10-04-2019 meeting w/ Camille

[gkiar]

Open Questions

Modelling/Dimensionality Reduction

• [ ] Decide how will we want to deal with 3D data (vectorized, 2D, or 3D matrices)
• [ ] Determine network capacity (N layers; size; type)
• [ ] Based on the above, what is a good dataset to test on?
• [x] Space
  • [x] resample to MNI152 space (FSL targets without another prior)
  • [x] resolution: all the same, maybe use 2x2x2
  • dump all data w/out complete headers
• [ ] Grouping
  • typically interested in group level for most tfMRI analyses
  • NV training will be largely group maps
  • [ ] identify portion of each
  • HCP training will be on individual maps and bootstrapped-groups
  • [ ] clarify

Data Processing

• [ ] What datasets?
  • [x] Training autoencoder: Neurovault (crawling in progress)
  • [ ] Removing tool artifacts: HCP Task-fMRI
  • Tasks: motor, gambling, (overlap with UKBB & HCP)
  • Convert to bids: https://github.com/suyashdb/hcp2bids
  • Req'd data: all functional sessions, T1w, event files, timing files
• [x] What should we use for storage?
  • Cedar on Compute Canada for now
• [x] What pipelines will we use? (are they all in boutiques/can they be? SPM maybe not...)
  • [x] AFNI and FSL command-lines taken care of and to be linked by Camille
  • [x] SPM: use octave version (under Camille's github), Alex's code for running it
  • [x] start from alex's pipelines in variability paper, move outwards from there if necessary
• [x] Where will we run them?
  • CC Cedar

Validation

• [ ] Currently, in lieu of GT: agreement with anatomy on salient task like motor
• [ ] Stability across many splits (groups, runs, noise, datasets, bootstrapped groups, etc...)

[cmaumet]

Greg to investigate:

• [ ] Duplicate image ids in NeuroVault?
• [x] Less images downloaded than the total count returned by the API at https://neurovault.org/api/images/ --> why? cf. 22125 => SOLVED

Camille:

• [ ] Think about phase encoding. Do we have a preference? Look at the paper.
• [ ] Send info about igrida and other machines in the team

Interesting links for the pipelines:

• Repo of examples in SPM/FSL/AFNI cf. Table
• Alex's code for the variability paper