This post provides a short theoretical overview of the concepts needed to understand FreeSurfer's Surface-Based Group Analyses. It describes General Linear Models and the differences between Different Offset, Different Slope and Different Offset, Same Slope models and how they are implemented in FreeSurfer.

Skip to your favourite section of the article.

A new semester has just begun, and you have signed up for a course on Network Neuroscience. The description promised that you would finally learn how to understand the brain with cool computational tools and after taking a couple of courses in both the Neuroscience and Computer science department, you feel confident that you can master this course. Enthusiastically you start learning FreeSurfer using the Harvard tutorial. Downloading the data? Easy. Visualizing some data? Check. Checking cortical segmentation? No problem at all. Running preprocessing on a new data set? Well, you needed a couple of break hours anyways. Correcting preprocessing errors? Alright, this is getting a bit difficult... Group Analysis and setting up GLMs? Group Analysis for what...?

If your experience of FreeSurfer and similar tools is anything like mine, you have noticed by now, that there seems to exist only two types of tutorials: explanations a toddler could follow and those that only a Ph.D. in the field seems to make accessible to you. The above-mentioned tutorial manages to start on the easy end and then swiftly brings in advanced terminology with little to no explanation. For an example, check out this section:

Chances are, that you know what most or many of the terms in this section mean.*** Perhaps you have a vague understanding of hypothesis testing being important and contrast being used for that? The tutorial certainly does not bother to explain itself and much worse, you will simply be provided with the code to perform this analysis and how to view the results. Who cares if you truly understand how you got the results...? Well, I do! It's incredibly easy to do fMRI data analysis wrong and most mistakes start with not understanding what analysis you are even exactly running. Don't despair, we are in this together. And I will try my very best to get you through the key theoretical concepts behind this tutorial alive (follow along here)!

Source. Think there will be a lot of gifs in this post? You bet there will be...

Throughout the article, I've provided toggle sections titled Pitfalls. They provide additional details on the methodology and why certain methodological choices might be problematic when interpreting your results. Feel free to read them for a deeper understanding of the analysis assumptions. They are, however, not necessary to understand the rest of the article.

Alright, let's get to it then! No time to waste.

*Note: And even if you do, could you explain all of them to someone who has little or no background in the field? If so, congratulations! You are free to proceed with the original tutorial. Feel free to drop the author (@Michelle Hackl) a note with better resources or explanations, they're always trying to learn more! If not, perhaps read on ;)