Search

To automatically detect and classify the lesions of diabetic retinopathy (DR) in fundus fluorescein angiography (FFA) images using deep learning algorithm through comparing 3 convolutional neural networks (CNNs). A total of 4067 FFA images from Eye Center at the Second Affiliated Hospital of Zhejiang University School of Medicine were annotated with 4 kinds of lesions of DR, including non-perfusion regions (NP), microaneurysms, leakages, and laser scars. Three CNNs including DenseNet, ResNet50, and VGG16 were trained to achieve multi-label classification, which means the algorithms could identify 4 retinal lesions above at the same time. The area under the curve (AUC) of DenseNet reached 0.8703, 0.9435, 0.9647, and 0.9653 for detecting NP, microaneurysms, leakages, and laser scars, respectively. For ResNet50, AUC was 0.8140 for NP, 0.9097 for microaneurysms, 0.9585 for leakages, and 0.9115 for laser scars. And for VGG16, AUC was 0.7125 for NP, 0.5569 for microaneurysms, 0.9177 for leakages, and 0.8537 for laser scars. Experimental results demonstrate that DenseNet is a suitable model to automatically detect and distinguish retinal lesions in the FFA images with multi-label classification, which lies the foundation of automatic analysis for FFA images and comprehensive diagnosis and treatment decision-making for DR.

Background/aimsFundus fluorescein angiography (FFA) is an important technique to evaluate diabetic retinopathy (DR) and other retinal diseases. The interpretation of FFA images is complex and time-consuming, and the ability of diagnosis is uneven among different ophthalmologists. The aim of the study is to develop a clinically usable multilevel classification deep learning model for FFA images, including prediagnosis assessment and lesion classification.MethodsA total of 15 599 FFA images of 1558 eyes from 845 patients diagnosed with DR were collected and annotated. Three convolutional neural network (CNN) models were trained to generate the label of image quality, location, laterality of eye, phase and five lesions. Performance of the models was evaluated by accuracy, F-1 score, the area under the curve and human-machine comparison. The images with false positive and false negative results were analysed in detail.ResultsCompared with LeNet-5 and VGG16, ResNet18 got the best result, achieving an accuracy of 80.79%–93.34% for prediagnosis assessment and an accuracy of 63.67%–88.88% for lesion detection. The human-machine comparison showed that the CNN had similar accuracy with junior ophthalmologists. The false positive and false negative analysis indicated a direction of improvement.ConclusionThis is the first study to do automated standardised labelling on FFA images. Our model is able to be applied in clinical practice, and will make great contributions to the development of intelligent diagnosis of FFA images.

Purpose: To establish quantitative profile of the morphologic changes among patients with active myopic choroidal neovascularization (mCNV) before and after anti-vascular endothelial growth factor (VEGF) therapy using optical coherence tomography angiography (OCTA) to assess the therapeutic response.Methods: Patients with active mCNV who received anti-VEGF injections between February 2017 to October 2020 and fit the study criteria were retrospectively reviewed. Quantitative analysis of their OCTA images were carried out to evaluate the morphologic features and vascular changes of mCNV lesions in response to anti-VEGF therapy. For further quantitative profiling, mCNV area, fractal dimension, vessel area, vessel density, vessel diameter, vessel length, vessel junction, junction density, and vessel tortuosity were obtained by means of advanced skeletonization postprocessing analyses.Results: Thirty-one eyes of 29 consecutive patients with OCTA-positive mCNV lesions (mean spherical equivalent: −12.55 ± 3.24 diopters) were included. The 31 cases were divided into two phenotypes at baseline: organized interlacing pattern (83.87%) and disorganized vascular loops pattern (16.13%). The values of mCNV area, fractal dimension, vessel area, vessel length, vessel junction, and junction density decreased remarkably 1 month after the initial anti-VEGF injection (p < 0.001). Although, vessel density, vessel diameter, and vessel tortuosity increased meanwhile, only vessel diameter displayed statistical significance (p = 0.027). Of note, relative ratio analysis showed that vessel junction was the most sensitive biomarker in response to anti-VEGF therapy, reflecting a mean decrease of 50.36%. Sensitivity lowered successively in biomarkers of vessel length, vessel area, junction density, mCNV area, and fractal dimension. In addition, percent change of mCNV area (r = 0.552, p = 0.002), fractal dimension (r = 0.446, p = 0.017), vessel area (r = 0.518, p = 0.005), and vessel length (r = 0.440, p = 0.019) were moderately associated with that of central retinal thickness.Conclusions: The study showed morphological as well as quantitative changes on OCTA responding to anti-VEGF treatment in mCNV patients, among which vessel junctions might be the most predictive biomarker. OCTA-based analysis, providing intuitive images and a large spectrum of quantitative data at the same time, could promote new insights into the therapeutic response assessment in mCNV patients.

