| 摘要: |
| 目的:探讨儿童磁共振成像检查中,苯巴比妥与水合氯醛复合镇静的影响因素,并比较镇静诱导时长预测模型的预测能
力。方法:收集170 例完成检查并使用两种镇静药物的患儿资料,采用弹性网络回归模型筛选影响镇静诱导时长的非共线性变
量,并建立随机森林和多元回归预测模型。通过测试集及新募集的85 例独立数据集对模型进行评估。结果:弹性网络回归模
型显示体质量、水合氯醛呕吐次数、苯巴比妥剂量、水合氯醛剂量为影响镇静诱导时长的独立因素。随机森林模型在测试集中
的平均绝对误差(MAE)为2. 9 min,均方根误差(RMSE)为3. 6 min;在独立验证集中的MAE 为3. 0 min,RMSE 为4. 5 min。多
元线性回归模型在测试集中的MAE 为3. 0 min,RMSE 为3. 6 min;在独立验证集中的MAE 为3. 8 min,RMSE 为4. 8 min。结论:
随机森林预测模型在镇静诱导时长的预测准确性高于多元线性回归模型,具有更强的临床转化意义。 |
| 关键词: 磁共振成像检查 苯巴比妥 水合氯醛 儿童 镇静预测 |
| DOI：doi:10.13407/j.cnki.jpp.1672-108X.2025.08.007 |
|
| 基金项目: |
|
| Establishment and Clinical Application of Prediction Model for Effect of Combined Sedation and SedationInduction Time in Children with Magnetic Resonance Imaging Examination |
| Xia Weiwei, Li Tingting |
| (Shanghai Children’s Children’s Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai
Children’s Hospital, Shanghai 200062, China) |
| Abstract: |
| Objective: To probe into the influencing factors of combined sedation with phenobarbital and chloral hydrate in children
undergoing magnetic resonance imaging (MRI) examination, and to compare the predictive ability of sedation induction time prediction
models. Methods: Data were collected from 170 children who underwent MRI with both sedative agents and completed the examination.
The elastic net regression model was employed to identify non-collinear variables affecting sedation induction time. Random forest and
multiple regression models were developed to predict sedation induction time, and the performance of these models was evaluated by
using testing dataset and additional independent dataset of 85 newly recruited cases. Results: The elastic net regression identified
weight, chloral hydrate vomiting frequency, phenobarbital dose, and chloral hydrate dose as independent factors influencing sedation
induction time. The random forest model achieved a mean absolute error (MAE) of 2. 9 min and a root mean square error (RMSE) of
3. 6 min in the testing dataset, with a MAE of 3. 0 min and a RMSE of 4. 5 min in the independent validation dataset. The multiple linear
regression model showed a MAE of 3. 0 min and a RMSE of 3. 6 min in the testing dataset, with a MAE of 3. 8 min and a RMSE of
4. 8 min in the validation dataset. Conclusion: The random forest prediction model is more accurate than the multiple linear regression
model in predicting the sedation induction time and has greater clinical applicability in predicting sedation induction time. |
| Key words: magnetic resonance imaging examination phenobarbital chloral hydrate children sedation prediction |
