fMRI measures the BOLD signal — a haemodynamic proxy for neural activity — with good spatial coverage and poor temporal resolution. Most CNS projects use it to derive functional connectomes: region-by-region matrices of how strongly brain areas co-activate over time.
Data¶
Datasets in this book that include fMRI:
PREVENT-AD — longitudinal, pre-symptomatic AD cohort
ADNI — Alzheimer’s Disease Neuroimaging Initiative
UK Biobank — large population cohort
ABCD — adolescent development
Portals that index open fMRI data: OpenNeuro, CONP, BrainCODE
Processing¶
The conventional path, in order:
Organize the raw data as BIDS
Check quality with MRIQC
Preprocess with fMRIPrep
Extract connectomes with giga_connectome
Nipoppy is useful for managing these pipelines across a whole dataset rather than one subject at a time. Nilearn is the usual next stop once preprocessing is done — masking, atlases, timeseries extraction, and plotting.
CNS Projects¶
Davor Curic — geometry-aware distances between functional connectomes
Josh Neudorf — how structural connectivity constrains functional connectivity
Mohamed Abdelhack — modelling cognition and psychiatric conditions from fMRI
Hao-Ting Wang — connectivity-based biomarker discovery
Julia-Katharina Pfarr — functional connectivity as a normative measure
Tom George — self-supervised pretraining across open neural datasets