Novel features extraction: pigment epithelial detachment detection using machine learning algorithms

Abstract

The majority of the retinal diseases have visual symptoms. Any area of the retina, a delicate layer of tissue on the interior the posterior wall of the human eye, can be impacted by retinal disorders. Optical coherence tomography (OCT) is the utmost commonly used imaging procedure for diagnosing retinal disorders such as age-related macular degeneration (ARMD), diabetic retinopathy, pigment epithelial detachment (PED), macular holes, and more. In this study, we put forth a brand-new technique for accurately extracting features from OCT images to identify PED diseases. For the preprocessing step, we examined the wiener filtering method. After that, we segmented the retinal pigment epithelium (RPE) layer used to the thresholding method, extracted the features from the RPE layer, and then gave the features to machine learning (ML) classifiers like the support vector machine (SVM), logistic regression (LR), k-nearest neighbors (KNN), decision tree (DT), random forest (RF), naive Bayes (NB), and artificial neural network (ANN). The total dataset about 200 images among 100 is normal and 100 is PED, we trained the dataset as an unbalanced and balanced group. The RF is the best outcome in comparison of other classifiers. The overall outcome of random forest is 100% accuracy.