Aim: To evaluate the correlation between inner retinal layer thinning and the foveal microvasculature in type 2 diabetes using optical coherence tomography angiography (OCTA).Methods: A cross-sectional study involved 155 diabetic eyes. All patients were divided into two groups based on diabetic retinopathy (DR) grade: no DR (NDR, n=80) and mild-to-moderate non-proliferative DR (NPDR, n=75). Foveal avascular zone (FAZ) area, FAZ circularity index, FAZ perimeter, vessel density and perfusion index of parafoveal and perifoveal area were calculated using OCTA. The thickness of the macular ganglion cell/inner plexiform layer (mGCIPL) was measured using OCT.Results: In both superficial and deep retinal capillary layers (SRL and DRL), FAZ areas in the NDR (0.38 mm2 , 0.49 mm2) and NPDR (0.38 mm2, 0.48 mm2) were greater than those in the control (0.33 mm2, 0.43 mm2). The FAZ circularity index, vessel density and perfusion index in the NDR (0.63, 17.8/mm, 0.32) and NPDR (0.63, 17.5/mm, 0.32) were smaller than those in the control (0.69, 19.6/mm, 0.39). mGCIPL thickness was significantly correlated with FAZ area in the SRL and DRL, as well as with FAZ circularity index, vessel density and perfusion index in the NDR and NPDR. In multivariate regression analysis, the FAZ circularity index (OR=12.2) and vessel density of the parafovea (OR=1.95) were correlated with mGCIPL thinning.Conclusion: OCTA revealed that early foveal microcirculatory alterations in diabetic eyes were related to mGCIPL thickness, regardless of the presence of DR. The decrease in FAZ circularity and parafoveal vessel density were highly correlated with mGCIPL thinning.
Objective: With the scenario of limited labeled dataset, this paper introduces a deep learning-based approach that leverages Diabetic Retinopathy (DR) severity recognition performance using fundus images combined with wide-field swept-source optical coherence tomography angiography (SS-OCTA). Methods: The proposed architecture comprises a backbone convolutional network associated with a Twofold Feature Augmentation mechanism, namely TFA-Net. The former includes multiple convolution blocks extracting representational features at various scales. The latter is constructed in a two-stage manner, i.e., the utilization of weight-sharing convolution kernels and the deployment of a Reverse Cross-Attention (RCA) stream. Results: The proposed model achieves a Quadratic Weighted Kappa rate of 90.2% on the small-sized internal KHUMC dataset. The robustness of the RCA stream is also evaluated by the single-modal Messidor dataset, of which the obtained mean Accuracy (94.8%) and Area Under Receiver Operating Characteristic (99.4%) outperform those of the state-of-the-arts significantly. Conclusion: Utilizing a network strongly regularized at feature space to learn the amalgamation of different modalities is of proven effectiveness. Thanks to the widespread availability of multi-modal retinal imaging for each diabetes patient nowadays, such approach can reduce the heavy reliance on large quantity of labeled visual data. Significance: Our TFA-Net is able to coordinate hybrid information of fundus photos and wide-field SS-OCTA for exhaustively exploiting DR-oriented biomarkers. Moreover, the embedded feature-wise augmentation scheme can enrich generalization ability efficiently despite learning from small-scale labeled data.
1 December 2018 in American Journal of Ophthalmology
Purpose: To investigate the patterns of fundus autofluorescence (FAF) abnormalities in patients with central serous chorioretinopathy (CSC). Methods: This cross-sectional observational study included 126 eyes of 118 patients who were diagnosed with central serous chorioretinopathy from December 2006 to April 2012 at Kyung Hee University Hospital, Seoul, Korea. Fundus autofluorescence patterns were analyzed with spectral domain optical coherence tomography and visual acuity. Results: Fundus autofluorescence patterns were grouped as blocked (38.9%), mottled (8.7%), hyper (31.0%), hyper/hypo (13.5%), or descending tract (8.0%). The duration of symptoms was 7.8 (±20.4), 28.3 (±31.8), 42.5 (±69.1), 163.8 (±183.5), and 174.5 (±162.3) days in the blocked, mottled, hyper, descending tract, and hyper/hypo groups, respectively (P < 0.001). The blocked FAF group had the best visual acuity (P = 0.011). The intact ellipsoid zone on the spectral domain optical coherence tomography was mostly found in the blocked FAF group (P < 0.001), and the disrupted ellipsoid zone was commonly exhibited in the hyper/hypo and descending tract groups. Disrupted external limiting membrane line on the spectral domain optical coherence tomography was seen in two patients of the descending tract group only. Conclusion: The FAF abnormalities in central serous chorioretinopathy show multiple patterns and are related with the chronicity and visual acuity. Fundus autofluorescence patterns in central serous chorioretinopathy are helpful when considering the timing of treatment and predicting the disease status.
1 December 2019 in International Journal of Medical Informatics