A Transfer Learning Based Approach for Alzheimer's Disease Classification

Alzheimer's disease (AD) is a degenerative neurological condition that impairs cognitive function and awareness in humans. It has a direct impact on the development of cognitive capacity and brain function. The incidence of Alzheimer's disease is progressively increasing, especially among persons aged 60 and older, finally resulting in mortality. It disrupts nearly all cognitive processes that the brain can perform. Over time, this leads to a reduction in brain size and volume. At now, there is no known therapy capable of reversing the neurological deterioration caused by Alzheimer's disease. Prompt recognition of symptoms is essential for enabling early detection and therapy. Over the years, extensive study has been carried out on the utilization of machine learning for the diagnosis and categorization of Alzheimer's disease. This research proposes a transfer learning-based approach that utilizes the VGG-16 deep learning architecture. The proposed framework offers a method for categorizing AD, consisting of the following steps: For data input, import the pre-trained VGG-16 model and exclude the fully linked layers. Revise the current fully connected layers or introduce new ones to align with the objective of classifying Alzheimer's disease. Implement weight freezing on the initial layers and do fine-tuning on the model using the OASIS dataset to tailor it for the specific characteristics of brain MRI scans. Divide the dataset into training, validation, and test sets. Continue the training process of the updated model using the training dataset. Monitor the model's performance on the validation set and adjust hyperparameters accordingly. The model's performance is assessed using accuracy, precision, recall, and F1 score as metrics on the OASIS datasets. The proposed method has achieved an accuracy of 94.4%, a precision of 90%, a recall of 99.9%, and an F1 score of 95%.