Swish-Enhanced Hybrid Model for Alzheimer’s Diagnosis Using VGG16 and ResNet50 on MRI Data

Alzheimer’s Disease is an extreme neurological illness that deteriorates over time, affecting memory, communication, and mental skills resulting in irreversible brain damage. Early detection is key, as timely treatment can help manage symptoms and slow down the progress of the disease. We created a hybrid deep learning model combining the VGG16 and ResNet50 architectures to classify brain MRI images into ‘Demented’ and ‘Nondemented’ groups. We also used the Swish activation function, which provides smoother gradients and improved performance than ReLU (Rectified Linear Unit). Through the use of k – fold cross – validation, the model’s average accuracy was 89.79% and final test accuracy was 91.44%. Other evaluation measures, like recall, precision and F1 – score, all showed an average of 0.91 for both classes, showing that the network works good on unseen data. While this work focuses on binary classification, future research will expand to multi – class classification to detect several phases of cognitive decline.