Brain MRI Image Analysis for Alzheimer’s Disease (AD) Prediction Using Deep Learning Approaches

Alzheimer’s disease (AD) is an irreversible, progressive neurodegenerative condition that causes memory impairment decline. Alzheimer’s disease (AD) stands as one of the most pervasive chronic ailments affecting the elderly population, presenting a considerably high prevalence rate. The significance of early detection in the treatment of AD cannot be overstated, owing to the potentially severe cognitive and neurological impairments that emerge as the disease progresses. Swift and accurate diagnosis holds a pivotal role in curbing the extent of brain deterioration that becomes more pronounced in the later stages of the disease’s course. Many strategies have been used to determine the pattern of disease to detect Alzheimer’s disease. The similarity of brain patterns in older people and different stages of Alzheimer’s disease makes categorization difficult. In recent times, the domain of medical imaging has witnessed a surge in the prominence and accomplishments of Deep Learning (DL) and Machine Learning (ML) techniques. These advancements have not only taken center stage in the evaluation of medical images but have also ignited substantial enthusiasm for enhancing the diagnosis of Alzheimer’s disease. Within this landscape, Deep Learning and machine models have risen to prominence, outpacing conventional Machine Learning methods in precision and efficiency, particularly in the realm of Alzheimer’s disease detection. A pivotal approach that has gained traction involves the utilization of pre-trained Convolutional Neural Network (CNN) models. These models, primed with extensive prior learning, exhibit an enhanced ability to categorize various stages of Alzheimer’s disease. Through the adept utilization of Deep Learning models, the categorization of six distinct phases of Alzheimer’s disease becomes attainable, offering a promising avenue for refining diagnostic accuracy and comprehension. Normal control (NC), Significant memory concern (SMC), Early mild cognitive impairment (EMCI), Late mild cognitive impairment (LMCI), and AD are all effectively diagnosed by the models (AD). The raw data were rigorously pre-processed before being used in any categorization approach. CNN models viz. EfficientNet, MobileNet, DenseNet, Resnet, AlexNet, InceptionV2, and NASNet are trained to classify the disease. For differentiating distinct phases of AD, our models evaluate average accuracies of 99.79%, 98.74%, 84.39%, 84.28%, 83.74%, 94.39%, and 88.28%, respectively. Other parameters such as recall, AUC, loss values, F1 score, precision, etc., are also determined in this paper. Later, we compare all these pre-trained models on the bases of their performance metrics and visualize their results.