文献速递气道管理术前评估与围术期医学

深度学习模型关联侧位颈椎X线特征与Cormack-Lehane 3或4级声门视野

Deep-learning model associating lateral cervical radiographic features with Cormack-Lehane grade 3 or 4 glottic view.
PubMed 文献 · Anaesthesia · 2022-10-05

本研究基于5939例甲状腺手术患者的侧位颈椎X线片,开发并测试卷积神经网络深度学习模型,预测直接喉镜下Cormack-Lehane 3或4级声门视野。253例(4%)出现3或4级视野。新模型Brier评分0.023(95%CI 0.021-0.025)优于其他六个模型(均p<0.001),AUC为0.965(95%CI 0.962-0.969)。舌骨、咽部和颈椎周围影像学特征与3和4级声门视野相关。

要点

  • 新深度学习模型Brier评分0.023,优于VGG、ResNet等六个模型(p<0.001)。
  • 模型AUC为0.965(95%CI 0.962-0.969),平衡准确率0.713(95%CI 0.684-0.742)。
  • 舌骨、咽部和颈椎周围影像学特征与Cormack-Lehane 3或4级视野相关。

核心结论

  • 基于侧位颈椎X线片的卷积神经网络深度学习模型可预测直接喉镜下Cormack-Lehane 3或4级声门视野,且Brier评分优于其他六个对比模型(均p<0.001)。

研究设计与样本

  • 回顾性分析本院5939例甲状腺手术患者的侧位颈椎X线片,其中253例(4%)出现Cormack-Lehane 3或4级声门视野。
  • 使用10个随机抽样数据集训练卷积神经网络模型,并与VGG、ResNet、Xception、ResNext、DenseNet和SENet六个模型比较。

主要发现

  • 新模型Brier评分0.023(95%CI 0.021-0.025),低于VGG 0.034(0.034-0.035)、ResNet 0.033(0.033-0.035)、Xception 0.032(0.031-0.033)、ResNext 0.033(0.032-0.033)、DenseNet 0.030(0.029-0.032)、SENet 0.031(0.029-0.032),均p<0.001。
  • 新模型R2为0.428(95%CI 0.388-0.468),均方误差0.023(95%CI 0.021-0.025),平均绝对误差0.048(95%CI 0.046-0.049)。
  • 新模型平衡准确率0.713(95%CI 0.684-0.742),受试者工作特征曲线下面积0.965(95%CI 0.962-0.969)。
  • 舌骨、咽部和颈椎周围的影像学特征与Cormack-Lehane 3和4级声门视野相关。

临床怎么用

  • 该模型基于侧位颈椎X线片预测困难喉镜视野,可作为术前气道评估的辅助工具,但需进一步外部验证。

局限性

  • 研究仅纳入本院甲状腺手术患者,样本来源单一,外部泛化性需验证。
  • 原文未报告模型在外部数据集上的验证结果。

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)。

原文

Unanticipated difficult laryngoscopy is associated with serious airway-related complications. We aimed to develop and test a convolutional neural network-based deep-learning model that uses lateral cervical spine radiographs to predict Cormack-Lehane grade 3 or 4 direct laryngoscopy views of the glottis. We analysed the radiographs of 5939 thyroid surgery patients at our hospital, 253 (4%) of whom had grade 3 or 4 glottic views. We used 10 randomly sampled datasets to train a model. We compared the new model with six similar models (VGG, ResNet, Xception, ResNext, DenseNet and SENet). The Brier score (95%CI) of the new model, 0.023 (0.021-0.025), was lower ('better') than the other models: VGG, 0.034 (0.034-0.035); ResNet, 0.033 (0.033-0.035); Xception, 0.032 (0.031-0.033); ResNext, 0.033 (0.032-0.033); DenseNet, 0.030 (0.029-0.032); SENet, 0.031 (0.029-0.032), all p < 0.001. We calculated mean (95%CI) of the new model for: R2 , 0.428 (0.388-0.468); mean squared error, 0.023 (0.021-0.025); mean absolute error, 0.048 (0.046-0.049); balanced accuracy, 0.713 (0.684-0.742); and area under the receiver operating characteristic curve, 0.965 (0.962-0.969). Radiographic features around the hyoid bone, pharynx and cervical spine were associated with grade 3 and 4 glottic views.

来源:Anaesthesia,查看原文。版权归原作者所有。

引用

  1. Cho HY, Lee K, Kong HJ, et al. Deep-learning model associating lateral cervical radiographic features with Cormack-Lehane grade 3 or 4 glottic view. Anaesthesia. 2022. PMID: 36198200. doi:10.1111/anae.15874

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。