Harvard FSL Workshop 2021
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 from the Apple Store. (This makes it easier to view and edit 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.
3. Download the Dataset
We will be using this dataset from openneuro.org for the practicals. This dataset uses the Flanker task, a robust measure of cognitive control.
4. 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.
MRIcroGL and a sample dataset: Used for converting DICOM to NIFTI files.
make_FSL_Timings.sh: A script for converting the timing files from openneuro into timing files that can be read by FSL.
run_1stLevel_Analysis.sh: A script for running 1st level preprocessing and statistical analysis for each subject.
Day 1: fMRI Fundamentals and an Introduction to FSL
Agenda
(1:00pm-1:30pm) Unix essentials (Practical, optional)
This session will review Unix commands and concepts used in this course (download this demo script to follow along):
The shell
Navigating directories and files
Environmental variables
Commands and options (or “switches”)
Shell scripts (in particular, loops and conditionals)
Please see this playlist for an introduction to the Unix terminal, as well as the Surrey Unix tutorial. I also recommend downloading TextWrangler for editing shell scripts.
(1:30pm-2:30pm) Review of fMRI Data Processing and Analysis (Lecture)
We will review what is done with fMRI data from start to finish in a typical pipeline. This lecture will cover:
Hemodynamics and the BOLD signal
The BOLD signal and linearity
Understanding preprocessing: motion correction, registration, normalization, and smoothing
From scanner to computer: Converting DICOM files to NIFTI with MRIcroGL (exercise dataset can be downloaded here)
(2:45pm-4:00pm) 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:
Overview of the GUI
Brain extraction / skull stripping
Registration of T1 and T2-weighted data
Boundary-based registration
User options: Slice timing correction, temporal derivatives, and smoothing size
Troubleshooting preprocessing failures
(4:00pm-5:00pm) 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 variability
Design matrices
Custom timing files, and how to make OpenNeuro timing files compatible with FSL. Download the timing formatting script here.
Day 3: ROI Analysis & Advanced Topics
(12:00pm-1:15pm) 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
ROI analysis with featquery and fslstats
Scripting ROI analyses from the command line
(1:15pm): Group Photo
(1:30pm-3:00pm) Other Statistical Scenarios (Lecture & Practical)
An overview of common statistical problems in neuroimaging, how to identify them, and how to avoid them.
How to properly test for double dissociations
Introduction to biased analyses, and how to avoid them
False positive rates: The dead salmon study and further discussion of Eklund et al. (2016) and more recent papers
Balancing false positive and false negative rates with threshold-free cluster enhancement
Non-parametric testing with randomise
(3:00pm-4:00pm) Looking Forward: Standardized Preprocessing Pipelines (Lecture)
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.
(4:00pm-5:00pm) General Q&A
Day 2: Group-Level Analysis, Scripting, & the FSL Imaging Viewer
Agenda
(12:00pm-1:00pm) Scripting your analysis (Practical)
Automating analyses is an indispensable skill for the neuroimaging researcher. This practical will introduce the power of loops, which will save you countless hours and reduce the probability of error. The analysis script can be downloaded here.
Review of shell scripting
Saving the analysis script from the GUI
Editing the .fsf file
Looping your analysis over subjects
(1:00pm-2:15pm) Group-level analysis (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).
Second-level vs. third-level analysis
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
(2:30pm-4:00pm) FSLeyes and viewing results (Practical)
We will tour FSL’s fsleyes data visualization tool, which is useful for understanding fMRI data conceptually – for example, the connection between the canonical HRF and beta weights. A playlist about FSLeyes can be found here for review outside of the workshop.
Fsleyes – the FSL data visualization tool
Overlays, underlays, and thresholds
Atlases
FEAT mode and cluster analysis
Tips for creating publication-quality figures
(4:00pm-4:30pm) Meta-analyses (Practical)
We will briefly review the meta-analysis website neurosynth.org in order to see whether our results correspond to what other studies have found, and how to properly use association tests and uniformity tests. This will also serve as a segue to ROI analysis, which is covered the following week.
(4:30pm-5:00pm) General Q&A
This is an opportunity to ask questions about any of the topics covered during the past two days.