Artificial Intelligence -Driven Neuroimaging and Omics Approaches for Alzheimer's disease Biomarker Discover - A Narrative Review of Deep Learning in Education

Abstract

Alzheimer's disease (AD) remains the leading cause of dementia worldwide, and the search for reliable, early, and biologically interpretable biomarkers has become a central research priority. This descriptive narrative review examines how artificial intelligence (AI) is being applied across two complementary evidence bases for AD biomarker discovery: the large and comparatively mature neuroimaging literature (MRI, PET, and related modalities) and the smaller but rapidly emerging molecular omics literature (genomics, transcriptomics, proteomics, and metabolomics). Genomic analyses of APOE, APP, and PSEN1/2 variants, informed by genome-wide association studies, continue to identify key genetic risk factors, and deep-learning genomics models have reported classification accuracies as high as 98.78% in distinguishing AD from healthy controls. Complementary transcriptomic, proteomic, and metabolomic studies using explainable deep learning and gradient-boosted models have reported accuracies and areas under the curve (AUC) in the 0.88-0.95 range on blood- and plasma-based samples, though typically on smaller cohorts than neuroimaging studies. In parallel, neuroimaging-based AI models applied to MRI and PET data have reported diagnostic accuracies frequently exceeding 95-99% on large, standardized datasets such as ADNI and OASIS. This review summarizes representative, independently verified results from both fields, identifies some common methodological trends (such as bioinformatics-driven feature selection, architectures based on attention mechanisms and explainable AI), explores the value of incorporation of omics and neuroimaging data into multimodal AI systems for precision diagnosis of Alzheimer's disease, and discusses the complementarity of both, the current limitations and potential for future applications.