Multimodal OCT-A and Oxidative Stress Biomarker Fusion Network for Early Choroidal Vascular Changes inAge-Related Macular Degeneration

Age-related macular degeneration (AMD) is a common degenerative disease in the elderly, and its early stages are often accompanied by imperceptible microvascular changes in the choroid and elevated levels of oxidative stress. Traditional unimodal detection methods have obvious limitations in spatial resolution or metabolic state reflection, which are difficult to meet the sensitivity and accuracy requirements for early screening. To this end, this paper proposes a multimodal neural network architecture that integrates optical coherence tomography angiography (OCT-A) and oxidative stress biomarkers to construct a synergistic identification mechanism under a unified feature space. The model adopts a dual-stream encoder structure, which combines the attention mechanism with the Transformer fusion module to improve the adaptability and expressiveness of different modal features. The model is trained and validated with real clinical datasets, and the experimental results show that this method outperforms the existing unimodal methods in terms of AUC, sensitivity, specificity, etc., and shows good robustness especially in early lesion identification. The study demonstrates the complementary advantages of structural information and molecular mechanism, proves the potential of multimodal fusion strategy in the early screening of AMD, and provides an effective technical support for individualized prediction and precise intervention.