University of Crete (UOC)
Ημι-αυτοματοποιημένη ποσοτικοποίηση του περιβοθρικού τριχοειδικού δικτύου σε ασθενείς με σακχαρώδη διαβήτη
Abstract
dc:descriptionPurpose: The perifoveal capillary network (PCN) constitutes an important, complex, retinal microstructure that covers the metabolic needs and plays a pivotal role in the maintenance of the central visual acuity. Therefore, the imaging and assessment of this structure is of great significance for monitoring the retinal function in various retinal vascular diseases. Hence, the purpose of the presented work was to develop and evaluate the performance of a novel semi-automated computerized method that detects and quantifies the PCN calculating metrics in fluorescein angiography (FA) images. Furthermore, the morphology of the PCN and the foveal avascular zone (FAZ) was assessed quantitatively applying this method on FA images of diabetic patients with variable severity of diabetic retinopathy (DR) and controls. Materials and methods: The work was developed and conducted in two parts. In the first part of the study, an algorithm that detects the perifoveal capillary bed as depicted in high resolution grayscale FA images and creates a one-pixel-wide skeleton of the PCN, was developed using MatLab software. Besides PCN detection, the capillary density and the branch point (BP) density in two circular areas centered on the center of the FAZ of 500μm and 750μm radius were calculated by the algorithm. Three consecutive FA images with distinguishable PCN from 40 subjects were used for analysis. Both manual and semi-automated detection of the PCN and branch points were performed and compared. Three different intensity thresholds were used for detection of the PCN to optimize the method defined as mean(I)+0.05*SD(I), mean(I) and mean(I)-0.05*SD(I), where I is the grayscale intensity of each image analyzed and SD the standard deviation. Limits of agreement (LoA), intraclass correlation coefficient (ICC) and Pearson’s correlation coefficient (r) were calculated to evaluate the proposed method. Thereafter, in the second part of the thesis, the algorithm using the intensity threshold with the highest repeatability and the greatest agreement with the manual method detected in the first part of the study was applied on high resolution FA images from non-diabetic subjects and diabetic patients with different stages of DR. The method provided semi-automated measurements of the PCN density and the BP density in two central circular areas of 500 and 750μm radius, as well as manual measurements of the FAZ including the surface, acircularity and perimeter of the FAZ, for all subjects. Results: In the first part of the study, using the mean(I)-0.05*SD(I) as threshold the average difference of PCN density between the semi-automated and manual method was calculated 0.219 deg-1 (0.335 deg-1) at the 500μm radius and 0.413 deg-1 (0.617 deg-1) at the 750μm radius. The LoA were found -0.546 to 0.986 and -0.905 to 1.748 deg-1 for each radius area, respectively. Regarding BP density, the average difference of BP density between the semi-automated and manual method was zero for both areas; LoA were -0.018 to 0.019 and -0.003 to 0.002 branch points/deg2, respectively. The other two intensity thresholds provided wider limits of agreement for both PCN and BP density. Furthermore, the semi-automated algorithm showed great repeatability for both PCN and BP density (ICC>0.91 for both metrics) in the 500μm radius area and fair repeatability (ICC>0.75 for both metrics) in the area of 750μm radius. The second step of the study included a group of 320 subjects; 78 non-diabetic controls, 71 diabetic patients without (DR) findings (no DR), 90 patients with non-proliferative DR (NPDR) and 81 patients with proliferative DR (PDR). Gradually lower values for PCN and BP density and inversely gradually higher values for all FAZ metrics were detected as DR evolves. Applying one-way ANOVA statistically significant differences were detected for all PCN and FAZ metrics among several groups except for PCN and BP density in the central circular zone of 500μm radius. The PCN density at 750μm did not differ significantly between controls and no DR subjects, and between NPDR and PDR patients. On the contrary, BP density was significantly different only between controls and PDR patients. Similarly with PCN density, FAZ area and perimeter was not found to differ significantly between controls and no DR subjects, and between NPDR and PDR patients. FAZ acircularity was significantly different between every compared group except for controls and no DR patients. Conclusion: The developed semi-automated algorithm seems to provide readings of PCN metrics in agreement with those of the manual capillary tracing in FA images. The parameters assessed in the second part of the thesis seem to reflect the PCN and FAZ abnormalities related to diabetes mellitus and vary accordingly across the different DR stages. Until today, literature and research have focused on optimization of algorithms on non-invasive imaging techniques of the PCN and thus, the analysis and quantification of PCN in FA images lacks progress and refinement, although FA still remains the gold standard imaging method for the diagnosis, classification and monitoring of several retinal vascular diseases including DR. Moreover, early detection of PCN ischemia even in subclinical DR cases could be beneficial for the maintenance of the retinal vasculature function.
Degree
thesis:*- Grantor dc:publisher
- University of Crete (UOC)
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kapsala, Zoi
- Καψάλα, Ζωή
Subjects
dc:subject × 14Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/58498