基于决策树的门脉高压脾切术并发门脉血栓的早期诊断与预测

The Early Diagnosis and Forecast of PVT After Splenectomy in Patient with Portal Hypertension Based on Decision Tree

利用生物医学统计方法,通过检测P-选择素(P-sel)、血栓前体蛋白(TpP)、D-二聚体(D-D)以及血小板(PLT)对门脉高压症患者术后是否会并发门脉血栓(PVT)进行早期预测.采用SPSS17.0中决策树的CHAID和CRT算法,对48例门脉高压症患者术后的病案数据,通过确定分类因变量、自变量、决策树生长方法和修剪规则,建立术后是否会并发PVT的预测模型.分别建立了术后1、3、5、7、14天的CHAID或CRT分类模型,并在术后第三天和第五天产生了相对较好的预测模型.在门静脉高压症患者术后第三天对P-sel、TpP、D-D、PLT指标进行检测,尤其检测P-sel、D-D的值有助于PVT的早期预测.

In this paper,by using biomedical statistical method,such as testing the P-sel,TpP,D-D and PLT,we propose an approach,which can early predict whether portal hypertension patients should be complicated by PVT after the operation.Based on the CRT and CHAID models of decision tree in SPSS17.0,and by fixing the classification dependent variable and independent variable,the decision tree growth and pruning rules were built from 48portal hypertension patients' medical record data.CHAID or CRT classification models were also built on the first day and 3,5,7,14days after splenectomy respectively,then were produced relatively good forecasting models on the third day and the fifth day.As a result,to test the portal hypertension patients' value of P-sel,TpP,D-D,and PLT on the third day after the operation,especially testing the value of P-sel and D-D can contribute to forecasting PVT.