A new plasmonic sensor combined with artificial intelligence could accurately distinguish osteoarthritis from rheumatoid arthritis by analyzing synovial fluid samples, according to a new study.
In the study, researchers described a device that achieved 98.1% accuracy, 97.3% sensitivity, and 100% specificity in diagnosing arthritis types.
The researchers developed the sensor by reducing gold onto cellulose acetate paper with ascorbic acid, forming a dense urchin-like gold nanoarchitecture. This structure amplified Raman signals, allowing detection of chemical differences in joint fluid without labels or dyes. The sensor was tested on samples from 40 patients with osteoarthritis (OA) and 80 patients with rheumatoid arthritis (RA), with most participants around 60 years old.
"By utilizing a support vector machine (SVM)-based machine learning model, the Raman signals of OA and RA groups were successfully classified with 98.1% accuracy," said lead study author Boyou Heo, of the Korea Institute of Materials Science, and colleagues.
The synovial fluid samples were analyzed using surface-enhanced Raman scattering technology. The researchers identified significant differences at 13 Raman shifts, including 442 cm⁻¹, 535 cm⁻¹, and 1,663 cm⁻¹, pointing to distinct biochemical profiles between OA and RA.
Further metabolite profiling revealed that 1,6-anhydro-beta-D-glucose, inosine, and mannitol were more abundant in patients with RA; while acetylspermine, adonitol, and D-proline were higher in patients with OA. Pearson correlation coefficients and nonnegative matrix factorization were applied to link specific Raman features with known metabolites, enhancing the biomarker discovery process.
The device was also evaluated for classifying RA severity based on white blood cell counts. Samples were grouped into three ranges: below 5,000, between 5,000 and 20,000, and above 20,000 cells per µL. The artificial intelligence–assisted platform classified severity with an accuracy of 98.1%. Receiver operating characteristic curves showed area under the curve values of 0.96, 0.95, and 1.00 for the respective groups.
Funding for the research came from the Bio&Medical Technology Development Program of the National Research Foundation and the Technology Innovation Program of the Korean government.
The researchers noted that the platform could support rapid, low-cost arthritis diagnosis and help clinicians differentiate between inflammatory and degenerative joint diseases using synovial fluid biomarkers. No conflicts of interest were reported.
Source: Small