可解释机器学习预测1型糖尿病患者胰岛素需求量
本研究基于神户大学医院2010年4月1日至2024年2月29日期间使用持续葡萄糖监测联合持续皮下胰岛素输注的110例1型糖尿病患者数据,回顾性开发机器学习模型预测每日胰岛素总剂量(TDD)。表现最佳的随机森林模型平均绝对百分比误差(MAPE)为19.8%,体重是最重要的预测特征,其次为腰围和碳水化合物摄入量。作者认为此类模型可能有助于胰岛素治疗患者的治疗,但需进一步开发。
要点
- 随机森林模型预测每日胰岛素总剂量的平均绝对百分比误差(MAPE)为19.8%。
- 预测每日胰岛素总剂量最重要的特征为体重,其次为腰围和碳水化合物摄入量。
- 作者认为该模型可能有助于胰岛素治疗患者的治疗,但需进一步开发。
核心结论
- 作者开发了基于临床信息预测每日胰岛素总剂量的机器学习模型,认为此类模型可能有助于许多接受胰岛素治疗患者的治疗,但需进一步开发。
研究设计与样本
- 本研究为单中心回顾性观察研究,纳入2010年4月1日至2024年2月29日期间就诊于神户大学医院并使用持续葡萄糖监测联合持续皮下胰岛素输注的1型糖尿病患者,共110例。
- 研究开发基于机器学习的每日胰岛素总剂量预测模型,并以平均绝对百分比误差(MAPE)评估模型性能,同时应用可解释人工智能框架提升模型可解释性。
主要发现
- 共纳入110例1型糖尿病患者。
- 表现最佳的模型为随机森林,其平均绝对百分比误差(MAPE)为19.8%。
- 预测每日胰岛素总剂量最重要的特征为体重,其次为腰围和碳水化合物摄入量。
临床怎么用
- 原文未据此给出临床推荐。
局限性
- 作者在结论中写明需进一步开发(further developments being warranted)。
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
CONTEXT: Adverse events related to insulin therapy remain common in individuals with type 1 diabetes (T1D). Optimization of insulin dose can reduce the frequency of these events and help to prevent macrovascular and microvascular complications. OBJECTIVE: The aim of the present study was to develop machine-learning models to predict the total daily dose (TDD) of insulin on the basis of data available in routine clinical practice, to evaluate the performance of the models, and to interpret the relation between its predictions and features. METHODS: This retrospective observational study conducted at a single center recruited individuals diagnosed with T1D who visited Kobe University Hospital and used continuous glucose monitoring in combination with continuous subcutaneous insulin infusion between April 1, 2010 and February 29, 2024. We developed TDD prediction models based on machine learning and evaluated its performance on the basis of the mean absolute percentage error (MAPE). Explainable artificial intelligence frameworks were applied to the machine-learning model to facilitate interpretability. RESULTS: A total of 110 individuals with T1D were included in the study. The best-performing model, the random forest, achieved a MAPE of 19.8%. The most important feature of the model for prediction of the TDD of insulin was body weight, followed by waist circumference and carbohydrate intake. CONCLUSION: We here developed machine-learning models that predict the TDD of insulin from clinical information. Such models could contribute to the treatment of many individuals undergoing insulin therapy, with further developments being warranted.
引用
- Yoshimura K, Hirota Y, Saito S, et al. Prediction of Insulin Requirements by Explainable Machine Learning for Individuals With Type 1 Diabetes. J Clin Endocrinol Metab. 2025. PMID: 39661460. doi:10.1210/clinem/dgae863
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。