روش های مختلف چند متغیره برای طبقه بندی خودکار اطلاعات MRI در بیماری آلزایمر و اختلال خفیف شناختی

Automated structural magnetic resonance imaging (MRI) processing pipelines and different multivariate techniques are gaining popularity for Alzheimer’s disease (AD) research. We used four supervised learning methods to classify AD patients and controls (CTL) and to prospectively predict the conversion of mild cognitive impairment (MCI) to AD from baseline MRI data. A total of 345 participants from the AddNeuroMed cohort were included in this study; 116 AD patients, 119 MCI patients and 110 CTL individuals. High resolution sagittal 3D MP-RAGE datasets were acquired and MRI data were processed using FreeSurfer. We explored the classification ability of orthogonal projections to latent structures (OPLS), decision trees (Trees), artificial neural networks (ANN) and support vector machines (SVM). Applying 10-fold cross-validation demonstrated that SVM and OPLS were slightly superior to Trees and ANN, although not statistically significant for distinguishing between AD and CTL. The classification experiments resulted in up to 83% sensitivity and 87% specificity for the best techniques. For the prediction of conversion of MCI patients at baseline to AD at 1-year follow-up, we obtained an accuracy of up to 86%. The value of the multivariate models derived from the classification of AD vs. CTL was shown to be robust and efficient in the identification of MCI converters.

Alzheimer’s disease (AD) is one of the most common forms of neurodegenerative disorders. The disease is related to pathological amyloid depositions and hyper-phosphorylation of structural proteins which leads to progressive loss of cognitive function, synaptic dysfunction and structural changes in the brain. Magnetic resonance imaging (MRI) has been extensively investigated in AD and, consistent with pathology, very early changes have been demonstrated in the hippocampus and entorhinal cortex. However, no imaging measure currently provides a reliable prediction of which patients with mild cognitive impairment (MCI) will rapidly progress to develop AD (O’Brien, 2007 and Ries et al., 2008). With MRI it is possible to measure both regional (hippocampus/entorhinal cortex) and global (whole brain) atrophy, which are considered sensitive surrogate markers, capable of quantifying the extent of brain degeneration in dementia (Apostolova et al., 2006). It is possible to obtain multiple volumetric and cortical thickness measures from high resolution MRI by automated segmentation techniques. It has previously been shown that a combination of global and prediction of conversion of MCI to AD as compared with using manual volumetric measures of the hippocampus (still considered to be the gold standard) (Westman et al., 2011c). Further, multivariate analysis using multiple regions in the brain as input gives better accuracy for AD classification and MCI prediction than visual assessment (Scheltens scale for medial temporal lobe atrophy) performed by an experienced radiologist (Westman et al., 2011a). Positive results have also been reported using a whole-brain grey-matter-based support vector machine (SVM) approach (Kloppel et al., 2008a). A large number of multivariate methods have been introduced in recent years for classifying individual patients with AD using structural MRI (Vemuri et al., 2008, Plant et al., 2010, Kloppel et al., 2008b, Teipel et al., 2007, Davatzikos et al., 2008 and Magnin et al., 2009: Fan et al., 2008). However, the lack of studies using multiple methods on the same data has made it difficult to directly compare the results of the different techniques. In this study we combine multiple morphometric measures derived from an automated pipeline to directly compare different multivariate classifiers. The specific aims were (1) to compare linear and non-linear multivariate methods for the classification of AD vs. cognitively normal controls (CTL) using an automated pipeline; (2) to test the resulting classifiers in predicting AD conversion from the prodromal stage of the disease, MCI; (3) to assess the effect of age, education and APOE genotype in the prediction of AD vs. CTL; and (4) to identify the optimal classifier(s).