文献速递骨肌技术与安全CT

基于深度学习的骶髂关节炎CT分级系统:一项多中心研究显示其准确性与效率优于人类阅片者

Deep learning based CT grading system for sacroiliitis: a multi-center studydemonstrating superior accuracy and efficiency compared to human readers.
PubMed 文献 · Eur J Radiol · 2025-11-22

本研究开发并验证了用于中轴型脊柱关节炎(axSpA)骶髂关节炎自动CT分级的3D-ResNet50深度卷积神经网络,共纳入1590例患者,其中1341例来自风湿病专科医院用于模型开发,并在130例内部验证集和两家三级医院的249例外部队列中评估。模型在外部验证集对I-IV级的AUC分别为0.93、0.89、0.92、0.88,均超过阅片者(AUC 0.65-0.93),且每例诊断时间显著缩短(2.74±0.15秒 vs 119.4±42.0秒)。

要点

  • 模型在外部验证集对I-IV级骶髂关节炎的AUC为0.93、0.89、0.92、0.88,均高于阅片者。
  • 模型每例诊断时间2.74±0.15秒,显著短于阅片者的119.4±42.0秒。
  • II级改善最显著(I-III级p<0.001),IV级与阅片者无显著差异(p>0.05)。

核心结论

  • 3D-ResNet50模型能准确进行骶髂关节炎分级,敏感性和效率优于风湿科医生和放射科医生,支持将其整合到临床实践中辅助axSpA诊断。

研究设计与样本

  • 开发并验证用于axSpA骶髂关节炎自动CT分级的深度卷积神经网络(DCNN),共纳入1590例患者,其中1341例axSpA患者来自风湿病专科医院用于3D-ResNet50模型开发,内部验证集130例,外部验证集来自两家三级医院共249例。

诊断标准与关键数值

  • 3D-ResNet50模型在训练集、测试集、内部验证集和外部验证集的总体样本级诊断准确率分别为89.6%、89.9%、88.5%和87.1%。

临床应用与阅片要点

  • 模型在外部验证集对Grade I、II、III、IV的AUC分别为0.93、0.89、0.92、0.88,均超过阅片者(AUC 0.65-0.93),其中Grade II改善最显著(Grade I-III p<0.001;Grade IV p>0.05)。模型每例诊断时间2.74±0.15秒,显著短于阅片者的119.4±42.0秒。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

英文原文(摘要)· 点击展开

BACKGROUND: To develop and validate a deep convolutional neural network (DCNN) for automated sacroiliitis grading in axial spondyloarthritis (axSpA) using CT images.

METHODS: A total of 1,590 patients were enrolled, including 1341 axSpA patients from a Rheumatology Specialist Hospital for 3D-ResNet50 model development. The model was then evaluated on an internal validation set of 130 patients and external sets from two tertiary hospitals (n = 249). Diagnostic ssensitivity and reading time were assessed for six readers (two rheumatologists, two junior radiologists, and two senior radiologists). Participant-assigned sacroiliitis Grades were compared with model predictions, and the area under the curve (AUC) was analyzed using the DeLong test.

RESULTS: The 3D-ResNet50 model achieved overall sample-level diagnostic accuracies of 89.6 %, 89.9 %, 88.5 %, and 87.1 % in the training, testing, internal validation, and external validation sets, respectively.The AUCs for Grade I, II, III, and IV in external validation set were 0.93, 0.89, 0.92 and 0.88, all exceeding those of readers (AUC: 0.65-0.93), with the most significant improvement at Grade II (p < 0.001 for Grades I-III; p > 0.05 for Grade IV). The model significantly reduced diagnostic time per case (2.74 ± 0.15 s) compared to readers (119.4 ± 42.0 s).

CONCLUSION: The 3D-ResNet50 model accurately grades sacroiliitis with superior sensitivity and efficiency compared to rheumatologists and radiologists. Its strong performance supports its integration into clinical practice to assist in axSpA diagnosis.

来源:Eur J Radiol,查看原文。版权归原作者所有。

引用

  1. Xu M, Li Z, Ding N, et al. Deep learning based CT grading system for sacroiliitis: a multi-center studydemonstrating superior accuracy and efficiency compared to human readers. Eur J Radiol. 2025. PMID: 41330312. doi:10.1016/j.ejrad.2025.112557

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。