Background/Purpose: Nailfold videocapillaroscopy (NVC) is a valuable non-invasive tool for assessing microvascular damage in autoimmune diseases. However, its application in idiopathic inflammatory myopathies (IIM) remains limited, partly due to variability in interpretation. Artificial intelligence (AI)-based software offers the potential for objective, standardized capillaroscopic analysis. The CapIAMI project aimed to explore the role of automated NVC evaluation in IIM patients across a large international cohort.

Methods: We conducted an ambispective, observational, international, multicenter study including patients diagnosed with IIM according to established criteria. Standardized NVC images were analyzed using Capillary.io, an AI-based platform extracting quantitative microvascular parameters. Clinical data, laboratory values (CPK, aldolase, LDH, CRP, ESR, and ferritin), and autoantibody profiles were collected at the time of NVC. Disease activity was determined based on clinical evaluation. Predictive models were developed using machine learning algorithms, with performance assessed by cross-validation metrics.

Results: A total of 272 IIM patients from 25 centers were included. The median age was 58.0 years (IQR 41.3–67.0), and 68.2 % were female. Ethnic distribution was 56.2 % Caucasian, 37.7 % Latin American, and 6.0 % other groups. Dermatomyositis was the predominant subtype (53.7 %), followed by antisynthetase syndrome (15.8 %) and overlap myopathy (11.8 %) (Table 1). Clinically active disease was present in 49.3 % of patients. The most frequent active manifestations were myositis (80.5 %), cutaneous involvement (42.7 %), articular symptoms (29.3 %) and interstitial lung disease (ILD) (24.26%) (Table 1).Patients with active disease exhibited significantly higher serum CK and aldolase, elevated CRP (all p < 0.05) and increased prevalence of ILD and late-scleroderma NVC pattern (both p < 0.001). Autoantibody profiles are detailed in Table 2. The most prevalent specificities were anti–Ro-52 (31.6 %), anti–Mi-2 (19.9 %) and anti–PM/Scl (13.2 %), with lesser frequencies for other myositis‐specific and myositis‐associated antibodies.Combining AI-derived capillaroscopic metrics with laboratory biomarkers yielded an overall precision of 0.746 in cross-validated predictive models. Capillary tortuosity, ramification abnormalities, megacapillary presence and hemorrhagic patterns constituted the principal predictive features, with capillary ramification (p = 0.005) and average limb width (p = 0.026) remaining independent predictors. UMAP embedding of these features demonstrated clear segregation of early-, active- and late-scleroderma–like patterns from non-specific/normal patterns (Fig. 1).

Conclusion: Automated NVC analysis combined with conventional laboratory biomarkers improves the prediction of disease activity in IIM patients. These preliminary findings underscore the added value of standardized capillaroscopic evaluation as a complementary tool in clinical practice. Further validation in larger cohorts is needed to confirm and expand these observations.

Figure 1. UMAP projection of capillaroscopic variables from CapIAMI patients. Each dot represents an individual capillaroscopy, coloured by capillaroscopic pattern and shaped by IIM subtype. The figure shows distinct clustering based on microvascular characteristics, supporting the potential of AI-assisted stratification in idiopathic inflammatory myopathies.

Table 1. Cohort & Key Clinical Features

Table 2. Autoantibody Positivity

« Back to ACR Convergence 2025

ACR Meeting Abstracts - https://acrabstracts.org/abstract/artificial-intelligence-based-capillaroscopy-and-laboratory-biomarkers-to-predict-disease-activity-in-idiopathic-inflammatory-myopathies-preliminary-data-from-the-capiami-cohort/