Background Doctors can detect symptoms of diabetic retinopathy (DR) early by using retinal ophthalmoscopy, and they can improve diagnostic efficiency with the assistance of deep learning to select treatments and support personnel workflow. Conventionally, most deep learning methods for DR diagnosis categorize retinal ophthalmoscopy images into training and validation data sets according to the 80/20 rule, and they use the synthetic minority oversampling technique (SMOTE) in data processing (e.g., rotating, scaling, and translating training images) to increase the number of training samples. Oversampling training may lead to overfitting of the training model. Therefore, untrained or unverified images can yield erroneous predictions. Although the accuracy of prediction results is 90%–99%, this overfitting of training data may distort training module variables. Results This study uses a 2-stage training method to solve the overfitting problem. In the training phase, to build the model, the Learning module 1 used to identify the DR and no-DR. The Learning module 2 on SMOTE synthetic datasets to identify the mild-NPDR, moderate NPDR, severe NPDR and proliferative DR classification. These two modules also used early stopping and data dividing methods to reduce overfitting by oversampling. In the test phase, we use the DIARETDB0, DIARETDB1, eOphtha, MESSIDOR, and DRIVE datasets to evaluate the performance of the training network. The prediction accuracy achieved to 85.38%, 84.27%, 85.75%, 86.73%, and 92.5%. Conclusions Based on the experiment, a general deep learning model for detecting DR was developed, and it could be used with all DR databases. We provided a simple method of addressing the imbalance of DR databases, and this method can be used with other medical images.

Background: Both central serous chorioretinopathy (CSCR) and heart failure (HF) are disorders with a complex pathogenesis, whereas the two diseases might share similar pathogenesis. This study aimed to evaluate whether patients with HF are exposed to potential risk of CSCR by using the National Health Insurance Research Database (NHIRD). Methods: Data were collected from the NHIRD over a 14-year period. Variables were analyzed with the Pearson chi-square test and Fisher’s exact test. The risk factors for disease development were examined by adjusted hazard ratio (aHR). Kaplan–Meier analysis was performed to compare the cumulative incidence of CSCR. Results: A total of 24 426 patients with HF were enrolled in the study cohort, and there were 24 426 patients without HF in the control cohort. The incidence rate of CSCR was higher in the study cohort than in the control cohort (aHR = 4.572, p < 0.001). CSCR occurred more commonly in males than in females. The overall incidence of CSCR was 30.07 per 100 000 person-years in the study cohort and 23.06 per 100 000 person-years in the control cohort. Besides, subgroup analysis revealed that no matter in gender or age group, HF patients were in an increased risk of CSCR diagnosis (male/female, aHR = 3.268/7.701; 20-59 years/≥60 years, aHR = 3.405/5.501, p < 0.001). Conclusion: HF is a significant indicator for CSCR. Patients with HF should stay alert for potential disorder of visual impairment. Further prospective studies to investigate the relationship between HF and CSCR could provide more information.

Background: Nephrotic syndrome (NS) is characterized by various etiologies that damage the glomerulus. Central serous chorioretinopathy (CSCR) is a retinal disease characterized by neurosensory detachment of the retina. Several case reports have described the relationship between both. Therefore, we try to analyze the epidemiological associations between NS and CSCR using the National Health Insurance Research Database in Taiwan. Methods: Data spanning 14 years were extracted from the National Health Insurance Research Database and sub-grouped. The variables were analyzed using Pearson’s chi-squared test and Fisher’s exact test. The risk factors for disease development with or without comorbidities were examined using an adjusted hazard ratio (aHR). Kaplan-Meier analysis was performed to evaluate the cumulative incidence of CSCR with or without NS. Results: A total of 14 794 patients with NS and 14 794 matched controls without NS were enrolled in this cohort study. The incidence rate of CSCR was higher in the study cohort than in the control cohort (aHR = 3.349, p < 0.001). The overall incidence of CSCR was 44.51 per 100 000 person-years in the study cohort and 33.39 per 100 000 person-years in the control cohort. In both groups, CSCR occurred more frequently in males than in females. Patients aged 40–49, 50–59, and ≥60 years in the study cohort had a significantly higher risk of developing CSCR than those in the control cohort (aHR = 3.445, 5.421, and 4.957, all p < 0.001). NS patient with a 4-week history of steroid usage has a higher risk of developing CSCR (aHR = 2.010, p < 0.001). Conclusion: Our data showed that patients with NS have an increased risk of developing subsequent CSCR. Physician should routinely refer their NS patients to ophthalmologist for ophthalmic evaluation. This is the first nationwide epidemiological study reporting the association between these two diseases. Further studies are needed to clarify this relationship.