Application Of CCTA-based Coronary Morphological Parameters And LASSO Algorithm In Coronary Physiology Evaluation

Background and Aims:Coronary ischemia can be detected based on the ratio of the coronary morphological parameter lesion length(LL)to the minimum lumen diameter(MLD)to the fourth power(LL/MLD~4)on coronary computed tomography angiography(CCTA).However,its diagnostic performance still has room for improvement,and the difference in diagnostic performance of coronary ischemia between LL/MLD~4 and CCTA-derived fractional flow reserve(CT-FFR)remains unclear.The inflow angle(α),a key parameter for evaluating blood flow energy loss in coronary stenosis,was significantly different in patients with fractional flow reserve(FFR)≤0.80 and>0.80.In addition,there is currently no clinical risk prediction model based on the least absolute shrinkage and selection operator(LASSO)algorithm for coronary functional assessment in patients.Therefore,the main purpose of this study were(1)to develop a new coronary morphological parameter,the product ofαand LL/MLD~4(α×LL/MLD~4),in order to improve the diagnostic performance of morphological parameters for coronary ischemia,and to explore whether there are differences in the performance ofα×LL/MLD~4,LL/MLD~4and CT-FFR in diagnosing coronary ischemia.(2)In order to further improve the diagnostic performance of morphological parameters for coronary functional ischemia,combining patient characteristic parameters,coronary morphological parameters and plaque characteristic parameters to generate a high-dimensional data set,and use the LASSO intelligent parameter selection algorithm to screen variables in this data set,and then use the screened variables to construct a risk predictive model to detect coronary ischemia.Methods:The patients with suspected or known coronary atherosclerotic heart disease(CAD)who underwent CCTA and invasive coronary angiography(ICA)and FFR at the National Heart Centre of Singapore and the National University Hospital of Singapore from September 2016 to March 2020 were retrospectively included,and the date of CCTA and ICA was not more than 6 months.Demographic characteristics data,clinical characteristics data,FFR measurement images and CCTA original image data of the enrolled patients were also collected.Coronary 3D reconstruction and hydrodynamic simulation calculations were performed using the Ansys software suite(Ansys Space Claim,Ansys Work Bench,Ansys Fluent,and CFD-Post)(ANSYS Corporation,USA),and then the position of the patient’s FFR measurement was reviewed through the images at the time of FFR measurement,then use the constructed coronary 3D model that can display CT-FFR values at any position to obtain CT-FFR results at the same position.Taking FFR as the gold standard,the differences in diagnostic performance of coronary ischemia(FFR≤0.80)betweenα×LL/MLD~4,LL/MLD~4 and CT-FFR were compared by the area under the receiver operating characteristic(ROC)curve(AUC).Use R software(version number:4.2.0,Lucent,USA)to call the LASSO algorithm in the"glmnet"package,screen out the variables related to coronary functional ischemia through tenfold cross validation,and use the screened variables to build a risk prediction model,and then the ROC curve is used to analyze the performance of the model.And,the diagnostic performance of this model will be compared with CT-FFR,α×LL/MLD~4 and diameter stenosis percentage(%DS).The‘resourceselection’package is used to test the goodness of fit of the model,and the bootstrap method is used to internally verify the model.In order to further evaluate the performance of the model,the‘rms’package is used to draw the decision curve analysis(DCA)and clinical impact curve of LASSO model,and the DCA result of this model will be compared with CT-FFR,α×LL/MLD~4 and%DS.Finally,a nomogram is drawn to visualize the model.Results:Finally,a total of 133 patients and 210 vessels were included,of which75 patients(56.4%)had FFR≤0.80,58 patients(43.6%)had FFR>0.80,and 89(42.4%)coronary arteries with FFR≤0.80,121(57.6%)coronary arteries with FFR>0.80.At the vessel level,α×LL/MLD~4 improved the diagnostic performance of coronary ischemia by 9.2%compared with LL/MLD~4(AUC:0.932 vs.0.840,p<0.001).Compared with CT-FFR,the diagnostic performance ofα×LL/MLD~4improved by 3.0%,but there was no statistical difference(AUC:0.932 vs.0.902,p=0.238),and there was also no significant difference between LL/MLD~4 and CT-FFR(AUC:0.840 vs.0.902,p=0.061).The patient-level comparisons amongα×LL/MLD~4,LL/MLD~4,and CT-FFR were similar to the results of vessel level.A total of 10 variables were determined to be related to coronary ischemia,they wereα,LL,MLD,%DS≥50%,positive re-modelling index,plaque burden,Agatston score of single coronary artery,LAP volume of lesion,LAP area of obstruction and LL/MLD~4.The results of the LASSO regression model constructed with these 10variables showed that the AUC value of the model reached 0.966(95%CI:0.932-0.986).Compared with%DS≥50%based on CCTA,the diagnostic performance of LASSO model was improved by 34.2%(AUC:0.966 vs.0.624,p<0.0001).Compared with CT-FFR,the diagnostic performance was improved by 6.4%(AUC:0.966 vs.0.902,p=0.002).Compared withα×LL/MLD~4,there is a 3.4%improvement in diagnostic performance(AUC:0.966 vs.0.932,p=0.01).The results of DCA showed that the net benefit of the LASSO regression model for clinical decision-making guided by all threshold probabilities was higher than that of CT-FFR,α×LL/MLD~4 and%DS≥50%.The results of goodness-of-fit test and calibration curve indicated that the LASSO model had stable and reliable predictive ability of coronary ischemia.Conclusions:Compared with the traditional coronary morphology parameter LL/MLD~4,the new coronary morphology parameterα×LL/MLD~4 developed by us for the first time not only achieved the diagnostic performance of coronary functional ischemia not inferior to CT-FFR,but also further improved the diagnostic ability of functional ischemia of simple coronary morphology parameters.In addition,the risk prediction model built by using LASSO intelligent parameter selection algorithm further improves the diagnostic ability of morphological parameters for coronary functional ischemia,which can better help patients suspected or known of CAD understand their coronary physiological function.