Jaundice Recognition in Newborn Face, Chest and Abdomen using Spatial and Spectral Domain Graph Neural Network

Abstract

Jaundice in newborns is common and generally no pain, but if it is not diagnosed and not handled with proper time, it will cause acute yellowing of the skin, which damages the brain and even death. Jaundice in a newborn manifest as yellowing of the infant’s face and chest. This is caused by the buildup of bilirubin in the blood of baby. Naturally, the liver of pregnant mother removes bilirubin from the baby, but adhering to delivery, thebody of baby does not begin to remove bilirubin, causing newborn jaundice. The infant’s face and chest turn yellow when bilirubin levels produce in the blood are too high and yellow coloration is present on the total serum bilirubin (TSB) level in the blood. Deep learning (DL) methods have been used to determine the degree of newborn jaundice using spectral and spatial graph neural networks (SSGNN). This jaundice prediction will improve the health and quality of life of a neonatal. It is a novel model based on graphical neural networks to extract information from photos of the face and the chest in the spatial and spectral domains.The image color information of the face and chest are used to predict the TSB levels. The combined impacts from spatial domain based on graph neural networks (SPAGNN) and the spectral domain based on graph neural networks (SPEGNN) with supplementary extraction will be carried out to maximize the intensity of the new model with higher accuracy. The performance of the SSGNN model is evaluated using recall,accuracy, specificity, and F1 score.