Author: Thibaud Mautuit (France)
Co-authors: Pierre Cunnac, Carol Cheung, Emanuele Trucco, Vincent Daien, Christophe Chiquet
Purpose
Retinal vascular diameter can be measured from retinal photographs using semi-automated software tools. We aim to compare the measurements from three most used packages in the literature: SIVA (Singapore I Vessel Assessment), VAMPIRE (Vascular Assessment and Measurement Platform for Images of the Retina) and IVAN (Integrative Vessel Analysis) and to generate conversion algorithms for CRAE and CRVE measurements, between SIVA and IVAN and between SIVA and VAMPIRE.
Setting/Venue
We analyzed 223 retinal photographs from 133 human subjects (CR-2 Canon fundus camera, Canon™) with each software tool, computing vessel widths measurements and the central retinal artery equivalent (CRAE), central retinal vein equivalent (CRVE) and arteriole-to-venule ratio (AVR).
Methods
The agreement was assessed using Bland-Altman plots and intra-class correlation coefficients. We propose a transformation between measurements from different system using linear regression, and validated it using bootstrapping and root-mean-square-error (RMSE).
Results
The agreement between VAMPIRE (version 3.1, 2017 and IVAN was poor to moderate: the mean difference was 20.2 µm (95% LOA, -12.2 to 52.6 µm) for CRAE, 21.0 µm (95% LOA, -17.5 to 59.5 µm) for CRVE, and 0.023 (95% LOA, -0.12 to 0.16) for AVR. The agreement between VAMPIRE and SIVA was also poor to moderate: the mean difference was 36.6 µm (95% LOA, -12.8 to 60.4 µm) for CRAE, 40.3 µm (95% LOA, 5.6 to 75.0 µm) for CRVE, and 0.037 (95% LOA, -0.071 to 0.15) for AVR. The agreement between IVAN and SIVA was good to excellent: the mean difference was 16.4 µm (95% LOA, -4.25 to 37.0 µm) for CRAE, 19.3 µm (95% LOA, 0.09 to 38.6 µm) for CRVE, and 0.015 (95% LOA, -0.09 to 0.11) for AVR. Using the conversion algorithm, the agreement was higher between SIVA measurements and IVAN-derivate SIVA approximates (RMSE was 4.88 µm for CRAE and 5.19 µm for CRVE) than between SIVA measurements and VAMPIRE-derivate SIVA approximates (RMSE was 7.27 µm for CRAE and 9.69 µm for CRVE).
Conlusions
Comparing CRAE, CRVE and AVR estimates from three widely used retinal vessel measurement software tools, we found an excellent agreement between SIVA and IVAN, a poor to moderate agreement between VAMPIRE (version 3.1, 2017) and the two other software tools. We propose an algorithm converting IVAN and VAMPIRE measurements into SIVA estimated measurements, that could be used to homogenize sets of vascular measurements obtained with different packages.
Financial Disclosure
none
Comments
-
