Optimizing Pretrained ResNet Models for Alzheimer’s Disease Detection in MRI Images

Alzheimer’s disease (AD) is a serious concern in healthcare because it affects cognitive skills and quality of life. Recent advances in medical imaging and machine learning techniques provide great opportunities for early identification and diagnosis of Alzheimer’s disease. This study explores the optimization of pretrained ResNet models for AD detection in MRI images, focusing on ResNet50, ResNet101, and ResNet152 architectures. The research comprises two phases: transfer learning and fine-tuning. Results show that deeper model architectures perform better in the transfer learning phase, with ResNet152 demonstrating superior performance, achieving an accuracy of 99.417%. Fine-tuning significantly enhances model accuracy, with all models achieving nearly perfect scores. ResNet152 emerges as the top-performing model, emphasizing the importance of fine-tuning in optimizing classification accuracy for AD detection. Future study might concentrate on fine-tuning model architectures and using sophisticated regularization approaches to improve AD classification. This work helps to the development of accurate diagnostic methods for early Alzheimer’s disease identification, allowing for prompt intervention and treatment techniques.