In order to be able to compare these fundamental properties across data structures, FieldTrip defines two fields in the geometrical data xxx pertaining to the interpretation of the geometrical units, xxx.unit, and to the interpretation of the coordinate system in which the coordinates are expressed, xxx.coordsys.

To compare the morphological similarity between the warped template, the original MRI and the original template were processed with Freesurfer (Dale et al., 1999) to get summary statistics of the surface area, grey matter volume, average cortical thickness, and mean curvature within cortical parcels based on the anatomical labels provided by Freesurfer automatic labelling (Destrieux et al., 2010).

However, as we still expect some correlation between morphological features, we rationalised that the similarity between the warped template and the original MRI should not be greater than the similarity between the warped template and the original template

2.3.5. Anatomical MRI features

The source spaces consisted of an evenly spaced 6 × 6 × 6 mm grid covering the entire volume defined in standard MNI coordinates.

2.3.1. Dipole analysis

SM

SIRP

AOD1

Review of multimodal fusion analysis in psychiatric disorders

Review of multivariate, multimodal fusion models

Moreover, the most recent parallel group ICA + ICA is able to deal with 4D fMRI and 3D sMRI features simultaneously (Qi et al., 2019).

Blind methods that have been developed for raw fMRI data (4D data),

Blind methods that

Overview of the current popular multimodal fusion approaches. The fusion models with model-driven (green), unsupervised (blue), and semi-supervised (orange) data-driven learning are listed in different categories, in which the models that can deal with 4D fMRI data are highlighted in yellow.

By contrast, emerging fusion approaches have been developed to directly handle the first-level 4D fMRI d

ifferent brain imaging data types are intrinsically dissimilar, making it difficult to analyze them together without making several assumption

All feature extraction, analyses

tfMRI z-scored COPEs

blind source separation

cortical thickness, resting state functional connectivity correlations, n-back working memory task COPE in the 2-back condition, and relational processing task COPE in the relational condition

Runs were then concatenated, and mean timeseries were extracted from a known cortical surface parcellation scheme (Gordon, Laumann et al. 2016), with the addition of several parcels from the cerebellum (Culbreth, Kandala et al. 2016) and subcortical regions defined by Freesurfer. Functional connectivity matrices were then computed as the Pearson correlation between all parcels. Given that these matrices are symmetric, only the lower triangle of each subject's correlation matrix was included in downstream analyses in order to avoid the inclusion of redundant data.

Next, the second pipeline processed participant data through a customized version of Freesurfer to generate subject-specific brain segmentations and parcellations that take advantage of the HCP's high-resolution structural data. Finally, the third and final pipeline converted Freesurfer output files into NIFTI, CIFTI, and GIFTI formats, as well as registered data to several different surface meshes, including the 32k surface mesh that was used as the standard space for all downstream analyses in the present report. The present study used Freesurfer-determined measures of cortical thickness at every vertex in the 32k surface mesh as the sMRI measure.

tfMRI data were analyzed using FSL to generate subject-specific spatial map COPEs corresponding to the desired task condition. In the present study, we used the working memory task: 2-back condition contrast; and the relational processing task: relational condition contrast (tasks described above in behavioral).

All imaging data were collected and pre-processed as part of the HCP

For each participant, we computed a composite measure of cognitive control as the summed z-score of four behavioral metrics, with each of these tasks described

While there are numerous benefits to mCCA+jICA, including the ability to detect both strong and weak links across modalities, its computational tractability and scalability to large datasets, and its robustness to noise and source estimation accuracy (Sui, Adali et al. 2012), there are several downsides. As an unsupervised learning method, mCCA+jICA is not guaranteed to identify all relevant patterns in the data. Furthermore, while there are methods for determining key model parameters (see supplementary methods for one such approach), there is no single correct choice and analysis results may vary with these parameters.

The independent analyses identified a component in each cohort that was highly similar to the other and significantly correlated with cognitive control performance.

in cognitive control, including dorsal prefrontal cortex, and frontal parietal and cingulo-opercular brain networks

Multimodal neuroimaging relationships in progressive supranuclear palsy

DTI images (fractional anisotropy (FA) and mean diffusivity (MD)) were processed as described in a previous study[17] and the median DTI scalar values (FA, MD) were measured within all of the specific white matter tract ROIs (27 in total, excluding lobar white matter regions) in the Johns Hopkins University atlas (Suppleme

Tissue volume (grey and white matter) was calculated in same set of ROIs and normalized with respect to the total intracranial volume.

Median flortaucipir PET value in each ROI was divided by median uptake in cerebellar crus grey matter to create standard uptake value ratios (SUVR)

-wise comparison was performed between patients and controls. Sparse canonical correlations analysis was applied on regional measurements of flortaucipir uptake, tissue volume, fractional anisotropy and mean diffusivity of the PSP population.

