CT scans may now examine regional morphological and textural defects in the pancreas thanks to accurate sub-regional segmentation of the pancreatic head, body, and tail. Manual sub-region mapping of the pancreas requires a lot of time and effort and might lead to mistakes. Current methods for zonal segmentation of various anatomical features make use of many deep learning networks. As a result, the current algorithms can only make limited use of the contextual data since the three sub areas are rarely visible together on the two-dimensional CT abdominal slices. Using computed tomography (CT) images of the pancreas, we offer a multi-stage approach for precise and automated 3D segmentation. The U-Net model is then used to perform the optimum sub-regional segmentation by computing the joint probability of the two maps. A healthy pancreas from the public NIH dataset and the datasets D1 and D2 of contrast-enhanced abdominal CT images were used to assess the model's accuracy