Comprehensive Prediction of Retinopathy in Preterm Infants Using Deep Learning Approaches

Abstract

Preterm newborns' visual condition is greatly endangered by retinopathy, which makes prompt and precise identification essential for successful treatment. Using the capabilities of ML (machine learning) computations, the current investigation suggests a combined strategy for retinopathy forecasting in preterm infants. To improve prediction accuracy, specificity, and sensitivity, the investigation also includes a gray-level co-occurrence matrix (GLCM) for collecting features and a median filter to reduce noise. The work uses the RF, SVM, and MLP algorithms for modeling predictions, using their different learning capacities. Together, the non-linear mapping of MLP strengthens the retinopathy forecasting framework. Experiments on a dataset consisting of retinal pictures from preterm infants with different levels of retinopathy show how effective the suggested combined strategy is; the tool used is Jupyter Notebook, and the language used is Python. The findings confirm the usefulness of the suggested method in healthcare settings, guaranteeing prompt and precise identification to avert vision-threatening problems. From the results obtained, the proposed MLP produces an accuracy of 90%, a sensitivity of 0.91, and a specificity of 0.86.