Deep learning for predicting circular retinal nerve fiber layer thickness from fundus photographs and diagnosing glaucoma.

Purpose: This study aimed to propose a new deep learning (DL) approach to automatically predict the retinal nerve fiber layer thickness (RNFLT) around optic disc regions in fundus photography trained by optical coherence tomography (OCT) and diagnose glaucoma based on the predicted comprehensive information about RNFLT. Methods: A total of 1403 pairs of fundus photographs and OCT RNFLT scans from 1403 eyes of 1196 participants were included. A residual deep neural network was trained to predict the RNFLT for each local image in a fundus photograph, and then a RNFLT report was generated based on the local images. Two indicators were designed based on the generated report. The support vector machines (SVM) algorithm was used to diagnose glaucoma based on the two indicators. Results: A strong correlation was found between the predicted and actual RNFLT values on local images. On three testing datasets, we found the Pearson r to be 0.893, 0.850, and 0.831, respectively, and the mean absolute error of the prediction to be 14.345, 17.780, and 19.250 µm, respectively. The area under the receiver operating characteristic curves for discriminating glaucomatous from healthy eyes was 0.860 (95 % confidence interval, 0.799-0.921). Conclusions: We established a novel local image-based DL approach to provide comprehensive quantitative information on RNFLT in fundus photographs, which was used to diagnose glaucoma. In addition, training a deep neural network based on local images to predict objective detail information in fundus photographs provided a new paradigm for the diagnosis of ophthalmic diseases.

Topographic Differences in Superficial Macular Vessel Density in Eyes with Early Primary Open-Angle Glaucoma and Normal Tension Glaucoma.

INTRODUCTION: The aim of the study was to compare macular vascular microcirculation in early primary open-angle glaucoma (POAG), normal tension glaucoma (NTG), and normal subjects. METHODS: 99 patients with early glaucoma (99 eyes: 60 POAG and 39 NTG) and 78 normal subjects were included. All subjects underwent optical coherence tomography angiography scan at 6 × 6 mm macular area. Macular vessel density (VD) and perfusion density (PD) and 9 sectors were compared between the controls, POAG, and NTG groups. Linear regression analysis was used to investigate the relationship between VD and other variables including macular PD, signal strength (SS), and mean macular ganglion cell-inner plexiform layer (mGCIPL) thickness. RESULTS: Significant losses in total area of VD and PD were detected in POAG and NTG groups compared to the controls (all p < 0.01). There were no significant differences in all inner sectors of macular VD and PD between POAG and controls (all p > 0.05). Except for outer-nasal sector, all other outer sectors of macular VD and PD were significantly lower in POAG than in the controls (all p < 0.01). The inferior-inner sector and all outer sectors of VD and PD were significantly lower in NTG than in the controls (all p < 0.01). Macular VD was significantly correlated with macular PD (r = 0.99, p < 0.001), SS (r = 0.60, p < 0.001), and mGCIPL thickness (r = 0.51, p < 0.001). CONCLUSIONS: Macular microcirculation declined significantly in early POAG and NTG patients. Macular microcirculation loss in the NTG group was more central and nasal compared with that in the POAG group. A decrease in macular VD was correlated with lower macular PD, lower SS, and thinner mGCIPL thickness.

Superficial macular vessel density in eyes with mild, moderate, and severe primary open-angle glaucoma.

PURPOSE: The aim of this study is to report changes in and associations of macular vessel density (VD) and perfusion density (PD) using optical coherence tomography angiography (OCTA) in mild, moderate, and severe open-angle glaucoma. METHODS: One hundred thirty-three patients with open-angle glaucoma (133 eyes: 47 mild, 33 moderate, and 53 severe glaucoma) and 73 normal subjects (right eyes) were included in this cross-sectional study. All subjects underwent Cirrus OCTA measurements. One-way analysis of variance (ANOVA) was used to compare macular VD and PD between the controls and mild, moderate, and severe glaucoma groups. Multiple linear regression was performed with OCTA parameters as the predicted variable and age, gender, spherical equivalent (SE), intraocular pressure (IOP), mean deviation (MD), signal strength (SS), and mean macular ganglion cell-inner plexiform layer (mGCIPL) thickness as the predictor variables. RESULTS: The total area of VD showed significant differences between the controls vs. mild (p < 0.001) and moderate vs. severe glaucoma (p = 0.003); no significant difference was found between mild and moderate glaucoma (p = 1.000). Macular VD was associated with age (ß = -0.02, p = 0.003), MD (ß = 0.04, p = 0.001), SS (ß = 1.43, p < 0.001), and mGCIPL thickness (ß = 0.04, p = 0.002) but not with gender, SE, and IOP (all p > 0.05). CONCLUSIONS: Macular microcirculation declined significantly in mild and severe glaucoma. No significant difference was found between mild and moderate glaucoma. Decrease macular VD was independently associated with age, severe MD, lower SS, and thinner mGCIPL thickness.

