mne-hcp-pitch-biomag-2018.pdf

These are the slides from my Biomag 2018 talk in the panel on open science.

Transcript

1. Lessons learned from processing the Human Connectome MEG data Denis A. Engemann denis-alexander.engemann@inria.fr www.denis-engemann.de github: dengemann Biomag 2018 Philadelphia
2. Q: Why would you care about a Python reader for the HCP MEG data?
3. Bridging resources • Dataset provider: make sure the data can be used with any tool. • Software provider: make sure the software can be used with any dataset • Researcher: use a baseline dataset in a most unconstrained way. Enabling cross-toolbox, cross- methods, cross-site comparisons
4. Example 1: validation of power envelope connectivity Soon in CTF Brainstorm Hipp et al. Nat Neuro 2012 CAMCAN HCP
5. Example 2: Benchmark against new MNE family solvers Challenge me!
6. Example 3: compare machine learning tools
7. Challenges we solved together • The ethics protocol may not allow sharing all aspects of the data • Derivative data may be stored differently from what your tools expects (anatomy, coordinate systems) • Knowing which information is which file and how to get the missing bits and pieces + +
8. Programmatic access to HCP data https:/mne-tools.github.io/mne-hcp
9. …/mne-hcp/auto_tutorials/index.html Add your example! Run all analyses supported by MNE on HCP data
10. Add your example! Load the evoked responses
11. Add your example! Get the forward model and compute the noise covariance
12. Add your example! Compute the inverse solution
13. Add your example! Be fascinated by visual cortex activity
14. The lessons learned • Principle of charity: there can be good reasons why some data are not easily available. • Work together with data providers • Document how a data resource can be used with your software • Mind your community’s software eco-system • Talk to each other!
15. Thank you • Robert Oostenveld • Jan-Mathijs Schoffelen • Giorgos Michalareas • Eric Larson • Alexandre Gramfort • Mainak Jas • Virginie van Wassenhove