AI-Assisted Plasmonic Diagnostics Platform for Osteoarthritis and Rheumatoid Arthritis With Biomarker Quantification Using Mathematical Models

Abstract

Osteoarthritis (OA) and rheumatoid arthritis (RA) are major causes of functional impairment, disability, and chronic pain, leading to a substantial rise in healthcare costs. Despite differences in pathophysiology, these diseases share overlapping features that complicate diagnosis, necessitating early, more accurate, and cost-effective diagnostic tools. This study introduces an innovative plasmonic diagnostics platform for rapid and accurate label-free diagnosis of OA and RA. The sensing platform utilizes a highly dense urchin-like gold nanoarchitecture (UGN), which enhances the surface plasmonic area to significantly amplify the Raman signal. The feasibility of the developed platform for arthritis diagnosis is demonstrated by analyzing the synovial fluid (SVF) of patients. Assisted by a machine learning model, Raman signals of OA and RA groups are successfully classified with high clinical sensitivity and specificity. Metabolic biomarkers are further investigated using mathematical models of combined Pearson correlation coefficient (PCC) and non-negative matrix factorization (NMF), suggesting valuable insights for arthritis biomarker quantification. In addition, RA severity is studied using the sensing platform by classifying results from the hematology test, achieving successful stage discrimination. This platform offers a versatile, affordable, and scalable in-clinic arthritis diagnostic solution with potential applications in diagnosing and monitoring other diseases through biofluid analysis.