Towards Next-Generation Artificial Intelligence Models for Integrated Diagnostics Embedding Eye-Brain Connections

Abstract The brain and eye share profound and homologous relationships in terms of development, anatomy, and microcirculatory structures, yet current analyses typically examine them in isolation, limiting cross-organ insights and the development of integrated artificial intelligence (AI) models. Such a fragmented paradigm also constrains clinical understanding despite growing evidence of shared, anatomically connected, and predictive relationships between ocular and cerebral phenotypes in both brain and eye diseases. In this context, this review introduces an Eye-Brain Connection Pyramid, a multiscale organizational framework for examining eye-brain relationships at three complementary biological levels: genetic and genomic susceptibility, molecular and protein processes, and neurovascular and tissue phenotypes. The pyramid does not represent a serial causal pathway. Across these levels, the framework distinguishes shared vulnerability and parallel pathology, direct brain-to-eye effects, retinal correlates and predictors of cerebral or systemic disease, and putative eye-to-brain causal contributions, which remain insufficiently established in humans. This framework may inform future clinical research and the development of AI models integrating multimodal cross-organ data, whose incremental diagnostic value over existing single-organ and clinical approaches requires prospective validation. Furthermore, this review presents a comprehensive AI research roadmap, spanning longitudinal prognostic models, explainable algorithms for clinical adoption, validation of incremental clinical value, fundus-based brain health screening setups, and multi-site collaboration. By outlining these priorities, the proposed Eye-Brain Connection Pyramid provides an organizational framework for investigating integrated eye-brain diagnostics and for evaluating the potential clinical value of cross-organ AI in risk stratification, longitudinal monitoring, and healthcare delivery.