SCCA was applied to investigate multivariate relationships between flortaucipir SUVR, tissue volume, DTI-FA and DTI-MD regional measurements within the PSP group

SPM12 was used to perform voxel-wise multiple regression analyses that assessed differences in flortaucipir PET uptake, MRI grey and white matter, DTI-FA and DTI-MD between PSP and controls.

Thirty-four patients with progressive supranuclear palsy and 29 normal controls underwent flortaucipir PET, MRI and DTI.

Progressive supranuclear palsy is characterized primarily by 4R tau inclusions, atrophy in the brainstem and basal ganglia, and neurodegeneration along the dentatorubrothalamic tract, which are measurable in vivo using flortaucipir PET, T1-weighted MRI, and MRI with diffusion tensor imaging (DTI).

SCCA was applied using the penalized multivariate analysis (PMA) R package (Witten et al., 2009) to investigate multivariate relationships between all the combinations of modalities (six possible pairs). In each analysis, a lasso penalty of 0.3 for PET and MRI data and 0.4 for DTI data was assigned to achieve the desired level of sparsity (Adams et al., 2018),

for all the imaging modalities to identify alterations in tau deposition, gray matter volume, glucose metabolism, and fractional anisotropy

n other words, it simultaneously finds the canonical weight vectors in each data that maximize the correlation of the projections of each ventricle measurement and cortical thickness input data onto their canonical weight vectors.

sCCA attempts to find relationships between two sets of multi-voxel or multi-vertex dimensional W-scores of DTI measures (i.e., FA,DA, DR) and cortical thickness independently obtained from DTI and T1 imaging modalities.

jICA assumes that two or more features share the same mixing coefficient matrix and maximizes the independence among joint components. It is suitable for examining the common connection between features.mCCA maximizes the inter-subject covariation across two sets of features and generates two linked variables, one from each dataset, i.e., canonical variants (CVs); which correlate each other only on the same indices (rows) and their corresponding correlation values are called canonical correlation coefficients (CCC). This scheme allows for common as well as distinct level of connection between two features, but the identified CVs may not be sufficiently sparse and the performance might suffer when the canonical correlation coefficients are not sufficiently distinct.Parallel ICA and CC-ICA both need prior information to solve a constrained optimization problem. They are semi-blind source separation approaches.PLS is based on the definition of a linear relationship between a dependent variable and predictor variables, and hence the goal is to determine which aspect of a set of observations (e.g., imaging data) are related directly to another set of data (e.g., experimental design, behavior). Even though PLS has some similarity to CCA in that it also finds between-set connections, PLS is based on the definition of a dependency such that one dataset is dependent on another in one aspect, e.g., the temporal signatures of the concurrent FMRI and EEG (Martinez-Montes et al., 2004). Thus, PLS is not suitable for our case, since the two tasks SM and SBP are temporally distinct so that we cannot define or assume such a dependency.

3.4. Data acquisitionFMRI data of 14 subjects (7 aMCI subjects and 7 normal controls) were acquired with Institutional Review Board approval on a 3T GE HDx MRI scanner equipped with an 8-channel head coil. The subjects in the two groups were matched by age, education and right-handedness. Acquisition parameters for the EPI sequence were: TR/TE=2000 ms/30 ms, parallel imaging factor=2, slices=25 (coronal oblique, perpendicular to the long axis of hippocampus), slice thickness/gap = 4.0 mm/1.0 mm, 288 time frames (total scan duration 9.6 min), in plane matrix 96 × 96 voxels, FOV=220 mm. The fMRI volumes were interpolated to have an isotropic voxel size of 2 mm × 2 mm × 2 mm. A conventional structural T1-weighted image (0.43 mm × 0.43 mm × 1 mm) and a standard T2-weighted image (coplanar to the EPI) but with higher resolution (0.43 mm × 0.43 mm × 2.5 mm) were also acquired.An episodic memory task was performed to obtain fMRI data for each subject. Resting-state data with eyes closed were collected with the same acquisition parameters. The episodic memory task contained visual stimuli which show a novel face paired with an occupation. The entire task consisted of six periods of encoding, distraction, recognition and brief instructions to remind subjects of the task ahead. Specifically, during the encoding task, the subject was asked to memorize 7 faces paired with occupations, displayed in sequential order for a duration of 3s each and 21s in total. A distraction task (duration 11s) then followed each encoding task, where the subject was instructed to press the right or left button as fast as possible when the letter “Y” or “N” randomly appeared on the screen (right button for “Y” and left button for “N”). The recognition task consisted of fourteen stimuli, half novel and half identical to the stimuli seen in the previous encoding task. The subject was instructed to press the right button when the stimulus was previously shown and the left button when the stimulus was new. Scan duration was 9 min 36 s, and 288 time frames were collected. The design matrix X was constructed by convolving the task design consisting of 4 regressors for Instruction, Encoding, Distraction and Recognition (see Fig. 2) with the canonical hemodynamic response function.

