A novel MRI-biomarker candidate for Alzheimer’s disease composed of regional brain volume and perfusion variables
Article first published online: 28 JUN 2010
© 2010 The Author(s). European Journal of Neurology © 2010 EFNS
European Journal of Neurology
Volume 17, Issue 12, pages 1437–1444, December 2010
How to Cite
Luckhaus, C., Jänner, M., Cohnen, M., Flüß, M. O., Teipel, S. J., Grothe, M., Hampel, H., Kornhuber, J., Rüther, E., Peters, O., Supprian, T., Gaebel, W., Mödder, U. and Wittsack, H.-J. (2010), A novel MRI-biomarker candidate for Alzheimer’s disease composed of regional brain volume and perfusion variables. European Journal of Neurology, 17: 1437–1444. doi: 10.1111/j.1468-1331.2010.03038.x
- Issue published online: 18 NOV 2010
- Article first published online: 28 JUN 2010
- Received 16 December 2009 Accepted 10 March 2010
- Alzheimer’s disease;
- biomarker combination;
- composite marker;
- mild cognitive impairment;
- perfusion-weighted magnetic resonance imaging;
- volumetric magnetic resonance imaging
Background: Earlier evidence indicates that regional cerebral volume (rVOL) and blood flow (rCBF) variables carry independent information on incipient and early Alzheimer’s disease (AD) and combining these modalities may increase discriminant performance. We compared single variables and combinations regarding their power for optimizing diagnostic accuracy.
Methods: Twelve cognitively normal elderly controls (CN), 30 subjects with mild cognitive impairment (MCI) and 15 with mild AD were examined by structural and perfusion-weighted magnetic resonance imaging (MRI) in single sessions at 1.5 Tesla. rVOLs were measured by manual volumetry, and rCBFs were calculated with a ROI-based co-localization technique.
Results: Applying single MRI variables for the differentiation of AD versus CN, the area under curve (AUC) of receiver operating characteristic curves (ROCCs) was highest for rVOL variables (maximum of 0.972 for right amygdala). A composite marker selected and weighted by logistic regression containing left amygdalar rCBF, left hippocampal and right amygdalar rVOLs gave a diagnostic accuracy for AD versus CN of 100%. Internal cross-validation revealed a reliability of 88.9%.
Conclusions: Whilst external revalidation is mandatory employing a naturalistic sample containing disease controls, our phase I/II findings demonstrate that deducing composite markers from multimodal MRI acquisitions can optimize diagnostic accuracy for AD.