MESA Cardiac MRI (Cardiac Atlas Project)

The Multi-Ethnic Study of Atherosclerosis (MESA) is a US population study of men and women aged 45 to 84 from White, African-American, Hispanic and Chinese backgrounds, recruited at six centres without known cardiovascular disease. Part of the cardiac MRI from its baseline examination was contributed to the Cardiac Atlas Project (CAP), a consortium led from the University of Auckland that shares cardiac imaging for shape and function atlases. This entry covers that CAP share. The MESA brain MRI from Exam 6 is a separate collection held by the MESA Coordinating Center.

Composition

CAP hosts 2,450 MESA baseline cases. All participants were considered asymptomatic at recruitment. The 2011 CAP database paper reported 2,864 MESA volunteers contributed at that time; the CAP website now lists 2,450. CAP lists no breakdown by sex, age or ethnicity for the shared cases.

Acquisition

MESA scanned on Siemens and GE 1.5 T systems. Cine images were acquired with a gradient recalled echo (GRE) sequence in short-axis slices from base to apex and in three long-axis planes. CAP stores images as de-identified DICOM. The CAP page warns that GRE cine differs from the SSFP cine used in current protocols.

Annotations

The MESA MRI core laboratory at Johns Hopkins traced ventricular contours with MASS 4.0 and contributed them to CAP, but CAP states that it does not share these contours or the MESA clinical variables. CAP itself built about 300 3D left ventricular models from the images semi-automatically with the Cardiac Image Modeller (CIM) software.

Known limitations

• Access is per research project: CAP and the MESA contributors review a proposal before a data distribution agreement is signed. The MESA-specific agreement users sign is not published.
• When this entry was verified, the CAP MESA page said the dataset was temporarily unavailable while the data sharing agreement with the contributor was being renewed.
• The GRE cine images may not transfer to methods built for SSFP cine.
• No manual contours, demographics or clinical parameters are distributed with the images.