Jaundice, also known as hyperbilirubinemia, is a frequent health problem that affects the human body. It is a pathological condition that affects the skin and the whites of the eyes, turning them yellow as a result of the deposition of bile pigments caused by excess bilirubin in the blood. An increase in the level of bilirubin in the blood leads to brain damage and sometimes death. Neonatal jaundice is one of the most common types of jaundice, occurring in the majority of premature and newborns within the first several weeks of their life. This thesis aims to estimate the level of bilirubin in the newborn non-invasively using digital images captured by smartphone and Artificial Neural Networks (ANN). The proposed system is used to eliminate the need for frequent finger pricks and blood tests for evaluation of the level of bilirubin in the newborn. The proposed system uses 145 images of newborn babies (50 normal and 95 abnormal) between the ages of one day and several weeks. The augmentation method is used to increase the number of images in the training dataset. The ANN algorithms that have been used in this study are Visual Geometric Group 16 (VGG16), 19 (VGG19), ResNet50, EfficientNet B0, and EfficientNet B7. The classifier is built to classify three classes (normal, low level (TSB1), and high level (TSB2)). The classification results show that ResNet50 has the highest accuracy of 84.091% compared to other algorithms. The low-cost, low-powered, and small-sized Raspberry Pi4 is used as the hardware platform to implement the proposed system

newborn, bilirubin, hyperbilirubinemia, jaundice, neural network algorithm