Premium
Semi‐automated quantification of the parafoveal capillary network in diabetic subjects
Author(s) -
KAPSALA Z,
PALLIKARIS A,
MOSCHANDREAS J,
TSILIMBARIS MK
Publication year - 2014
Publication title -
acta ophthalmologica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.534
H-Index - 87
eISSN - 1755-3768
pISSN - 1755-375X
DOI - 10.1111/j.1755-3768.2014.4745.x
Subject(s) - foveal avascular zone , diabetic retinopathy , ophthalmology , medicine , capillary action , diabetic macular edema , foveal , diabetes mellitus , biomedical engineering , nuclear medicine , retinal , materials science , endocrinology , composite material
Purpose The detection and quantification of the parafoveal capillary network in diabetic subjects using a novel semi‐automated computerized method. Methods Using the MatLab R2011a we developed an algorithm that detects automatically the parafoveal capillary bed in fluorescein angiography (FA) images. The detection starts after delineating manually the foveal avascular zone (FAZ) in a cropped 1500μm subimage and hence excluding this area from the process. The algorithm calculates the capillary density in a 1000μm area measured from the centroid of the FAZ. The method was applied on high resolution FA images from 49 subjects (age=43±19years); 13 controls, 13 subjects with diabetes mellitus (type I or II) without diabetic retinopathy (DR) findings (no DR), 15 subjects with non‐proliferative DR (NPDR) and 8 subjects with proliferative DR (PDR). Using the statistical software SPSS (version 20) we assessed the mean capillary density for the mentioned subject groups. Results Capillary density in the mentioned area was found 0.86±0.17, 1.01±0.22, 0.67±0.25 and 0.76±0.30 degrees‐1 for the studied groups, respectively (p=0.003). The no DR group did not differ significantly from the controls and the only statistically significant differences were those between the following three groups; control‐NPDR, no DR‐NPDR and no DR‐PDR. Conclusion This semi‐automated algorithm could serve as a potential tool for detecting, classifying and monitoring automatically capillary abnormalities in this area. Our preliminary results indicate that the capillary density in the central 1000μm area varies among different DR stages. However, a larger sample is required to confirm these initial data.