Introduction & Data Modalities

Simon Vandekar

Welcome

  • Syllabus
  • There will be group projects
    • Most of the things you learn will come from the projects
  • Auditors – decide if you want to participate in the group projects.

Homework

  • Map the Box folder for the course to your own Box account and setup Box to sync on your laptop
  • Install FSL

Data repositories

Reproducible Brain Charts

Reproducible Brain Charts

  • We will use structural, resting state functional MRI (rs-fMRI), and “task” fMRI (t-fMRI)

ABCD Study

Software

Analysis software

  • Statistical Parametric Mapping (SPM; in MATLAB)
  • FSL (linux/terminal based) – Primary software for this course
  • AFNI
  • Neuroconductor – modular analysis packages in R

Getting setup with FSL

echo $FSLDIR
  • Visualize the FSL template image
  • Visualize my brain

Visualization software

Visualization is as important in image analysis as univariate data analysis (Always look at your data)

Brain anatomy and terminology

Anatomy and terminology

1introduction_data_modalities/figures/simon.nii.gz

Directory paths will be relative to the course directory on Box.

  • Gray Matter: Brain tissue composed mainly of neuronal cell bodies. It is involved in processing and interpreting information and is found in regions like the cortex and deep brain nuclei.
  • White Matter: Tissue made up of axons coated with myelin, which appears white. White matter connects regions of gray matter, allowing communication across the brain.
  • Ventricles: Through the middle of the brain filled with cerebrospinal fluid.

More brain anatomy and terminology

  • Cortex: The outer layer of the brain’s gray matter. It is highly folded and responsible for higher cognitive functions like perception, decision-making, and voluntary movement.
  • Subcortex: Brain regions located beneath the cortex, including structures like the thalamus, basal ganglia, and hippocampus. These areas are important for functions such as emotion, memory, and movement regulation.
  • Cerebellum: A separate structure located at the back of the brain, underneath the cortex. It plays a key role in coordinating movement, balance, and fine motor skills.

Even more brain anatomy and terminology

  • Lobes: The brain is divided into four main lobes — frontal, parietal, temporal, and occipital — each associated with specific functions.
  • Gyri (singular: gyrus): Ridges on the brain’s surface.
  • Sulci (singular: sulcus): Grooves between the gyri.
  • Anterior / Posterior / Superior / Inferior: Anatomical directions — front/back/top/bottom.

Neuroimaging data

  • Neuroimaging provide a way to study brain structure and function in vivo in humans
  • Magnetic resonance imaging (MRI) is noninvasive
  • “Possibly, the most promising technology to link human emotion, cognition, and behavior with the brain.” – me

What is an image?

  • A function from I(v): \mathbb{R}^3 \mapsto \mathbb{R}.
  • Numeric values in the image represent biological values.
  • Visually brighter pixels are higher values.
  • MRI are collected in the frequency domain, but few people analyze their data there

Medical imaging data formats

  • DICOM: Medical image format, which includes a lot of metadata in the image header. Raw output from MRI scanner. .dcm
  • NIfTI: File format for 3D or 4D neuroimaging data with less (but still a lot of) metadata in the image header. .nii or .nii.gz.

You don’t need to gunzip.nii.gz files, all software can open them directly.

Neuroimaging terminology

  • Voxel: A 3D pixel.
  • TR: Repetition time - time required to obtain an entire volume (typically for fMRI data).
  • TE: Echo time - time between application of a radiofrequency (RF) pulse and the peak of the signal (echo) generated.
  • EPI: Echo planar imaging - the type of imaging used to collect fMRI data.
  • T1-weighted / T2-weighted images: Types of MRI scans for different tissue properties. T1 show anatomy with good contrast for gray and white matter; T2 highlight fluid and often used to detect pathology.

Orientation terminology

  • Axial View: A horizontal slice of the brain, viewed from above (as if looking down from the top of the head).
  • Coronal View: A vertical slice viewed from the front (as if looking straight at the face).
  • Sagittal View: A vertical slice viewed from the side (as if looking at the brain in profile).

Image space

introduction_data_modalities/figures/MNI152_T1_2mm_brain.nii.gz
  • MNI Space / Talairach Space: Standardized brain coordinate systems used to align and compare brain scans across individuals.
  • Radiological versus neurological convention
    • Radiological - (left on the right)
    • Neurological - (right on the right)
    • Nice overview on this stuff
    • Radiological is more common

Atlases

Atlases define regions of the brain based on anatomical, microstructural, functional, or other features

  • Common Atlases
    • MNI

Data modalities

Structural imaging

Measures the anatomy of the brain

  • Derived from T1/T2 MRI scans
  • Structural measures
    • Volume, Cortical thickness, Surface area
  • Data structures
    • Images
    • Regions
    • Gray matter surface
    • Structural-based “similarity” networks

Freesurfer: structural analysis software

  • Cortical surface estimation and analysis
  • Now over 25 years old
  • Many tools now built on analyzing data on the surface

Fischl and Dale 2000

Freesurfer visualization

Diamond et al. 2020

Irimia et al. 2012

Functional imaging

Measures brain functioning (inferred by deoxygentated blood)

  • Blood oxygen level dependent (BOLD) activity
  • History of functional MRI (Bandettini 2012)
  • functional Magnetic Resonance Imaging (fMRI) are time-series data

Early fMRI Bandettini 1991

Types of fMRI data

  • task fMRI – acquired doing a task
  • resting state fMRI (rs-fMRI) – acquired doing nothing

Task fMRI overview

  • Measures brain activity during a task using BOLD signal.
    • Block design: repeated task periods alternating with baseline (e.g. 30s task, 30s rest).
    • Event-related design: timing randomized, short stimuli.
    • Naturalistic: E.g., viewing a movie.
  • Analysis fits a time-series model and compares conditions.
  • Data structures
    • Images, Regions, Gray matter surface

Task fMRI design illustrations

Block versus task designs Petersen & Dubis 2011

Notes on illustration

  • Known stimulus time series are convolved with an assumed hemodynamic response function (HRF)
  • HRF is well studied, often assumed, and likely inaccurate for many brain regions REFs
  • Hemodynamic (blood) response to stimuli is slow.
  • fMRI sampled every 1.5-3 seconds or so.
  • RBC fMRI parameters

RBC task data

Despicable Me

rs-fMRI

  • What it is? Correlation between fMRI time-series in the brain at rest (doing nothing).
  • History: Biswal 1995 first study to do this
  • Now, things are quite a bit more sophisticated
  • Data structures
    • Images
    • Regions
    • Gray matter surface
    • Networks

First resting state networks (De Luca)

Published in 2006 using independent components analysis

Early resting state networks De Luca 2006

Resting state networks (Damoiseaux)

Published the same year, using a different ICA method.

Early resting state networks Damoiseaux 2006

Diffusion imaging

Measures features of water “diffusion” in the brain. Typically, to estimate biological features of the white matter

  • Will not focus on this data type in this course

Diffusion imaging in Autism Wilkinson 2016

Other measures

  • Perfusion – Quantitative measure of blood flow
  • Spectroscopy
  • Many others actively in development
  • Many “derivatives” exist and in development
    • Derivatives are quantitative values derived from existing modalities