Classification of Rib Fracture Formation Time Using Deep Learning Models Based on 2.5D Technology

doi:10.3969/j.issn.1671-2072.2026.02.006

Chinese Journal of Forensic Sciences ›› 2026 ›› Issue (2): 46-52.DOI: 10.3969/j.issn.1671-2072.2026.02.006

• Special Topic Discussion: New Quality Productive Forces Empower Multi-Scenario Applications in Foren • Previous Articles Next Articles

Classification of Rib Fracture Formation Time Using Deep Learning Models Based on 2.5D Technology

SUN Yaning^1,2, YU Xiaoying¹, SHENG Yanliang², WAN Lei¹, XIA Wentao¹

1. Shanghai Key Laboratory of Forensic Medicine, Key Laboratory of Forensic Science, Ministry of Justice, Shanghai Forensic Service Platform, Academy of Forensic Science;2. Key Laboratory of Microecology-Immune Regulatory Network and Related Diseases, School of Basic Medicine, Jiamusi University

Received:2025-09-04 Published:2026-03-15 Online:2026-03-25

基于2.5D技术的深度学习模型用于肋骨骨折形成时间段分类的研究

孙亚宁^1，2，俞晓英¹，盛延良²，万雷¹，夏文涛¹

1.司法鉴定科学研究院上海市法医学重点实验室司法部司法鉴定重点实验室上海市司法鉴定专业技术服务平台；
2. 佳木斯大学基础医学院微生态-免疫调节网络与相关疾病重点实验室

Abstract

Abstract: Objective Based on 2.5D deep learning technology, chest computed tomography （CT） images are used to assist forensic estimation of the time interval of rib fracture injury formation. Methods A total of 331 patients with rib fractures accepted by the Academy of Forensic Science between 2017 and 2024 were enrolled, and 1 290 CT images of rib fractures obtained from the day of injury to 90 days post-injury were included in the dataset. Among them, 464 fractures had an injury formation time ≤21 days and 826 had an injury formation time > 21-90 days. The dataset was divided into training, validation, and test sets at an approximately ratio of 7∶2∶1. After ex-tracting rib fracture regions of interest (ROI) using a segmentation model, the slice with the largest area was taken as the center. Two adjacent slices anterior and posterior to this reference were selected to form a five-slice stack. These slices were merged into a five-channel ResNet-101 deep learning model. The model’s performance on the classification task of rib fracture injury formation time interval was evaluated using indicators including area under the curve, accuracy. Results In the test set, the model achieved an area under the curve of 0.88, an accuracy of 81.54%. Conclusion The deep learning model based on 2.5D technology demonstrates good application potential in determining the time of rib fracture injury formation, effectively integrating multi-slice feature information, and can provide an objective and quantitative method for forensic clinical appraisal.

Key words: forensic clinical medicine, rib fracture, injury formation time, deep learning, CT

摘要： 目的基于2.5D深度学习技术，利用胸部计算机体层成像（computed tomography，CT）影像辅助法医推断肋骨骨折损伤形成时间范围。方法收集司法鉴定科学研究院2017年至2024年间受理的涉及肋骨骨折的331名伤者，将受伤当天到伤后90 d范围内的1 290处肋骨骨折CT图像纳入数据集，其中损伤形成时间≤21 d和21 d<损伤形成时间≤90 d的肋骨骨折分别为464处和826处。将数据集按约7∶2∶1比例划分为训练集、验证集和测试集等。通过分割模型提取肋骨骨折感兴趣区域（region of interest，ROI）后，以最大面积的切片为中心，选取其前后各2张相邻切片（共5张），构建5通道的ResNet-101深度学习模型。采用曲线下面积、准确率等指标，评估模型对肋骨骨折损伤形成时间段分类任务的性能。结果在测试集中，模型的曲线下面积为0.88、准确率为81.54% 。结论基于2.5D技术的深度学习模型在肋骨骨折损伤形成时间的判定任务中展现出良好的应用潜力，有效融合多切片特征信息，可为法医临床鉴定提供一种客观量化的方法。

关键词: 法医临床学, 肋骨骨折, 损伤形成时间, 深度学习, CT

CLC Number:

DF795.4

SUN Yaning, YU Xiaoying, SHENG Yanliang, WAN Lei, XIA Wentao. Classification of Rib Fracture Formation Time Using Deep Learning Models Based on 2.5D Technology[J]. Chinese Journal of Forensic Sciences, 2026(2): 46-52.

孙亚宁, 俞晓英, 盛延良, 万雷, 夏文涛. 基于2.5D技术的深度学习模型用于肋骨骨折形成时间段分类的研究[J]. 中国司法鉴定, 2026(2): 46-52.