Vision loss caused by diabetic macular edema (DME) can be prevented by early detection and laser photocoagulation. As there is no comprehensive detection technique to recognize NPA, we proposed an automatic detection method of NPA on fundus fluorescein angiography (FFA) in DME. The study included 3,014 FFA images of 221 patients with DME. We use 3 convolutional neural networks (CNNs), including DenseNet, ResNet50, and VGG16, to identify non-perfusion regions (NP), microaneurysms, and leakages in FFA images. The NPA was segmented using attention U-net. To validate its performance, we applied our detection algorithm on 249 FFA images in which the NPA areas were manually delineated by 3 ophthalmologists. For DR lesion classification, area under the curve is 0.8855 for NP regions, 0.9782 for microaneurysms, and 0.9765 for leakage classifier. The average precision of NP region overlap ratio is 0.643. NP regions of DME in FFA images are identified based a new automated deep learning algorithm. This study is an in-depth study from computer-aided diagnosis to treatment, and will be the theoretical basis for the application of intelligent guided laser.

Purpose To predict the anti-vascular endothelial growth factor (VEGF) therapeutic response of diabetic macular oedema (DME) patients from optical coherence tomography (OCT) at the initiation stage of treatment using a machine learning-based self-explainable system. Methods A total of 712 DME patients were included and classified into poor and good responder groups according to central macular thickness decrease after three consecutive injections. Machine learning models were constructed to make predictions based on related features extracted automatically using deep learning algorithms from OCT scans at baseline. Five-fold cross-validation was applied to optimize and evaluate the models. The model with the best performance was then compared with two ophthalmologists. Feature importance was further investigated, and a Wilcoxon rank-sum test was performed to assess the difference of a single feature between two groups. Results Of 712 patients, 294 were poor responders and 418 were good responders. The best performance for the prediction task was achieved by random forest (RF), with sensitivity, specificity and area under the receiver operating characteristic curve of 0.900, 0.851 and 0.923. Ophthalmologist 1 and ophthalmologist 2 reached sensitivity of 0.775 and 0.750, and specificity of 0.716 and 0.821, respectively. The sum of hyperreflective dots was found to be the most relevant feature for prediction. Conclusion An RF classifier was constructed to predict the treatment response of anti-VEGF from OCT images of DME patients with high accuracy. The algorithm contributes to predicting treatment requirements in advance and provides an optimal individualized therapeutic regimen.

The study aimed to evaluate the effect of drooped eyelid on corneal tomography in congenital blepharoptosis patients. Sixty-four patients with congenital blepharoptosis and 64 age- and sex- matched healthy subjects were included. According to the eyelid margin to corneal light reflex distance (MRD), eyes with congenital blepharoptosis were categorized as mild, moderate, or severe. The eyes were scanned using the rotating Scheimpflug camera. Increased topometric parameters were observed in moderate and severe blepharoptosis. Back corneal elevations at the thinnest point were significant higher for mild (P = 0.009), moderate (P P P P P

OBJECTIVE To evaluate the degree of microvascular impairment in DR 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. METHODS This retrospective cross-sectional study included a total of 143 eyes of 94 patients with different stages of DR. The retinal microvasculature was imaged by projection removed OCTA. We examined the degree of association between fractal metrics of the retinal microvasculature and DR severity. The area under the ROC curve was used to estimate the diagnostic performance. RESULTS With increasing DR severity, the multifractal spectrum shifted toward the left bottom and exhibited less left skewness and asymmetry. The vessel density, multifractal features, and lacunarity measured from the DCP were strongly associated with DR severity. The multifractal feature D5 showed the highest diagnostic ability. The combination of multifractal features further improved the discriminating power. CONCLUSIONS Multifractal and lacunarity analyses can be potentially valuable tools for assessment of microvascular impairments in DR. Multifractal geometric parameters exhibit a better discriminatory performance than Euclidean measures, particularly for detection of the early stages of DR.

