Computer-aided classification of Alzheimer's disease based on support vector machine with combination of cerebral image features in MRI

13th South-East Asian Congress of Medical Physics 2015

物理学;医药卫生

Jongkreangkrai, C.^1 ; Vichianin, Y.^1 ; Tocharoenchai, C.^1 ; Arimura, H.^2

Department of Radiological Technology, Faculty of Medical Technology, Mahidol University, Thailand^1

Faculty of Medical Science, Kyushu University, Fukuoka, Japan^2

关键词: Alzheimer's disease;    Brain hemispheres;    Brain images;    Cerebral images;    Computer Aided Classification;    Entorhinal cortex;    Receiver operating characteristic analysis;    Statistically significant difference;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/694/1/012036/pdf
DOI  :  10.1088/1742-6596/694/1/012036

学科分类：卫生学

来源: IOP

【 摘 要 】

Several studies have differentiated Alzheimer's disease (AD) using cerebral image features derived from MR brain images. In this study, we were interested in combining hippocampus and amygdala volumes and entorhinal cortex thickness to improve the performance of AD differentiation. Thus, our objective was to investigate the useful features obtained from MRI for classification of AD patients using support vector machine (SVM). T1-weighted MR brain images of 100 AD patients and 100 normal subjects were processed using FreeSurfer software to measure hippocampus and amygdala volumes and entorhinal cortex thicknesses in both brain hemispheres. Relative volumes of hippocampus and amygdala were calculated to correct variation in individual head size. SVM was employed with five combinations of features (H: hippocampus relative volumes, A: amygdala relative volumes, E: entorhinal cortex thicknesses, HA: hippocampus and amygdala relative volumes and ALL: all features). Receiver operating characteristic (ROC) analysis was used to evaluate the method. AUC values of five combinations were 0.8575 (H), 0.8374 (A), 0.8422 (E), 0.8631 (HA) and 0.8906 (ALL). Although "ALL" provided the highest AUC, there were no statistically significant differences among them except for "A" feature. Our results showed that all suggested features may be feasible for computer-aided classification of AD patients.