Madvit: multiview Alzheimer’s disease detection based on vision transformer networks

Alzheimer’s disease (AD) is marked with a progressive cognitive decline as well as memory loss. It stands for a universal health challenge, corroborating the high significance of early and accurate diagnosis for timely interventions and improved patient outcomes. To face this challenge, we introduce a method that leverages advanced deep learning techniques to promote the detection and classification of AD based on magnetic resonance neuroimaging data. Our method preprocesses structural MRI (sMRI) scans and extracts key slices from the axial, coronal and sagittal planes. These slices are individually processed through vision transformers (ViT) networks, which capture complex spatial relationships and patterns. The final classification is achieved using an ensemble learning mechanism, ensuring robust and reliable diagnostic results. As revealed experimentally, our method demonstrates a higher efficiency compared to leading methods reported in the literature.