2. NiMARE is an open source alternative to existing
tools
- https://github.com/neurostuff/NiMARE
- Pure Python
- Collaboratively built
- Common syntax across tools
- Joins ecosystem of Python tools for neuroimaging, including nipype,
nibabel, nilearn, and nistats
2
3. nimare.decode
✔ BrainMap
✔ Neurosynth
✔ Correlation
✔ Correlation Distribution
✔ GCLDA Continuous
✔ GCLDA Discrete
✔ GCLDA Encoding
nimare.workflows
o MACM
o CBP
o Meta-Analytic Clustering
NiMARE will support a range of analyses
nimare.extract
✔ NeuroVault
✔ Neurosynth
✔ Brainspell
nimare.annotate
✔ Cognitive Paradigm
Ontology
✔ Cognitive Atlas
✔ LDA
✔ GCLDA
o GloVe
o Deep Boltzmann
3
nimare.meta
✔ ALE/SCALE
✔ MKDA/KDA Density
✔ MKDA Chi-square
✔ MFX/FFX GLM
✔ RFX GLM
✔ Contrast Permutation
✔ Z Permutation
✔ Fisher’s
✔ Stouffer’s
✔ Weighted Stouffer’s
o SDM
o BHICP
4. Meta-analytic algorithms are already
implemented
4
21 pain studies from NeuroVault (https://neurovault.org/collections/1425/)
5. Next steps
- Recruit contributors
- Code
- Documentation
- Testing
- Examples
- Design input
- Stable API for datasets and metadata
- Extraction, annotation, and decoding tools
- Logo!
5
6. Outside Contributors
Dr. Tal Yarkoni
Dr. Camille Maumet
Dr. Thomas Nichols
Dr. Chris Gorgolewski
Acknowledgements
6
Neuroinformatics and Brain Connectivity Lab
Dr. Angela Laird
Dr. Matthew Sutherland
Dr. Michael Riedel
Dr. Michael Tobia
Dr. Veronica Del Prete
Chelsea Greaves
Rosario Pintos Lobo
Laura Ucros
Jessica Bartley
Katherine Bottenhorn
Jessica Flannery
Ranjita Poudel
Lauren Hill
Diamela Arencibia
Jennifer Foreman
Ariel Gonzalez
NSF 1631325
NIH U24 DA039832
NIH R01 DA041353
NIH U01 DA041156
NSF REAL DRL-1420627
NSF CNS 1532061
NIH K01DA037819
NIH U54MD012393
Editor's Notes
Hello everyone. I’m Taylor Salo, a graduate student in Dr. Laird’s lab, and I’ll be talking about a Python package I’ve been working on called NiMARE, which stands for “neuroimaging meta-analysis research environment.”
The goal of NiMARE is to build interfaces for, or translate to Python, as many of the tools neuroimagers use for meta-analyses as possible.
NiMARE is written in pure Python, a general programming language that emphasizes readability and which is commonly used among neuroimagers. It is open source, meaning that anyone can access the core code and report bugs if necessary. It is collaboratively built, meaning that the developers, including myself, are open to outside contributions. NiMARE also aims to provide a common syntax for these tools, which will make it easier to switch to the most appropriate tool as necessary. Finally, NiMARE joins an ever-growing ecosystem of tools written in Python, with the goal of making it easy to use tools from any of these packages together in concert.
Development of NiMARE started in January, and we’ve been able to add a large number of interfaces so far, with many more in development.
We also plan to support common workflows like meta-analytic coactivation modeling, coactivation-based parcellation, meta-analytic clustering analysis, and meta-analytic ICA.
We have tested our implementations of both coordinate- and image-based meta-analytic algorithms on a dataset of 21 pain studies shared on NeuroVault, which had previously been used with a MATLAB image-based meta-analysis package, and have found results consistent with expectations. While there are potentially minor bugs in these implementations, we feel confident about them and will be incorporating unit testing and continuous integration soon.
Among our next steps, we will focus on recruiting contributors to the project. Contributions can come in many forms, from adding code to documentation to simple input on the design. We’ll also be working on developing a stable API for dataset objects and metadata (such as which images/data are included and how annotations are organized), which may require substantial input from the neuroimaging community.
I have already started on several tools for pulling data from open datasets like Brainspell, NeuroVault, and Neurosynth, as well as tools for automatically annotating studies and decoding maps using a variety of topic models and standardized ontologies.
Oh, and we also need a logo!
I would like to thank everyone involved in this project, including the Neuroinformatics and Brain Connectivity Lab led by Drs. Laird and Sutherland, as well as a number of contributors from outside the lab, including Drs. Tal Yarkoni, Camille Maumet, Thomas Nichols, and Chris Gorgolewski.
Thank you all for listening. Does anyone have any questions?