Kim Joohui, Lee Seungeun, Ahn So Min, Choi Gayoung, Je Bo-Kyung, Park Beom Jin, Cho Yongwon, Oh Saelin
Department of Radiology, Anam Hospital, Korea University College of Medicine, Seoul, Republic of Korea.
Department of Mathematics, Korea University, Seoul, Republic of Korea.
Front Bioeng Biotechnol. 2025 Sep 2;13:1613417. doi: 10.3389/fbioe.2025.1613417. eCollection 2025.
Orbit fractures under 20 years are a medical emergency requiring urgent surgery with the gold standard modality being high-resolution CT. If radiography could be used to identify patients without fractures, the number of unnecessary CT scans could be reduced. The purpose of this study was to develop and validate a deep learning-based multi-input model with a novel cross-sequence learning method, which outperforms the conventional single-input models, to detect orbital fractures on radiographs of young patients. Development datasets for this retrospective study were acquired from two hospitals (n = 904 and n = 910). The datasets included patients with facial trauma who underwent orbital rim view and CT. The development dataset was split into training, tuning, and internal test sets in 7:1:2 ratios. A radiology resident, pediatric radiologist, and ophthalmic surgeon participated in a two-session observer study examining an internal test set, with or without model assistance. The area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and 95% confidence intervals (CIs) were obtained. Our proposed model detected orbital fractures with an AUROC of 0.802. The sensitivity, specificity, PPV, and NPV of the model achieved 65.8, 86.5, 70.9, and 83.5%, respectively. With model assistance, all values for orbital fracture detection improved for the ophthalmic surgeon, with a statistically significant difference in specificity (P < 0.001). For the radiology resident, specificity exhibited significant improvement with model assistance (P < 0.001). Our proposed model was able to identify orbital fractures on radiographs, reducing unnecessary CT scans and radiation exposure.
20岁以下的眼眶骨折是一种需要紧急手术的医疗急症，其金标准检查方式是高分辨率CT。如果能利用X线摄影来识别无骨折的患者，就可以减少不必要的CT扫描次数。本研究的目的是开发并验证一种基于深度学习的多输入模型，该模型采用一种新颖的跨序列学习方法，在检测年轻患者X线片上的眼眶骨折方面优于传统的单输入模型。这项回顾性研究的开发数据集来自两家医院（分别为n = 904和n = 910）。数据集包括接受眼眶边缘视图和CT检查的面部创伤患者。开发数据集按7:1:2的比例分为训练集、调整集和内部测试集。一名放射科住院医师、一名儿科放射科医生和一名眼科外科医生参与了一项两阶段的观察者研究，对内部测试集进行检查，有无模型辅助均可。获得了受试者操作特征曲线下面积（AUROC）、灵敏度、特异度、阳性预测值（PPV）、阴性预测值（NPV）以及95%置信区间（CI）。我们提出的模型检测眼眶骨折的AUROC为0.802。该模型的灵敏度、特异度、PPV和NPV分别达到65.8%、86.5%、70.9%和83.5%。在模型辅助下，眼科外科医生检测眼眶骨折的所有指标均有所改善，特异度有统计学显著差异（P < 0.001）。对于放射科住院医师，在模型辅助下特异度有显著提高（P < 0.001）。我们提出的模型能够在X线片上识别眼眶骨折，减少不必要的CT扫描和辐射暴露。
