Alzheimer’s Disease Recognition Applying Non-Negative Matrix Factorization Characteristics from Brain Magnetic Resonance Images (MRI)

Department of Information Technology, GRIET, India

^* Corresponding author: sivamangarajuboramchu@gmail.com

Abstract

To more accurately depict Alzheimer’s disease (AD) and projecting clinical outcomes while taking into account advancements in clinical imaging and substantial learning, several experts are gradually using ConvNet (CNNs) to remove deep intensity features from gathering images. A small deep learning algorithm called the principal component analysis network (PCA-Net) creates multi-faceted channel banks to verify the accuracy of voluminous head part assessments. After binarization, block wise histograms are constructed to obtain picture properties. PCANet is less adaptable because multi-facet channel banks are built with test data, resulting in PCA-Net features with thousands or even millions of aspects. The non-negative matrix factorization tensor decomposition network, or NMF-TD-Net, is an information-free organization based on PCA-Net that we present in this study to address these issues. Instead of PCA, staggered channel banks are made to test nonnegative matrix factorization (NMF). By applying tensor decomposition (TD) to a higher-demand tensor derived from the learning results, the input’s dimensionality is reduced, resulting in the final image features. The support vector machine (SVM) in our technique uses these properties as input to diagnose, predict clinical score, and categorize AD.