Enhancing Early Alzheimer's Detection: A Framework Combining Data Augmentation and Transfer Learning

Abstract

The early detection of Alzheimer's disease (AD) is crucial for timely treatment and management. However, the development of accurate machine learning models is challenged by the limited availability of labeled medical imaging data and inherent variability in brain scans. This study introduces a framework that integrates data augmentation with transfer learning to enhance the performance of deep learning models in early AD detection. By employing advanced data augmentation techniques on MRI and PET images and refining pre-trained models, this approach addresses the challenges of small datasets and overfitting, thereby improving the model robustness and efficiency. Experimental results demonstrate that the proposed framework achieves state-of-the-art performance in early AD detection, offering a promising solution for real-world clinical applications, including the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. Data augmentation methods, such as random rotations, flipping, scaling, and intensity adjustments, have been utilized to diversify training data and enhance model generalization. Pre-trained models, including ResNet-50, VGG-16, and EfficientNet-B0, were fine-tuned using the augmented dataset. The models were evaluated using metrics, such as accuracy, precision, recall, F1-score, and AUC-ROC. These findings indicate that the proposed framework surpasses the baseline model and other contemporary methods. EfficientNet-B0 achieved the highest performance, with an accuracy of 93.8% and an AUC-ROC of 0.97. High AUC-ROC values (≥ 0.94) reflected the models' capability to effectively distinguish between AD and non-AD cases. The combination of data augmentation and transfer learning significantly enhances the accuracy and efficiency of early AD detection, making the framework suitable for clinical applications. Future research should aim to expand the framework to incorporate multimodal data, explore advanced augmentation techniques, and develop methods to interpret model predictions, thereby increasing their reliability for clinical practitioners