A NEW artificial intelligence-powered system, CAPI-Detect, has demonstrated the ability to significantly improve the diagnosis of systemic sclerosis (SSc) and differentiate it from other vascular conditions using nailfold videocapillaroscopy (NVC), according to a study published in Rheumatology. By leveraging machine learning (ML), this model surpasses traditional methods in identifying capillary patterns associated with SSc while reducing examiner-related bias.
NVC is the gold standard for diagnosing SSc and differentiating primary from secondary Raynaud’s phenomenon. Until now, the CAPI-Score algorithm has been widely used, but it relies on a limited set of capillary variables. Researchers sought to enhance classification accuracy by developing a more sophisticated ML model incorporating 24 capillary architecture-related variables extracted through automated NVC analysis.
The study involved 1,780 capillaroscopies randomly and blindly assessed by multiple trained observers. Using the CatBoost machine learning model, the research team trained their system on capillaroscopies where observers reached either partial or full consensus. The model demonstrated impressive accuracy in different classification tasks: 91.2% for distinguishing SSc from non-SSc cases, 81.2% for differentiating among SSc capillary patterns, and 74.6% for identifying normal versus non-specific patterns. When trained on fully agreed classifications, accuracy improved further to over 90% across all tasks.
Unlike previous diagnostic models, CAPI-Detect identified novel capillary variables that significantly influence SSc classification, highlighting the power of ML in uncovering previously overlooked diagnostic markers. By offering an objective, quantitative approach to capillary structure assessment, CAPI-Detect reduces reliance on human interpretation, potentially leading to earlier and more accurate SSc diagnoses in clinical practice.
Aleksandra Zurowska, EMJ
Reference
Lledó-Ibáñez GM et al. CAPI-detect: machine learning in capillaroscopy reveals new variables influencing diagnosis, Rheumatology, 2025;DOI: 10.1093/rheumatology/keaf073.
