The background of the study is rooted in deep learning and e-health for diagnosis and prediction of spinal bifida, a congenital abnormality that affects 1–5% of the population. Basically, Ultrasonography can detect about 60–70% of the abnormalities, with the remaining 30–40% detected after delivery. The effect of these abnormalities when disregarded and medically untouched can prevent walking and making it unpleasant. The objective of the study is to develop and generate evidence-based infant congenital abnormalities that focus on deep learning. The model is evaluated for prediction using a standard parameter matrix such as Accuracy, Precision, Recall Score, and F1 Score. The deep learning pretrained architecture was deployed during the implementation of the model. The outcome of prediction in the model utilizes three algorithms which are Deep Belief Network (DBN), U-Network (U-Net), and Deep Residual Network (ResNet). The dataset anomalies detected were those related to Spinal bifida; of around 6100 ultrasound datasets, 80% had abnormal status and 20% had normal status during training and testing of the model. The developed model in this research has the potential in assisting medical professionals to forecast and manage the causes of spinal Bifida mortality rate. The global causes of this problem will require earlier detection as provided in the result of this model.