Abstract:
The development of a new system that aims at the detection of Alzheimer's disease using patient's Magnetic Resonance Imaging (MRI) scans featuring deep learning as a technological part has been done by researchers due to the escalation in the number of Alzheimer's disease around the world and the challenges in healthcare systems. The need for precise and high-performing diagnostics tools committed two deep learning algorithms to find a way forward. This research takes advantage of the EfficientNetB3 architecture, which is an appropriate choice between the general computational requirements and precision of performance. The model reaches a high level of accuracy in distinguishing different levels of dementia through transfer learning with pre-trained weight. It has a special ability to identify cases that involve actual disorders. We get a model that can handle datasets that represent different demographics and modality sensitivity using our proposed system and data selection; this is a model that is adaptable and robust - the qualities of practical implementation. In this paper, a thorough analysis of the experimental design, model training, as well as methods for data pre-training are conducted. The measures of assessment, such as the confusion matrix report for classification, determine the good performance of the model. First, it develops an effective identification component across Alzheimer's patients but also unveils relevant details regarding dropout prediction, proving that machine learning is indeed a universal and important category of algorithms having broader application spheres. This contribution involves building a classifying system that would highlight merits and flaws. The final stage would be to improve the medical care as well as diagnostic efficiency of area hospitals.