.md .pdf

repository open issue suggest edit

Contents

Welcome

“Introduction to brain decoding & encoding in fMRI”

This jupyter book presents an introduction to brain decoding & encoding using fMRI. It was developed within the educational courses, conducted as part of the Montreal AI and Neuroscience (MAIN) conference in November 2022.

Building upon the prior sections of the educational courses, the here presented resources aim to provide an overview of how decoding & encoding models can be applied to fMRI data in order to investigate brain function. Importantly, the respective methods cannot only be utilized to analyze data from biological agents (e.g. humans, non-human primates, etc.) but also artificial neural networks, as well as presenting the opportunity to compare processing in both. They are thus core approaches that are prominently used at the intersection of neuroscience and AI.

Fig. 1 To test the consistency of representations in artificial neural networks (ANNs) and the brain, it is possible to encode brain activity based on ANN presented with similar stimuli, or decode brain activity by predicting the expected ANN activity and corresponding annotation of cognitive states. Figure from Schrimpf et al. (2020) [SKH+20], under a CC-BY 4.0 license.

The tutorials make heavy use of nilearn concerning manipulating and processing fMRI data, as well as scikit-learn and pytorch to apply decoding & encoding models on the data.

We used the Jupyter Book framework to provide all materials in an open, structured and interactive manner. ALl pages and section you see here are built from markdown files or jupyter notebooks, allowing you to read through the materials and/or run them, locally or in the cloud. The three symbols on the top right allow to enable full screen mode, link to the underlying GitHub repository and allow you to download the respective sections as a pdf or jupyter notebook respectively. Some sections will additionally have a little rocket in that row which will allow you to interactively rerun certain parts via cloud computing (please see the Binder section for more information).

Brain decoding vs. encoding

In short, encoding and decoding entail contrary operations that can yet be utilized in a complementary manner. Encoding models applied to brain data, e.g. fMRI, aim to predict brain responses/activity based on annotations or features of the stimuli perceived by the participant. These can be obtained from a multitude of options, including artificial neural networks which would allow to relate their processing of the stimuli to that of biological agents, ie brains. Decoding models on the other hand comprise models with which we aim to estimate/predict what a participant is perceiving or doing based on recordings of brain responses/activity, e.g. fMRI.

Fig. 2 Encoding and decoding present contrary, yet complementary operations. While the former targets the prediction of brain activity/responses based on stimulus percepts/features (e.g. vision & audition), cognitive states or behavior, the latter aims to predict those aspects based on brain activity/responses.

More information and their application can be found in the respective sections of this resource. You can either use the ToC on the left or the links below to navigate accordingly.

An overview of the Haxby Dataset

Explore and prepare the tutorial dataset.

Explore this tutorial

Brain decoding with SVM

Utilizing an SVM classifier to predict percepts from fMRI data.

Explore this tutorial

Brain decoding with MLP

Brain decoding using a basic artificial neural network.

Explore this tutorial

Brain decoding with GCN

Graph convolutional networks for brain decoding.

Explore this tutorial

Brain encoding

Brain encoding using ridge regression.

Explore this tutorial

Setup

There are two ways to run the tutorials: local installation and using free cloud computing provided by Binder. As noted below, we strongly recommend the local installation, as the Binder option comes with limited computational resources, as well as the missed possibility to directly further explore the approaches presented in this tutorial on your own machine.

Local installation (Recommended)

For the local installation to work, you need two things: the fitting python environment and the content. Concerning python, please have a look at the hint below.

Install python

You need to have access to a terminal with Python 3. If you have setup your environment based on instructions of MAIN educational installation guide, you are good to go 🎉

If it not already the case, here is a quick guide to install python 3 on any OS.

After making sure you have a working python installation, you need to get the content that is going to presented during the tutorial. In more detail, this is done via interactive jupyter notebooks which you can obtain by following the steps below:

1. Clone/download this repository to your machine and navigate to the directory.

   git clone https://github.com/main-educational/brain_encoding_decoding.git
   cd brain_encoding_decoding
   

2. We encourage you to use a virtual environment for this tutorial (and for all your projects, that’s a good practice). To do this, run the following commands in your terminal, it will create the environment in a folder named env_tuto:

   python3 -m venv env_tuto
   

   Then the following command will activate the environment:

   source env_tuto/bin/activate
   

   Finally, you can install the required libraries:

   pip install -r binder/requirements.txt
   

3. Navigate to the content of the jupyter book:

   cd content/
   

   Now that you are all set, you can run the notebooks with the command:

   jupyter notebook
   

   Click on the .md files. They will be rendered as jupyter notebooks 🎉

Cloud computing with Binder

If you wish to run the tutorial in Binder, click on the rocket icon 🚀 in the top right of a given notebook to launch it on Binder.

Warning

The computing resource on Binder is limited. Some cells might not execute correctly, or the data download will not be completed. For the full experience, we recommend using the local set up instruction.

Instructors

This tutorial was prepared and presented by

Peer Herholz

Shima Rastegarnia

Bertrand Thirion

Isil Bilgin

Alexandre Pasquiou

Pravish Sainath

Thanks and acknowledgements

Parts of the tutorial are directly adapted from a nilearn tutorial on the so-called Haxby dataset.

It was adapted from a prior version which was prepared and presented by Pravish Sainath Shima Rastegarnia, Hao-Ting Wang Loic Tetrel and Pierre Bellec.

Furthermore, some images and code are used from a previous iteration of this tutorial, prepared by Dr Yu Zhang.

We would like to thank the Jupyter community, specifically, the Executable/Jupyter Book and mybinder project for enabling us to create this tutorial. Furthermore, we are grateful for the entire open neuroscience community and the amazing support and resources it provides. This includes the community driven development of data and processing standards, as well as unbelievable amount of software packages that make the here introduced approaches possible to begin with.

The tutorial is rendered here using Jupyter Book.

References

SKH+20

Martin Schrimpf, Jonas Kubilius, Ha Hong, Najib J Majaj, Rishi Rajalingham, Elias B Issa, Kohitij Kar, Pouya Bashivan, Jonathan Prescott-Roy, Franziska Geiger, Kailyn Schmidt, Daniel L K Yamins, and James J DiCarlo. Brain-Score: which artificial neural network for object recognition is most Brain-Like? bioRxiv, pages 407007, January 2020.

---

next

An overview of the Haxby dataset