/Uploads/file/20190226/20190226190783618361.pdf
Individual identifcation using multi-metric of DTI in Alzheimer’s disease and mild cognitive impairment
Ying-Teng Zhang(张应腾) and Shen-Quan Liu(刘深泉)*
School of Mathematics, South China University of Technology, Guangzhou 510640, China
Abstract: Accurate identifcation of the most relevant white matter (WM) tracts linked to Alzheimer’s disease (AD) and mild cognitive impairment (MCI) is crucial so as to improve diagnosis techniques and to better understand the neurodegenerative process. In this work, we aim to apply machine learning method for individual identifcation and identify the discriminate features associated with AD and MCI. DTI scans of 48 patients with AD, 39 patients with late MCI (LMCI), 75 patients with early MCI (EMCI) and 51 agematched healthy controls (HC) are acquired from the ADNI database. In addition to the common fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (DA) and radial diffusivity (RD) metrics, there are two novel metrics, named local diffusion homogeneity used spearman’s rank correlation coefcient (LDHs) and Kendall’s coefcient concordance (LDHk), which are taken as classifcation metrics. The recursive feature elimination (RFE) method for support vector machine (SVM) and logistic regression (LR) combined with leave-one-out cross validation (LOOCV) are applied to determine the optimal feature dimensions. Then SVM and LR are implemented the classifcation and compared the classifcation performance each other. The results can be well proved that multi-type and multi-regional features combination received higher accuracy than single metric. In addition, the permutation test, receiver operating characteristic (ROC) curves and area under the curve (AUC) are validated the robustness of the classifers. Finally, the uncinated fasciculus, cingulum, corpus callosum, corona radiate, external capsule and internal capsule have been regard as the most important WM tracts to identify AD, MCI and HC. Our fndings reveal a guidance role for machine-learning based image analysis on clinical diagnosis.
Keywords: Alzheimer’s disease, mild cognitive impairment, magnetic resonance imaging, classification