This study proposed a 3D multi-tasking segmentation neural network to address difficulties in image segmentation of the hip for the diagnosis of hip osteoarthritis. The task for this convolutional neural network (CNN) is to segment images of the hip using CT scans and to reconstruct them as 3D models for clinicians to assess. This study employed a modified U-Net-based neural network that included a Transformer module and a boundary optimization algorithm. 125 single-leg CT cases were analyzed, with the data split into training and test groups at a 9:1 ratio. Dice coefficient, combined with Precision, Recall, ASD, and HD95, was used to construct a loss function assessment system. With all additions combined, the final constructed model was able to score a Dice coefficient of (0.95 ± 0.03). This data demonstrates that the constructed CNN is more effective than most other models represented in other literature. This study overall provides a valuable new model for segmentation and 3D reconstruction of the periarticular bone. With both high accuracy and fast generation times, this study’s model has high potential, but still requires clinical implementation for further improvement.