Anomalies in the pancreas regional morphology and texture may now be examined by accurately segmenting the organs head, body, and tail on CT images. Hand-drawn pancreatic subregion mapping is labor-intensive, slow, and prone to mistakes. For the zonal segmentation of various anatomical properties, many deep learning networks have been utilized in the present approaches. The three subregions are almost ever visible together in the two-dimensional CT abdominal slices, which limits how the contextual data may be used by the current methods. In this study, we offer a multistage method that uses CT images of pancreatic subregions to accurately and automatically segment 3D objects. The U-Net model is then used to calculate the combined probability of the two maps to perform the best sub regional segmentation. The datasets D1 and D2 of contrast-enhanced abdominal CT images were used to assess the models performance together with a healthy pancreas from the public NIH dataset.