The Study Based On New Three-dimensional Quantitative Analysis Technique To Improve The Diagnostic Accuracy Of Coronary CTA

Background:Coronary heart disease with the high morbidity and mortality threatens public health.The magnetic resonance imaging as well as coronary CTA made it possible for the early diagnosis.However,coronary angiography remains the golden standard for evaluation of coronary artery lesions stenosis.Because of its complex operation and high cost,coronary angiography has difficulty in detecting the early coronary changes.Accurate evaluation has a significant effect on the further treatment management.Coronary CTA not only can visualize the coronary artery anatomy quickly and accurately,but also provides information on the plaque characteristics.Compared with other methods,Coronary CTA is considered as a"gatekeeper" with high diagnostic sensitivity and specificity before Cath lab.It has become an important method in clinical practice.The coronary artery is a three-dimensional structure,evaluating from a 2D image can directly result in overestimation or underestimation.Moreover,more overestimation of stenosis can be ascribed to the limitations of CT technique and the volume effect of calcified plaques,which in turn triggers subsequent unnecessary invasive coronary angiography and economic burdens on patients.Therefore,accurate evaluation of lesions is beneficial to the formulation of treatment strategies and improve the prognosis of patients.Three-dimensional quantitative coronary analysis based on coronary CT plays an increasingly important role in our clinical work.Objective:To elaborate the algorithm theory and the methods of technique implementation and explore 3DctQCA which is collaborated with the School of Life Science and Technology of Xidian University college in the lesions so as to improve the diagnostic accuracy of coronary stenosis degree.Methods:1.Coronary CT data of 44 patients are selected for theoretical learning of the algorithm.Blooming artifacts of calcified plaque are removed based on binary deconvolution.After the segmentation of CT angiography images is performed using deep learning method,the refinement algorithm of Zhang-Suen extracts the centerline of the segmented blood vessel.Finally the proximal and distal of the diseased vessel segment are selected to complete the 3DctQCA quantitative calculation.2.We retrospectively included patients suspected with coronary artery undergoing coronary computed tomography angiography and Coronary artery angiography within one week at the PLA Hospital from 2016-10-01 to 2017-01-31 patients with suspected coronary heart diseases.Totally 103 patients and 117 coronary target lesions were analyzed in our study.3According to Agatston calcification scores,we perform angiography QCA and 3DctQCA to analyze the lesion if the target with calcification score ?10 in the selected patients.Blooming artifacts s correction were performed on calcified plaques with calcification score>10 in the 2D images.After correction,3DctQCA was performed and compared with QCA results.Results:1.We have finished algorithm research and then developed 3D quantitative coronary analysis technique based on coronary CTA which can improve the accuracy of coronary artery segmentation and calculation2.For non-calcified and slightly calcified plaques,the diagnostic threshold of ROC curves to detect coronary heart disease was the coronary stenosis>45.68%(sensitivity and specificity were 86.89%and 92.86%,positive predictive value and negative predictive value was 96.4%and 76.5%,the area under the AUC curve was 0.944,95%CI 0.874-0.981 P<0.001).There is an excellent correlation between diagnosis of stenosis quantified with MedisQCA and 3DctQCA of the same non-calcified target lesions(R=0.785,P<0.001).For calcification score>10 calcified plaques,,the diagnostic threshold of ROC curves to detect coronary heart disease was the coronary stenosis?50.19,(sensitivity and specificity were90.00%and 77.78%,positive predictive value and negative predictive value was 69.21%and 93.37%,the area under the AUC curve was 0.894,95%Cl 0.720-0.978 P<0.001).There is an mild correlation between diagnosis of stenosis quantified with QCA and 3DctQCA of the same non-calcified target lesions(R=0.667,P<0.001).3.A weighted Kappa of 0.635 was found between the 2 measurements.It's demonstrated that the 3DctQCA is mild in line with QCA for the diagnosis of the same coronary stenosis.5.Bland-Altman analysis showed the concordance of measurements was satisfactory.Conclusions:1.Compared with the golden standard software,three-dimensional coronary artery quantitative analysis technique can improve the diagnostic accuracy of CTA in evaluating coronary artery stenosis.2.Based on the method of binary deconvolution to remove the blooming artifacts of the calcified plaque,the accurate evaluation of CTA can be improved and a proportion of further CAG may be cancelled.