Exploration of Neural Activity under Cognitive Reappraisal Using Simultaneous EEG-fMRI Data and Kernel Canonical Correlation Analysis

choices are further made based on the number of observations and number of features within each variable, known prior knowledge about the variable, such as feature structures, and specific questions of interest for research studies

and discover covariated patterns among more than two variables (dashed orange box)

Uncover common patterns across multiple modalities Multiset CCASparse multiset CCA Multiset constrained CCADeep CCA

5.5. AbstractTo summarize, conventional CCA uncovers joint multivariate linear relationships between two modalities and can be quickly and easily applied. In neuroscience research, due to the existing multiple modalities and nonlinear intermodality relationships, multiset CCA and nonlinear CCA have their own advantages when applied accordingly to appropriate variables. Constraints can be applied in these three methods to stabilize results, remove noninformative features, and produce supervised meaningful results. However, optimization expertise and prior knowledge about the data are required to select the appropriate constraints.

TABLE 6Advantages and limitations of each CCA‐related technique

As a supervised technique, multiset CCA can be applied to uncover covariated patterns across multiple variables of special interest.

Multiset CCA with reference

Sparse multiset CCA

Multiset CCA has also been applied to group analysis

TABLE 4Multiset CCA applications

.4. Multiset CCA: More than two modalitie

Constrained CCA penalizes canonical coefficients u1 and u2 to satisfy certain requirements and more specifically, to avoid overfitting and unstable results caused by insufficient observations in Y1 or Y2

Practical considerations and data reduction steps As we stated in Section 3.1, only if numbers of observations are more than numbers of features in both Y1 and Y2, that is, N ≫ pk, k = 1, 2, conventional CCA can produce statistically stable and meaningful results. However, in neuroscience applications, this requirement is not always fullfilled, especially when Y1 or Y2 represents brain activities where each brain voxel is considered a feature individually. In this case, any feature can be picked up and learned by the CCA process and directly applying Equation (1) to two sets will produce overfitted and unstable results. Therefore, additional data‐reduction steps applied before CCA or constraints incorporated in the CCA algorithm are necessary to avoid overfitting in CCA applications. In this section, we focus on data reduction steps applied before conventional CCA.

Here, we denote Yk∈RN×pk,k=1,2<math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="nlm-math-3"><msub><mi mathvariant="bold-italic">Y</mi><mi>k</mi></msub><mo>∈</mo><msup><mi mathvariant="double-struck">R</mi><mrow><mi>N</mi><mo>×</mo><msub><mi>p</mi><mi>k</mi></msub></mrow></msup><mo>,</mo><mi>k</mi><mo>=</mo><mn>1</mn><mo>,</mo><mn>2</mn></math> as collected samples of these two variables, where N represents the number of observations (samples) and pk, k = 1, 2 represent the number of features in each variable. CCA determines the canonical coefficients u1∈Rp1×1<math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="nlm-math-4"><msub><mi mathvariant="bold-italic">u</mi><mn>1</mn></msub><mo>∈</mo><msup><mi mathvariant="double-struck">R</mi><mrow><msub><mi>p</mi><mn>1</mn></msub><mo>×</mo><mn>1</mn></mrow></msup></math> and u2∈Rp2×1<math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="nlm-math-5"><msub><mi mathvariant="bold-italic">u</mi><mn>2</mn></msub><mo>∈</mo><msup><mi mathvariant="double-struck">R</mi><mrow><msub><mi>p</mi><mn>2</mn></msub><mo>×</mo><mn>1</mn></mrow></msup></math> for Y1 and Y2, respectively, by maximizing the correlation between Y1u1 and Y2u2:

CCA is designed to maximize the correlation between two latent variables

Multiple CCA variants, including kernel CCA, constrained CCA, deep CCA, and multiset CCA, also have been applied in neuroscience research.

4.2.2. Constraints in CCA algorithm: Constrained CCA to stabilize results

Prior knowledge about Y1 and Y2 might also be available in neuroscience data

4.2.1. Constraints in CCA algorithms: Sparse CCA to remove noninformative features

Kernel CCA: Focusing on a nonlinear relationship between two modalities

The most commonly u

Canonical granger causality

Denoising neuroscience data

Brain activation in response to task stimuli

In normal healthy subjects, using

Combine phenotypes and brain activities To date, the most common CCA application in neuroscience is to find joint multivariate linear associations between phenotypic features and neurobiological activities. Phenotypic features usually include one or more measurements from demographics, genetic information, behavioral measurements, clinical symptoms, and performances of neuropsychological tests. Neurobiological activities are generally summarized with brain structural measurements, functional activations during specific tasks, both static and dynamic resting‐state functional connectivity measurements, network topological measurements, and electrophysiological recordings (Table ​(Table11).

3.2.5. Statistical inferences of CCA variants Nonparametric permutation tests have been widely performed in CCA variant techniques to determine the statistical significance of each canonical correlation value and the corresponding canonical coefficients. In these permutation tests, as we described in Section 3.1, observations of one variable are randomly shuffled (Y1 becomes Y1ˆ<math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="nlm-math-74"><mover accent="true"><msub><mi mathvariant="bold-italic">Y</mi><mn>1</mn></msub><mo stretchy="true">^</mo></mover></math>), while observations of the other variable are kept intact (Y2 remains). This random shuffling is repeated multiple times (~5,000), and the exact same CCA variant technique is applied to each shuffled data. The obtained canonical correlation values from these randomly shuffled data form the null distribution. Statistical significances (p‐values) of true canonical correlation values are determined by comparing true values to this null distribution.Besides permutation tests, a null distribution can also be built by creating null data input to CCA variant techniques. The null data are usually generated based on the physical properties of input variables. For instance, when applying CCA‐variant technique to link task fMRI data and the task stimuli, the null data of task fMRI can be obtained by applying wavelet‐resampling to resting‐state fMRI data (Breakspear, Brammer, Bullmore, Das, & Williams, 2004; Zhuang et al., 2017). The null hypothesis here is that task fMRI data are not multivariately correlated with task stimuli, and the wavelet resampled resting‐state fMRI data fits the requirements of the null data in this case.

Tensor CCA for multiset data Another way to handle input variables with high‐dimensional feature spaces is to generalize conventional CCA by analyzing constructed covariance tensors (Luo, Tao, Ramamohanarao, Xu, & Wen, 2015). This method requires random variables to be vectorized and is similar to multiset CCA since both of them deal with more than two input modalities. The differences between tensor CCA and multiset CCA in this case lie in that tensor CCA constructs a high‐order covariance tensor for all input variables (Luo et al., 2015), whereas multiset CCA finds pair‐wise covariance matrices. In addition, tensor CCA (Luo et al., 2015) does not maximize the pairwise correlation as in multiset CCA; instead, it directly maximizes the correlation over all canonical variables,

A technical review of canonical correlation analysis for neuroscience applications

Multiset CCA with constraints In constrained multiset CCA, penalty terms can be added to each ui individually. Here we give examples of two commonly incorporated constraints in multiset CCA: sparse multiset CCA and multiset CCA with reference.

Besides maximizing SUMCOR, Kettenring (1971) summarizes four other possible objective functions in multiset CCA optimization: (a) SSQCOR, maximizing sum of squared pairwise correlations ∑Ki,j∑ˆ2ij<math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="nlm-math-36"><msubsup><mrow><msubsup><mo>∑</mo><mrow><mi>i</mi><mo>,</mo><mi>j</mi></mrow><mi>K</mi></msubsup><mover accent="true"><mi mathvariant="bold">∑</mi><mo stretchy="true">^</mo></mover></mrow><mi mathvariant="italic">ij</mi><mn>2</mn></msubsup></math>; (b) MAXVAR, maximizing largest eigenvalue of correlation matrix λmax(∑ˆ)<math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="nlm-math-37"><msub><mi>λ</mi><mi>max</mi></msub><mfenced open="(" close=")"><mover accent="true"><mi mathvariant="bold">∑</mi><mo stretchy="true">^</mo></mover></mfenced></math>; (c) MINVAR, minimizing smallest eigenvalue of correlation matrix λmin(∑ˆ)<math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="nlm-math-38"><msub><mi>λ</mi><mi>min</mi></msub><mfenced open="(" close=")"><mover accent="true"><mi mathvariant="bold">∑</mi><mo stretchy="true">^</mo></mover></mfenced></math>; and (d) GENVAR, minimizing the determinant of correlation matrix det(∑ˆ)<math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="nlm-math-39"><mi mathvariant="italic">det</mi><mfenced open="(" close=")"><mover accent="true"><mi mathvariant="bold">∑</mi><mo stretchy="true">^</mo></mover></mfenced></math>. In practice, SUMCOR multiset CCA is most commonly used followed by MAXVAR and SSQCOR multiset CCA.

Constraints

3.2.3. Multiset CCA

Kernel and deep CCA are designed to uncover nonlinear correlations between modalities by projecting the original variables to new nonlinear feature spaces. Multiset CCA is proposed to find multivariate associations among more than two modalities.

In the following sections, we present technical details (Section 3) and neuroscience applications for each category (Section 4). In Section 5, we discuss technical differences and summarize advantages and limitations of each CCA‐related technique. We finally provide an experimental example and guidance in Section 6 to researchers who are interested in applying multivariate CCA‐related techniques in their work.

among different modalities and avoids multiple correction steps, which would be more appropriate to disentangle joint relationship between modalities and guarantees full utilization of all common information.

Finally, a general guideline in how to select the most appropriate CCA‐related technique based on the properties of available data sets and particularly targeted neuroscience questions is provided

Therefore, in this study, a comprehensive review of CCA and its variant techniques is provided.

