Quantitative Analysis Of Optical Coherence Tomography Angiography Based Retinal Microvascular Morphology In Diabetic Retinopathy

Part Ⅰ Binarization of optical coherence tomography angiography based images and its measurement repeatbilityPurposeOptical coherence tomography angiography(OCTA)is a novel technique and a functional extension of OCT that detect retinal microvasculature invasively.Several commercial OCTA devices have been applied in clinical practice,but limitations have been noticed in the aspects of vessel map segementation and quantitative methods.This part of our study is aim to binarize the retinal microvasculature in OCTA image,and test its measurement repeatability.MethodsFor OCTA image binarization,we developed a complex method with multilevel thresholding and vessel filtering for image Precessing using MATLAB.Vessel perfusion density(VPD)was calculated on binarized OCTA image,and intraclass correlation coefficient(ICC)and repeatbiltiy coefficient were used for assessing measurement agreement.ResultsThe custom-bulit software in MATLAB binarized OCTA images with both normal and diseased retinas at the level of superficial and deep levels.Superficial and deep VPD from binarized OCTA images were significantly correlated with commercial vessel density(both P<0.05).Intraobserver and interobserver ICCs of VPD mesurements on binarized OCTA image with normal and diseased retina ranged from 0.978 to 0.993 and from 0.939 to 0.982,respectively.The ICC results indicated a good agreement of VPD measurement on our binarized OCTA image.ConclusionsThe developed complex method could successfully binarized OCTA images with either normal or diseased retina,and showing a high repeatability for VPD measurement.Part Ⅱ Multifractal and Lacunarity Analyses of Microvascular Morphology in Eyes with Diabetic Retinopathy Using Optical Coherence Tomography AngiographyPurposeTo evaluate the degree of microvascular impairment in diabetic retinopathy using multifractal and lacunarity analyses and to compare the diagnostic ability between traditional Euclidean measures(fovea avascular zone area and vessel density)and fractal geometric features.MethodsThis retrospective cross-sectional study included a total of 143 eyes of 94 patients with different stages of diabetic retinopathy.The retinal microvasculature was imaged by projection removed optical coherence tomography angiography.We examined the degree of association between fractal metrics of the retinal microvasculature and diabetic retinopathy severity.The area under the receiver operating characteristic curve was used to estimate the diagnostic performance.ResultsWith increasing diabetic retinopathy severity,the multifractal spectrum shifted towards the left bottom and exhibited less left skewness and asymmetry.The vessel density,multifractal features and lacunarity measured from the deep capillary plexus were strongly associated with diabetic retinopathy severity.The multifractal feature D5 showed the highest diagnostic ability.The combination ot multifractal features further improved the discriminating power.ConclusionMultifractal and lacunarity analyses can be potentially valuable tools for assessment of microvascular impairments in diabetic retinopathy.Multifractal geometric parameters exhibit a better discriminatory performance than Euclidean measures,particularly for detection of the early stages of diabetic retinopathy.Part Ⅲ Comparison of optical coherence tomography angiography based metrics for early detection of retinal microvascular impairments in diabetes mellitusPurposeTo determine the ability of nonperfusion,vessel density and morphological measurements using projection-resolved optical coherence tomography angiography(OCTA)to detect early retinal microvasculature impairments in diabetes mellitus(DM).MethodThis cross-sectional study included 66 eyes from 66 patients with type 2 DM and 35 control eyes from 35 healthy subjects.Foveal avascular zone-related metrics and extrafoveal avascular area were measured by an OCTA system.Vessel density and fractal dimension were calculated in OCTA images with and without a skeletonization process.The vessel diameter index and vessel tortuosity were computed.Receiver operating characteristic curve analysis estimated diagnostic performances.ResultsDilated capillary diameter was observed in the deep capillary plexus in the diabetic groups.Compared to control eyes,eyes with no diabetic retinopathy(DR)(both P<0.05)and with early DR(both P<0.001)showed significant stepwise decreases in vessel density and fractal dimension in skeletonized OCTA deep capillary plexus images.The skeletonization process improved vessel density and fractal dimension measurement discrimination powers.ConclusionVessel density and fractal dimension metrics in the skeletonized deep capillary network may be better biomarkers than those in the nonskeletonized network for preclinical DR,showing higher diagnostic performance for detecting early DR.