: The use of computed tomography (CT)-based navigation systems has been shown to improve surgical accuracy in total hip arthroplasty. However, there is limited literature available about the application of CT-based navigation systems in severe hip dysplasia. This study aimed to evaluate the accuracy of a CT-based navigation system in patients with severe hip dysplasia using three-dimensional (3D)-printed bone models. : 3D-printed bone models were generated from CT data of patients with severe hip dysplasia (Crowe type II, 10 hips; type III, 10 hips; and type IV, 10 hips). The accuracy of automatic segmentation, success rate, point-matching accuracy across different registration methods, and deviation values at reference points after registration were assessed. : For the combined cohort of Crowe II, III, and IV cases ( = 30), the Dice Similarity Coefficient and Jaccard Index were 0.99 ± 0.01 and 0.98 ± 0.02, respectively. These values indicate a high level of segmentation accuracy. The "Matching with true and false acetabulum + iliac crest" method achieved a 100% success rate across all groups, with mean deviations of 0.08 ± 0.28 mm in the Crowe II group, 0.12 ± 0.33 mm in the Crowe III group, and 0.14 ± 0.50 mm in the Crowe IV group ( = 0.572). In the Crowe IV group, the anterior superior iliac spine deviation was significantly lower using the "Matching with true and false acetabulum + iliac crest" method compared to the "Matching with true and false acetabulum" method (0.28 ± 0.49 mm vs. 3.29 ± 2.56 mm, < 0.05). : This study demonstrated the high accuracy of automatic AI-based segmentation, with a Dice Similarity Coefficient of 0.99 ± 0.01 and a Jaccard Index of 0.98 ± 0.02 in the combined cohort of Crowe type II, III, and IV cases ( = 30). The matching success rate was 100%, with additional points on the iliac crest, which improved matching accuracy and reduced deviations, depending on the case.