Add to citation manager EndNote|Ris|BibTeX

URL: http://www.chsfjd.cn/EN/10.3969/j.issn.1671-2072.2026.02.006

http://www.chsfjd.cn/EN/Y2026/V0/I2/46

References

[ 1 ] 何俊，王东建，王泽飞，等. 影像检查鉴别肋骨陈旧与新鲜骨折1例[J]. 中国法医学杂志，2025，40（S1）：277-278.
[ 2 ] 任智慧，陈佳音. 肋骨骨折鉴定分析1例[J]. 中国法医学杂志，2025，40（S1）：272-273.
[ 3 ] 郭伟，杨泽瑞. 肋骨骨折鉴定引起争议分析1例[J]. 中国法医学杂志，2025，40（S1）：147-148.
[ 4 ] 杨斐，姚永杰，李杰，等. 双能量CT Bone Marrow Edema定量评估肋骨骨折演变时间节点的价值[J]. 中国法医学杂志，2025，40（2）：188-193.
[ 5 ] 林园园，杜运智，崔虹飞，等. 隐匿性肋骨骨折多层螺旋CT检查的最佳时间窗[J]. 中国法医学杂志，2017，32（6）：563-566.
[ 6 ] 倪伟勇，钱昭军. 肋骨骨折鉴定时间研究[J]. 刑事技术，2015，40（2）：114-117.
[ 7 ] JIN L，YANG J C，KUANG K M，et al. Deep-learning-assisted detection and segmentation of rib fractures from CT scans：Development and validation of FracNet[J]. eBioMedicine，2020，62：103106.
[ 8 ] 赵婷婷，郭宇，才佳，等. 医学图像检测和分割模型在肋骨骨折诊断中的应用进展[J]. 天津科技大学学报，2025，40（1）：1-12.
[ 9 ] ZHOU Q Q，TANG W，WANG J S，et al. Automatic detection and classification of rib fractures based on patients’ CT images and clinical information via convolutional neural network[J]. European Radiology，2021，31（6）：3815-3825.
[10] 白洁，孙晶，程晓光，等. 基于YOLOv3算法的肋骨骨折诊断模型的构建及应用[J]. 法医学杂志，2023，39（4）：343-349.
[11] BRINK F W，ADLER B，BAMBACH S，et al. Using deep learning for estimation of time-since-injury in pediatric accidental fractures[J]. Pediatric Radiology，2025，55：1257-1269.
[12] LI Y M，WEI D，LIU X，et al. Molecular subtyping of diffuse gliomas using magnetic resonance imaging：Comparison and correlation between radiomics and deep learning[J]. European Radiology，2022，32（2）：747-758.
[13] CHOI Y S，BAE S，CHANG J H，et al. Fully automated hybrid approach to predict the IDH mutation status of gliomas via deep learning and radiomics[J]. Neuro-oncology，2021，23（2）：304-313.
[14] KIM H，LEE D，CHO W S，et al. CT-based deep learning model to differentiate invasive pulmonary adenocarcinomas appearing as subsolid nodules among surgical candidates：Comparison of the diagnostic performance with a size-based logistic model and radiologists[J]. European Radiology，2020，30（6）：3295-3305.
[15] HU J，ZHENG Z F，WANG S H，et al. Missed rib fractures on initial chest CT in trauma patients：Time patterns，clinical and forensic significance[J]. European Radiology，2021，31（4）：2332-2339.
[16] TANG Y L，JIN L，JI W B，et al. Determining rib fracture age from CT scans with a radiomics-based combined model：A multicenter retrospective study[J]. Insights into Imaging，2023，14（1）：214.
[17] JIN L，SUN Y L，MA Z J，et al. Radiomics-based machine learning for predicting the injury time of rib fractures in gemstone spectral imaging scans[J]. Bioengineering，2022，10（1）：8.
[18] 张怀瀚，曹永杰，张吉，等. CT三维重建技术结合深度学习算法推断成人坐骨年龄[J]. 法医学杂志，2024，40（2）：154-163.
[19] 江洁清，张继宗，刘旭丹. X线数字图像法推断前臂骨折形成时间[J]. 中国法医学杂志，2010，25（3）：165-168.
[20] ATWAN Y，SCHEMITSCH E H. Radiographic evaluations：Which are most effective to follow fracture healing[J]. Injury，2020，51：S18-S22.
[21] 李三保，唐光健，王山林，等. 肋骨隐匿性骨折多排螺旋CT征象分析[J]. 中国CT和MRI杂志，2018，16（6）：141-144.
[22] FERRARI R，TRINCI M，CASINELLI A，et al. Radiomics in radiology：What the radiologist needs to know about technical aspects and clinical impact[J]. La Radiologia Medica，2024，129（12）：1751-1765.

Classification of Rib Fracture Formation Time Using Deep Learning Models Based on 2.5D Technology

基于2.5D技术的深度学习模型用于肋骨骨折形成时间段分类的研究

PDF

Knowledge

Abstract

Cite this article

share this article

References

Recommended Articles

Metrics