An MRI based histogram oriented gradient and deep learning approach for accurate classification of mild cognitive impairment and Alzheimer’s disease

Abstract

Alzheimer’s disease (AD) is a common form of dementia that affects the central nervous system, causing progressive cognitive decline, particularly in memory. Early, non-invasive diagnosis is critical for improving patient care and treatment outcomes. This study proposes a robust feature extraction approach combined with three classifiers to achieve optimal classification of AD stages. T1-weighted brain MRI scans were used as input data. Features were extracted using Harris Corner interest points and the Histogram of Oriented Gradients (HOG) method. Classification was performed using Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and a Deep Neural Network (DNN)-based pipeline. The proposed system classified three AD stages—Control Normal (CN), Mild Cognitive Impairment (MCI), and AD—with high accuracy: KNN (88%), SVM (91.5%), and DNN (95.6%). The DNN approach outperformed other classifiers and was further compared with state-of-the-art deep learning models, demonstrating competitive performance. These results highlight the potential of the proposed framework for early, accurate AD diagnosis using non-invasive imaging.