基于XGBoost算法的出血性脑卒中患者血肿扩张风险预测

Risk Prediction of Hematoma Expansion in Hemorrhagic Stroke Patients Based on XGBoost Algorithm

摘要: 出血性脑卒血肿扩张事件的发生在医学上是一个重要的治疗时间点，如果发现及时可以减少脑卒中患者发生不良预后结果的概率。本文使用XGBoost算法对出血性脑卒中患者发生血肿扩张的风险进行预测，模型训练完成后，使用混淆矩阵、ROC曲线、AUC值、准确率、精确率、召回率和F1_score等指标进行模型的性能评估。通过对130名患者的数据进行研究后发现，基于XGBoost算法的血肿扩张预测模型的预测准确率为80%。其中特征向量数为32时，预测模型的表现最佳，各项性能指标均优于其他特征向量数。本研究纳入模型建立的患者样本数量较少，后续可考虑增加样本量从而进一步提高模型的预测准确率。综上所述，基于XGBoost算法的血肿扩张预测模型可为医疗人员在临床医治时提供参考，具有一定的临床应用价值。

Abstract: The occurrence of hemorrhagic stroke hematoma dilatation events is an important therapeutic time point in medicine, and if detected in a timely manner can reduce the probability of adverse prognostic outcomes in stroke patients. In this paper, the risk of hematoma dilatation in hemorrhagic stroke patients was predicted using the XGBoost algorithm, and after the model training was completed, the performance of the model was evaluated using metrics such as confusion matrix, ROC curves, AUC values, accuracy, precision, recall, and F1_score. After studying the data of 130 patients, it was found that the prediction accuracy of hematoma expansion prediction model based on XGBoost algorithm was 80%. Among them, the best performance of the prediction model was achieved when the number of feature vectors was 32, and all the performance metrics were better than the other number of feature vectors. The number of patient samples included in this study for modeling is small, and increasing the sample size can be considered to further improve the prediction accuracy of the model. In conclusion, the hematoma expansion prediction model based on the XGBoost algorithm can provide reference for medical personnel in clinical treatment and has certain clinical application value.