Vanderbilt University FSL Workshop, 2024

Below is an annotated agenda for the workshop. To prepare for the course, do the following steps:

1. Get Started with Unix

This workshop requires you to be familiar with Unix. Watch this playlist for an introduction to Unix, and go through the tutorials located here. It is also recommended that you install Xcode and TextWrangler from the Apple Store, which are useful for creating and editing scripts.

2. Install FSL

Use this link to install FSL. Follow the instructions for downloading and installing FSL on a Macintosh. A tutorial video for installing FSL and testing your installation can be found here. Note that FSL needs to be run from either a Unix operating system, or a Unix emulator (such as Macintosh’s Terminal application, or Windows’ Cygwin application). Although FSL can run on Windows emulators, it is not well-supported, and it is not guaranteed that it will work on your particular machine. We will also be attempting to use Neurodesk.org as a container for using FSL, but have a backup installed on your local machine.

4. Download the Data

We will be using this dataset from openneuro.org for the practicals. This dataset uses the Flanker task, a robust measure of cognitive control.

5. Download Programs and Scripts

Some of the practical sessions require downloading an application or analysis script. Here is a list of links to the relevant applications and analysis scripts, which can also be found in the agenda below.

Day 1: fMRI Fundamentals of FSL, Preprocessing, and QA Checks

AFNI_Viewer.png

Agenda

(9:00am-10:30am) Review of fMRI Data Processing and Analysis, with Review of Unix Basics (Lecture)
This will be a brief overview of what is done with fMRI data from start to finish in a typical pipeline. We will also review some of the basics of Unix. This lecture will cover:

  • Brief review of history and current trends in fMRI

  • Hemodynamics and the BOLD signal

  • The BOLD signal and linearity

  • Overview of BIDS Apps and an introduction to Neurodesk

(10:30am-12:00pm) FSL Compared to Other Packages: Preprocessing the individual subject (Practical)
This first practical will be a guided hands-on tutorial about how to process fMRI data. We will review the following topics:

  • Skull Stripping

  • Registration of T1 and T2-weighted data

  • Slice timing correction, registration, and smoothing size

  • Non-linear warping

  • Troubleshooting preprocessing failures

(12:00pm-1:00pm) LUNCH BREAK

(1:00pm-2:00pm) First-level analysis and the general linear model (Lecture & Practical)
How to set up the GLM for an individual subject and generate parameter estimates.

  • Overview of the GLM

  • How the GLM relates to fMRI data

  • Beta values, parameter estimates, and collinearity

  • Design matrices

(2:00pm-3:00pm) Group-level analysis and Pitfalls to Avoid (Lecture & Practical)
An overview of how to set up group-level analyses, as well as caveats to be aware of. The lecture will cover the basic mechanisms of group analysis, and correction issues unique to fMRI data. We will also briefly discuss the findings of Eklund et al. (2016).

  • Setting up group-level analyses

  • T-tests and F-tests: How to set them up and when to use them

  • Correction mechanisms: FWE, FDR, and cluster-forming thresholds

  • Introduction to Threshold-Free Cluster Enhancement (TFCE)

  • Pitfalls to avoid with multiple comparisons correction and ROI selection

  • Longitudinal Analyses

(3:00pm-4:00pm) Region of Interest (ROI) analysis (Lecture & Practical)
This expands upon the group-level analysis lecture by demonstrating different methods for performing inferential statistics.

  • Anatomical vs. Spherical ROIs

  • Testing for double dissociations

  • ROI analysis

  • What not to do: Biased & Circular Analysis

(4:00pm-4:30pm) FSLeyes and how to present results
In this session we will briefly review some of the norms in figure presentation, along with current recommendations for how to highlight results (e.g., Taylor et al., 2023). We will also discuss the presentation of effect sizes, how to use FSLeyes to its full potential, and how the finished product is displayed in published papers.

(4:30pm-5:00pm) General Q & A and Exercises
This session will be dedicated to answering any outstanding questions, along with providing practice scenarios that the attendees can use to consolidate what they have learned.

Day 2: Standardization, BIDS, and Machine Learning

Agenda

(9:00am-10:30am) Standardized Preprocessing Pipelines (Lecture & Practical)

Recently, standardized organization and preprocessing has become more popular. We will talk about BIDS format (which happens to be how the data for this workshop was organized), and how that enables you to use preprocessing tools such as MRIQC and fMRIPrep. Since you may encounter these in the future, we will discuss their advantages and disadvantages, and further training materials that you can use on your own. We will also discuss how to include covariates that are generated from fMRIPREP to deal with special populations, such as children and clinical populations.

(10:30am-12:00pm) Basics of Machine Learning (Lecture)
This session will provide an overview of MVPA, a popular multivariate tool for neuroimaging data.

  • Overview of Machine Learning and hyperplanes

  • Uses of MVPA

  • Experimental design considerations for studies that will use MVPA: Timing schemes, masks, and how to validate the results

(12:00pm-1:00pm) LUNCH BREAK and Group Photo

(1:00pm-2:00pm) Representational Similarity Analysis (Lecture & Practical)
Representational similarity analysis (RSA) exploits the correlation similarity structure of voxels, using different distance metrics to illustrate the representational distance of different conditions.

  • Measurement of both content and format of representations

  • How this is used across modalities

  • Editing The Decoding Toolbox template scripts for RSA

  • Analysis of sample dataset

(2:00pm-3:00pm) Introduction to Python and Jupyter Notebooks (Practical)
More advanced analyses such as Hyperalignment require Python. In this lecture, we will review the fundamentals of Python, including packages, methods, and how to use Jupyter notebooks.

(3:00pm-5:00pm) Hyperalignment (Lecture & Practical)
Hyperalignment is a relatively new classification technique developed by Jim Haxby’s lab, which aligns subjects' brain data in a high-dimensional space of voxels/features.

  • Introduction to hyperalignment

  • Benefits of hyperalignment vs. traditional MVPA: Alignment of functional topographies

  • Hybrid hyperalignment: Combining task and functional connectivity profiles

  • Analysis of movie datasets: Classifying which part of a movie a subject was watching

  • Application to other datasets