Multimodal Neuroimaging Fusion for Alzheimer's Disease: An Image Colorization Approach With Mobile Vision Transformer

• DOI: 10.1002/ima.23158
• Publication Date: 2024-08-26T08:09:57Z
• Authors: Modupe Odusami; Robertas Damasevi...; Egle Milieskaite‐...; Rytis Maskeliunas
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

ABSTRACT Multimodal neuroimaging, combining data from different sources, has shown promise in the classification of Alzheimer's disease (AD) stage. Existing multimodal neuroimaging fusion methods exhibit certain limitations, which require advancements to enhance their objective performance, sensitivity, and specificity for AD classification. This study uses use a Pareto‐optimal cosine color map performance visual clarity fused images. A mobile vision transformer (ViT) model, incorporating swish activation function, is introduced effective feature extraction Fused images Disease Neuroimaging Initiative (ADNI), Whole Brain Atlas (AANLIB), Open Access Series Imaging Studies (OASIS) datasets, obtained through optimized transposed convolution, are utilized model training, while evaluation achieved using that have not been same databases. The proposed demonstrates high accuracy across achieving 98.76% Early Mild Cognitive Impairment (EMCI) versus LMCI, 98.65% Late (LMCI) AD, 98.60% EMCI 99.25% Normal (CN) ADNI dataset. Similarly, on OASIS AANLIB, precision CN 99.50% 96.00%, respectively. Evaluation metrics showcase model's precision, recall, F1 score various binary classifications, emphasizing its robust performance.