Neonatal Jaundice Identification Over the Face and Sclera using Graph Neural Networks

Neonatal jaundice is normal and generally harmless, but it is not a diagnosis and gets not treated with time. Severe jaundice leads to brain damage ends with death. Symptoms of neonatal jaundice showyellowing in the face and sclera in the baby’s eyes. Buildup of bilirubin in baby’s blood creates jaundice. Naturally, pregnant motherliver removes bilirubin to the baby but after delivery, the baby’s body not begins to remove bilirubin causes neonatal jaundice. When bilirubin exceeds in the baby’s blood, yellow ill appear in the face and sclera in eyes. Presence of yellow coloration built on Total Serum Bilirubin (TSB) level in the blood. Computer vision techniques and deep learning algorithms were performed to identify the intensity of neonatal jaundice to calculate the accuracy, recall, Flscore, and specificity. Spatial and spectral graph neural network (SSGNN), an innovative model based on graphical neural networks, is intended to extract spatial and spectral domain facts from face and sclera images. The pixel color values of face and sclera to predict the Total Serum Bilirubin (TSB) levels. The combination results of SPAGNN, SPEGNN and supplementary extraction would be carried out to maximum intensity of neonatal jaundice with higher accuracy.