ARIC-NCS brain MRI

The Atherosclerosis Risk in Communities (ARIC) study has followed adults from four US communities since 1987-89, when they were 45 to 64 years old: Washington County (Maryland), Forsyth County (North Carolina), Jackson (Mississippi) and suburban Minneapolis (Minnesota). At the fifth examination in 2011-13, the ARIC Neurocognitive Study (ARIC-NCS) scanned the brains of a subset of participants. This entry covers those visit 5 MRI scans and the reads made on them, which are often used to study cerebral microbleeds and small vessel disease in a biracial ageing population.

Composition

Visit 5 MRI targeted everyone with cognitive impairment or a prior ARIC brain MRI, plus an age- and site-stratified random sample of cognitively normal participants. Of these, 1,928 completed the scan. A 2017 microbleed analysis kept 1,677 non-demented participants with usable data: 1,072 cognitively normal and 605 with mild cognitive impairment, 60% women, mean age 76 years, about a quarter Black participants from Jackson. In that sample 398 people had definite microbleeds, 262 strictly lobar and 136 deep or mixed. The sampling scheme oversamples cognitive impairment, so the scanned group is not representative of the full cohort without weights.

Acquisition

All sites used 3 T Siemens scanners. The protocol included ADNI-style 3D MPRAGE and FLAIR and an axial 2D T2* gradient echo (TR/TE 200/20 ms, 3.3 mm slices, 256 x 224 matrix, about 5 minutes).

Annotations

Trained image analysts marked microbleeds (hypointense lesions up to 10 mm) on the T2* GRE as definite or possible, and a radiologist (K. Kantarci or C. R. Jack) confirmed them. Agreement on definite versus not definite was 85% (kappa 0.68). Microbleeds were classed as lobar, deep or infratentorial. Infarcts were read on FLAIR the same way, white matter hyperintensities were measured semi-automatically, and regional volumes come from FreeSurfer. Reads are counts and locations per participant; no lesion masks are described.

Known limitations

• Images are not openly available; access needs an ARIC investigator partnership and an approved proposal.
• Counts by sex, age and cognitive status are published only for the 1,677-person analysis sample.
• The 2017 analysis used only definite microbleeds; possible ones are rated but not counted above.