Segmentation of retinal fluid based on deep learning:application of three-dimensional fully convolutional neural networks in optical coherence tomography images

AIM: To explore a segmentation algorithm based on deep learning to achieve accurate diagnosis and treatment of patients with retinal fluid.METHODS: A two-dimensional(2D) fully convolutional network for retinal segmentation was employed. In order to solve the category imbalance in retinal optical coherence tomography(OCT) images, the network parameters and loss function based on the 2D fully convolutional network were modified. For this network, the correlations of corresponding positions among adjacent images in space are ignored. Thus, we proposed a three-dimensional(3D) fully convolutional network for segmentation in the retinal OCT images.RESULTS: The algorithm was evaluated according to segmentation accuracy, Kappa coefficient, and F1 score. For the 3D fully convolutional network proposed in this paper, the overall segmentation accuracy rate is 99.56%, Kappa coefficient is 98.47%, and F1 score of retinal fluid is 95.50%. CONCLUSION: The OCT image segmentation algorithm based on deep learning is primarily founded on the 2D convolutional network. The 3D network architecture proposed in this paper reduces the influence of category imbalance, realizes end-to-end segmentation of volume images, and achieves optimal segmentation results. The segmentation maps are practically the same as the manual annotations of doctors, and can provide doctors with more accurate diagnostic data.

Comprehensive assessment of growth factors,inflammatory mediators, and cytokines in vitreous from patients with proliferative diabetic retinopathy

AIM: To assess alterations in growth factors, inflammatory mediators, and cytokines associated with vitreous-retinal diseases in vitreous humor from patients with proliferative diabetic retinopathy(PDR), and to identify potential new treatment targets and strategies.METHODS: Control vitreous samples were collected from patients with macular hole, epiretinal membranes, or rhegmatogenous retinal detachments, and PDR samples from patients with complications of PDR, who required pars plana vitrectomy. Specimens were stored at-80℃ and then investigated by Luminex multi-factor assay. Parametric and nonparametric analyses of demographic characteristics and cytokine expression levels were conducted using SPSS.RESULTS: There were no significant differences in demographic characteristics between patients with and without PDR. Expression levels of growth factors [plateletderived growth factor(PDGF)-AA, glial cell line-derived neurotrophic factor(GDNF), and vascular endothelial growth factor A(VEGFA)], inflammatory mediators [interleukin(IL)-8, IL-11, and tumor necrosis factor-α(TNF-α)] and cytokines [chemokine C-X-C ligand(CXCL)10, interferon-γ(IFN-γ), and granulocyte macrophage-colony stimulating factor(GM-CSF)] were significantly elevated in vitreous humor from patients with PDR compared with those in the control group(all P<0.05). Further, VEGFA levels were lower in patients with PDR treated with anti-VEGF injection than those who were not(P<0.05), and there was no difference between the PDR group treated with anti-VEGF and controls(P>0.05).CONCLUSION: This proof-of-concept study demonstrates the potential for combinational therapeutic strategies to ameliorate diabetic retinopathy progression by targeting growth factors, inflammatory factors, and cytokines, in addition to VEGFA.

A quantitative comparison of five optical coherence tomography angiography systems in clinical performance

AIM： To compare the clinical performance of 4 spectraldomain（SD） optical coherence tomography angiography（OCTA） systems： AngioVueTM, AngioPlexTM, Spectralis？ OCTA, AngioScan, and 1 swept-source（SS） OCTA SS OCT AngioTM. METHODS： Twenty-seven undilated right eyes of 27 participants underwent OCTA examination using five different systems respectively for both 3×3 and 6×6 mm^2 scan pattern（Spectralis OCTA for 3×3 mm^2 scan only）. Image quality, including vessel valid visibility and the number of motion artifacts, and acquisition time were evaluated. Repeated measures analysis of variance（ANOVA） with Bonferroni＇s post-test and Friedman test with Dunn＇s posttest were used to compare measurements.RESULTS： The age of the subjects was 28.19±5.55 y（range, 23-49 y）. The spherical equivalent refraction was-2.55±1.84 D（range, 0.00 to-5.25 D）. Significant difference was observed in the evaluation of vessel valid visibility（Angio Vue the highest： 0.111±0.031 for 3×3 mm^2 scan and 0.128±0.020 for 6×6 mm^2 scan）, number of motion artifacts（AngioVue the fewest： 0.778±1.086 for 3×3 mm^2 scan and 0.333±0.620 for 6×6 mm^2 scan） and acquisition time（AngioPlex the shortest： 8.537±1.921 s for 3×3 mm^2 scan and 8.298±1.741 s for 6×6 mm^2 scan; all P〈0.001）. CONCLUSION： There is poor agreement of measurements among systems. AngioVue provides images with the highest vessel valid visibility and the fewest motion artifacts. AngioPlex achieves the shortest acquisition.

Quantitative analysis of retinal intermediate and deep capillary plexus in patients with retinal deep vascular complex ischemia

AIM: To quantitatively analyze the retinal intermediate and deep capillary plexus(ICP and DCP) in patients with retinal deep vascular complex ischemia(RDVCI), using 3D projection artifacts removal(3D PAR) optical coherence tomography angiography(OCTA).METHODS: RDVCI patients and gender-and agematched healthy controls were assessed and underwent OCTA examinations. The parafoveal vessel density(PFVD) of retinal deep vascular complex(DVC), ICP, and DCP were analyzed, and the percentage of reduction(PR) of PFVD was calculated.RESULTS: Twenty-four eyes in 22 RDVCI patients(20 in acute phase and 4 in chronic phase) and 24 eyes of 22 healthy subjects were enrolled as the control group. Significant reduction of PFVD in DVC, ICP, and DCP was observed in comparison with the controls(DVC: acute: 43.59%±6.58% vs 49.92%±5.49%, PR=12.69%;chronic: 43.50%±3.33% vs 51.20%±3.80%, PR=15.04%. ICP: acute: 40.28%±7.91% vs 46.97%±7.14%, PR=14.23%;chronic: 41.48%±2.87% vs 46.43%±3.29%, PR=10.66%. DCP: acute: 45.44%±8.27% vs 51.51%±9.97%, PR=11.79%;chronic: 37.78%±3.48% vs 51.73%±5.17%, PR=26.97%;all P<0.05). No significant PR difference was found among DVC, ICP, and DCP of RDVCI in acute phase(P=0.812), but significant difference in chronic phase(P=0.006, DVC vs DCP, ICP vs DCP). No significant difference in PR between acute and chronic phases in the DVC(P=0.735) or ICP(P=0.681) was found, but significant difference in the DCP(P=0.041).CONCLUSION: The PFVD of DVC, ICP, and DCP in RDVCI is significantly decreased in both acute and chronic phases. ICP impairment is stabilized from acute to chronic phase in RDVCI, whereas subsequent DCP impairment is uncovered and can be explained by ischemia-reperfusion damage.

A challenging case of tuberculosis-associated uveitis after corticosteroid treatment for Vogt-Koyanagi-Harada disease

Dear Editor,I am Dr.Tian-Wei Qian,from Shanghai General Hospital,Shanghai,China.I write to present one case report of tuberculosis（TB）-associated uveitis after corticosteroid treatment for Vogt-Koyanagi-Harada（VKH）disease.