Comparative Analysis of CNN Algorithms for Alzheimer's Detection

There is an estimation that around 60% of dementia sufferers and their care takers are affected by Alzheimer's disease, which can adversely affect people psychological and physiological well-being, and their economic and social stability. Since medical staff numbers are scarce, automatic Alzheimer's disease diagnosis is becoming increasingly significant for relieving medical staff workloads and increasing diagnostic accuracy. Recent years, numerous researchers have shown that deep learning can perform detection and prediction at high speeds with exceptional accuracy without compromising. This research study proposing the contribution of the advanced deep learning method comparison to improve the diagnostic accuracy of Alzheimer's disease in this study. With the help of MRI scans, EfficientNetB4, VGG19 and DenseNet121 models are used for Alzheimer's disease detection. Transfer learning (TL), is one of the advanced techniques that uses domain knowledge from heterogeneous datasets and is proposed to improve the fine-tuning of hyperparameters and, therefore, detection accuracy. During the experimentation observations the DenseNet121 architecture produced better results compared with other architectures. The proposed methods VGG19, EfficientNetB4, and DenseNet121 achieved validation accuracies of 97.25, 76.56, and 99.8 percent.