Improved Alzheimer's Disease Classification from MRI Scans Using Deep Learning

The global rise in life expectancy has increased the elderly population, leading to a higher prevalence of Alzheimer's disease. Early detection is crucial, as timely intervention can slow disease progression and improve patients' quality of life. Deep learning provides an effective solution for automating brain MRI classification by identifying complex patterns within medical images. This study investigates the implementation of the VGG-16 deep learning architecture in conjunction with two data augmentation techniques: Albumentations and CutMix for the classification of Alzheimer's disease using MRI images. The classification task encompasses four categories: Mild Demented, Moderate Demented, Very Mild Demented, and Non-Demented. Model performance was evaluated using accuracy, precision, recall, and F1-score as assessment metrics. The VGG-16 model achieved an accuracy of 87.89% when employing a combination of Albumentations and CutMix, compared to 78.12% with CutMix as a single technique, 73.44% with Albumentations as a single technique, and 35.16% without augmentation. These results demonstrate that combining Albumentations and CutMix enhances data diversity and model generalization, thereby improving classification accuracy.