Background: Nailfold videocapillaroscopy (NVC) is the established gold standard for the non-invasive assessment of microangiopathy in Raynaud phenomenon, systemic sclerosis (SSc) and related spectrum diseases. Despite its high diagnostic value, the traditional protocol involving the examination of 8 fingers is time-consuming, creating a significant bottleneck in busy rheumatology outpatient clinics. Validating streamlined protocols that maintain diagnostic accuracy while reducing examination time is essential for the broader implementation of NVC in routine care.

Objectives: The primary objective of this study was to evaluate the diagnostic performance of various reduced-finger protocols compared to the standard 8-finger visual analysis. Additionally, we aimed to assess the integration of unsupervised artificial intelligence (AI) as a support tool and determine whether patient-specific clinical variables influence the reliability of these simplified approaches.

Methods: We performed a retrospective observational study of 49 consecutive patients (77.6% female; mean age 57±12.7 years) referred for NVC due to Raynaud’s phenomenon or suspected autoimmune connective tissue disease. The gold standard was defined as the expert visual evaluation of 8 fingers (2nd to 5th digits bilaterally). This was compared against single-hand protocols (right vs. left), specific fingers pairs, and fully automated, unsupervised AI analysis of 8 fingers (capillary.io software). A univariate analysis was conducted to examine the impact of clinical predictors, including age at onset and duration of Raynaud’s phenomenon on the diagnostic outcome to evaluate the necessity of complex multivariable adjusted models.

Results: The prevalence of a definitive scleroderma pattern (Early, Active, or Late) was 22.4% (n=11). Among the simplified protocols, the four-finger examination of the right hand demonstrated exceptional diagnostic utility, achieving a Sensitivity (S) of 90.9% and a Specificity (E) of 100%, with an almost perfect Cohen’s kappa agreement (kappa=0.94, p<0.001). The 4th finger bilateral protocol also showed high performance, matching the 90.9% sensitivity, though with a slight decrease in specificity (97.4%, kappa=0.88).

In contrast, unsupervised AI analysis showed a good sensitivity of 81.8% but a notably lower specificity (63.2%, kappa=0.32), primarily due to the over-identification of non-specific changes as scleroderma patterns. Crucially, the univariate analysis revealed that clinical variables—age (p=0.48) and Raynaud’s duration (p=0.55)—were not significant predictors of the scleroderma pattern in this cohort. This indicates that the diagnostic accuracy of the reduced right-hand protocol is highly robust and independent of the patient’s individual clinical profile.

Conclusions: NVC analysis can be significantly optimized by focusing on the right hand (fingers 2-5 ) or the 4th finger bilaterally without compromising diagnostic validity. This reduction in fingers could decrease examination time, facilitating its use in daily clinical practice. Regarding digital health integration, unsupervised AI emerges as a promising complementary tool for preliminary screening ; however, at its current stage of development, expert clinical correlation remains indispensable to ensure diagnostic precision and avoid false-positive results.

REFERENCES: [1] Smith, V. et al. Autoimmun Rev. 2019 Nov;18(11):102394.

Acknowledgments: NIL.

Disclosure of Interests: None declared.