ICA-based automatic classification of PET images from ADNI database

Authors:
Yang Wenlu;He Fangyu;Chen Xinyun;Huang Xudong
Affiliations:
College of Information Engineering, Shanghai Maritime University, Shanghai, China;College of Information Engineering, Shanghai Maritime University, Shanghai, China;College of Information Engineering, Shanghai Maritime University, Shanghai, China;Department of Radiology, Brigham and Women's Hospital and Harvard Medical School, Boston, MA
Venue:
ICONIP'11 Proceedings of the 18th international conference on Neural Information Processing - Volume Part I
Year:
2011

Citing 2
Cited 0

18F-FDG PET imaging analysis for computer aided Alzheimer's diagnosis

Information Sciences: an International Journal
Principal component analysis-based techniques and supervised classification schemes for the early detection of Alzheimer's disease

Neurocomputing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Due to the unknown pathogenesis and pathologies of Alzhei-mer's Disease(AD) that it brings about the serious of social problems, it is urgent to find appropriate technology for early detection of the Alzheimer's Disease. As a kind of function imaging, FDG-PET images can display lesions distribution of AD through the glucose metabolism in brain, directly reflect lesions of specific areas and the metabolic features, to diagnose and identify AD. In the paper, we propose a novel method combining Independent Component Analysis(ICA) and voxel of interest in PET images for automatic classification of AD vs healthy controls(HC) in ADNI database. The method includes four steps: preprocessing, feature extraction using ICA, selection of voxel of interest, and classification of AD vs healthy controls using Support Vector Machine(SVM). The experimental results show that the proposed method based on ICA is able to obtain the averaged accuracy of 86.78%. In addition, we selected different number of independent component for classification, achieving the average accuracy of classification results with the biggest difference only 1.47%. According to the experimental results, we can see that this method can successfully distinguish AD from healthy controls, so it is suitable for automatic classification of PET images.