Enhanced Detection of Alzheimer’s and Mild Cognitive Impairment: Leveraging Advanced Preprocessing and Convolutional Neural Networks

Alzheimer’s disease (AD) and moderate cognitive impairment (MCI) are neurodegenerative disorders that cause significant impairment in brain function. This study examines the effectiveness of combining preprocessing methods, including Gaussian and median filters, with techniques such as CLAHE, image resizing, normalization, and skull stripping. The goal is to improve the visibility of critical brain areas and enable accurate diagnosis of AD and MCI using Magnetic Resonance Imaging (MRI) and Convolutional Neural Networks (CNNs) while addressing the challenges associated with these conditions. The Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset images were crucial for our analysis since they included a three-class classification of AD, MCI, and Cognitively Normal (CN) cases. The research employs sophisticated preprocessing techniques and CNN architectures to thoroughly assess each class’s unique characteristics, ensuring precise identification. The CNN architectures under investigation, namely ResNet-50, DenseNet-121, and EfficientNet-B0, undergo a comprehensive MRI preprocessing procedure to enhance image clarity and accuracy. The results indicate that the meticulous combination of Gaussian and median filters significantly improved CNN’s efficiency in identifying Alzheimer’s and MCI. The DenseNet-121 model exhibited outstanding performance across all categories, with an overall accuracy of 99.71% and an F1-score of 94.68%.