IVUS And OCT-Based Patient-Specific Coronary Plaque Fluid-Structure-Interaction Models For Plaque Progression And Vulnerability Prediction

Heart attacks are usually associated with the growth and sudden rupture of atherosclerotic coronary plaques.Effective prediction of plaque growth and rupture is essential for early diagnosis,prevention and treatment of cardiovascular diseases.Currently,using computational methods to construct patient-specific coronary plaque models and perform morphological and biomechanical analyses is an important method for vulnerable plaque research.However,in vivo medical image resolution of coronary plaques seriously affects the accuracy of plaque stress/strain calculation and the reliability of closely related studies on plaque growth and vulnerability.This paper introduced an innovative three-dimensional fluid-structure interaction(FSI)mathematical modeling approach by combining intravascular ultrasound(IVUS)and optical coherence tomography(OCT)to get accurate plaque morphology and stress/strain calculations.This provided a possibility for more accurate prediction of plaque growth and vulnerability.The thesis included the follow 5 parts:The bottleneck of patient-specific coronary plaque model construction is the resolution of in vivo medical imaging.The threshold of cap thickness of vulnerable coronary plaques is 65 microns,while the resolution of in vivo coronary IVUS images used in IVUS-based models is 150-200 microns,which is not enough to identify vulnerable plaques and construct accurate biomechanical plaque models.Combining with angiography,OCT and IVUS image data and in vivo material properties,a new and more accurate coronary fluid-structure interaction(FSI)model with cycle bending was introduced in this paper.High resolution OCT(5-10 μm)and IVUS with better penetration depth were merged together to obtain accurate plaque morphology.IVUS+OCT method provided a mechanical model for accurate thin fibrous cap thickness quantification,stress/strain calculation,and laid down a more accurate and reliable foundation for further vulnerable plaque studies.OCT image segmentation is a hot spot and challenge in medical image research.In this paper,least squares support vector machine(LS-SVM)and deep learning methods were used to segment OCT images,extract contours and determine cap thickness,respectively,which could provide accurate morphological information for modeling.Preliminary results showed that the method based on deep learning gave better the prediction accuracy(accuracy=96%),and the error of the average fiber cap thickness obtained by the segmentation based on deep learning is only 3.8%.Material properties are one of the important factors affecting model calculation results.A new method was proposed in this paper for quantification of in vivo material properties based on thin slice model by using Cine IVUS and virtual histological IVUS(VH-IVUS)data.The method can determine patient-specific on-site in vivo vessel material properties.Preliminary results showed that the young’s modulus measured by ex vivo samples in the existing literature is 470% of the young’s modulus of in vivo coronary.The difference in material properties could lead to an error of 40-180% in plaque stress and strain calculations.Patient-specific coronary plaque models based on IVUS,OCT and angiographic follow-up data were constructed and mechanical and morphological risk factor values were obtained for plaque progression analysis.The generalized linear mixed model(GLMM)and LS-SVM methods were used to predict plaques progression.The results showed that the combination of morphological and mechanical factors was beneficial to the prediction of plaque progression(the accuracy was about 90%).Compared with the prediction using IVUS-based models(accuracy at 70%),IVUS+OCT model had more accurate data and has the potential to improve the accuracy of plaque progression prediction.Available follow-up studies of patients with coronary atherosclerosis showed that the incidence of clinically significant events within two years was 2-5%.It is obvious that,for small size research projects,it is almost impossible to obtain enough plaque rupture or major clinical cases to serve as the gold standard for predicting plaque vulnerability by tracking patients.Using fibrous cap,lipid core and plaque morphology from the accurate IVUS+OCT data,three IVUS+OCT-based morphological plaque vulnerability indices were introduced and their values at follow-up were used as goldstandard for vulnerability prediction.Using IVUS+OCT morphology and stress/strain baseline data,5 machine-learning-based predictive methods were used to predict plaque vulnerability changes.Preliminary results indicated that LS-SVM method had the best accuracy for predicting lipid index change,with accuracy at 95%.The prediction accuracy for fibrous cap thickness increase using random forest method was 80%,while the prediction accuracy for morphological plaque vulnerability index change using Discriminant Analysis classifier was 74%。Innovations of this paper included: combining IVUS and OCT to make plaque models;using machine-learning-based method for OCT segmentation;using Cine-IVUS to determine in vivo coronary vessel material properties,and using machinelearning-based method to predict plaque progression and vulnerability change.Data used in this paper were provided by Cardiovascular Research Foundation at Columbia University,Emory University,and Southeast University Affiliated Zhongda Hospital with written consent obtained from patients.