中国组织工程研究 ›› 2026, Vol. 30 ›› Issue (15): 3929-3935.doi: 10.12307/2026.534

• 骨与关节图像与影像Bone and joint imaging • 上一篇    下一篇

数字算法骨折CT影像识别软件识别AO-C2型桡骨远端骨折的精准性及稳定性

刘  飞1,2,邓新恒3,成永忠1,3,尹晓冬1,2,李晓敏1,朱书朝3,王朝鲁 1,2   

  1. 1中国中医科学院望京医院,北京市  100102;2南阳市中医骨伤生物力学重点实验室,河南省南阳市  473000;3南阳市中医院,河南省南阳市  473000
  • 接受日期:2025-01-20 出版日期:2026-05-28 发布日期:2025-11-07
  • 通讯作者: 王朝鲁,主任医师,中国中医科学院望京医院,北京市 100102;南阳市中医骨伤生物力学重点实验室,河南省南阳市 473000 共同通讯作者:成永忠,主任医师,中国中医科学院望京医院,北京市 100102;南阳市中医院,河南省南阳市 473000
  • 作者简介:第一作者:刘飞,男,1996年生,山东省泰安市人,博士,主要从事骨伤科疾病的临床与基础研究。 共同第一作者:邓新恒,男,1984年生,河南省南阳市人,硕士,主治医师,主要从事创伤骨科疾病的临床研究。
  • 基金资助:
    中国中医科学院望京医院高水平中医医院建设临床循证研究专项(WJYY-XZKT-2023-14),项目负责人:王朝鲁;首都临床特色诊疗技术研究及转化应用(Z2211007422075),项目负责人:成永忠;河南省中医药科学研究专项(2024ZY2197),项目负责人:王朝鲁;南阳市基础与前沿技术研究计划项目(GJGG121),项目负责人:邓新恒

Accuracy and stability of digital algorithm-based CT imaging recognition software in identifying AO-C2 type distal radius fractures

Liu Fei1, 2, Deng Xinheng3, Cheng Yongzhong1, 3, Yin Xiaodong1, 2, Li Xiaomin1, Zhu Shuchao3, Wang Chaolu1, 2   

  1. 1Wangjing Hospital, China Academy of Chinese Medical Sciences, Beijing 100102, China; 2Key Laboratory of Biomechanics for Nanyang Traditional Chinese Medicine Orthopedics, Nanyang 473000, Henan Province, China; 3Traditional Chinese Medicine Hospital of Nanyang, Nanyang 473000, Henan Province, China
  • Accepted:2025-01-20 Online:2026-05-28 Published:2025-11-07
  • Contact: Wang Chaolu, Chief physician, Wangjing Hospital, China Academy of Chinese Medical Sciences, Beijing 100102, China; Key Laboratory of Biomechanics for Nanyang Traditional Chinese Medicine Orthopedics, Nanyang 473000, Henan Province, China Co-corresponding author: Cheng Yongzhong, Chief physician, Wangjing Hospital, China Academy of Chinese Medical Sciences, Beijing 100102, China; Traditional Chinese Medicine Hospital of Nanyang, Nanyang 473000, Henan Province, China
  • About author:Liu Fei, MD, Wangjing Hospital, China Academy of Chinese Medical Sciences, Beijing 100102, China; Key Laboratory of Biomechanics for Nanyang Traditional Chinese Medicine Orthopedics, Nanyang 473000, Henan Province, China Deng Xinheng, MS, Attending physician, Traditional Chinese Medicine Hospital of Nanyang, Nanyang 473000, Henan Province, China Liu Fei and Deng Xinheng contributed equally to this article.
  • Supported by:
    Clinical Evidence-Based Research Project for the Construction of High-Level TCM hospitals in Wangjing Hospital, China Academy of Chinese Medical Sciences, No. WJYY-XZKT-2023-14 (WCL); Research and Transformation Application of Clinical Characteristic Diagnosis and Treatment Technologies in the Capital, No. Z2211007422075 (to CYZ); Henan Province TCM Scientific Research Project, No. 2024ZY2197 (to WCL); Nanyang Basic and Frontier Technology Research Plan Project, No. GJGG121 (to DXH)

摘要:
文题释义
数字算法:是指用于处理数字数据的计算方法,广泛应用于各个领域,通常涉及数值计算、数据处理、分析和优化等方面,以提高效率和准确性。数字算法的设计和实现通常需要考虑效率和准确性,以确保在处理大规模数据时能够快速且精确地得出结果。
AO-C2型桡骨远端骨折:是根据AO/OTA分类系统对桡骨远端骨折进行的一种分类,指骨折线涉及桡骨远端关节面,伴随干骺端的骨折,形成一个不稳定的骨折类型。

摘要
背景:传统的骨折CT影像阅片主要依赖于医生的经验,存在主观性强和误差较大的问题。因此,开发基于数字算法的骨折CT影像识别软件能够有效辅助医生进行骨折类型及位移、旋转等特征的准确识别,具有重要的临床意义。
目的:验证自主开发骨折CT影像识别软件在AO-C2型桡骨远端骨折中的诊断准确性、骨折点识别稳定性,对比软件与医师测量的骨折块位移、旋转角度的差异,探讨CT影像识别软件的临床应用前景。
方法:收集2024年1-6月南阳市中医院收治的25例AO-C2型桡骨远端骨折患者的CT影像,应用骨折CT影像识别软件进行了一系列验证,包括软件在骨折类型、骨折点识别、骨折移位方面的测量,对比骨折CT影像识别软件与医师医疗影像存储与传输系统影像识读测量数据的差异;应用变异系数、双向组内相关系数一致性分析、Bland-Altman分析评估两种方案测量结果的稳定性及一致性。
结果与结论:①骨折CT影像识别软件对骨折类型识别准确率达92%;总骨折点识别的变异系数均小于19%,关节面骨折点变异系数均小于25%,骨干部骨折点变异系数均小于18%,骨折点识读稳定性良好;②组内相关系数分析表明,不同级别医师应用骨折CT影像识别软件测量骨折块移位、旋转值均具有较高的一致性;③Bland-Altman分析表明软件测量与医师医疗影像存储与传输系统测量在骨折位移中无显著差异,软件在骨折块旋转测量中具有较高的精准性;④提示基于数字算法的骨折CT影像识别软件在骨折点识别中具有较好的稳定性,在骨折移位、旋转识别上具有较好的一致性与精准性,对骨折旋转的识别明显优于医疗影像存储与传输系统测量,在AO-C2型桡骨远端骨折的应用中具有良好的临床应用前景,能够辅助医师更快地做出治疗决策。



中国组织工程研究杂志出版内容重点:人工关节;骨植入物;脊柱;骨折;内固定;数字化骨科;组织工程

关键词: 桡骨远端骨折, CT影像, 数字算法, 识别软件, 骨折类型, 骨折点识别, 骨折移位

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

中图分类号: