基于血管内超声(IVUS)的人体冠状动脉随心动周期形变的流固耦合数值模拟与分析

人体动脉血管内粥样斑的受损和破裂与血管内的受力等力学情况密切相关.基于各种医学成像数据建立的血管数值模拟模型能很好的考察血管内的受力及血流情况,已成为对血管内粥样斑的受损和破裂作出评估和预测的有力工具.本文利用一位患者的血管内超声(IVUS)图像数据建立了其冠状动脉血管的三维流固耦合数值模拟模型,并根据患者CT造影的动态影像数据模拟了冠状动脉在人体内随心脏搏动而产生的周期性运动过程.模型中冠状动脉血管采用各向异性Mooney-Rivlin材料模型,其材料参数是由冠状动脉血管样本的双轴加载拉伸实验获得的拉伸比率———应力数据拟合得到的,血压数据也采用了患者自身的数据,因而本文建立的数值模型真实模拟了人体内冠状动脉的实际运动情况.本文给出了血管内的应力应变分布、血流速度、血液最大剪切应力等数值模拟结果,并比较了血管的周期弯曲和血压变化的相位差的影响.结果表明:血管内部的脂肪斑对血管的应力应变分布有显著的影响;就该患者个体而言,血管弯曲程度对血管内部的应力应变分布的影响要强于血压的影响.而相比于血管的弯曲程度,血管内的血流速度和流通量受血压的影响较为显著.本文建立的模拟冠状动脉周期运动的数值方法可以进一步地应用于大量冠状动脉粥样硬化患者的病例研究.随着患者数量的积累,通过分析可以更精确地评估和预测粥样斑的受损性和破裂发生可能性.

Patient-Specific Echo-Based Fluid-Structure Interaction Modeling Study of Blood Flow in the Left Ventricle with Infarction and Hypertension

Understanding cardiac blood flow behaviors is of importance for cardiovascular research and clinical assessment of ventricle functions.Patient-specific Echo-based left ventricle(LV)fluid-structure interaction(FSI)models were introduced to perform ventricle mechanical analysis,investigate flow behaviors,and evaluate the impact of myocardial infarction(MI)and hypertension on blood flow in the LV.Echo image data were acquired from 3 patients with consent obtained:one healthy volunteer(P1),one hypertension patient(P2),and one patient who had an inferior and posterior myocardial infarction(P3).The nonlinear Mooney-Rivlin model was used for ventricle tissue with material parameter values chosen to match echo-measure LV volume data.Using the healthy case as baseline,LV with MI had lower peak flow velocity(30%lower at beginejection)and hypertension LV had higher peak flow velocity(16%higher at begin-filling).The vortex area(defined as the area with vorticity>0)for P3 was 19%smaller than that of P1.The vortex area for P2 was 12%smaller than that of P1.At peak of filling,the maximum flow shear stress(FSS)for P2 and P3 were 390%higher and 63%lower than that of P1,respectively.Meanwhile,LV stress and strain of P2 were 41%and 15%higher than those of P1,respectively.LV stress and strain of P3 were 36%and 42%lower than those of P1,respectively.In conclusion,FSI models could provide both flow and structural stress/strain information which would serve as the base for further cardiovascular investigations related to disease initiation,progression,and treatment strategy selections.Large-scale studies are needed to validate our findings.

Bioprosthetic Valve Size Selection to Optimize Aortic Valve ReplacementSurgical Outcome: A Fluid-Structure Interaction Modeling Study

Aortic valve replacement(AVR)remains a major treatment option for patients with severe aortic valve disease.Clinical outcome of AVR is strongly dependent on implanted prosthetic valve size.Fluid-structure interaction(FSI)aortic root models were constructed to investigate the effect of valve size on hemodynamics of the implanted bioprosthetic valve and optimize the outcome of AVR surgery.FSI models with 4 sizes of bioprosthetic valves(19(No.19),21(No.21),23(No.23)and 25 mm(No.25))were constructed.Left ventricle outflow track flow data from one patient was collected and used as model flow conditions.Anisotropic Mooney–Rivlin models were used to describe mechanical properties of aortic valve leaflets.Blood flow pressure,velocity,systolic valve orifice pressure gradient(SVOPG),systolic cross-valve pressure difference(SCVPD),geometric orifice area,and flow shear stresses from the four valve models were compared.Our results indicated that larger valves led to lower transvalvular pressure gradient,which is linked to better post AVR outcome.Peak SVOPG,mean SCVPD and maximum velocity for Valve No.25 were 48.17%,49.3%,and 44.60%lower than that from Valve No.19,respectively.Geometric orifice area from Valve No.25 was 52.03%higher than that from Valve No.19(1.87 cm2 vs.1.23 cm2).Implantation of larger valves can significantly reduce mean flow shear stress on valve leaflets.Our initial results suggested that larger valve size may lead to improved hemodynamic performance and valve cardiac function post AVR.More patient studies are needed to validate our findings.

Erosion-Induced Inflammation on Coronary Plaque Stress/Strain and Flow Shear Stress Calculations Using OCT-Based FSI Computational Model

Plaque erosion,together with plaque rupture,is a common cause for acute coronary syndrome(ACS).Plaque erosion alone is responsible for about one third of the patients with ACS.Eroded plaque is defined as thrombosed,endothelium-absent and non-ruptured but often-inflamed plaques based on histological findings.Even though there is efficient imaging technologies to detect the eroded plaque in vivo and tailored treatment strategy has also been developed for ACScaused by erosion in clinics,the pathogenesis mechanisms that cause plaque erosion are not fully understood.It is widely postulated that thrombus formation and endothelial apoptosis(the precursors of plaque erosion)have closed association with biomechanical conditions in the coronary vessel.Revealing of the mechanical conditions in the eroded plaque could advance our knowledge in understanding the formation of plaque erosion.To this end,patient-specific OCT-based fluid-structure interaction(FSI)models were developed to investigate the plaque biomechanical conditions and investigate the impact of erosioninduced inflammation on biomechanical conditions.In vivo OCTand Biplane X-ray angiographic data of eroded coronary plaque were acquired from one male patient(age:64). OCT images were segmented manually with external elastic membrane contour and the trailing edge of the lipid-rich necrotic core(lipid)assumed to have positive remodeling ratio 1.1.Locations with luminal surface having direct contact with intraluminal thrombus on OCT images were identified erosion sites.Fusion of OCT and biplane X-ray angiographic data were performed to obtain the 3D coronary geometry.OCT-based FSI models with pre-shrink-stretch process and anisotropic material properties were constructed following previously established procedures.To reflect tissue weakening caused by erosion-induced inflammation,the material stiffness of plaque intima at the erosion site was adjust to one tenth of un-eroded fibrous plaque tissue.Three FSI models were constructed to investigate the impacts of inflammation and lipid component on plaque biomechanics:M1,without erosion(this means plaque intima at the erosion sites were not softened)and without inclusion of lipid component;M2,with erosion but no lipid;M3,with erosion and inclusion of lipid.FSI models were solved by ADINA to obtain the biomechanical conditions at peak blood pressure including plaque wall stress/strain(PWS/PWSn)and flow wall shear stress(WSS).The average values of three biomechanical conditions at the erosion sites and at the fibrous cap overlaying lipid component were calculated from three models for analysis.The results of M1 and M2 were compared to investigate the impact of erosion-induced inflammation on plaque biomechanics.Mean PWS value decreases from 49.98 kPa to 18.83 kPa(62.32%decrease)while Mean PWSn value increases from 0.123 1 to 0.138 4(12%increase)as the material stiffness becomes 10times soft.Comparing M2 and M3 at the cap sites,M3(with inclusion of lipid)will elevates mean PWS and PWSn values by48.59%and 16.09%,respectively.The impacts of erosion and lipid on flow shear stress were limited(<2%).To conclude,erosion-induced inflammation would lead to lower stress distribution but larger strain distribution,while lipid would elevate both stress and strain conditions.This shows the influence of erosion and lipid component has impacts on stress/strain cal-culations which are closely related to plaque assessment.

