Result summary of three different methods on ADNI dataset.

“*” denotes the GMM+MS is significantly better than t-test approach at a statistical level of 0.05. The p-value is calculated by the corrected paired t-test tailored for comparing learning algorithms [41]). LIBSVM [40] is used to build the SVM models. A linear kernel is used, with a grid search for parameter optimization. Grid search considers only the optimization of the penalty parameter C in the linear SVM, selecting the value of C yielding the best classification result based on the training data. After the best value of C is found, we apply it to the test data. AUC: area under ROC curve. Each experiment was repeated 10 times with a 10-fold cross-validation. P: results using “primary sensorimotor cortex” region for intensity normalization. G: results using “grand mean” method for intensity normalization. Results from AIC are noted in brackets, following the BIC results.Result summary of three different methods on ADNI dataset.