDS was tested on two neuroimaging problems using de-identified data sets: (1) the prediction of antidepressant response from task-based fMRI (original data set n = 163), and (2) the prediction of Parkinson's disease (PD) symptom trajectory from baseline resting-state fMRI regional homogeneity (original data set n = 43).

BLENDS readily generates hundreds of new fMRI from existing images, with unique anatomical variations from the source images, that significantly improve prediction performance. For antidepressant response prediction, augmenting each original image once (2 × the original training data) significantly increased prediction R ² from 0.055 to 0.098 ( p < 1 e − 6 ), whereas at 10 × augmentation R ² increased to 0.103. For the prediction of PD trajectory, 10 × augmentation R ² increased from −0.044 to 0.472 ( p < 1 e − 6 ).

Augmentation of fMRI through nonlinear transformations with BLENDS significantly improved the performance of deep learning models on clinically relevant predictive tasks. This method will help neuroimaging researchers overcome data set size limitations and achieve more accurate predictive models.

Given deep learning success across life science, there is interest in their application to neuroimaging. Deep learning requires a large training dataset, hence data augmentation has been proposed to bolster model training, however, there is no validated augmentation for synthesizing anatomically realistic 4D images. To remedy, we propose BLENDS, which generates new fMRI by combining intersubject coregistrations into a warp field which is applied to existing 4D fMRI to create new images. Tests on predicting: (1) antidepressant response and (2) Parkinson's trajectory demonstrate significant performance improvements. The proposed general method helps researchers overcome limited data and achieve accurate predictive models.

Multivoxel pattern analysis (MVPA) has emerged as a powerful unbiased approach for generating seed regions of interest (ROIs) in resting-state functional connectivity (RSFC) analysis in a data-driven manner. Studies exploring RSFC in multiple sclerosis have produced diverse and often incongruent results.

The aim of the present study was to investigate RSFC differences between people with relapsing-remitting multiple sclerosis (RRMS) and healthy controls (HC).

We performed a whole-brain connectome-wide MVPA in 50 RRMS patients with expanded disability status scale ≤4 and 50 age and gender-matched HCs.

Significant group differences were noted in RSFC in three clusters distributed in the following regions: anterior cingulate gyrus, right middle frontal gyrus, and frontal medial cortex. Whole-brain seed-to-voxel RSFC characterization of these clusters as seed ROIs revealed network-specific abnormalities, specifically in the anterior cingulate cortex and the default mode network.

The network-wide RSFC abnormalities we report agree with the previous findings in RRMS, the cognitive and clinical implications of which are discussed herein.

This study investigated resting-state functional connectivity (RSFC) in relapsing-remitting multiple sclerosis (RRMS) people with mild disability (expanded disability status scale ≤4). Whole-brain connectome-wide multivoxel pattern analysis was used for assessing RSFC. Compared with healthy controls, we were able to identify three regions of interest for significant differences in connectivity patterns, which were then extracted as a mask for whole-brain seed-to-voxel analysis. A reduced connectivity was noted in the RRMS group, particularly in the anterior cingulate cortex and the default mode network regions, providing insights into the RSFC abnormalities in RRMS.

Video game playing is most often a perceptually and cognitively engaging activity. Players enter into sensory-rich competitive environments, which require them to go from trivial tasks to making active decisions repeatedly and could lend themselves to improve sensorimotor decision-making capabilities. Since video game playing requires moment-to-moment switching of attention from one aspect of sensory information and task to another, enhanced attention control and attention-switching mechanism in the brain can be thought as the neural basis for such improvements. Previous studies have suggested that attention switching is mediated by the salience network (SN). However, how SN interacts with the dorsal attention network (DAN) in active decision-making tasks and whether video game playing modulates these networks remain to be investigated.

Using a modified version of the left–right moving dot motion task in a functional magnetic resonance imaging experiment, we examined the decision response times (dRTs) and functional interactions within and between SN and DAN for video game players (VGPs) and nonvideo game players (NVGPs).

We found that VGPs had lower response times for all task conditions and higher decision accuracy for a medium speed setting of moving dots. Associated with this improved task performance in VGPs compared with NVGPs was an increase in DAN to SN connectivity. This SN-DAN connectivity was negatively correlated with dRT.

These results suggest that enhanced influence of DAN over SN is the brain basis for improved sensorimotor decision-making performance as a result of engaging long term in cognitively challenging and attention-demanding activities such as video game playing.

Being able to flexibly direct attention is a key factor in sensorimotor decision-making. Video game playing, an attentionally and cognitively engaging activity, can have a beneficial effect on attention and decision-making. Through this study, we examined whether video game players (VGPs) have improved decision-making skills and investigated the brain basis for improvements in a functional magnetic resonance imaging experiment. Brain connectivity from dorsal attention network regions to salience network regions was higher in VGPs and negatively correlated with decision response time for both groups. These results suggest that video game playing can enhance the top–down interaction to improve sensorimotor decision-making.

For decades, predicting response to the antidepressant medication has been a critical unmet need in depression treatment in clinic, and a technical challenge in depression research.

In this study, a recently developed functional brain network controllability (fBNC) analysis approach was employed to identify the antidepressant treatment responders and nonresponders from depression patients at the pretreatment period. The fBNC, which captures the ability of brain regions to guide the brain's behavior from an initial state to a desired state with suitable choice of inputs, may provide valuable features for antidepressant response prediction. The performance of prediction was evaluated using resting-state functional magnetic resonance imaging data collected from a 6-week longitudinal clinical trial with escitalopram in treating unmedicated depression patients (n = 20). Treatment outcomes were assessed using the Hamilton Depression Rating Scale (HAMD) scores. Patients were considered as the treatment responders if their post-treatment HAMD scores were decreased by 50% or more at 6 weeks post-treatment.

Results showed significantly larger global average controllability and lower global modal controllability, greater regional average controllability, and smaller regional modal controllability of default mode network in treatment responders compared with the treatment nonresponders at the pretreatment period. By performing optimal control analysis, our results showed no significant difference of the neuromodulation effects between the treatment responders and nonresponders.

Our results suggest that the fBNC measures may be utilized as novel biomarkers to predict antidepressant response on depression and provide theoretical support to employ neuromodulation for treating antidepressant nonresponders.

In this study, by employing the novel functional brain controllability analysis on top of the brain connectivity network, we identified a set of biomarkers to identify the groups of depressive patients who responded to the antidepressant treatments from those who did not. We further provided the theoretical support to utilize neuromodulation for treating antidepressant nonresponders. These findings have clinical implications as accurate identification of antidepressant treatment response before starting the treatment may reduce patients' suffering and costs and increase the treatment outcomes by adjusting and personalizing the treatment protocol.