Big Linear Modelling and Big Linear Mixed Modelling

[TomMaullin]

Title

Big Linear Modelling and Big Linear Mixed Modelling

Short description and the goals for the OHBM BrainHack

Large-scale, shared datasets are becoming increasingly commonplace in fMRI, challenging existing tools both in terms of overall scale and complexity of the study designs. As sample sizes grow, new opportunities arise to detect and account for grouping factors and covariance structures present in large experimental designs. To facilitate large sample analysis, we have created two Python toolboxes for use on HPC clusters:

• “Big” Linear Models (BLM); a toolbox for large-scale distributed fMRI Linear Model analyses.
• “Big” Linear Mixed Models (BLMM); a toolbox for large-scale distributed fMRI Linear Mixed Model analyses.

At present, both tools are functioning and can be used for the analysis of tens of thousands of fMRI images. However, there is plenty that could be improved. Some of the goals we hope to address during the hackathon include:

1. Developing a rigorous testing suite, potentially with continuous integration via Travis CI.
2. Using Dask to streamline the current code base (at present there are a lot of bash scripts for 'qsub'bing).
3. Package releases. Neither of the toolboxes are currently on the Python Package Index.
4. Adding customized covariance support. In previous work, we showed how the underlying methods BLMM uses could model custom covariance structures (e.g. AR, Diagonal, Toeplitz etc). However, at present BLMM does not support analyses with these features.

Link to the Project

https://github.com/TomMaullin/BLMM

Image for the OHBM brainhack website

No response

Project lead

Thomas Maullin-Sapey
Github: TomMaullin
Discord: Tom Maullin

Main Hub

Glasgow

Other Hub covered by the leaders

• Glasgow
• Asia / Pacific
• Europe / Middle East / Africa
• Americas

Skills

• Familiarity with the Python programming language (All goals).
• Familiarity with Travis CI (Goal 1 - recommended but not a necessity).
• Familiarity with Dask (Goal 2).
• Understanding of Linear Mixed Models and/or statistics (Goal 4).

Recommended tutorials for new contributors

• Set up a brainhack friendly computing environment
• A introduction to Bash
• Python: Writing a script
• Python: Data analysis with Python
• Python: Visualisation
• Machine learning basics
• BIDS
• Machine learning for neuroimaging
• VCS: Using Git and Github
• VCS: Data management with Datalad
• Virtualization
• High Performance Computing: Oracle cloud resource

Good first issues

Potential good first issues include:

• BLMM #28 Use Dask
• BLMM #32 Record cause of voxel dropout
• BLM #82 Dask Integration

Please let me know if you would like more information (@TomMaullin).

Twitter summary

BLMM - A toolbox for large-scale distributed fMRI Linear Mixed Model analyses.
https://github.com/TomMaullin/BLMM
@TomMaullin
#OHBMHackathon #Brainhack #OHBM2022

Short name for the Discord chat channel (~15 chars)

BLMM

Please read and follow the OHBM Code of Conduct

• I agree to follow the OHBM Code of Conduct during the hackathon

[stebo85]

Dear @TomMaullin ,

Exciting project :)

It sounds like our brainhack cloud project (#50) could be very useful for your developments:

1. You could run dask on Kubernetes: https://brainhack.org/brainhack_cloud/docs/kubernetes/
2. You could run a development HPC during brainhack: https://brainhack.org/brainhack_cloud/tutorials/hpc/

If you think this could be useful, make sure to sign up before the hackathon so we can create the accounts: https://brainhack.org/brainhack_cloud/docs/request/

Cheers
Steffen

[djarecka]

Thank you for submitting the project! We have 35 projects right now, woohoo! But that means the projects pitches will have to be short. We will give you tomorrow 2 minutes to pitch your project, you can have one slide or no slides!
If you decide to use a slide, please include the link to the slide here.

And don't worry, you will still have more time to talk about your project during the BrainHack :-)

[TomMaullin]

https://docs.google.com/presentation/d/1RYomiqBTLcY2O19kRyqXn-ZJD4ERAV2QleOVUCtcDsc/edit?usp=sharing