High-order correlation detecting in features for diagnosis of Alzheimer's disease and mild cognitive impairment. (August 2019)

• Record Type: Journal Article
• Authors: Ding, Yi; Luo, Chuanji; Li, Chang; Lan, Tian; Qin, ZhiGuang
• Is Part Of: Biomedical signal processing and control. Volume 53(2019)
• Publication Date: 2019-08
• Subjects: Alzheimer's disease diagnosis -- Mild cognitive impairment diagnosis -- Hypergraph -- High-order correlation -- Feature selection; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2019.101564
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: It can be found that there are various high-order correlations among features. Detecting the high-order correlations among features can effectively improve the feature selection method. The high-order correlations among features can be detected by the hypergraph method to effectively diagnose the AD/MCI. The proposed method with high-order correlation has more obvious advantages on classifying the multi-class task. Abstract: As shown in the literature, the identification of discriminative features of Alzheimer's Disease (AD) and Mild Cognitive Impairment (MCI) can improve the diagnostic accuracy. Besides, many researches have proven that the feature selection method, which also considers the relationship among features, shows improvements in performance for AD diagnosis. However, most existing feature selection methods only consider the pairwise correlation between features. Instead, when adopting these features to diagnose the AD, the high-order correlated relationships among features need to be taken more into consideration. In this paper, a novel classification framework for diagnosing the AD and MCI has been proposed to address these problems. This framework mainly consists of three processes: feature extraction, feature selection and classification. The feature extraction is mainly used to extract texture features and morphometric features from brain MR images. The feature selection process is to select discriminative features. This process firstly measures the high-order correlation among features by borrowing the idea of hypergraph theory and then to generate the optimal feature subset based on the high-order correlation. The classification process will adopt the feature subset to finish the classification task. The main contribution of this framework is to consider the high-order correlation among features instead of pairwise correlation when classifying the AD. Experiments on the Alzheimer's Disease Neuroimaging Initiative ADNI database shows that the proposed method obtains better performance than other state-of-the-art counterpart methods. Overall, the experiment result demonstrates the effectiveness of the proposed framework.