主管:中华人民共和国司法部
主办:司法鉴定科学研究院
ISSN 1671-2072  CN 31-1863/N

中国司法鉴定 ›› 2024 ›› Issue (3): 49-54.DOI: 10.3969/j.issn.1671-2072.2024.03.007

• 专题研讨:司法鉴定领域新质生产力 • 上一篇    下一篇

股骨颈骨折影像人工智能快速诊断方法研究

马文静1,刘  凡1,赵  亮2,刘  华1,裴京哲3,张  睿4,施  维4   

  1. 1. 北京市公安司法鉴定中心,北京 100192; 2. 山西医科大学,山西 太原 030600; 3. 北京积水潭医院,北京 100035;
    4. 中国科学院苏州生物医学工程技术研究所,江苏 苏州215163
  • 收稿日期:2023-10-05 出版日期:2024-05-15 发布日期:2024-05-16
  • 作者简介:马文静(1985—),女,副高级警务,硕士,主要从事法医临床学研究。E-mail: 549836639@qq.com
  • 基金资助:
    2019年度国家自然科学基金(201903-24);公安部“双十计划”重点攻关项目(2021SSGG03)。

Study on the Rapid Imaging Diagnosis of Femoral Neck Fracture by Artificial Intelligence

MA Wenjing1, LIU Fan1, ZHAO Liang2, LIU Hua1, PEI Jingzhe3, ZHANG Rui4, SHI Wei4   

  1. 1. Beijing District Pubic Security Identification Center, Beijing 100192, China; 2. Shanxi Medical University, Taiyuan 030600, China; 3. Beijing Jishuitan Hospital, Beijing 100035, China; 4. Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, Suzhou 215163, China
  • Received:2023-10-05 Published:2024-05-15 Online:2024-05-16

摘要: 目的 旨在构建人体股骨颈骨折影像人工智能(artificial intelligence,AI)快速诊断方法,从而实现股骨颈骨折的自动化评估。方法 采集1 018例股骨颈骨折的病例作为研究样本,分为训练集(676例)、验证集(112例)和测试集(230例)。应用训练集和验证集样本通过自动化影像预处理、骨折诊断模型构建和模型评估三个步骤进行股骨颈骨折评估模型构建,应用测试集对模型进行测试。 结果 建立了1 018例股骨颈骨折样本数据库。建立的智能评估模型识别股骨颈骨折精准率达到78.7%。 结论 该模型将为股骨颈骨折自动化评估软件的研发提供技术支撑。

关键词: 法医影像学, 股骨颈骨折, 人工智能, 图像识别, 深度学习

Abstract: Objective To develop an artificial intelligence (AI) method for the rapid imaging diagnosis of human femoral neck fractures and realize automatic evaluation. Methods A total of 1 018 cases of femoral neck fractures were collected and divided into training set (676 cases),validation set (112 cases),and test set (230 cases). The training set and validation set were used to construct the evaluation model of femoral neck fracture through three steps: automatic image pretreatment,fracture diagnosis model construction and model evaluation. The test set was used to validate the model. Results A database of 1 018 cases of femoral neck fracture was established. The accuracy of the established intelligent assessment model in identifying femoral neck fractures reached 78.7%. Conclusion The established intelligent evaluation model will provide technical support for the development of automatic evaluation software of femoral neck fractures.

Key words: forensic imaging, femoral neck fracture, artificial intelligence(AI), image recognition, deep learning

中图分类号: