Packages for fMRI Analysis

Date: December 8, 2023 11:50 AM —

Table of Contents

  1. Packages for fMRI Analysis
    1. Preprocessing:
    2. Functional Connectivity Analysis:
    3. Statistical Analysis:
    4. Post-processing and Visualization:

Table of Tools for fMRI Analysis (by Sid Chopra, 2023)

Table by Kaley:

Preprocessing:

  • Reasons for distortion
    • motion artifacts
    • dropout (magnetic field not-consistent)
      • can somewhat correct using a field map
    • ghosting / electrical signal distortion
  • Slice Timing Correction: Correct for differences in acquisition time between slices.
    • Data are collected 1 slice at a time
      • can make diff areas collected many seconds apart
  • Motion Correction: Correct for head motion during the scan.
    • methods:
      • ICA -within scan effects of head-moving etc.
  • Spatial Normalization: Transform the images to a standardized anatomical space.
  • Spatial Smoothing: Apply a spatial filter to improve the signal-to-noise ratio.

Functional Connectivity Analysis:

  • [Seed Correlation Analysis](https://fcp-indi.github.io/docs/latest/user/sca) (aka contrast values or parameter estimates)
    • A region/voxel activation weights get extracted (eg beta-weights, or other HR-adjusted activation measure) and correlated:
      • to a specific target region (between subjects)
        • diff regions activate more, lots of 3rd factor compounding variables… this isn’t enough to = connectivity
        • neg correlation might be more sig?
      • across the brain (within subjects)
        • most correlations here will be activation induced, see below vv

  • how to take out motion-activation, hemodynamic fluctuations and other activation-induced correlations?

    methods:

    • extract + concatenate timepoints, then remove first 6 seconds to adjust for HDF
    • nuisance model
    • beta-series correlation
      • find the size of evoked response, and then correlate all the trials — fit a model using a separate regressor for each trial.
    • PPI
      • GLM + interactions between seed region x time and seed region x time x task
    • Multivariate Decomposition:
      • decomposing matrix into separate compontents → find active networks (aka “Matrix factorization” methods)
        • will find groups even if there aren’t groups

      Types of MFA:

      • Principal Components Analysis
        • assumption: groups gaussian & orthogonal
        • first group: direction thru data w/ most amt variance. 2nd group = second amt variace
        • pros: simple + easy to implement
        • cons: looking for orthogonal groups & will find even if they’re not there, only sensitive to gaussian distribution signals
        • may be used for data reduction (only use first few principal components)
        • method: format data into 2 dimensional matrix (voxels in one, timepoints in other), run
      • Independent Components Analysis
  • ROIs + Atlas

Statistical Analysis:

  • General Linear Model (GLM): Model the hemodynamic response function for each voxel.
  • Group Analysis: Combine data across subjects for group-level inferences.
  • Correction for Multiple Comparisons: Adjust p-values for multiple tests to control for false positives.

Post-processing and Visualization:

  • Result Visualization: Visualize activation maps, connectivity matrices, or other relevant outputs.
  • Data Interpretation: Interpret the results in the context of your research question.

Tools

FSL* (Carrisa)

SPM* (Carrisa)

AFNI* (Carrisa)

We generally use the HCP preprocessing pipelines, which also implements a combination of the above and the whole pipeline is implemented either via shell scripting (linux & bash seems to be the vast preference), python, or MATLAB. See here:  https://github.com/Washington-University/HCPpipelines

May be accessed via:

  • fMRI Prep
  • Nipype
  • FSL/SPM/Freesurfer Directly
  • Via Jupyter notebook pushing BASH to these programs
  • shell scripting files (.sh)

Packages/methods from:

FSL*

SPM*

AFNI*

ANTs

FreeSurfer

MRTrix

May be accessed via:

  • fMRI Prep
  • Nipype
  • FSL/SPM/Freesurfer Directly
  • Via Jupyter notebook pushing BASH to these programs
  • shell scripting files (.sh)

Python Data Science Packages (Sklearn, Scipy) (Carrisa)

May be accessed via:

  • fMRI Prep
  • Nipype
  • FSL/SPM/Freesurfer Directly
  • Via Jupyter notebook pushing BASH to these programs
  • shell scripting files (.sh)

Find tools via: https://sidchop.shinyapps.io/braincode_selector/

HCP Workbench

FSL Eyes

Pysurfer

Nilearn

Surplt

HCPUtils

May be accessed via:

  • HCP
  • fMRI Prep
  • Nipype
  • FSL/SPM/Freesurfer Directly
  • Via Jupyter notebook pushing BASH to these programs
  • shell scripting files (.sh)

  • Shell Scripting

    linux & bash (majority of lab)

  • GUI/IDE

    Jupyter Lab (Carrisa)

    R (half the lab)

    Nipype & C-PAC

  • Preprocessing

    FSL* (Carrisa)

    SPM* (Carrisa)

    AFNI* (Carrisa)

    We generally use the HCP preprocessing pipelines, which also implements a combination of the above and the whole pipeline is implemented either via shell scripting (linux & bash seems to be the vast preference), python, or MATLAB. See here:  https://github.com/Washington-University/HCPpipelines

  • Functional Connectivity Analysis

    ANTs

    FreeSurfer

    MRTrix

  • Statistical Analysis

    Tracula

    Python Data Science Packages (Sklearn, Scipy) (Carrisa)

  • Post-processing and Visualization

  • Diffusion MRI

    = measures the diffusion (movement direction & magnitude) of water molecules in the brain; primary method of studying macroscale brain activity

    works because axons = water barriers, so the water must flow along the axon

  • Voxel

    A point in space and time in a 3D region

  • hemodynamic response (HR)

    hemodynamic fluctuation in the BOLD response, need to control it out to get real measures of activation

  • Region of Interest (ROI)

    Define specific brain regions or regions of interest for connectivity analysis.

  • Seed-Based Analysis (SCA)

    Examine connectivity patterns from a seed region.

  • Independent Component Analysis (ICA)

    Identify independent components representing functional networks.

  • nuisance model

    method to remove unwanted activation-based correlation. basically fitting a model that explains the variation from the task and the activation. then control based on that model, and analyze other regions for their residuals

  • beta series model

    response to each event estimated w/ a separate regressor for each trial

    correlation between these regressors shows correlations between-trials of diff regions

    *need 8-10secs between trials to control for HR

  • Psychophysiological Interaction (PPI)

    models how activity in seed region is modulated by some other factor (task)

  • General Linear Model (GLM)

    estimates the coefficient of the task interaction over the timecourse

  • deconvolution

    most likely neuronal signal that would have given rise to that fMRI signal

  • principal components analysis (PCA)
    • PCA is a method for reexpressing a dataset in terms of a set of components that are uncorrelated,

  • matrix factorization analysis (MFA)

    decomposing matrix into separate compontents → find active networks

  • orthogonal

    uncorrelated

  • large-scale dynamical circuit model:

    computational framework that aims to simulate and understand the complex interactions and dynamics of neural circuits on a large scale (ie many circuits and neurons)

  • dynamic

    time dynamics, patterns of activity over time

  • dynamic connectivity

    signal correlation between 2 regions over time

  • stochastic

    random

  • neuronal dynamics

    changes in electrical and chemical activity of neurons and network as a whole

  • synchrony:

    simultaneous action / occurrence

  • spontaneous brain activity aka resting state analysis

    analyzing which areas are firing simultaneously/connectedly while the body is at rest in the fMRI scanner

  • local circuit properties

    average properties of the neural circuitry in different regions of the brain. ie synaptic strength, firing rate, and other anatomical and functional properties

  • index of

    serve as a representative, proportional or indicative measure of another property

  • principal resting-state functional connectivity (FC) gradient

    the gradual change or variation in functional connectivity across the brain

    strength observed between different brain regions

  • parcellation

    division of something into subsections

  • temporal resolution

    how accurate it is to real-time— for neuroimaging, measuring every millisecond = high temporal res (EEG is this), measuring every 2 seconds = low (fMRI is this)

  • Pearson correlation

    Common correlation metric to measure how correlated two data series are

  • sliding window correlation (SWC) analysis

    approach for evaluating dynamic functional connectivity

    computes a succession of pairwise correlation matrices using the time series from a given parcellation of brain regions

    method:

    1. define window size (aka time points within the whole scan, aka temporal resolution)
    2. for each time point:
      1. create correlation matrix for each brain regions to all of the others (correlation measure like Pearson correlation
      2. stack onto all other matrices
    3. → creates set of matrices = connectivity over time (aka dynamic connectivity)
    4. compute statistics for this set of matrices

  • tensor

    multilayer matrix— matrices tend to have only 2 dimensions, X and Y, but a tensor can have infinite dimensions, to build more and more metadata and higher-order analysis into the nodes and links characterized by the connections matrix

  • functional connectivity vs diffusion MRI

    Diffusion MRI (dMRI) and fcMRI have recently emerged as promising tools for mapping the connectivity of the human brain, each with distinct strengths and weaknesses. dMRI measures the diffusion of water, thus allowing direct noninvasive mapping of white matter pathways (Basser et al. 1994). However, dMRI is presently limited to resolving major fiber tracts. By contrast, fcMRI measures intrinsic functional correlations between brain regions (Biswal et al. 1995) and is sensitive to coupling between distributed as well as adjacent brain areas (e.g., see Sepulcre et al. 2010 for discussion). Although not a direct measure of anatomical connectivity, the functional couplings detected by fcMRI are sufficiently constrained by anatomy to provide insights into properties of circuit organization (for reviews, see Fox and Raichle 2007; Van Dijk et al. 2010). When describing these correlations, we use the term functional connectivity as coined by Karl Friston (1994) to denote “temporal correlations between remote neurophysiological events” for which the causal relation is undetermined.

    There are important limitations of fcMRI, including sensitivity to indirect anatomical connectivity and functional coupling that changes in response to recent experience and the current task being engaged (Buckner 2010). For these reasons, some discussions of fcMRI have emphasized that intrinsic activity measured by fcMRI reflects the prior history of activity through brain systems and not simply static anatomical connectivity (Power et al. 2010). fcMRI also does not presently provide information about whether connections are feedforward (ascending) or feedback (descending). These limitations constrain how analyses are conducted and results can be interpreted.

  • FreeSurfer

    FreeSurfer is software that constitutes a suite of automated algorithms for reconstructing accurate surface mesh representations of the cortex from individual subjects’ T1 images (Fig. 1, B and C) and the overlay of fMRI on the surfaces for group analysis

  • silhouette

    the silhouette of a data point measures the similarity of the data point to other data points of the same cluster compared with data points belonging to the next closest cluster

  • large-scale dynamical circuit model:

    computational framework that aims to simulate and understand the complex interactions and dynamics of neural circuits on a large scale (ie many circuits and neurons)

  • dynamic

    time dynamics, patterns of activity over time

  • dynamic connectivity

    signal correlation between 2 regions over time

  • stochastic

    random

  • neuronal dynamics

    changes in electrical and chemical activity of neurons and network as a whole

  • synchrony:

    simultaneous action / occurrence

  • spontaneous brain activity aka resting state analysis

    analyzing which areas are firing simultaneously/connectedly while the body is at rest in the fMRI scanner

  • local circuit properties

    average properties of the neural circuitry in different regions of the brain. ie synaptic strength, firing rate, and other anatomical and functional properties

  • index of

    serve as a representative, proportional or indicative measure of another property

  • principal resting-state functional connectivity (FC) gradient

    the gradual change or variation in functional connectivity across the brain

    strength observed between different brain regions

  • parcellation

    division of something into subsections

  • temporal resolution

    how accurate it is to real-time— for neuroimaging, measuring every millisecond = high temporal res (EEG is this), measuring every 2 seconds = low (fMRI is this)

  • Pearson correlation

    Common correlation metric to measure how correlated two data series are

  • sliding window correlation (SWC) analysis

    approach for evaluating dynamic functional connectivity

    computes a succession of pairwise correlation matrices using the time series from a given parcellation of brain regions

    method:

    1. define window size (aka time points within the whole scan, aka temporal resolution)
    2. for each time point:
      1. create correlation matrix for each brain regions to all of the others (correlation measure like Pearson correlation
      2. stack onto all other matrices
    3. → creates set of matrices = connectivity over time (aka dynamic connectivity)
    4. compute statistics for this set of matrices

  • tensor

    multilayer matrix— matrices tend to have only 2 dimensions, X and Y, but a tensor can have infinite dimensions, to build more and more metadata and higher-order analysis into the nodes and links characterized by the connections matrix

  • tractography

    involves defining white matter connections at the macroanatomical scale, which permits the measurement of connectivity strengths by counting the streamlines that link each pair of regions. Streamlines are constructed to pass through multiple adjacent voxels in DTI data, when the principal diffusion tensor per voxel aligns well with some of its direct neighbors (9). Tractography therefore produces subject-specific measures of regional interconnectivity that are ideally suited for brain network-level analysis.

  • Diffusion tensor imaging (DTI)
    • enables in vivo noninvasive study of white matter in the brain (6). This technique characterizes the diffusion of water molecules, which occurs preferentially in parallel to nerve fibers due to constraints imposed by axonal membranes and myelin sheaths (7). Metrics commonly derived from DTI, such as fractional anisotropy or mean diffusivity, reflect white matter microstructure and can index its integrity (7, 8).

  • MNI Space

    defines the boundaries around the brain, expressed in millimeters, from a set origin

  • White Matter

    White matter structures are largely composed of axonal fibres and myelin sheaths. Communication of neurons with far-away other neurons

  • Grey Matter (Gray)

    Pyramidal and other types of neuronal cell bodies, and their dendrites. Communication of neurons and close-by neighbors

  • Diffusion tensor imaging (DTI)

    = model for parsing diffusion MRI data, the measurements acquired for the different directions are put into a 3x3 matrix called a tensor, which can then be defined by eigenvectors and eigenvalues.

  • a priori approach

    have them do a task, that area that lights up = the area related to that task

    • not always translatable to differrent modalities

  • field map

    uses a image at two separate echo times to compute the difference in the magnetic field, and tehn apply that to the area with dropout, to estimate what the og reading should have looked like

  • independent component analysis (ICA)

    “finds a set of components that are independent of one another” ex: boundaries identified in the brain through the analysis of functionally specialized areas

  • validated parcellation scheme

    research-based mapping of functional sections of the brain