Clinical Application Of Two-dimensional Echocardiography Combined With Three-dimensional Speckle-tracking Imaging To Construct A Prediction Model For Early Identification Of FHCM Mutation Gene Carriers

Objective To use 3D-STI to evaluate the changes of left ventricular systolic function and synchrony in FHCM mutation gene carriers,and to explore the echocardiographic parameters for early identification of FHCM mutation gene carriers.Method 129 first-degree relatives of FHCM patients in our hospital from November 2017 to August 2022 were enrolled.Whole exome sequencing and Sanger sequencing were used for genetic testing,and they were divided into FHCM mutation gene carrier group(G+P-: n=54)and the control group(G-P-: n=75)according to the genetic testing results.Using Philips i E33 ultrasonic diagnostic instrument and Tom Tec off-line software to obtain relevant ultrasonic parameters.Compare the differences between the two groups in conventional echocardiographic measurement parameters,left ventricular three-dimensional global strain parameters and synchronization parameters,and the diagnostic efficacy of each parameter for early identification of G+P-were evaluated.Result 1.Compared with G-P-group,LVMWT,LVOT-V,LVOT-PG,LVOTVTI,EDT,E/e’ in G+P-group increased significantly,the difference was statistically significant(P < 0.05);2.Compared with the G-P-group,the GLS and GRS in the G+P-group were significantly reduced,and the TCS-diff,TLS-diff,TRS-diff,TLS-SD,and TRS-SD were significantly increased.The differences were statistically significant(P < 0.05);3.The results of ROC curve analysis showed that among many echocardiographic parameters,GLS had the largest area under the curve(0.824),followed by LVMWT(0.804).When the GLS cut-off value was-22.09,the sensitivity of diagnosing G+P-was 0.852,and the specificity was 0.680.Conclusion 1.3D-STI can indicate the reduction of global left ventricular systolic function and the incoordination of peak systolic time in FHCM mutation gene carriers before left ventricular hypertrophy and LVEF are still within the normal range.2.The results of ROC curve analysis showed that GLS is the parameter with the highest diagnostic efficiency in the early identification of G+P-among many echocardiographic parameters,and it can be applied to the early identification of FHCM mutation gene carriers.Objective Explore the use of machine learning methods to build a parametric prediction model based on 2D-TTE and 3D-STI related parameters for predicting FHCM mutation gene carriers,providing a new diagnostic method for early identification of G+P-patients.Method The echocardiographic parameters and genetic testing results of 129 first-degree relatives of FHCM patients were retrospectively analyzed.Using the genetic test results as the gold standard,the subjects were divided into a training set(n=90)and a test set(n=39)at a ratio of 7:3.The Lasso algorithm was used to screen the features on the training set,and based on the selected features,three models of Lasso,random forest and support vector machine for early prediction of G+P-were constructed.ROC curves were used to evaluate the diagnostic performance of the three models in the training set and test set,and model calibration and decision curve analysis were performed on the optimal model.The characteristics of Lasso screening were analyzed by multivariate binary Logistic regression to obtain independent risk factors for predicting G+P-,and a nomogram for individualized prediction of G+P-was constructed.Result 1.9 key features were selected through the Lasso algorithm to build three machine learning models for early prediction of G+P-,and the three models were compared using ROC curve analysis and Delong test.The Lasso model was the optimal model,and it was in the training set AUC value,accuracy,sensitivity and specificity were: 0.932(95%CI: 0.883 ~ 0.982),0.856,0.895 and 0.846(critical value-0.414);AUC value,accuracy,sensitivity and specificity in the test set were: 0.948(95%CI: 0.886-1.000),0.897,0.812 and 0.957.2.The results of the calibration curve showed that the risk of G+P-predicted by the Lasso model was basically consistent with the actual risk(training set P=0.965,test set P=0.919);the clinical decision curve shows that the use of this model can bring benefits to patients.3.The results of multi-factor binary Logistic regression showed that among the 9 variable characteristics screened by Lasso,GLS and LVOT-VTI can be used as independent risk factors for predicting G+P-.The Cindices of the nomogram drawn based on independent risk factors in the training set and test set were 0.885(95%CI: 0.816-0.954)and 0.878(95%CI: 0.764-0.992),respectively.Through the Delong test,the AUC value of the nomogram compared with the Lasso model in the training set(P=0.278)and the test set(P=0.290)has decreased,but there is no statistical difference,which can be used as a simpler method for individualization predict G+P-.Conclusion 1.Among the three machine learning models for early identification of FHCM mutation gene carriers,the Lasso model has good discrimination,calibration,and clinical benefits,and is the optimal model for early identification of G+P-.2.Among the 9 key parameter characteristics of Lasso screening,GLS and LVOT-VTI can be used as independent risk factors for early identification of G+P-.3.The nomogram established based on independent risk factors can be used as a simpler method for early accurate identification of FHCM mutation gene carriers by echocardiography.