Clinical value of a nomogram based on ultrasonic radiomics and clinical characteristics for predicting the prognosis of patients with left ventricular aneurysm after myocardial infarction
Objective To construct a nomogram model based on ultrasonic radiomics and clinical characteristics,and to explore its clinical application value in predicting the long-term prognosis of patients with left ventricular aneurysm after acute myocardial infarction(PI-LVA).Methods A total of 153 patients with PI-LVA from the cardiovascular centre of Suining Central Hospital were selected.They were divided into event group(38 cases)and non-event group(115 cases)according to the followed up with the endpoint event of cardiac death at 24th,36th,60th months,and the differences in general data and ultrasonographic results were compared.Multifactorial Cox regression analysis was used to screen the influencing factors related to the long-term prognosis of patients with PI-LVA.Images of ventricular wall tumour from the apical of four-chamber cardiac sections were captured,regions of interest were segmented and acquired by ITK-SNAP software,and ultrasonic radiomics features were extracted by Pyradiomics,the feature subset and corresponding non-zero coefficients were screened by least absolute systolic and selection operator(LASSO)regression.The prediction model was constructed and plotted as a nomogram based on the ultrasonic radiomics and influencing factors screened by multifactorial Cox regression analysis,and the differentiation of the model for predicting the endpoint event at 24th,36th,60th months in patients with PI-LVA was analyzed by receiver operating characteristic(ROC)curve.The calibration curve and decision curve were applied to assess the calibration degree and clinical applicability of the model.Results There were statistical significance in proportion of left ventricular ejection fraction(LVEF)<35％,severe right coronary artery stenosis and Tp-Te interval between event and non-event groups(all P<0.05).Multifactorial Cox regression analysis showed that LVEF<35％,severe right coronary artery stenosis and Tp-Te interval were independent influencing factors on the long-term prognosis of patients with PI-LVA.A total of 107 ultrasonic radiomics features were extracted,three significant features were screened and selected by LASSO regression analysis.A prediction model was constructed based on LVEF<35％,severe right coronary artery stenosis,Tp-Te interval and ultrasonic radiomics features,then a nomogram was plotted for visualisation.ROC curve analysis showed that the area under the curve of nomogram model for predicting the endpoint event at 24th,36th,60th months in patients with PI-LVA were 0.850,0.831 and 0.898,respectively.Calibration curve showed that the predicted probability was consistent with the actual probability,and decision curve analysis showed that the nomogram had good clinical applicability.Conclusion The nomogram model based on ultrasonic radiomics and clinical characteristics can effectively predict the long-term prognosis of patients with PI-LVA,and has good clinical application value.
NETL
NSTL
万方数据