A Fast-Fractional Flow Reserve Simulation Method in A Patient with Coronary Stenosis Based on Resistance Boundary Conditions

Fractional flow reserve(FFR)is the gold standard to identify individual stenosis causing myocardial ischemia in catheter laboratory.The purpose of this study is to present a fast simulation method to estimate FFR value of a coronary artery,which can evaluate the performance of vascular stenosis,based on resistance boundary conditions.A patient-specific 3-dimensional(3D)model of the left coronary system with intermediate diameter stenosis was reconstructed based on the CTA images.The resistance boundary conditions used to simulate the coronary microcirculation were computed based on anatomical reconstruction of coronary 3D model.This study was performed by coupling the 3D coronary tree model with the lumped parameter model(0D model).The flow rate and pressure of coronary tree were calculated in twenty minutes.In addition,the effect of inlet pressure and myocardial mass on FFRss values has been investigated.The results showed that the effect of myocardial mass was greater than the effect of inlet pressure on FFRss.This FFRss simulation method can quickly and accurately assess the influence of coronary stenosis in aid clinical diagnosis.

Computational Modeling of Human Bicuspid Pulmonary Valve Dynamic Deformation in Patients with Tetralogy of Fallot

Pulmonary valve stenosis(PVS)is one common right ventricular outflow tract obstruction problem in patients with tetralogy of Fallot(TOF).Congenital bicuspid pulmonary valve(BPV)is a condition of valvular stenosis,and the occurrence of congenital BPV is often associated with TOF.Dynamic computational models of normal pulmonary root(PR)with tri-leaflet and PR with BPV in patients with TOF were developed to investigate the effect of geometric structure of BPV on valve stress and strain distributions.The pulmonary root geometry included valvular leaflets,sinuses,interleaflet triangles and annulus.Mechanical properties of pulmonary valve leaflet were obtained from biaxial testing of human PV leaflet,and characterized by an anisotropic Mooney-Rivlin model.The complete cardiac cycle was simulated to observe valve leaflet dynamic stress/strain behaviors.Our results indicated that stress/strain distribution patterns of normal tri-leaflet pulmonary valve(TPV)and the BPV were different on valve leaflets when the valve was fully open,but they were similar when valves were completely closed.When the valve was fully open,the BPV maximum stress value on the leaflets was 197.2 kPa,which was 94.3%higher than of the normal TPV value(101.5 kPa).During the valve was fully open,the stress distribution in the interleaflet triangles region of the PR was asymmetric in the BPV model compared with that in the TPV model.The geometric orifice area value in the completely opened position of BPV model was reduced 55.6%from that of the normal PV.Our initial results demonstrated that valve geometrical variations with BPV may be a potential risk factor linked to occurrence of PVS in patients with TOF.Computational models could be a useful tool in identifying possible linkage between valve disease development and biomechanical factors.Large-scale clinical studies are needed to validate these preliminary findings.

Anisotropic Models of Human Pulmonary Root with Bicuspid Pulmonary Valve in Patients with Tetralogy of Fallot: Pulmonary Root Function Assessment and Mechanical Stress Analysis

Background Tetralogy of Fallot(TOF)is the most common cyanotic heart defect,accounting for 10%of all congenital defects.Pulmonary valve stenosis(PVS)is one common right ventricular outflow tract obstruction problem in patients with TOF.Congenital bicuspid pulmonary valve(BPV)is a condition of valvular stenosis,which morphologic feature is the presence of only two pulmonary leaflets instead of the normal tri-leaflet.Congenitally BPV are uncommon and the occurrence is often associated with TOF.Methods The three-dimensional geometric reconstruction of pulmonary root(PR)were based on well-accepted mathematical analytic models with physiological parameters obtained from a typical sample of the pulmonary root used in clinical surgery.The PR geometry included valvular leaflets,sinuses,interleaflet triangles and annulus.The dynamic computational models of normal PR with tri-leaflet and PR with BPV in patients with TOF were developed to investigate the effect of geometric structure of BPV on valve stress and strain distributions and the geometric orifice area.Mechanical properties of pulmonary valve leaflet were obtained from biaxial testing of human pulmonary valve left leaflet,and characterized by an anisotropic Mooney-Rivlin model.The complete cardiac cycle was simulated to observe valve leaflet dynamic stress and strain behaviors.Results Our results indicated that stress/strain distribution patterns of normal tri-leaflet pulmonary valve(TPV)and the BPV were different on valve leaflets when the valve was fully open,but they were similar when valves were completely closed.When the valve was fully open,the BPV maximum stress value on the leaflets was 218.1 kPa,which was 128.0%higher than of the normal TPV value(95.6 kPa),and BPV maximum strain value on the leaflets was 70.7%higher than of the normal TPV.The location of the maximum stress from TPV and BPV were also different,which were found at the bottom of the valve near the leaflet attachment for TPV and the vicinity of cusp of the fusion of two leaflets for BPV,respectively.During the valve was fully open,the stress distribution in the interleaflet triangles region of the PR was more asymmetric in the BPV model compared with that in the normal TPV model,and the largest change on the PR with the geometrical variations in the two models was 39.6%in maximum stress.This stress asymmetry indicates that BPV may be one of the causes of post-stenotic pulmonary artery dilatation and aneurysm in patients with TOF.The cusp of the BPV model showed significant eccentricity during peak systolic period,and its geometric orifice area value in the completely opened position of valve was reduced 57.5%from that of the normal TPV model.Conclusions Our initial results demonstrated that valve geometrical variations with BPV may be a potential risk factor linked to occurrence of PVS in patients with TOF.Computational models could be used as an effective tool to identifying possible linkage between pulmonary valve malformation disease development and biomechanical factors,better design of artificial valves and new surgical procedures without testing those on patients.Large-scale clinical studies are needed to validate these preliminary findings.

