King's College London

Research portal

AddNeuroMed and ADNI: similar patterns of Alzheimer's atrophy and automated MRI classification accuracy in Europe and North America

Research output: Contribution to journalArticle

Eric Westman, Andrew Simmons, J-Sebastian Muehlboeck, Patrizia Mecocci, Bruno Vellas, Magda Tsolaki, Iwona Kloszewska, Hilkka Soininen, Michael W. Weiner, Simon Lovestone, Christian Spenger, Lars-Olof Wahlund, AddNeuroMed Consortium, Alzheimer's Dis Neuroimaging Initi

Original languageEnglish
Pages (from-to)818-828
Number of pages11
Issue number3
Publication statusPublished - 1 Oct 2011

King's Authors


The European Union AddNeuroMed program and the US-based Alzheimer Disease Neuroimaging Initiative (ADNI) are two large multi-center initiatives designed to collect and validate biomarker data for Alzheimer's disease (AD). Both initiatives use the same MRI data acquisition scheme. The current study aims to compare and combine magnetic resonance imaging (MRI) data from the two study cohorts using an automated image analysis pipeline and a multivariate data analysis approach. We hypothesized that the two cohorts would show similar patterns of atrophy, despite demographic differences and could therefore be combined. MRI scans were analyzed from a total of 1074 subjects (AD=295, MCI=444 and controls=335) using Freesurfer, an automated segmentation scheme which generates regional volume and regional cortical thickness measures which were subsequently used for multivariate analysis (orthogonal partial least squares to latent structures (OPLS)). OPLS models were created for the individual cohorts and for the combined cohort to discriminate between AD patients and controls. The ADNI cohort was used as a replication dataset to validate the model created for the AddNeuroMed cohort and vice versa. The combined cohort model was used to predict conversion to AD at baseline of MCI subjects at 1 year follow-up. The AddNeuroMed, the ADNI and the combined cohort showed similar patterns of atrophy and the predictive power was similar (between 80 and 90%). The combined model also showed potential in predicting conversion from MCI to AD, resulting in 71% of the MCI converters (MCI-c) from both cohorts classified as AD-like and 60% of the stable MCI subjects (MCI-s) classified as control-like. This demonstrates that the methods used are robust and that large data sets can be combined if MRI imaging protocols are carefully aligned.

View graph of relations

© 2018 King's College London | Strand | London WC2R 2LS | England | United Kingdom | Tel +44 (0)20 7836 5454