构建基于普美显增强MRI的肝切除术后肝衰综合预测模型

Construction of a comprehensive prediction model for post-hepatectomy liver failure based on Gd-EOB-DTPA enhanced MRI

目的:基于普美显增强MRI信号强度并结合临床变量,构建肝切除术后肝衰(post-hepatectomy liver failure,PHLF)的预测模型。方法:回顾性分析我院2016年01月至2020年12月接受肝切除术196例患者的临床资料,并测量普美显增强MRI图像的信号强度。通过Logistic回归确定PHLF的独立危险因素,以多因素Logistic回归分析结果构建列线图预测模型,并使用一致性指数、校正曲线和决策曲线分析验证模型的预测能力。最后,将模型与其他常用肝功能评估方法进行比较。结果:62例(31.6%)患者发生了PHLF,多因素Logistic回归分析发现凝血酶原活动度(PTA)、手术时间、肝胆期肝脏信号强度(SI_(HBP))是PHLF的独立危险因素。预测模型的曲线下面积(AUC)为0.810。校正曲线和决策曲线显示模型具有较高预测能力和临床实用性。AUC显示该模型均显著优于ICGR-15等肝功能评估方法。结论:基于普美显增强MRI肝胆期信号强度和临床变量的PHLF预测模型,具有较佳的预测性能。

Objective:To construct a predictive model for post-hepatectomy liver failure(PHLF)based on the signal intensity of Gd-EOB-DTPA enhanced MRI and clinical variables.Methods:The clinical data of 196 patients who underwent hepatectomy from January 2016 to December 2020 in our hospital were retrospectively analyzed,and the signal intensity of Gd-EOB-DTPA enhanced MRI images was measured.Independent high-risk factors for PHLF were identified by Logistic regression.A nomogram prediction model was constructed with the results of multivariable Logistic regression analysis,and the predictive ability of the model was validated using the concordance index,calibration curve,and decision curve analysis.Finally,the model was compared with other commonly used liver function assessment methods.Results:PHLF occurred in 62 patients(31.6%).Multivariable Logistic regression analysis identified prothrombin activity(PTA),operative time,and signal intensity of hepatobiliary phase(SI_(HBP))were independent risk factors for PHLF.The area under the curve(AUC)of the nomogram model was 0.810.The calibration curve and decision curve showed that the model had high predictive ability and clinical utility.The AUC showed that the model was significantly better than the ICGR-15 and other liver function assessment methods.Conclusion:The PHLF prediction model based on signal intensity in the hepatobiliary phase of Gd-EOB-DTPA enhanced MRI and clinical variables has better predictive performance.