Canonical correlation analysis (CCA) is one of the powerful multivariate tools to jointly investigate relationships among multiple data sets, which can uncover disease or environmental effects in various modalities simultaneously and characterize changes during development, aging, and disease progressions comprehensively.

Multimodal CCA allows a different mixing matrix for each modality and is used to find a transformed coordinate system that maximizes inter-subject covariation across the two data sets

Neural responses to negative facial emotions: Sex differences in the correlates of individual anger and fear traits

Tensor network factorizations: Relationships between brain structural connectomes and traits

We find that semantic memory performance is influenced by the number of streamlines of the left superior longitudinal fasciculus III (SLF-III), and emotion perception performance is influenced by the number of streamlines of the right SLF-III. Additionally, we find that performance on both semantic memory & emotion perception is influenced by the FA of the left arcuate fasciculus (AF).

subcomponents of the language and ToM networks,

Performance on neuropsychological assessments of semantic memory (NIH Toolbox Picture Vocabulary Test, TPVT) and emotion perception (Penn Emotion Recognition Test, PERT)

White matter association tracts underlying language and theory of mind: An investigation of 809 brains from the Human Connectome Project

Relationship between BMI and alcohol consumption levels in decision making

Decision-making is often assessed using the Iowa Gambling Task (IGT), which requires inhibition of impulsive responses by factoring in uncertainty, reward, and punishment. Interpretation of IGT performance is challenging since several cognitive constructs are assessed simultaneously, including memory, reward sensitivity, and inhibitory control. Nonetheless, decision-making behaviors have been measured with high ecological validity [23, 24]

Gambling task functional MRI in Human connectome project (HCP) dataset was used to investigate neural activation differences associated with reward or punishment (a key component of decision-making behavior) in 418 individuals with obesity (high BMI) and without obesity (lean BMI) and either at high (HR) or low (LR) risk of AUD based on their alcohol drinking levels.

Callosal gradient map (left), which was computed as the weighted average of callosal connectivity across all percentage maps of the gradient with the 7 regions of the Wittelson parcellation superimposed.

F) Scatterplots showing the relation between callosal connectivity (left) or number of non-zero connectivity voxels (right) and the residual between the gradient corrected for the number of voxels in the respective gradient percentage map. By callosal connectivity, we mean interindividual probability of connectio

This resulted in 100 distinct gradient percentage maps, each one presenting a 1% range of the gradi

The gradient intensities range between −7.52 and +10.07. A central value of 0 denoted areas not defined along the gradient. For computational reasons, we transposed the gradient values by multiplying the minimum intensity value by −1 and adding the result to all values, thus shifting all values into positive values. We subsequently masked the results with a binarized map of the principal gradient.

D) Midsagittal display of callosal connectivity based on the tractograms for the 20th, 40th and 60th gradient percentages. Callosal areas with low probabilities (red) indicate that fewer participants showed connections to the particular gradient percentage

Percentage-specific tractograms

B) Examples for percentage maps (20–40–60%)

A) Visualisation of the principal gradient

In order to assess callosal connections to cortical regions along the principal gradient, we segregated voxels within the cortex based on their location along the principal gradient. We parcellated the principal gradient map into 100 units where each unit represents one percentage of the principal gradient

averaged map of the principal gradient

All 7T rsfMRI and diffusion-weighted data used in this study were based on the Human Connectome Project (HCP; Van Essen et al., 2013), which is publicly available and anonymized. For rsfMRI data, we used an averaged map of the principal gradient published in Margulies et al. (2016), which can be found under: https://neurovault.org/collections/1598/

Data acquisition

This study investigated crystallized, fluid, and general intelligence composite scores’ relationship with brain network differences during a working memory task as compared resting state.

to investigate how brain network changes between working memory and resting state relate to intelligence score

we extracted a set of the main topological network features, including global efficiency, degree, leverage centrality, modularity, and clustering coefficient

Each subject’s data included composite intelligence scores, and fMRI during resting state and a 2-back working memory task.

fMRI Data Acquisition

Function in the Human Connectome: Task-fMRI and Individual Differences in Behavior

Our choice of tfMRI tasks was driven by the following considerations.

Individual Differences in the Human Connectome Project

Indeed, an extensive empirical literature already provides evidence for impairments in both structural and functional connectivity in psychiatric disorders such as autism

As another example, developmental research is increasingly suggesting that maturation of functional and structural networks in the human brain underlies key aspects of cognitive and emotional development

For example, higher IQ among healthy adults is associated with shorter path length and higher global efficiency in measures of brain functional connectivity

. However, we know that there are important individual differences in such patterns of connectivity even among persons with no diagnosable neurological or psychiatric disorders, and there is increasing evidence that this variability is associated with alterations in cognitive and behavioral variables that constrain real world function

he HCP is using task-fMRI (tfMRI) to help delineate the relationships between individual differences in the neurobiological substrates of mental processing and both functional and structural connectivity

understanding the relationship between brain connectivity and human behavior.

