Baby Connectome Project (BCP)

The UNC/UMN Baby Connectome Project (BCP) is the infant and preschool arm of the NIH Lifespan Human Connectome Project. Teams at the University of North Carolina at Chapel Hill and the University of Minnesota image typically developing children from birth to 5 years with an HCP-style protocol adapted to small, mostly sleeping children, and pair the scans with behavioral assessments. The aim is to chart how structural and functional connectivity emerge in early childhood and how they relate to behavior. Imaging data are shared through NIMH Data Archive collection 2848, which lists 420 subjects with shared data.

Composition

The study plans 500 children in an accelerated longitudinal design: 285 contribute 4 to 6 visits and 215 a single cross-sectional visit. Eligible children were born at term (37 to 42 weeks) with appropriate birth weight and no major pregnancy or delivery complications; children with a first-degree relative with autism, intellectual disability, schizophrenia or bipolar disorder are excluded. Behavioral measures include the Mullen Scales of Early Learning, Vineland Adaptive Behavior Scales and several parent questionnaires; saliva is collected for genetics. The NDA record does not break the 420 shared subjects down by age, sex or visit.

Acquisition

Both sites use a Siemens 3T Prisma with a 32-channel head coil. The protocol has T1w MPRAGE and T2w SPACE at 0.8 mm isotropic, resting-state BOLD at 2 mm (multiband 8, AP and PA runs), and a 6-shell diffusion scan at 1.5 mm, with a 2-shell diffusion scan and extra BOLD runs when the child stays asleep. Children up to 36 months are scanned in natural sleep; older children may be scanned awake after a mock scanner session.

Annotations

No manual labels are described. Structural scans get a visual motion rating and a radiological review for incidental findings; diffusion and BOLD scans pass automated QC.

Known limitations

• Only typically developing children, so no clinical groups.
• The protocol paper reports that by February 2018 only 43% of datasets had all four of T1w, T2w, BOLD and diffusion, so modality coverage varies between visits.
• Release dates and per-release contents are not published on the study pages.