Using Multiple Risk Factors and Generalized Linear Mixed Models with 5-Fold Cross-Validation Strategy for Optimal Carotid Plaque Progression Prediction

Background Cardiovascular diseases are closely linked to atherosclerotic plaque development and rupture.Plaque progression prediction is of fundamental significance to cardiovascular research and disease diagnosis,prevention,and treatment.Generalized linear mixed models(GLMM)is an extension of linear model for categorical responses while considering the correlation among observations.Methods Magnetic resonance image(MRI)data of carotid atheroscleroticplaques were acquired from 20 patients with consent obtained and 3D thin-layer models were constructed to calculate plaque stress and strain for plaque progression prediction.Data for ten morphological and biomechanical risk factors included wall thickness(WT),lipid percent(LP),minimum cap thickness(MinCT),plaque area(PA),plaque burden(PB),lumen area(LA),maximum plaque wall stress(MPWS),maximum plaque wall strain(MPWSn),average plaque wall stress(APWS),and average plaque wall strain(APWSn)were extracted from all slices for analysis.Wall thickness increase(WTI),plaque burden increase(PBI)and plaque area increase(PAI) were chosen as three measures for plaque progression.Generalized linear mixed models(GLMM)with 5-fold cross-validation strategy were used to calculate prediction accuracy for each predictor and identify optimal predictor with the highest prediction accuracy defined as sum of sensitivity and specificity.All 201 MRI slices were randomly divided into 4 training subgroups and 1 verification subgroup.The training subgroups were used for model fitting,and the verification subgroup was used to estimate the model.All combinations(total1023)of 10 risk factors were feed to GLMM and the prediction accuracy of each predictor were selected from the point on the ROC(receiver operating characteristic)curve with the highest sum of specificity and sensitivity.Results LA was the best single predictor for PBI with the highest prediction accuracy(1.360 1),and the area under of the ROC curve(AUC)is0.654 0,followed by APWSn(1.336 3)with AUC=0.6342.The optimal predictor among all possible combinations for PBI was the combination of LA,PA,LP,WT,MPWS and MPWSn with prediction accuracy=1.414 6(AUC=0.715 8).LA was once again the best single predictor for PAI with the highest prediction accuracy(1.184 6)with AUC=0.606 4,followed by MPWSn(1. 183 2)with AUC=0.6084.The combination of PA,PB,WT,MPWS,MPWSn and APWSn gave the best prediction accuracy(1.302 5)for PAI,and the AUC value is 0.6657.PA was the best single predictor for WTI with highest prediction accuracy(1.288 7)with AUC=0.641 5,followed by WT(1.254 0),with AUC=0.6097.The combination of PA,PB,WT,LP,MinCT,MPWS and MPWS was the best predictor for WTI with prediction accuracy as 1.314 0,with AUC=0.6552.This indicated that PBI was a more predictable measure than WTI and PAI. The combinational predictors improved prediction accuracy by 9.95%,4.01%and 1.96%over the best single predictors for PAI,PBI and WTI(AUC values improved by9.78%,9.45%,and 2.14%),respectively.Conclusions The use of GLMM with 5-fold cross-validation strategy combining both morphological and biomechanical risk factors could potentially improve the accuracy of carotid plaque progression prediction.This study suggests that a linear combination of multiple predictors can provide potential improvement to existing plaque assessment schemes.

