Perifoveal capillary network quantification in young diabetic patients with subclinical or no retinopathy

To assess the morphology of perifoveal capillary network with quantitative parameters in young patients with diabetes mellitus type I (DM I) using an algorithm. Fifty-three images (33 eyes of 33 DM I patients and 20 eyes of 20 non-DM controls) were chosen retrospectively from the University Hospital of Heraklion digital fluorescein angiography database. An additional group consisting of patients with advanced DR abnormalities was included in our analysis to investigate whether our method detects alterations when they are present. The developed algorithm allows the user to manually trace the perifoveal capillary network by selecting with the cursor in a 5° × 5° subimage field of the original image, including the foveal avascular zone (FAZ), and provides measurements of the capillary density, the branch point density, and the FAZ surface in this subarea. The age in the patient group was 19 ± 5 years; age was 21 ± 8 years for the control group. Patients had a history of DM I for 11 ± 5 years. The mapping revealed a perifoveal capillary density of 2.494 ± 0.559 deg−1 in the DM I group versus 2.974 ± 0.442 deg−1 in the control group (p = 0.005). The branch point density was 3.041 ± 0.919 branch points/deg2 and 3.613 ± 1.338 branch points/deg2 in each group, respectively (p = 0.128). The FAZ area was 0.216 ± 0.061 deg2 in the diabetic group and 0.208 ± 0.060 deg2 in the control group (p = 0.672). The selected quantitative parameters tend to increase or decrease in diabetic patients, in agreement with previous studies. Among the parameters, capillary density may represent the most sensitive metric for the detection of very early diabetic changes. Further improvement of the method could contribute to the development of an automated processing tool for capillary network quantitative assessment. Évaluer la morphologie du réseau capillaire périfovéal d’après des paramètres quantitatifs chez le jeune diabétique de type 1 (DT1) au moyen d’un algorithme. Cinquante-trois images (33 yeux de 33 patients DT1 et 20 yeux de 20 témoins sans diabète) ont été choisies de manière rétrospective dans la base de données des angiographies numériques à la fluorescéine de l’hôpital universitaire d’Héraklion. Un troisième groupe de patients présentant une rétinopathie diabétique avancée a été inclus dans notre analyse afin de vérifier si notre méthode permet de mettre au jour des altérations déjà présentes. L’algorithme ainsi élaboré permet à l’utilisateur de délimiter manuellement le r