Premium
P4‐298: Combining MRI and CSF measures for classification of Alzheimer's disease and prediction of mild cognitive impairment conversion
Author(s) -
Westman Eric,
Muehlboeck JSebastian,
Lovestone Simon,
Simmons Andrew
Publication year - 2011
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2011.09.043
Subject(s) - ctl* , magnetic resonance imaging , neuroimaging , cerebrospinal fluid , cognitive impairment , biomarker , alzheimer's disease neuroimaging initiative , medicine , positron emission tomography , psychology , multivariate analysis , disease , nuclear medicine , neuroscience , radiology , immunology , biochemistry , chemistry , antigen , cd8
Background: Dubois and coworkers have suggested a revision of the NINCDS-ADRDA criterion for the diagnosis of Alzheimer’s disease (AD). This new criterion is still centered on a clinical core of early and significant episodic memory impairment, but also includes at least one abnormal biomarker among magnetic resonance imaging (MRI), positron emission tomography (PET) and cerebrospinal fluid (CSF). The aim of this project is to investigate the potential of combining MRI and CSF using multivariate data analysis for classifying AD subjects and subjects with mild cognitive impairment (MCI) compared to healthy controls (CTL). Methods: A total of 373 subjects from the Alzheimer’s disease Neuroimaging Initiative (ADNI) are included in the current study (AD 1⁄4 96, MCI 1⁄4 166 and CTL 1⁄4 111). Freesurfer was used to generate regional volume and regional cortical thickness measures. A total of 60 variables were used for multivariate analysis (34 cortical thickness measures, 23 volumetricmeasures and 3CSFmeasures: Aß42, t-tau and p-tau).Orthogonal partial least square to latent structures (OPLS) models were created for CSF, MRI and CSF+MRI measures to distinguish between AD vs. CTL andMCI vs. CTL.Results: Combining CSF andMRI gave the best predictive accuracy for distinguishing between AD andCTL.We received an accuracy of 91.8% for the combined model compared to 81.6% for CSF measures and 87.0% for MRI measures. The combined model also gave the best accuracy when comparing MCI and CTL (75.8%). Within the MCI cohorts 36 subjects converted at 18 month follow-up, 88.9% of which were correctly classified. Conclusions: Combining MRI and CSF measures in a multivariate model gave better predictive accuracy, than using them separately. This shows that combining different measures are advantageous for the early diagnosis of AD.Themethodalso showspotential for selectingpopulations for clinical trials.