Classification of age-related macular degeneration using convolutional-neural-network-based transfer learning.

BACKGROUND: To diagnose key pathologies of age-related macular degeneration (AMD) and diabetic macular edema (DME) quickly and accurately, researchers attempted to develop effective artificial intelligence methods by using medical images. RESULTS: A convolutional neural network (CNN) with transfer learning capability is proposed and appropriate hyperparameters are selected for classifying optical coherence tomography (OCT) images of AMD and DME. To perform transfer learning, a pre-trained CNN model is used as the starting point for a new CNN model for solving related problems. The hyperparameters (parameters that have set values before the learning process begins) in this study were algorithm hyperparameters that affect learning speed and quality. During training, different CNN-based models require different algorithm hyperparameters (e.g., optimizer, learning rate, and mini-batch size). Experiments showed that, after transfer learning, the CNN models (8-layer Alexnet, 22-layer Googlenet, 16-layer VGG, 19-layer VGG, 18-layer Resnet, 50-layer Resnet, and a 101-layer Resnet) successfully classified OCT images of AMD and DME. CONCLUSIONS: The experimental results further showed that, after transfer learning, the VGG19, Resnet101, and Resnet50 models with appropriate algorithm hyperparameters had excellent capability and performance in classifying OCT images of AMD and DME.

Automatic screening of tear meniscus from lacrimal duct obstructions using anterior segment optical coherence tomography images by deep learning.

PURPOSE: We assessed the ability of deep learning (DL) models to distinguish between tear meniscus of lacrimal duct obstruction (LDO) patients and normal subjects using anterior segment optical coherence tomography (ASOCT) images. METHODS: The study included 117 ASOCT images (19 men and 98 women; mean age, 66.6 ± 13.6 years) from 101 LDO patients and 113 ASOCT images (29 men and 84 women; mean age, 38.3 ± 19.9 years) from 71 normal subjects. We trained to construct 9 single and 502 ensemble DL models with 9 different network structures, and calculated the area under the curve (AUC), sensitivity, and specificity to compare the distinguishing abilities of these single and ensemble DL models. RESULTS: For the highest single DL model (DenseNet169), the AUC, sensitivity, and specificity for distinguishing LDO were 0.778, 64.6%, and 72.1%, respectively. For the highest ensemble DL model (VGG16, ResNet50, DenseNet121, DenseNet169, InceptionResNetV2, InceptionV3, and Xception), the AUC, sensitivity, and specificity for distinguishing LDO were 0.824, 84.8%, and 58.8%, respectively. The heat maps indicated that these DL models placed their focus on the tear meniscus region of the ASOCT images. CONCLUSION: The combination of DL and ASOCT images could distinguish between tear meniscus of LDO patients and normal subjects with a high level of accuracy. These results suggest that DL might be useful for automatic screening of patients for LDO.

Difference of central foveal thickness measurement in patients with macular edema using optical coherence tomography in different display modes.

Purpose: To assess the differences in the measurement of central foveal thickness (CFT) in patients with macular edema (ME) between two display modes (1:1 pixel and 1:1 micron) on optical coherence tomography (OCT). Design: This is a retrospective, cross-sectional study. Methods: Group A consisted of participants with well-horizontal OCT B-scan images and group B consisted of participants with tilted OCT B-scan. We manually measured the CFT under the two display modes, and the values were compared statistically using the paired t-test. Spearman's test was used to assess the correlations between the OCT image tilting angle (OCT ITA) and the differences in CFT measurement. The area under the curve (AUC) was calculated to define the OCT ITA cutoff for a defined CFT difference. Results: In group A, the mean CFT in the 1:1 pixel display mode was 420.21 ± 130.61 µm, similar to the mean CFT of 415.27 ± 129.85 µm in the 1:1 micron display mode. In group B, the median CFT in the 1:1 pixel display mode is 409.00 µm (IQR: 171.75 µm) and 368.00 µm (IQR: 149.00 µm) in the 1:1 micron display mode. There were significant differences between the two display modes with the median (IQR) absolute difference and median (IQR) relative difference of 38.00 µm (75.00 µm) and 10.19% (21.91%) (all p = 0.01). The differences in CFT measurement between the two display modes were correlated with the OCT ITA (absolute differences, r = 0.88, p < 0.01; relative differences, r = 0.87, p < 0.01). The AUC for a predefined CFT difference was 0.878 (10 µm), 0.933 (20 µm), 0.938 (30 µm), 0.961 (40 µm), 0.962 (50 µm), and 0.970 (60 µm). Conclusion: In patients with DM, when the OCT B-scan images were well-horizontal, manual CFT measurements under the two display modes were similar, but when the B-scan images were tilted, the CFT measurements were different under the two display modes, and the differences were correlated to the OCT ITA.

Evaluation of the recurrence risk of macular hole in the high myopia after removing silicone oil by spectral domain optical coherence tomography / 中华实验眼科杂志

Background Macular hole in high myopia has been paid tremendous attention in clinical research due to its high recurrence rate and unpredictable prognosis.Spectral domain optical coherence tomography (SD-OCT) is considered to be of a good evaluating value for macular hole,yet its application is compromised in patient with high myopia.High myopia might lead to retinal detachment which consequently prevents an accurate OCT.Therefore,it is important to assess the effectiveness of OCT on prognosis in these patients after surgical restoration of the detached retina.Objective The aim of this study was to evaluate the risk of reopening of a macular hole in highly myopic patient after removing silicone oil by SD-OCT.Methods A case-observational study was designed.Twenty-five highly myopic patients with monocular macular holes who underwent vitrectomy combined with internal limiting membrane peeling and silicone oil tamponade were included in this study.Four patients were male and 21 patients were female,with the average age (61.4±9.0) years old and diopter (-14.14 ±6.86)D.Regular ocular examination,axial length measurement,fundus photography and OCT were performed at the day before removing silicone oil and every month till the 6 months after operation.Retinal thickness,macular height index (MHI) and choroidal thickness were measured by OCT.Written informed consent was obtained from each patient.Results All 25 patients finished the follow-up of 6 months.Macular holes remained closure 6 months after operation in 19 patients (76％) and reopened in 6 patients macular holes were in 1-5 months (24％),with a mean time at (3.3± 1.4) months.The average foveal retinal thickness,MHI and the choroidal thickness at the middle of the macular hole were (216.5±95.6) μm,0.30 ± 0.09 and (122.9 ± 20.5) μm in the closed group,respectively,and those in the unclosed group were (113.5±28.7) μm,0.58±0.27 and (96.8±22.9) μm,with significant differences between the two groups (t=2.577,-4.143,2.669,P ＜ 0.05).The percentage of macular hole closure was 85.7％ in the MHI＜0.5 group and 25.0％ in the MHI ≥ 0.5 group,showing a significant difference (P =0.031).The base diameter in the unclosed hole group was significantly larger than that in the closed hole group (1070.2±393.6 μm versus 533.3±277.7 μm) (t =-3.700,P =0.001).Conclusions The measurements of the retinal thickness,MHI and the choroidal thickness at the central area of the macular hole are helpful for the evaluation of reopening risk of macular hole after removing silicone oil.