0
selected
-
1.
Hyper-reflective foci segmentation in SD-OCT retinal images with diabetic retinopathy using deep convolutional neural networks.
Yu, C, Xie, S, Niu, S, Ji, Z, Fan, W, Yuan, S, Liu, Q, Chen, Q
Medical physics. 2019;(10):4502-4519
Abstract
PURPOSE The purpose of this study was to automatically and accurately segment hyper-reflective foci (HRF) in spectral domain optical coherence tomography (SD-OCT) images with diabetic retinopathy (DR) using deep convolutional neural networks. METHODS An automatic HRF segmentation model for SD-OCT images based on deep networks was constructed. The model segmented small lesions through pixel-wise predictions based on small image patches. We used an approach for discriminative features extraction for small patches by introducing small kernels and strides in convolutional and pooling layers, which was applied on the state-of-the-art deep classification networks (GoogLeNet and ResNet). The features extracted by the adapted deep networks were fed into a softmax layer to produce the probabilities of HRF. We trained different models on a dataset with 16 HRF eyes by using different sizes of patches, and then, we fused these models to generate optimal results. RESULTS Experimental results on 18 eyes demonstrated that our method is effective for the HRF segmentation. The dice similarity coefficient (DSC) for the foci area in B-scan, projection images, and foci amount in B-scan images reaches 67.81%, 74.09%, and 72.45%, respectively. CONCLUSIONS The proposed segmentation model can accurately segment HRF in SD-OCT images with DR and outperforms traditional methods. Our model may provide reliable segmentations for small lesions in SD-OCT images and may be helpful in the clinical diagnosis of diseases.
-
2.
Automated choroid segmentation of three-dimensional SD-OCT images by incorporating EDI-OCT images.
Chen, Q, Niu, S, Fang, W, Shuai, Y, Fan, W, Yuan, S, Liu, Q
Computer methods and programs in biomedicine. 2018;:161-171
Abstract
BACKGROUND AND OBJECTIVE The measurement of choroidal volume is more related with eye diseases than choroidal thickness, because the choroidal volume can reflect the diseases comprehensively. The purpose is to automatically segment choroid for three-dimensional (3D) spectral domain optical coherence tomography (SD-OCT) images. METHODS We present a novel choroid segmentation strategy for SD-OCT images by incorporating the enhanced depth imaging OCT (EDI-OCT) images. The down boundary of the choroid, namely choroid-sclera junction (CSJ), is almost invisible in SD-OCT images, while visible in EDI-OCT images. During the SD-OCT imaging, the EDI-OCT images can be generated for the same eye. Thus, we present an EDI-OCT-driven choroid segmentation method for SD-OCT images, where the choroid segmentation results of the EDI-OCT images are used to estimate the average choroidal thickness and to improve the construction of the CSJ feature space of the SD-OCT images. We also present a whole registration method between EDI-OCT and SD-OCT images based on retinal thickness and Bruch's Membrane (BM) position. The CSJ surface is obtained with a 3D graph search in the CSJ feature space. RESULTS Experimental results with 768 images (6 cubes, 128 B-scan images for each cube) from 2 healthy persons, 2 age-related macular degeneration (AMD) and 2 diabetic retinopathy (DR) patients, and 210 B-scan images from other 8 healthy persons and 21 patients demonstrate that our method can achieve high segmentation accuracy. The mean choroid volume difference and overlap ratio for 6 cubes between our proposed method and outlines drawn by experts were -1.96µm3 and 88.56%, respectively. CONCLUSIONS Our method is effective for the 3D choroid segmentation of SD-OCT images because the segmentation accuracy and stability are compared with the manual segmentation.
-
3.
Thin Rigid Contact Lens Used in Vitreous-Retinal Surgery for Corneal Protection: A Randomized Controlled Trial.
