Optimized Alzheimer’s Detection from Brain Scans using CNN-Based Approach

Abstract

A progressive neurological disease, Alzheimer's disease affects millions of individuals globally. For Alzheimer's disease to be effectively treated and managed, early detection and stage classification are essential. Our proposal in this paper is to use brain MRI scans to classify Alzheimer's disease stages using a Convolutional Neural Network (CNN) based method. Using Bayesian and Grid Search approaches, we optimize the models and investigate both simple and complex CNN architectures, including transfer learning with VGG16. Notably, the Grid Search Advanced CNN model has a validation accuracy of 0.5269, the Bayesian Basic CNN model has a validation accuracy of 0.5044, and the Grid Search Basic CNN model has a validation accuracy of 0.4203. The most accurate validation, 54.84%, is obtained by the sophisticated CNN model with Bayesian optimization, according to our data.  Based on validation accuracy, this data is useful for choosing the best model for a particular task.

We also go over the implications of our findings for early Alzheimer's disease detection and present a thorough comparison of the models.