However, we know that there are important individual differences in such patterns of connectivity, with evidence that this variability is associated with alterations in important cognitive and behavioral variables that affect real world function.

For both analysis streams

Visual Processing

Non-Toolbox Behavioral Measures

NIH Toolbox Behavioral Measures

Third, if our pilot analyses suggested that activation of a set of brain regions associated with a specific function could be identified within the context of another task, we did not include a separate task to isolate those regions. For example, our piloting included a task using point-light walkers (Antal et al., 2008) to assess regions associated with biological motion. However, our phase I results revealed that these same brain regions were also activated in the social cognition task that involved objects moving in biologically plausible ways. Thus, our final battery did not include a separate biological motion task.

data as validation for parcellation schemes used on the resting state connectivity data or diffusion data (e.g., do peaks fall in areas identified as low transition points between areal boundaries (Cohen et al., 2008; Nelson et al., 2010); 2) using peaks identified in the task data to subdivide regions identified in the resting state connectivity data (e.g., when there are different peaks from different tasks domains located within a larger “region” identified with resting state connectivity data); 3) examining whether boundaries of regional activations identified in the tfMRI data map to boundaries identified by other methods (e.g., rsfMRI, myelin maps, etc.); 4) examining whether parcellation results from task-based connectivity data correspond to results from resting state data or diffusion data; or 4) using peaks from task data as input to seed-based connectivity or tract tracing approaches. We are confident that other investigators will identify additional creative and innovative ways in which the tfMRI data can be used to help guide, validate and interpret the functional and structural connectivity data.

There are numerous ways in which the regions of activation identified in the tfMRI data could be used to facilitate the examination and interpretation of the functional and structural connectivity data.

to relate signatures of activation magnitude or location in key network nodes to individual differences in performance, psychometric measures, or other phenotypic traits.

allow a comparison of network connectivity in a task context to connectivity results generated using R-fMRI

help identify as many “nodes” as possible that can guide, validate and interpret the results of the connectivity analyses that will be conducted on resting state fMRI (R-fMRI), resting state MEG (R-MEG) and diffusion data;

To illustrate how these data might be used to examine the behavioral relevance of individual differences in functional or structural connectivity, investigators will be able to (for example) examine how variation in scores on the NIH Toolbox working memory task relates to variation in: 1) the amplitude of spontaneous resting-state fluctuations in timeseries associated with individual functional parcels from whole-brain parcellation; 2) connection strengths between network nodes (parcels), such as will be estimated via a) full or partial correlation matrices derived from the timeseries associated with whole-brain parcellation of rfMRI data, and/or b) probabilistic tractography estimated between different nodes from dMRI data; 3) ICA component spatial maps identified in the resting state data, or task based activation data during the working memory task; 4) connectivity metrics associated with specific regions of interest to working memory (e.g., superior parietal cortex, etc.), or 5) connectivity metrics associated with “hub” or “rich club” regions (Buckner et al., 2009; Collin et al., 2013; Harriger et al., 2012; van den Heuvel and Sporns, 2011). As another example, investigators will be able to examine how variation in personality variables such as extroversion or neuroticism relate to variation in the kinds of connectivity measures described above, including connectivity metrics associated with specific regions of interest to neuroticism or extroversion (e.g., amygdala, caudate, etc.).

To clean the data, all variables were tested for skewness using the R package e1071. Any variable with a skew above 1 or below −1 was transformed with a cubed root or squared, respectively. Then outliers were removed using the Outliers package from R.

H was significantly higher in movie than rest in the visual, somatomotor and dorsal attention networks, but was significantly lower during movie in the frontoparietal and default networks.

Previous fMRI studies have observed lower H during conventional tasks relative to resting state conditions, and shown that H is negatively correlated with task difficulty and novelty. To date, no study has investigated the fractal dynamics of BOLD signal during naturalistic conditions.

Fractal-Based Analysis of fMRI BOLD Signal During Naturalistic Viewing Conditions

a measure of the correlation structure in a signal (Eke et al., 2000), where H < 0.5 indicates anticorrelation in the signal, H = 0.5 indicates there is no correlation (it is a random white noise or walk), and H > 0.5 indicates positive correlation or the presence of long-memory in the proces

In monofractal (as opposed to multifractal) analysis, fractal signals are described by a single parameter, the Hurst exponent (H), that reflects the global scaling behavior of a syste

fractal analysis is a signal-based method that describes the correlation structure of a process across temporal scales (Eke et al., 2002

In contrast to previous work using conventional tasks, we found higher H values for movie relative to res

We performed fractal analysis on Human Connectome Project 7T fMRI data (n = 72, 41 females, mean age 29.46 ± 3.76 years) to compare H across movie-watching and rest.

