使用深度学习模型提升放射科医生检测脑动脉瘤的表现:一项多中心研究
本研究基于11个临床中心3829例患者数据训练深度学习模型,并在三个机构484例患者上测试,由10名放射科医生(4名初级、6名高级)分别以单独阅片、单独深度学习模型及两者结合三种方式判读。深度学习辅助使初级医生AUC从0.842升至0.881,高级医生从0.853升至0.895,敏感度在病灶水平和患者水平均显著提高,阅片时间缩短37.2%,后处理时间缩短90.8%。
要点
- 深度学习辅助使初级医生AUC从0.842升至0.881,高级医生从0.853升至0.895。
- 病灶水平敏感度初级医生68.9%升至81.6%,高级医生72.4%升至83.5%。
- 联合深度学习模型使阅片时间缩短37.2%,后处理时间缩短90.8%。
核心结论
- 深度学习模型提升了放射科医生检测脑动脉瘤的诊断表现,对初级医生尤为明显,并加快了工作流程。
研究设计与样本
- 多中心研究,深度学习模型训练使用11个临床中心3829例患者数据,测试使用三个机构484例患者;由10名放射科医生(4名初级、6名高级)、深度学习模型单独及两者结合三种方式判读。
诊断标准与关键数值
- 联合深度学习模型后,初级医生AUC从0.842升至0.881(P=0.008),高级医生从0.853升至0.895(P<0.001)。
- 联合深度学习模型后,病灶水平敏感度初级医生从68.9%升至81.6%(P=0.011),高级医生从72.4%升至83.5%(P<0.001)。
- 联合深度学习模型后,患者水平敏感度初级医生从76.2%升至86.9%(P=0.011),高级医生从80.1%升至88.2%(P<0.001)。
- 联合深度学习模型后,患者水平特异度初级医生从82.6%升至82.7%(P=0.005),高级医生从82.6%升至86.1%(P=0.021)。
临床应用与阅片要点
- 深度学习辅助可提高初级和高级放射科医生检测脑动脉瘤的敏感度与AUC。
- 联合深度学习模型可缩短阅片时间37.2%和后处理时间90.8%。
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
RATIONALE AND OBJECTIVES: This study aimed to develop a deep learning (DL)-based model for detecting and diagnosing cerebral aneurysms in clinical settings, with and without human assistance.
MATERIALS AND METHODS: The DL model was trained using data from 3829 patients across 11 clinical centers and tested on 484 patients from three institutions. Image interpretations were conducted by 10 radiologists (four junior, six senior), the DL model alone, and a combination of radiologists with the DL model. Time spent on post-processing and reading was recorded. The analysis of the area under the curve (AUC), sensitivity, and specificity for the above-mentioned three reading modes was performed at both the lesion and patient levels.
RESULTS: Combining the DL model with radiologists reduced image interpretation time by 37.2% and post-processing time by 90.8%. With DL model assistance, the AUC increased from 0.842 to 0.881 (P = 0.008) for junior radiologists (JRs) and from 0.853 to 0.895 (P < 0.001) for senior radiologists (SRs). With DL model assistance, sensitivity significantly improved at both lesion (JR: 68.9% to 81.6%, P = 0.011; SR: 72.4% to 83.5%, P < 0.001) and patient levels (JR: 76.2% to 86.9%, P = 0.011; SR: 80.1% to 88.2%, P < 0.001). Specificity at the patient level showed improvement (JR: 82.6% to 82.7%, P = 0.005; SR: 82.6% to 86.1%, P = 0.021).
CONCLUSIONS: The DL model enhanced radiologists' diagnostic performance in detecting cerebral aneurysms, especially for JRs, and expedited the workflow.
引用
- Zhuo L, Zhang Y, Song Z, et al. Enhancing Radiologists' Performance in Detecting Cerebral Aneurysms Using a Deep Learning Model: A Multicenter Study. Acad Radiol. 2024. PMID: 39406577. doi:10.1016/j.acra.2024.09.038
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。