nibabies
nibabies copied to clipboard
Processing tools for magnetic resonance images of infant brains
### A short summary of what you would like to see in *NiBabies*. Brains on the smaller size include excess background within reports. View should be zoomed in accordingly. ###...
Current approach to CIFTI resampling is to resample BOLD to age-specific template, and use an affine registration from the subcortical structures in age-specific template to adult template. This should be...
It would be good to include the T2w images in the html report as well for infants because the T2w has better contrast than the T1w in some studies and...
### Is your feature request related to a problem? Please describe. T2w can show greater contrast than T1w ### Describe the solution you'd like (Either as a default or flag)...
Current implementation: https://github.com/nipreps/nibabies/blob/bfe9bcbe3c83d5b949476afcb721241f45367dee/nibabies/workflows/anatomical/segmentation.py#L156-L178 - The validity of the segmentation directory should happen as soon as inputs are processed. - We should implement a more BIDS-like structure.
When running without `sloppy` & `debug`, TOPUP can take a very long time, likely due excessive grid searching. Potential optimization: - clipping the image (potentially masking the brain and then...
Some troublesome nodes to monitor. I have not been able to profile the specifics here yet. - aCompCor https://github.com/nipreps/nibabies/blob/b5d16d3aaba44dc42ee6215f17770f9cb0b933b9/nibabies/workflows/bold/confounds.py#L217-L221 - SignalExtraction https://github.com/nipreps/nibabies/blob/b5d16d3aaba44dc42ee6215f17770f9cb0b933b9/nibabies/workflows/bold/confounds.py#L248-L249 - ROIsPlot https://github.com/nipreps/nibabies/blob/b5d16d3aaba44dc42ee6215f17770f9cb0b933b9/nibabies/workflows/bold/confounds.py#L296-L297 When all run in parallel,...
This might be tricky now that we are generating an average bold reference prior to individual run processing.
This follows the patch in #20 to allow processing despite a lack of labelled templates. FSL FAST segmentations are far from ideal - instead we should probably: - [ ]...