Nanostructured SERS-AI integration for noninvasive biochemical detection and disease classification in retinal vascular and chorioretinal diseases

Retinal and choroidal vascular diseases, such as retinal vein occlusion, diabetic macular edema, central serous chorioretinopathy, and wet age-related macular degeneration, are major causes of irreversible vision loss across diverse age groups. Conventional imaging provides structural but not molecular information, often delaying the opportunity to modify ineffective regimens. We developed an integrated diagnostic platform combining surface-enhanced Raman spectroscopy (SERS) with machine learning algorithms for label-free biochemical profiling of retinal diseases. Reproducible AI-interpretable spectral features were extracted from microliter-scale aqueous humor (AH) samples using gold-coated ZnO nanorod SERS chips. The nanostructured substrate acted as an electromagnetic field enhancer, amplifying Raman signals to enable data-driven disease classification directly from trace biofluids. We implemented a multi-stage AI framework incorporating support vector machine (SVM), linear (LDA), and quadratic discriminant analysis (QDA) to optimize clinical workflows. The PC-SVM model achieved an outstanding primary screening accuracy of 96.45%, with a 10-fold cross-validation (CV) accuracy of 92.93% and an area under the curve (AUC) of 0.994. PC-LDA and PC-QDA models reached stable 10-fold CV scores of 87.63% and 86.45%, respectively, demonstrating high resistance to overfitting and strong generalizability across disease phenotypes. Responder prediction for anti-VEGF therapy exceeded 90% accuracy before visible anatomical improvement, confirming that biochemical alterations precede structural recovery. This AI-assisted SERS platform provides a rapid, minimally invasive strategy for biochemical profiling and therapeutic monitoring of retinal diseases. This technology bridges the gap between structural biomarkers and true biochemical disease activity, offering a point-of-care tool for personalized ophthalmic care and real-time treatment guidance in clinical settings. • SERS-AI enables rapid, personalized monitoring of retinal vascular diseases. • Label-free biochemical profiling overcomes limits of structural imaging. • Gold-coated ZnO nanorods sense Raman signals from 90% accuracy before structural changes.