الفهرس 
ACKNOWLEDGEMENTSOnly 14 pages are availabe for public view
 |  | from | 157 |  |  |
| | | from | 157 | | |
Abstract
Alzheimer’s disease (AD) is a progressive neurological disorder described by memory loss, cognitive impairment, and difficulty performing daily tasks. AD is a global condition that mainly affects those over 65 years old, and early diagnosis of AD is very important. Manual AD diagnosis is time consuming and error prone. Deep learning, machine learning, and computer vision algorithms significantly impact AD identification and categorization. This research proposes a system for AD multigrade diagnosis. The system has been applied and tested on the ADNI dataset and worked on two modalities of images : magnetic resonance imaging (MRI) and positron emission tomography (PET). Most diagnoses today rely on a single imaging modality, necessitating multimodal imaging. Technological advancements significantly impact clinical diagnosis using structural, diffusion, functional MRI, and functional imaging. Automated clinical diagnosis and prognosis have already greatly benefited from these recent advancements, and they are expected to continue rapidly improving with ML and DL techniques. The proposed system in this thesis has been applied to classify AD into three classes : AD, NC, and MCI. The novelty of this thesis depends on using a vision transformer ViT-B16 with a pre-trained DL model (MobileNetV1). These used models have many advantages that help extract the most significant features, such as the depthwise convolution of MobileNetV1 and the division of images into patches with ViT-B16. These models have been applied as feature extractors only, and finally, SVM classification has been applied. There are many comparisons that have been tested, and our proposed methodology achieves 98.5% for accuracy, 98.5% for recall, 98.6% for precision, and 98.5% for F1-score. In the future, different modalities with various classes from the ADNI dataset will be used.