Hu, Z, Ding, Y, Zheng, X, Yuan, S, Li, J, Xie, P, Liu, Q
Eye & contact lens. 2018;:S355-S360
Abstract
PURPOSE To design a rigid contact lens (CL) to be used in combination with a wide-angle viewing system and analyze its protection for corneal epithelial during vitreous-retinal surgery. METHODS A thin and lightweight rigid CL was designed and constructed. The impact of the CL on the visualized fundus range was evaluated using a concrete eye model. Patients with severe proliferative diabetic retinopathy (PDR) were randomized to either the CL group, corneal protective agent (CPA) group, or balanced salt solution (BSS) group. All patients underwent phacoemulsification and a standard 23-gauge three-port vitrectomy. Surgery time and corneal fluorescein staining score (FSS) postoperatively were mainly measured. RESULTS In the eye model, a larger area of fundus was visualized with the use of our CL under 128 D or 60 D Resight lens. The mean surgery time was 51.36±8.06 min, 50.89±8.26 min, and 55.46±9.14 in CL, CPA, and BSS group, respectively (F=2.325, P=0.105). In eight eyes in the BSS group, corneal epithelial layer was peeled off because the dryness of the cornea could not maintain a clear fundus image. The FSS in BSS group was markedly higher than that of CL and BSS group 1 day (P<0.001), 3 days (P<0.001), and 7 days (P=0.002) postoperatively. There was no statistical significance of the FSS between CL and CPA group at each follow-up endpoint. CONCLUSIONS The CL that we designed can slightly enlarge the visible fundus range and efficiently protect corneal epithelium during vitrectomy for patients with PDR.
-
4.
Multimodality analysis of Hyper-reflective Foci and Hard Exudates in Patients with Diabetic Retinopathy.
Niu, S, Yu, C, Chen, Q, Yuan, S, Lin, J, Fan, W, Liu, Q
Scientific reports. 2017;(1):1568
Abstract
To investigate the correlations between hyper-reflective foci and hard exudates in patients with non-proliferative diabetic retinopathy (NPDR) and proliferative diabetic retinopathy (PDR) by spectral-domain optical coherence tomography (SD OCT) images. Hyper-reflective foci in retinal SD OCT images were automatically detected by the developed algorithm. Then, the cropped CFP images generated by the semi-automatic registration method were automatically segmented for the hard exudates and corrected by the experienced clinical ophthalmologist. Finally, a set of 5 quantitative imaging features were automatically extracted from SD OCT images, which were used for investigating the correlations of hyper-reflective foci and hard exudates and predicting the severity of diabetic retinopathy. Experimental results demonstrated the positive correlations in area and amount between hard exudates and hyper-reflective foci at different stages of diabetic retinopathy, with statistical significance (all p < 0.05). In addition, the area and amount can be taken as potential discriminant indicators of the severity of diabetic retinopathy.
-
5.
Association between transcription factor 7-like 2 rs7903146 polymorphism and diabetic retinopathy in type 2 diabetes mellitus: A meta-analysis.
Ding, Y, Hu, Z, Yuan, S, Xie, P, Liu, Q
Diabetes & vascular disease research. 2015;(6):436-44
Abstract
As one of the vascular complications of type 2 diabetes mellitus, the incidence of diabetes retinopathy is greatly increasing worldwide. Both genetic and environmental factors are involved in the pathologies. A meta-analysis was conducted to assess the association between transcription factor 7-like 2 polymorphism (rs7903146) and type 2 diabetic retinopathy. Published literature from PubMed, Web of Science and China National Knowledge Infrastructure were retrieved. Pooled odds ratios with 95% confidence intervals were calculated to estimate the strength of the association. Eight studies including 6422 participants were included in the final meta-analysis. Our analysis provides substantial evidence that the rs7903146 variant is significantly associated with the risk of diabetic retinopathy in Caucasian populations while not in East Asian populations. The variant of rs7903146 appeared more likely to be a promising genetic biomarker of diabetic retinopathy in Caucasians.