Exploring Brain Imaging and Genetic Risk Factors in Different Progression States of Alzheimer's Disease Through OSnetNMF-Based Methods
Abstract
Alzheimer's disease (AD) is a neurodegenerative disease with multiple factors for which there is currently no effective treatment. Mild cognitive impairment (MCI) is an early disease that may progress to AD. Multimodal imaging genetics can integrate imaging data and genetic data to gain a deeper understanding of the progression of complex diseases and the individual variations. Exploring the pathogenic mechanisms from normal progression to MCI and ultimately leading to AD will contribute to understanding the pathogenesis of AD and providing insights for early diagnosis. As an effective joint feature extraction and dimensionality reduction method, non-negative Matrix Factorization (NMF) and its improved methods, especially the netNMF method, have been widely used in the field of multimodal analysis for mining brain imaging and genetics data by considering the interactions between different features. However, many of these methods overlook the importance of the coefficient matrix and do not address issues related to data accuracy and feature redundancy. To cope with this problem, we propose an orthogonal sparse network nonnegative matrix factorization (OSnetNMF) algorithm by adding orthogonal and sparse constraints based on netNMF. By establishing linear relationships between structural magnetic resonance imaging (sMRI) and corresponding gene expression data and with orthogonal and sparsity constraints as conditions, the proposed OSnetNMF is able to reduce the redundancy of features and decrease the correlation between the data, resulting in more accurate and reliable biomarker extraction. Experiments showed that the OSnetNMF algorithm allows us to accurately identify the risk regions of interest (ROIs) and key genes that characterize the progression of AD, as well as some obvious trends of ROI pairs, such as l4thVen-HIF1A, rBst-MPO, rBst-PTK2B, etc. Meanwhile, comparative experiments showed that the proposed algorithm can find more disease-related biomarkers with better reconstruction performance.
Domains
Medical ImagingOrigin | Files produced by the author(s) |
---|---|
Licence |