深度学习技术在重症监护室患者住院死亡风险预测中的应用

目的运用深度学习方法构建ICU患者住院死亡风险智能化预测模型,并将该模型与ICU患者中普遍使用的简化急性生理评分(SAPS-II)进行预测性能比较。方法采用深度学习算法建立一种智能化的死亡风险预测模型,使用十二折交叉验证法对模型性能进行比较,选取准确率、灵敏度、特异度、约登指数、召回率等5种评价指标。结果实验结果显示,基于深度学习模型的方法比SAPS-Ⅱ在准确率、灵敏度、特异度、约登指数和召回率上分别提高1.77%、1.02%、0.49%、0.15%和1.10%。结论针对ICU患者住院死亡风险数据的非线性、复杂性和无规律性,深度学习模型比SAPS-Ⅱ表现出更好的泛化能力,具有更高的准确率,更适合对ICU患者住院死亡风险进行预测,可为医院的智能化预测提供了一种新的方向。

Application of deep learning technology in risk prediction of death in ICU

【Objective】To construct the intelligent prediction model of death risk of ICU patients by using deep learning algorithm,and to compare the prediction performance of the model with that of the Simplified Acute Physiology Score II(SAPS-II)commonly used in ICU patients.【Methods】A kind of intelligent death risk prediction model was established by using deep learning algorithm.The performance of the model was compared by using 12 fold cross validation method.Five evaluation indexes,such as accuracy,sensitivity,specificity,Jordan index and recall rate,were selected.【Results】The experimental results showed that the accuracy,sensitivity,specificity,Jordan index and recall of the proposed method were 1.77%,1.02%,0.49%,0.15% and 1.10% higher than those of SAPS-II.【Conclusion】In view of the nonlinearity,complexity and irregularity of death risk data in ICU,deep learning model has better generalization ability and higher accuracy than SAPS-II.Therefore,the in-depth learning model proposed in this paper is more suitable for predicting the risk of death in ICU,and provides a new direction for the intelligent prediction of the hospital.