An MRI based Histogram Oriented Gradient (HOG) and Deep Learning Approach for Accurate Classification of Mild Cognitive Impairment and Alzheimer's Disease

Shakir Khan^1,2*J Arunnehru³Anwar Basha⁴Abdullah M Albarrak²Aleem Ali⁵

• ¹College of Computer and Information Sciences, Imam Muhammad ibn Saud Islamic University, Riyadh, Saudi Arabia
• ²Imam Muhammad ibn Saud Islamic University, Riyadh, Saudi Arabia
• ³SRM Institute of Science and Technology, Chennai, Tamil Nadu, India
• ⁴Rajalakshmi Institute of Technology (RIT), Chennai, Tamil Nadu, India
• ⁵Chandigarh University, Mohali, Punjab, India

Alzheimer's disease (AD) is a prevalent kind of dementia that primarily impacts the central nervous system, leading to a deterioration in several cognitive functions, including memory impairment. The current focus of research is on non-invasive methods for early identification of Alzheimer's disease (AD), as the timely diagnosis of this illness has significance in enhancing patient care and treatment effectiveness. This research develops a robust feature extraction method and a set of three classifiers to identify the optimum classification results. The proposed system utilized a t1-weighted brain MRI (Magnetic Resonance Image) as the input data. The system focused on the deep extraction of data with Harris Corner interest points and the Histogram-Oriented Gradients (HOG) method and stratification with a Support Vector Machine (SVM), K-Nearest Neighbour (KNN), and a deep learning-based pipeline architecture called a Deep Neural Network. The proposed system focused on classifying the major three stages of AD as control normal (CN), Mild Cognitive Impairment (MCI), and AD. The frameworks were able to classify reasonably enough, with an accuracy of KNN at 88%, SVM at 91.5%, and DNN at 95.6%. In addition, our approach with DNN has been compared with cutting-edge deep learning models to validate our framework's performance.