Standard 7T MR session structure

Retinotopy task fMRI

Retinotopy stimuli and experimental details

Movie watching stimuli and experimental details

For analysis of the movie-watching data, members of the laboratory of Jack Gallant createdfeatures labels for the four movies used. Two types of labels were created: motion-energy labelsthat describe the low-level structural features of the movies, and semantic-category labels thatdescribe the high-level semantic features of the movies

A zip archive with two CSV files, 1) one detailing the sources of each clip and frames used ineach exerpt, and 2) one that gives the exact timings of movie clip block start, end, and durationblock,

Movie-watching fMRI

below. The “Pre_20140821_version” wasused for subjects collected from the beginning of Phase 2 scanning through August 21, 2014and the "Post_20140821_version", or v2, was used for subjects from August 22, 2014 to theend of 7T data collection.

Resting-state fMRI (rfMRI)

with eyes open with relaxed fixation on a projectedbright cross-hair on a dark background (and presented in a darkened room). Am

Task-evoked fMRI (tfMRI)

Conclusion

Connectivity-Based Topographical Changes of the Corpus Callosum During Aging

This study has several limitations. Firstly, due to the intrinsic limitation of DTI in the regions with complex fiber heterogeneity, we used the SS3T-CSD algorithm to estimate the fiber orientation distributions for whole-brain tractography. We applied the most used DTI measurements to study the aging pattern in the CC subregions. Advanced diffusion models with multi-shell protocol should be used to capture the microstructure (Raffelt et al., 2017). Secondly, the CC masks were generated by segmenting the T1 images and registered to the DWI images, the multi-modal registration is challenging because DTI is susceptible to both affine/linear (e.g., eddy-current and head motion) and non-linear echo planar image field distortions. The registration results in each step were all visually inspected and showed their accuracy. Finally, there were relatively small samples in groups from 21–50 years and 81–90 years, which may bias the results for our imaging measures. In this work, we tried to analyze as much as possible data from our center and therefore we combined two imaging datasets with identical imaging scanner and protocols. Further study should recruit more participants to explore the microstructural changes of the CC during the earlier lifespan.Go to:

Curve Fitting of Age-Related Diffusion Indices Changes of Corpus Callosum Subregions

Subregional Connection Probability Differences Between Age Groups

Differences in Diffusion Indices Between Corpus Callosum Subregions

Population-Based Probabilistic Corpus Callosum Connection Maps

Reproducibility of Probabilistic Corpus Callosum Connection Maps

ences in the diffusion indices between the CC subregions were compared by the repeated-measures analysis of variance (RM-ANOVA) with age, sex, education years and TIV being regresse

For each CC subregion, the connection probability differences among the seven age groups were compared using the analysis of covariance (ANCOVA) with the factors of sex, education years and TIV as covariates.

For each subregion, the quadratic curve was used to model the age-related changes in each diffusion indices (including FA, MD, RD, and AD) as follows:

Curve Fitting of Age-Related Diffusion Indices Changes of Corpus Callosum Subregions

Subregional Connection Probability of Age Groups

Construction of Population-Based Probabilistic Corpus Callosum Connection Maps

With the CC masks and functional atlases constrained, the callosal fibers that pass through the CC and connect each functional network were obtained

Construction of Individual Corpus Callosum Connection Maps

Whole Brain Tractography

DWIs were preprocessed using the FSL software

The T1w images

Image Acquisition and Preprocessing

Participants

In this study, we first presented the tractography-based segmentation of the CC that is related to functionally organized networks. Then, we investigate the age-related microstructural changes and connectivity of each callosal segment. To achieve this, we analyzed the dMRI from a large number of healthy subjects (n = 1,086) within a wide age range (21–90 years). Afterward, we used the diffusion tractography-based approach to track the CC fibers that interconnect functionally organized networks, including the visual, sensory-motor functional networks and higher-order association networks (Yeo et al., 2011). Then we parcellated the CC by the neural projections. The participants were divided into seven groups by age to compare the connection probability of each callosal segment among different age ranges. The diffusion indices of the CC segments were established to estimate the aging patterns of each CC segment and compare the microstructural differences among CC segments. We hypothesized that the callosal segments had different compositions and showed region-specific changes along the normal aging process.

Several tractography-based partitioning approaches have been proposed to divide the CC structure based on the reconstructed streamlines that pass through it and connect it to specific cortical terminations (Huang et al., 2005; Park et al., 2008; Chao et al., 2009). For example, Park et al. (2008) used diffusion tensor imaging (DTI) to present CC population connectivity maps according to 47 semi-automatically partitioned cortical subregions. Similarly, another study presented a cortical cytoarchitectural subregions-based parcellation of the CC using HARDI-based tractography (Chao et al., 2009). These studies and others show that tractography-based partitioning methods provide a more rational subdivision of the CC to link the association between the CC segments and cerebral subsystems concerning distinct functions of the human brain.