Diagnostic ability of ganglion cell complex thickness to detect glaucoma in high myopia eyes by Fourier domain optical coherence tomography.

AIM: To evaluate the ability of macular ganglion cell complex (GCC) thickness using Fourier domain optical coherence tomography (FD-OCT) to detect glaucoma in highly myopic eyes. METHODS: Cross-sectional study. A total of 114 participants, consecutively were enrolled. Macular GCC thickness and peripapillary retinal nerve fiber layer (RNFL) thickness were obtained with RTVue FD-OCT. Receiver operating characteristics curves were constructed for each measurement parameter, and areas under the curves (AUCs) were compared. RESULTS: Both the average GCC and average RNFL thickness showed negative correlations with axial length (rGCC=-0.404, P=0.001; rRNFL=-0.561, P<0.001). The largest AUCs from GCC, and RNFL parameters were 0.968 [global loss volume (GLV)], and 0.855 (average RNFL), respectively. GLV was significantly better for detecting high myopic glaucoma than average RNFL (P<0.001). CONCLUSION: Macular GCC thickness has higher diagnostic power than peripapillary RNFL thickness to discriminate glaucoma patients from non-glaucoma subjects in high myopia.

The short-term effects of exercise on intraocular pressure, choroidal thickness and axial length.

PURPOSE: To explore ocular changes in healthy people after exercise. METHODS: Twenty five volunteers underwent exercise for 15 minutes on a treadmill. Measurements of choroidal thickness, intraocular pressure (IOP), ocular biometry, and blood pressure were taken before and after exercise. Enhanced Depth Imaging optical coherence tomography (EDI-OCT) was used to measure choroidal thickness at the fovea. Intraocular pressure (IOP) was measured by Goldmann applanation tonometry. Ocular biometric measures were collected using A scan ultrasound. Blood pressure was measured concurrently with the acquisition of the scans. RESULTS: Twenty five volunteers (25 eyes) with a mean age of 25.44±3.25 years were measured. There was a significant increase in systolic and diastolic pressure after exercise (P<0.05). The IOP showed a significant decrease after exercise (P<0.05). However there was no significant difference in the mean choroidal thickness, ocular axial length, anterior chamber depth, lens thickness, or vitreous length before and after exercise measurements (P>0.05). CONCLUSION: There was a significant decrease in IOP from exercise without a change in choroidal thickness and ocular biometric measures. IOP and choroidal thickness were not correlated, suggesting that the IOP decrease from exercise is not due to changes in choridal thickness.

CD4 positive T helper cells contribute to retinal ganglion cell death in mouse model of ischemia reperfusion injury.

Neuron degeneration is a common pathological process associated with many disease conditions in the central nervous system including retina. Although immune responses have been proposed as one potential element in triggering neural damage, the mechanism of action of specific immune components underlying the pathogenesis is unclear. In this study we focus on adaptive immune activities to evaluate CD4 positive helper cells in the retinal ganglion cell (RGC) degeneration in response to transient retinal ischemic/reperfusion (I/R) injury. Transient retinal ischemia was induced in four mouse strains with different immune backgrounds, including wild type mice from C57BL/6 and BABL/c strains, severe combined immunodeficient (SCID) mice lacking T and B lymphocytes, SCID mice with transferred wild type CD4+ T cells, and the STAT6 deficient mice without T helper 2 (TH2) cells. In SCID mice RGCs showed a strong resistance to cell death in response to I/R injury (89% ± 3% of the survival cells in contralateral eye) compared with C57BL/6 (p = 0.018) and BALB/C (p = 0.038) wild types. By transferring the mature CD4+ T cells from matched wild type into SCID mice, the resistance of RGCs to injury was significantly compromised (p < 0.05). Furthermore a significant resistance of RGCs to cell death (p < 0.05) accompanied with an overexpression of STAT1 and STAT3 was confirmed in STAT6 deficient mice in response to I/R injury compared with the wild type controls, indicating that TH2 cells maturation might be involved in RGC damage. Adaptive immunity carried by CD4 T cells plays an essential role in RGC degeneration.