Progression to end-stage renal disease (ESRD) in childhood is associated with an increased risk of cardiovascular disease, metabolic bone disease, and death. Congenital abnormalities of the kidney and urinary tract (CAKUT), including posterior urethral valves, account for 50-60% of chronic kidney disease (CKD) in children and are the most common cause of ESRD in this age group. In children with CAKUT, kidney injury is often already established at birth due to renal dysplasia. Some children, however, maintain preserved kidney function into adulthood while others progress to ESRD in childhood. Our ability to effectively implement therapies to slow CKD progression is limited by our lack of understanding of which patients are at greatest risk for CKD progression and therefore would be most likely to benefit from early intervention. Thus there is a need for biomarkers that can identify children with CAKUT early in life who are at high risk of future CKD progression. To identify children with CAKUT early in life who are at high risk of future CKD progression, we will develop novel computational methods to derive clinically informative biomarkers from ultrasound (US) imaging data using deep convolutional neural networks (CNNs) and effectively integrate them with established clinical measures for early prediction of CKD progression. To achieve reliable and accurate kidney segmentation, Aim 1 develops an automatic kidney segmentation method by adopting fully CNNs, conditional random fields, and active contour models to simultaneously learn informative high-level image features and inter-voxel relationship under shape regularizations to improve the segmentation accuracy and robustness to imaging noise. To achieve improved prediction of CDK progression based on US imaging data, deep CNNs will be adopted to learn informative imaging features in a multi-instance learning framework to predict which children with CAKUT will develop CKD progression and their timing of progression in Aim 2. These techniques will be applied to a dataset of patients followed at the Children's Hospital of Philadelphia, in order to derive individualized predictive indices of CKD progression. The proposed new techniques will allow us to early differentiate patients with distinct disease progression patterns.