Contents

Welcome

neuroscout logo


Jupyter Book Badge Binder GitHub size GitHub issues GitHub PR License

Welcome

Welcome! This is an interactive walkthrough of our publication “Neuroscout: a unified platform for generalizable and reproducible fMRI research”. Here you can visualize and re-run the code we used to create analyses and figures.

In the paper, we validate the Neuroscout platform by replicating established effects from cognitive neuroscience using automatically extracted features in over 30 naturalistic datasets. We then use meta-analysis to synthesize single dataset findings, resulting in more robust and generalize estimates. In addition, we also showcase more exploratory applications in two domains (face processing & natural language perception) that demonstrate how Neuroscout can be used to run more generalizable naturalistic fMRI research.

These analyses require specifying and estimating models at the level of individual datasets/tasks, and the outputs of these analyses are used as inputs to meta-analyses. This is reflected by the structure of the GitHub repository and of this book.

Re-running the analyses

The analyses follow the structure of figures in the manuscript. Most analyses require first running single dataset results (using Neuroscout) and then performing a meta-analysis (using NiMARE). You can use this resource to simply visualize the analyses, or to re-run them and recreate the figure.

Note that, if you want to re-run meta-analyses, you do not need to re-run the dataset-level models. All statistical maps are uploaded to NeuroVault and can be downloaded using our meta-analysis code. If you wish to recreate and re-estimate dataset-level models, you will have to do so locally.

  • Cloud computing

    Notebooks can be re-run on the cloud using mybinder by clicking on the rocket icon at the top of the notebook page. This is potentially the easiest option as you don’t have to install/download anything. You can also easily access all analyses at:

    Binder


  • Software containers

    If you want to re-run the analyses and recreate the figures locally, you can use our software containers to recreate a suitable environment. More precisely, you can obtain the corresponding Docker image via:


    docker pull neuroscout/neuroscout-paper:preprint

    and then start it:


    docker run -it --rm -p 8888:8888 neuroscout-paper

    Subsequently, start a jupyter notebook server via:


    jupyter-notebook --port=8888 --no-browser --ip=0.0.0.0

    which should provide you with a link that looks roughly like this:


    http://127.0.0.1:8888/?token=d47d101bcb9d1233471aa4fb21240ff74d520887d4c0e0b6


    If you click on this link or copy-paste it in your browser, you should see a jupyter notebook server that allows you to navigate these resources.


  • local python environment

    Finally, if you want to re-run analyses and re-create figures locally without sotware containers, you can do so via using a python environment. For this to work, you initially need to download the repository with the notebooks and other necessary files from GitHub.

    It is recommend to create a new python environment through e.g. conda to avoid installation and dependencies issues. For example:


    conda create -n neuroscout_analyses python==3.8

    which you then can activate and after navigating to the downloaded repository, install the required libraries via:


    conda activate neuroscout_analyses
cd /path/to/neuroscout-paper
pip install -r requirements.txt

    (NB: you need to run the above code line-by-line and exchange the /path/to part above to the path you downloaded the neuroscout-paper repository to.)

    Subsequently, start a jupyter notebook server via:


    jupyter-notebook

    which should provide you with a link that looks roughly like this:


    http://127.0.0.1:8888/?token=d47d101bcb9d1233471aa4fb21240ff74d520887d4c0e0b6


    If you click on this link or copy-paste it in your browser, you should see a jupyter notebook server that allows you to re-run the analyses and re-create the figures through the dedicated python environment created above.


Feedback & Questions

If you have any feedback, don’t hesitate to get in touch! We also support public reviews and comments through an hypothes.is plugin with, which you can interact by clicking on the arrow at the top right side of the page.

next

Individual dataset/task analyses