@article{BasAlcolea:111355,
author = "Bas Alcolea, Paula and Gracia Tello, Borja del Carmelo",
title = "{Reproducibilidad de un nuevo sistema automático
(CAPILLARY.IO) en el análisis de imágenes
capilaroscópicas}",
year = "2020",
note = "INTRODUCTION: Nailfold Capillaroscopy is a simple,
inexpensive and non-invasive technique that allows
microvascular damage to be observed, gaining recent
importance in the diagnosis, monitoring and prognosis of
many diseases with microangiopathy. However, the
variability in the results interpretation has led to the
development of new computerized systems that allow the
automatic analysis of capillaroscopic images. OBJECTIVES:
to compare the degree of agreement between the automatic
system Capillary.io and a gold standard obtained from the
agreement of 9 expert capillaroscopists and to know the
degree of the interobserver reliability. To demonstrate the
validity of the system to detect normal and enlarged
capillaries, hemorrhages, megacapillaries, ramifications
and tortuosities. METHODS: a cross-sectional study was
performed in which 300 random and anonymous nailfold
capillaroscopic images (1165 capillaries) were analyzed by
9 experienced observers. The degree of interobserver
agreement was calculated from the 5 users. Likewise, the
system performed an automatic assessment of the images and
their agreement with the gold standard was calculated
(interobserver agreement greater than 5, 6, 7, 8 and 9
successively). The validity of the program for each
variable was also analyzed using sensitivity and
specificity, positive and negative predictive values, and
likelihood ratios, as well as their degree of agreement
using the weighted kappa statistic (95% CI, p <0.05). The
programs used for statistical calculations were SPSS 22.0
and EPIDAT 3.0. RESULTS: the degree of interobserver
agreement was 76.5% for the agreement of 5 or more
observers, progressively decreasing to 15.4% for the 9
observers. Capillary.io obtained higher levels of
agreement, reaching 97.7% for the 9 observers.
Statistically significant results were obtained in the
automated detection of all the morphological alterations
analyzed. CONCLUSIONS: Capillary.io demonstrated a high
degree of agreement with the gold standard, stronger with
greater consensus among observers. He was able to detect
with great sensitivity and specificity hemorrhages and
megacapillaries, very relevant alterations in
microangiopathies. More studies will be necessary to
analyze its functionality to detect dilatations,
tortuosities and ramifications, although the results have
been promising.",
}