Abstract: BACKGROUND: Traditional interpretation of fracture CT images primarily depends on physician experience, leading to substantial subjectivity and potential for significant error. Consequently, developing fracture CT image recognition software based on digital algorithms can effectively assist physicians in accurately identifying fracture types, displacement, rotation, and other features, which is of great clinical significance.
OBJECTIVE: To validate the diagnostic accuracy and stability of fracture point recognition of self-developed CT imaging recognition software for AO-C2 type distal radius fractures, compare the differences in displacement and rotation data of fracture fragments measured by the software and physicians, and explore the clinical application prospects of the CT imaging recognition software. 
METHODS: CT images were collected from 25 cases of AO-C2 type distal radius fractures treated at Nanyang Traditional Chinese Medicine Hospital between January and June 2024. A series of validations were performed using the fracture CT imaging recognition software, assessing its performance in fracture type identification, fracture point recognition, and displacement measurement. Differences between the software and physician measurements based on the picture archiving and communication system were compared. The stability and consistency of the results were analyzed using the coefficient of variation, intraclass correlation coefficient, and Bland-Altman analysis.  
RESULTS AND CONCLUSION: (1) The fracture CT imaging recognition software achieved a 92% accuracy rate in identifying fracture types. The overall coefficient of variation for fracture point recognition was below 19%, with coefficient of variation for joint surface fracture points under 25% and for bone shaft fracture points under 18%, demonstrating good stability in fracture point recognition. (2) Intraclass correlation coefficient analysis revealed high consistency in displacement and rotation measurements of fracture fragments by physicians of varying experience levels using the fracture CT image recognition software. (3) Bland-Altman analysis showed no significant differences in fracture displacement measurements between the software and those obtained by physicians using the picture archiving and communication system, while the software exhibited high precision in measuring the rotation of fracture fragments. (4) This study indicates that the fracture CT imaging recognition software, based on digital algorithms, shows good stability in recognizing fracture points and strong consistency and precision in identifying fracture displacement and rotation. The recognition of fracture rotation is significantly better than that of picture archiving and communication system. It holds promising clinical application prospects for AO-C2 type distal radius fractures, enabling physicians to make quicker treatment decisions. 

Key words:  distal radius fracture, CT imaging, digital algorithms, recognition software, fracture type, fracture point identification, fracture displacement