Angle of Attack between Blood Flow and Mitral Valve Leaflets in Hypertrophic Obstructive Cardiomyopathy: An In Vivo Multipatient CT-based FSI Study

The mechanisms of systolic anterior motion(SAM)of the mitral valve in hypertrophic obstructive cardiomyopathy(HOCM)remain unclear.To investigate the angle of attack between blood flow and mitral valve leaflets at pre-SAM time point,patient-specific CT-based computational models were constructed for 5 patients receiving septal myectomy surgery to obtain pre-and post-operative 2D vector flow mapping.The comparisons between pre-and post-operative angles of attack based on 2D vector flow mapping of 5 patients were performed.It was found that there was no statistically significant difference between pre-and post-operative angles of attack(61.1±t wa o vs.56.2±56.o,p=0.306,n=5).Therefore,we propose that the angle of attack might not play an important role in the initiation of SAM.

Improving Right Ventricle Cardiac Function for Repaired Tetralogy of Fallot Patient with Contracting Bands: A Modelling Study

Objective Patients with repaired tetralogy of Fallot(rTOF)account for the majority of cases with late onset right ventricle(RV)failure.The current surgical approach,including pulmonary valve replacement/insertion(PVR),has yielded mixed results with some patients recover RV function and some do not.An innovative surgical approach was proposed to help ventricle to contract and improve RV function qualified by ejection fraction with one or more active contracting bands.Computational biomechanical modelling is a widely used method in cardiovascular study for investigation of mechanisms governing disease development,quantitative diagnostic and treatment strategies and improving surgical designs for better outcome.Muscle active contraction caused by zero-load sarcomere shortening leads to change of zero-load configurations.In lieu of experimenting using real surgery on animal or human,computational simulations(virtual surgery)were performed to test different band combination and insertion options to identify optimal surgery design and band insertion plan.Methods Cardiac magnetic resonance(CMR)data were obtained from one rTOF patient(sex:male,age:22.5 y)before pulmonary valve replacement surgery.The patient was suffering from RV dilation and dysfunction with RV end-systole volume 254.49ml and end-diastole volume 406.91 mL.A total of 15 computational RV/LV/Patch/Band combination models based on(CMR)imaging were constructed to investigate the influence of different band insertion surgery plans.These models included 5 different band insertion models combined and 3 different band contraction ratio(10%,15%and 20%band zero-stress length reduction).These models included 5 different band insertion models:Model 1 with one band at anterior to the middle of papillary muscle;Model 2 with one band at posterior to the middle of papillary muscle;Model 3 with 2 bands which are the ones from Models 1&2 combined;Model 4 with a band at the base of the papillary muscle;Model 5 with 3 bands which is a combination of Models 3&4.A pre-shrink process was performed on in-vivo begin-filling and end-systole MRI data to obtain diastole and systole zero4oad ventricle geometries.An extra 5%-8%shrinkage was applied to obtain corresponding systole zero-load geometry reflecting myocardium sarcomere shortening.The zero-load band length in systole was 10%,15%and 20%shorter than that in diastole according to their corresponding contraction ratio.The nonlinear Mooney-Rivlin model was used to describe the ventricle material properties with their material parameter values adjusted to match measured data with CMR.The band material properties were in the same scale with healthy right ventricle.The RV/LV/Band model construction and solution procedures were the same as described.Results Model 5 with band contraction ratio of 20%has the ability to improve RV ejection fraction to 41.07%,which represented a 3.61%absolute improvement,or 9.6%relative improvement using pre-PVR ejection fraction as the baseline number.The ejection fractions for Models 1-4 with band contraction ratio of 20%were 39.28%,39.47%,38.87%and 40.34%respectively.Compared to models with band contraction ratio15%and 20%,models with band contraction ratio 10%has the least ability on RV ejection fraction improvement with ejection fraction 38.28%,38.00%,38.81%,38.50%and 39.36%corresponding to Models 1-5.Conclusions This pilot work demonstrated that the band insertion surgery may have great potential to improve post-PVR RV cardiac function for patients with repaired TOF.More band contraction ratio and inserted band number may lead to better post-surgery outcome.Further investigations using in-vitro animal experiments and final patient studies are warranted.

