3D-DWT Improves Prediction of AD and MCI

摘要

In order to predict subjects with Alzheimer's disease (AD) and mild cognitive impairment (MCI) from normal elder controls (NC) more accurately, we compared two different kinds of discrete wavelet transform (DWT) based feature extraction techniques: multi-slice 2D-DWT and 3D-DWT. The dataset contained the magnetic resonance (MR) images of 178 subjects consisting of 97 NCs, 57 MCIs, and 24 ADs. We constructed two multiclass kernel support vector machine (MKSVM) classifiers based on multislice 2D-DWT features and 3D-DWT features, respectively. 5-fold cross validation was employed to obtain the out-of-sample estimate. Each classifier runs 10 times. Welch's t-test showed that the mean of the overall accuracy by 3D-DWT was higher than that of multislice 2D-DWT, and the difference was statistically significant (p=0.0146).