Enhancing Alzheimer's Disease Detection: A Comparative Study of Deep Learning Techniques with Transfer Learning and Custom CNN Models

Alzheimer's disease is a progressive neurological disorder that primarily impacts memory, cognition, and behavior, making it the leading cause of dementia in older adults. Timely and accurate diagnosis is essential for effective treatment and management. This study explores the application of deep learning methods to detect Alzheimer's disease using a dataset of 12,000 medical images, divided into four groups of 3,000 each. The images underwent processing with Histogram of Oriented Gradients (HOG) to identify key features. Various machine learning techniques were applied, including Principal Component Analysis (PCA), Standard Scaling, and Stratified K-Fold cross-validation, to enhance the analysis across four algorithms. For the deep learning approach, we employed data augmentation and developed a custom Convolutional Neural Network (CNN), which achieved an accuracy of 89%. However, utilizing transfer learning with the VGG16 model significantly improved performance, reaching an accuracy of 94%. This comparative study highlights the superiority of the VGG16 model over the custom CNN for detecting Alzheimer's disease in medical images, emphasizing the transformative potential of advanced deep learning techniques, especially those leveraging transfer learning, in enhancing diagnostic accuracy for this condition.