Objectives Evaluation of the clinical performance of ultra-low-dose computed tomography (CT) images of the shoulder joint on image-based diagnosis and three-dimensional (3D) printing surgical planning. Materials and methods A total of 93 patients with displaced shoulder fractures were randomly divided into standard-dose, low-dose, and ultra-low-dose groups. Three-dimensional printing models of all patients’ shoulder joints were fabricated. The subjective image quality and 3D-printing model were evaluated by two senior orthopedic surgeons who were blinded to any scanning setting. A 3-point scale system was used to quantitatively assess the image quality and 3D printing model, where more than 2 points meant adequate level for clinical application. Results Compared with the standard dose protocol, ultra-low-dose technique reduced the radiation dose by 99.29% without loss of key image quality of fracture pattern. Regarding the subjective image quality, the assessment scores for groups of standard, low, and ultra-low doses were 3.00, 2.76, 2.00 points on scapula and humerus, and 3.00, 2.73, 2.44 points on clavicle. Scores of the three groups for the assessment of 3D printing models were 3.00, 2.80, 1.34 on scapula and humerus, and 3.00, 2.90, 2.06 on clavicle. In the ultra-low-dose group, 24 out of 33 (72.7%) 3D printing models of scapula and humerus received lower than 2 points of the evaluation score, while nearly 94% of the clavicle models reached the adequate level. Conclusion An ultra-low-dose protocol is adequate for the diagnosis of either displaced or non-displaced fractures of the shoulder joint even though minor flaws of images are present. Three-dimensional printing models of shoulder joints created from ultra-low-dose CT scans can be used for surgical planning at specific bone like the clavicle but perform insufficiently in the overall surgical planning for shoulder injuries due to the significant geometric flaws.