3D-HOG Features –Based Classification using MRI Images to Early Diagnosis of Alzheimer’s Disease

Alzheimer's is categorized as one severe dementia with a shrinking brain shape and reduced brain volume overall. The correlation between shrinking of the brain shape and decreasing volume also affects the change in texture shape. In this proposed study, a new feature descriptor called the Histogram of Oriented Gradients from Three Orthogonal of Planes (HOG-TOP) is proposed to extract the dynamic texture features of 3D MRI brain images. The extension of the local binary pattern is the complete local binary pattern of sign magnitude (CLBPSM) as a feature extraction method also introduced. Because the features were in a high dimensional space, then probabilistic principal component analysis (PPCA) is used as one method of dimensionality reduction method. Furthermore, the random forest classifier is used for binary classification of Alzheimer's, Mild Cognitive Impairment (MCI) and normal. In the experimental results show that the 3D HOG-TOP features provide the highest sensitivity value compared to CLBPSM-TOP and hybrid feature for all classifications.