This study revealed the callosal subregions related to functional networks and uncovered an overall “anterior-to-posterior” region-specific changing trend during aging

STEAM &HARDI

We created some regions of interest (ROIs) from the JHU ICBM 81 white matter labels atlas included in FSL (Mori et al., 2005), visually checked the location of each ROI on each map

Upper limb motor rehabilitation impacts white matter microstructure in multiple sclerosis

Corpus Callosum Integrity Relates to Improvement of Upper-Extremity Function Following Intensive Rehabilitation in Children With Unilateral Spastic Cerebral Palsy

Regions of interest (ROIs) were determined using anatomical location (XYZ) through orientation-based color-coding maps (red for fibers with medial-lateral orientation).

Brain Plasticity following Intensive Bimanual Therapy in Children with Hemiparesis: Preliminary Evidence

A single ROI was used to extract the CC via a color coded midsagittal FA image [9, 25, 26].

Specifically, the use of individual fMRI, or atlas based fMRI if the former is not available, is recommended over the application of geometric subdivision schemes.

In 2019, in vivo histology at high resolution was used to suggest callosal partitioning. More specifically, by applying T1 relaxometry, Lee et al. (Citation2019Lee, B. Y., Zhu, X. H., Li, X., & Chen, W. (2019). High-resolution imaging of distinct human corpus callosum microstructure and topography of structural connectivity to cortices at high field. Brain Structure & Function, 224(2), 949–960. https://doi.org/10.1007/s00429-018-1804-0 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]) determined T1 values across the length of the CC, reflecting myelinated axonal size and myelin density.

Finally, it is noteworthy to highlight a newly available atlas that includes partitioning of the CC in seven subregions in vivo in humans (Radwan et al., Citation2022Radwan, A. M., Sunaert, S., Schilling, K., Descoteaux, M., Landman, B. A., Vandenbulcke, M., Theys, T., Dupont, P., & Emsell, L. (2022). An atlas of white matter anatomy, its variability, and reproducibility based on constrained spherical deconvolution of diffusion mri. NeuroImage, 254, 119029. https://doi.org/10.1016/j.neuroimage.2022.119029 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]). Segments refer to prefrontal, premotor, motor, sensory, parietal, occipital and temporal areas and are obtained through the application of probabilistic spherical deconvolution tractography.

n our recently published work (Gooijers et al., Citation2021Gooijers, J., De Luca, A., Zivari Adab, H., Leemans, A., Roebroeck, A., & Swinnen, S. P. (2021). Indices of callosal axonal density and radius from diffusion mri relate to upper and lower limb motor performance. NeuroImage, 241, 118433. https://doi.org/10.1016/j.neuroimage.2021.118433 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]), we have made significant steps in this direction. Specifically, we performed interhemispheric WM tractography based on seeds obtained from individual fMRI activations in response to hand and foot flexion-extension movements. In line with Domin and Lotze (Citation2019Domin, M., & Lotze, M. (2019). Parcellation of motor cortex-associated regions in the human corpus callosum on the basis of human connectome project data. Brain Structure & Function, 224(4), 1447–1455. https://doi.org/10.1007/s00429-019-01849-1 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]), although in subject-space, we revealed callosal hand and foot fibers in the posterior body with callosal hand fibers localized more inferior relative to callosal foot fibers (see Figure 1) (Gooijers et al., Citation2021Gooijers, J., De Luca, A., Zivari Adab, H., Leemans, A., Roebroeck, A., & Swinnen, S. P. (2021). Indices of callosal axonal density and radius from diffusion mri relate to upper and lower limb motor performance. NeuroImage, 241, 118433. https://doi.org/10.1016/j.neuroimage.2021.118433 [Crossref], [PubMed], [Web of Science ®], [Google Scholar]).

fMRI- and Atlas-Informed Callosal Tractography

The Role of the Corpus Callosum (Micro)Structure in Bimanual Coordination: A Literature Review Update

High-resolution Imaging of Distinct Human Corpus Callosum Microstructure and Topography of Structural Connectivity to Cortices at High Field

Comparison of individualized behavioral predictions across anatomical, diffusion and functional connectivity MRI

Structural and functional connectivity mapping of the human corpus callosum organization with white-matter functional networks

The functional parcellation of corpus callosum

Assessing the significance of overlap between DTI and RSFC results

We randomly permuted 10,000 times

Specifically, all callosal voxels were included in the simulation, assigning each callosal voxel a binary value. The proportion of voxels assigned a value of 1 was respectively determined by the number of voxels exceeding 15% threshold in DTI result and passing FWE correction (p < 0.05) in RSFC result

The structural parcellation of corpus callosum

averaged time series

Voxels in the corpus callosum were binarized and their value set at 1 if they were parts of the tracts in at least 15% of the

these fiber bundles corresponding to each WM-FN were accumulated across all subjects to establish a population-based probabilistic map of commissure tracts using Park's approach