Predicting Coronary Plaque Morphology Changes Based on Multimodality FSI Models Using Follow-Up IVUS and OCT Data

Background Current 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 intravascular ultrasound(IVUS)images is 150-200 microns,which is not enough to identify vulnerable plaques with thin caps and construct accurate biomechanical plaque models.Optical coherence tomography(OCT)with a 15-20μm resolution has the capacity to identify thin fibrous cap.IVUS and OCT images could complement each other and provide for more accurate plaque morphology,especially,fibrous cap thickness measurements.A modeling approach combining IVUS and OCT was introduced in our previous publication for cap thickness quantification and more accurate cap stress/strain calculations.In this paper,patient baseline and follow-up IVUS and OCT data were acquired and multimodality image-based Fluidstructure interaction(FSI)models combining 3D IVUS,OCT,angiography were constructed to better quantify human coronary atherosclerotic plaque morphology and plaque stress/strain conditions and investigate the relationship of plaque vulnerability and morphological and mechanical factors.Methods Baseline and 10-Month follow-up in vivo IVUS and OCT coronary plaque data were acquired from one patient with informed consent obtained.Co-registration and segmentation of baseline and follow-up IVUS and OCT images were performed for modeling use.Baseline and follow-up 3D FSI models based on IVUS and OCT were constructed to simulate the mechanical factors which integrating plaque morphology were employed to predict plaque vulnerability.These 3D models were solved by ADINA(ADINA R&D,Watertown,MA,USA).The quantitative indices of cap thickness,lipid percentage were classified according to histological literatures and denoted as Cap Index and Lipid Index.Cap Index,Lipid Index and Morphological Plaque Vulnerability Index(MPVI)were chosen to quantify plaque vulnerability,respectively.Random forest(RF)which was based 13 extracted features including morphological and mechanical factors was used for plaque vulnerability classification and prediction.Over sampling scheme and a 5-fold crossvalidation procedure was employed in all 45 slices for training and testing sets.Single and all different combinations of morphological and mechanical risk factors were used for plaque progression prediction.Results When Cap Index was used as the measurement,minimum cap thickness(MCT)was the best single predictor which area under curve(AUC)is 0.782 0;the combination of MCT,critical plaque wall strain(CPWSn),critical wall shear stress(CWSS)and cap wall shear stress(CapWSS)was the best predictor with ACU=0.868 6.When Lipid Index was used as the measurement,the lipid percentage(LP)was the best single predictor which AUC value is 0.857 8;the combination of Mean cap thickness(MeanCT),LP,CWSS and cap plaque wall stress(CapPWS)and was the best predictor with ACU=0.9821.When MPVI was used as the measurement,MCT was the best single predictor which AUC value is 0.782 9;the combination of MCT,LP,plaque area(PA),CPWSn and CapWSS was the best predictor with ACU=0.872 9.Conclusions Combinations of morphological and mechanical risk factors had higher prediction accuracy,compared to the prediction of single factors and other combination of morphological factors.

Automatic Segmentation for Intracoronary OCT Image Based on Convolutional Neural Network and Support Vector Machine Methods

