Acute lymphocytic leukemia, a blood cancer, is most common among children under five and is treated with chemotherapy. Neutropenic enterocolitis (NEC) is a life-threatening complication following chemotherapy with high mortality rates in newborns with an incidence rate of up to 46%. CT is used to diagnose NEC but has limited availability, reduced sensitivity, and specificity. Moreover, continuous CT monitoring is not feasible due to the risks involving multiple exposures to ionizing doses. Ultrasound is a viable low-cost, and safe alternative. However, its adoption suffers from several challenges that include the sonography protocol’s complexity, high inter and intra-user variability, and inability to track longitudinal changes throughout treatment. We propose ODIN, a guidance, detection, auto-capture, and measurement system that guides and captures high-quality images from POCUS devices to automate the detection of thickened bowel walls. It minimizes measurement variability across different patient cohorts and due to operators. Moreover, it utilizes deep learning models with interpretable features to generate explainable decisions and provide longitudinal insights.