Retinopathy of prematurity (ROP) is a disease that can cause blindness in premature infants. It is characterized by immature vascular growth of the retinal blood vessels. However, early detection and treatment of ROP can significantly improve the visual acuity of high-risk patients. Thus, early diagnosis of ROP is crucial in preventing visual impairment. However, several patients refrain from treatment owing to the lack of medical expertise in diagnosing the disease; this is especially problematic considering that the number of ROP cases is on the rise. To this end, we applied transfer learning to five deep neural network architectures for identifying ROP in preterm infants. Our results showed that the VGG19 model outperformed the other models in determining whether a preterm infant has ROP, with 96% accuracy, 96.6% sensitivity, and 95.2% specificity. We also classified the severity of the disease; the VGG19 model showed 98.82% accuracy in predicting the severity of the disease with a sensitivity and specificity of 100% and 98.41%, respectively. We performed 5-fold cross-validation on the datasets to validate the reliability of the VGG19 model and found that the VGG19 model exhibited high accuracy in predicting ROP. These findings could help promote the development of computer-aided diagnosis.

Retinopathy of prematurity (ROP) is a retinal disorder that occurs in preterm infants with low birth weight and is the leading cause of preventable blindness in children. Early identification of high-risk patients and early diagnosis and timely treatment of ROP can substantially improve patients' visual outcomes. However, manual screening consumes both time and resources. Telescreening using retinal fundus images has the potential to reduce the burden engendered by the necessity of on-site screening. Recently, substantial progress has been made in using computer-aided diagnosis with retinal fundus images, and this approach has attracted considerable attention for the diagnosis of eye diseases. Abnormalities of and alterations in retinal blood vessels may relate to the occurrence and progression of ROP. In this study, we examined the hypothesis that ROP severity may be associated with the angle and width of arteries and veins. We computationally determined the artery–artery and vein–vein angles in the temporal quadrants—the temporal artery angle (TAA) and temporal vein angle (TVA)—under normal conditions and in different ROP stages. We also estimated retinal vessel width—temporal artery width (TAW) and temporal vein width (TVW)—by applying the Radon transform method to fundus images. Our results revealed significant decreases in TAA and TVA and increases in TAW and TVW with increasing ROP severity (all P < 0.0001).In addition, we observed positive TAA–TVA and TAW–TVW correlations (both P < 0.0001). The TAA was negatively correlated with the TAW (r = −0.162, P = 0.0314). These retinal vessel features may be useful in assisting ophthalmologists in the early detection of ROP and its progression.

No abstract

Diabetic retinopathy (DR) has been the most frequently occurring complication in the patients suffering from a long-term diabetic condition, that ultimately leads to blindness. Early detection of the disease through biomarkers and effective treatment has been proposed to prevent/delay its occurrence. Several biomarkers have been explored, to help understand the incidence and progression of DR. These included the presence of microaneurysms, exudates, hemorrhages, etc. in the retina of the patients, which contributes to the disease. Investigation of the retinal images from time to time has been proposed as a strategy to prevent blindness. Evaluating the retinal images manually is time-consuming and demands great expertise in the diagnosis of DR. To circumvent such issues computer-aided diagnosis are very promising in the detection of DR. In the present study, we used a DR dataset and applied different classification algorithms in machine learning to predict the occurrence of the DR. The classifiers employed herein, included Knearest neighbor, random forest classifier, support vector machine, regression tree classifier, logistic regression and the Naïve Bayes theorem. Our results showed that the random forest classification model provided the significant detail of attributes in terms of their importance in the diagnosis of the DR. More importantly, our supervised classification models provided the prediction accuracy of the disease and Naïve Bayes classifier demonstrated highest accuracy of 80.15% in the prediction of DR compared to the others. Additionally, receiver operating characteristics (ROC) analysis, with the classifiers and the area under curve (AUC) represented the fitting results of each classifier. The presented approach can prove to be a potential tool for the ophthalmologist in the early diagnosis tool for DR.V. Spandana et al.