ADNI

ADNI is a long-running observational study, run like a clinical trial, that tracks how imaging, fluid biomarkers and cognition change from normal ageing through mild cognitive impairment (MCI) to Alzheimer's dementia. It started in 2004 as a public-private partnership led by Michael Weiner. This entry covers the phases ADNI1, ADNI GO, ADNI2 and ADNI3; ADNI4 is ongoing.

Composition

A 2023 review counts 868 cognitively unimpaired participants, 1,090 with MCI and 408 with Alzheimer's dementia enrolled since 2004. Each phase added new people and kept many earlier ones: ADNI1 began with 819 people (229 controls, 398 MCI, 192 AD, mean age about 75), ADNI GO added an early MCI group and ADNI2 added a subjective memory concern group. Most have several imaging sessions. Clinical, cognitive, genetic and CSF/blood biomarker data come with the images.

Acquisition

More than 60 sites in the US and Canada take part, on GE, Philips and Siemens scanners with harmonised protocols. ADNI1 scanned everyone at 1.5 T (T1-weighted and dual-echo PD/T2), a quarter also at 3 T. From ADNI GO on, imaging moved to 3 T and added FLAIR and an axial T2*-weighted gradient echo (GRE) for microbleeds (ADNI3: 0.85 x 0.85 x 4 mm, TE 20 ms; a 3-echo version on Siemens). Diffusion, resting-state fMRI and ASL were added per vendor in ADNI2 and for nearly all participants in ADNI3. ADNI4 (outside this entry) adds a multi-echo GRE at capable sites, from which SWI is to be derived. PET started with FDG and a small PiB pilot, then used florbetapir for amyloid and, from ADNI3, flortaucipir for tau.

Annotations

There are no voxel-level labels. Mayo Clinic reads the GRE scans for cerebral microbleeds and superficial siderosis and releases each finding with its location (table MAYOADIRL_MRI_MCH). The MRI core reports 641 participants with one or more microbleeds and 57 with superficial siderosis in ADNI GO/2 and ADNI3. Diagnoses, cognitive scores, QC ratings, FreeSurfer volumes and amyloid and tau SUVRs are also released.

Known limitations

• Diagnostic criteria and group names changed between phases (early/late MCI were merged again in ADNI3).
• Scanner upgrades over the years can break longitudinal comparability; the MRI core advises treating data across hardware changes as incompatible.
• Participants are mostly white and highly educated, which limits generalisability; ADNI4 aims to change this.
• Sequences beyond T1, FLAIR and T2* are not available for all participants in ADNI2.