Raman spectroscopy is being explored as a new way to identify monosodium urate (MSU) and calcium pyrophosphate (CPP) crystals in synovial fluid. However, interpreting these spectra needs particular expertise. Thus, the method does not apply to clinicians.

Researchers conducted a retrospective study automating the identification process of MSU and CPP crystals in synovial fluid using machine learning techniques.

They tested the system in a real-world setting at the outpatient rheumatology department and collected synovial fluid from 446 patients with various rheumatic conditions across 3 centers. The Raman spectroscope analyzed samples, using 246 for training and 200 for validation. Trained observers classified spectra, and two one-against-all classifiers – PCA and SVM – were designed 1 for MSU and CPP detection. 

The results showed that the classification accuracy for CPP, based on the 2023 ACR/EULAR/CPPD criteria, was 96.0% (95% CI: 92.3-98.3), and for MSU, using the 2015 ACR/EULAR gout criteria, was 92.5% (95% CI: 87.9-95.7). Overall accuracy for identifying pathological crystals was 88.0% (95% CI: 82.7-92.2). The model distinguished pathological crystals from artifacts and other particles like microplastics.

Investigators concluded that machine learning can accurately classify complex Raman spectra from clinical samples, providing an objective diagnosis that does not rely on the medical examiner’s opinion. This approach offers a promising alternative for diagnosing crystal-related conditions

Source: academic.oup.com/rheumatology/advance-article/doi/10.1093/rheumatology/keae472/7747839