Transplantation of gene transfected endothelial progenitor cells (EPCs) has provided a novel method for neovascularization (NV) therapy but not for ocular diseases. This study investigated the efficacy of endostatin (ES) transfected EPCs in retinal NV therapy. EPCs were transfected with ES-lentivirus (LV) recombinant construct to obtain ES-LV-EPCs. Oxygen-induced retinopathy rats received the same volume of intravitreal injections of empty-LV, ES-LV, EPCs, and ES-LV-EPCs. Retinal NV was compared using fundus fluoresce in angiography and hematoxylin-eosin staining method. Retinal expression of ES, vascular endothelial growth factor (VEGF), and CD31 were evaluated by immunohistochemistry method. QRT-PCR showed a high expression of ES in ES-LV-EPCs. The NV leakage area and the number of pre-retinal neovascular cell nuclei significantly decreased in ES-LV and ES-LV-EPCs groups, and the effects of ES-LV and ES-LV-EPCs on inhibiting retinal NV were almost the same. ES-LV group and ES-LV-EPCs group showed much more retinal distribution of ES. Intravitreal injections of ES-LV-EPCs inhibited retinal NV, VEGF or CD31 expression, and increased ES expression in vivo. ES-LV-EPC targeted and prevented pathologic retinal NV, and exerted a much safer gene therapy by using lentivirus transfected EPCs. Thus, gene combined EPCs represent a potential new therapeutic agent for the treatment of neovascular eye diseases. Funding Statement: This work was supported by National Natural Science Foundation of China (81570822, 81571819); Zhejiang Key Laboratory Fund of China (2011E10006); Zhejiang Province Key Research and Development Program (2015C03042); Zhejiang Provincial Natural Science Foundation of China (LZ18H180001, LY16H120002); National S&T Major Project of China ( NO.SQ2018ZX100301) Declaration of Interests: The authors declare that there are no conflicts of interest. Ethics Approval Statement: All procedures were performed in accordance with the ARVO statement for the use of Animals in Ophthalmic and Vision Research. The Institutional Animal Care and Ethics Committee of Zhejiang University (Hangzhou, China) approved all animal experiments.

Choroidal neovascularization (CNV) is common in various retinal and choroidal diseases, and may result in severe and irreversible loss of vision. Our previous studies suggested that Endostar, a novel recombinant endostatin, is able to inhibit the proliferation and migration of choroid‑retinal endothelial cells. To further evaluate the effect of Endostar on the formation of CNV in vivo, a rat model of laser‑induced CNV was constructed and Endostar or phosphate‑buffered saline treatment was administered intravitreally every other day. Using fluorescein angiography (FA), reduced CNV incidence and leakage grade was observed in the Endostar group. In addition, CNV area and maximal thickness were prominently reduced in the Endostar group measured by choroid flat mounts and sections. Furthermore, vascular endothelial growth factor (VEGF), hypoxia‑inducible factor 1α and chemokine C‑X‑C motif ligand 1 were markedly reduced in the Endostar group as determined by quantitative polymerase chain reaction and downregulation of VEGF was also verified by western blot analysis at the protein level. This study demonstrates that Endostar suppressed CNV in a rat model, which may be largely mediated by the downregulation of VEGF and other angiogenic molecules.

Idiopathic choroidal neovascularization (ICNV) is a disorder that primarily affecting patients younger than 50 years and can cause severe loss of vision. Choroidal abnormalities, especially choroidal inflammation, have been thought to be involved in the pathophysiology of ICNV. However, the exact pathogenesis of ICNV remains unclear. The aim of our study was investigate the levels of 27 inflammatory cytokines in the aqueous humor of eyes with ICNV and to determine the effect of intravitreal injection of ranibizumab (IVR) on cytokine levels. Significantly higher levels of IL-2, IL-10, IL-15, IL-17, basic FGF and GM-CSF were observed in patients with ICNV compared with controls. However, only IL-17 levels were significantly higher in patients with ICNV compared with controls after adjusting for axial length. Furthermore, there were significant correlations between the levels of IL-10, IL-17, GM-CSF and VEGF and the lesion area. Significant changes in visual acuity and central retinal thickness were observed after IVR. Besides VEGF, IVR also significantly reduced the levels of IL-2, IL-10, basic FGF and IL-12, however, the IL-6 levels were significantly increased. Our results suggest that there may be an involvement of IL-17-related inflammatory processes in the etiology of ICNV.