Background Cardiovascular diseases are closely associated with atherosclerotic plaque development and rupture.Traditional medical imaging techniques such as magnetic resonance imaging(MRI)and intravascular ultrasound(IVUS)were unable to identify vulnerable plaques due to their limited resolution.Fortunately,optical coherence tomography(OCT)is an advanced intravascular imaging technique developed in recent years which has high resolution approximately 10 microns and could provide more accurate morphology of coronary plaque.In particular,it has the ability to identify plaques with fibrous cap thickness<65μm,an accepted threshold value for vulnerable plaques.However,segmentation of OCT images in clinic is still mainly performed manually by physicians which is time consuming and subjective.To overcome time consumption,several methodologies have been proposed for automatic segmentation of OCT images but most of these methods were still limited by intricate image preprocessing and expensive computation.In this research,two automatic segmentation methods for intracoronary OCT image based on support vector machine(SVM)and convolutional neural network(CNN)were performed to identify the plaque region and characterize plaque components.Methods In vivo IVUS and OCT coronary plaque data from 5 patients were acquired at Emory University with patient’s consent obtained.OCT were obtained from ILUMIEN OPTIS System(St.Jude,Minnesota,MN).The OCT catheter was traversed to the segment of interest and the catheter pullback was limited at a rate of 20 mm/sec.Following the OCT image acquisition,the IVUS catheter was traversed distally though the artery to the same coronary segment(Volcano Therapeutics,Rancho Cordova)and the catheter pullback speed was at a standard rate of 0.5 mm/sec.Seventy-seven matched IVUS and OCT slices with good image quality and lipid cores were selected for our segmentation study.Manual OCT segmentation was performed by experts and used as gold standard in the automatic segmentations.VH-IVUS was used as references and guide by the experts in the manual segmentation process.Three plaque component tissue classes were identified from OCT images in this work:lipid tissue(LT),fibrous tissue(FT)and background(BG).Procedures using two machine learning methods(CNN and SVM)were developed to segment OCT images,respectively.For CNN method,the U-Net architecture was selected due to its good performance in very different biomedical segmentation and very few annotated images.For SVM method,local binary patterns(LBPs),gray level co-occurrence matrices(GLCMs)which contains contrast,correlation,energy and homogeneity,entropy and mean value were calculated as features and assembled to feed SVM classifier.The accuracies of two segmentation methods were evaluated and compared using the OCT dataset.Segmentation accuracy is defined as the ratio of the number of pixels correctly classified over the total number of pixels.Results The overall classification accuracy based CNN method reached 95.8%,and the accuracies for LT,FT and BG were 86.8%,83.4%,and 98.2%,respectively.The overall classification accuracy based SVM was 71.9%,and per-class accuracy for LT,FT and BG was 75.4%,78.3%,and67.0%,respectively.Conclusions The two methods proposed can automatically identify plaque region and characterize plaque compositions for OCT images and potentially reduce the time spent by doctors in segmenting and evaluating coronary plaque OCT images.CNN provided better segmentation accuracies compared to those achieved by SVM.

Image-Based Modeling for Atherosclerotic Coronary Plaque Progression and Vulnerability Research

China Cardiovascular Disease Report 2017(Summary)pointed out that at present,cardiovascular diseases(CVD)account for the highest number of deaths among urban and rural residents.In the middle or later stages of atherosclerosis,the plaques become increasingly unstable with high chance to rupture,which may lead acute death from coronary heart diseases.Medical imaging and image-based computational modeling have been used in recent years to quantify ather-osclerotic plaque morphological and biomechanical characteristics and predict the coronary plaque growth and rupture processes.Analyzing the vulnerability of plaques effectively could lead to better patient screening strategies and enable physicians to adopt timely and necessary intervention or conservative treatment.Earlier investigations of vulnerable plaques were mostly based on histopathological data.With the accumulation of experience in pathology and the gradual enrichment of autopsy materials,the criteria for the diagnosis of vulnerable plaques appeared in 2001,mainly manifested as the necrotic lipid nuclei,fibrous caps that are infiltrated by a large number of macrophages,and fibrous cap thickness less than 65μm.Because of the obvious importance of the thin fibrous cap in the study of plaque vulnerability,it has been a focus of attention by many investigations.Watson,M.G.et al.are concerned about the formation of early fibrous caps in recent years.The presentation of local maximum stress on plaque further confirmed the importance of thin fibrous cap.The development of medical images has greatly promoted the study of coronary atherosclerosis.Compared with autopsy ex vivo,medical image could provide plaque data under in vivo conditions and greatly promote the study of coronary atherosclerosis.Huang XY et al.used ex vivo magnetic resonance imaging(MRI)to study the relationship between plaque wall stress(PWS)and death caused by coronary artery disease.Due to technical limitations and the accessibility of the coronary artery in the body,MRI is not widely used for in vivo coronary studies.Interventional intravascular ultrasound(IVUS),with an image resolution of 150-200μm,has been used in research and clinical practice to identify plaques,quantify plaque morphology,and characterize plaque components.More recently,optical coherence tomography(OCT),with its resolution of 5-10μm,has emerged as an imaging modality which can be used to detect thin fibrous caps and improve diagnostic accuracy.It is commonly believed that mechanical forces play an important role in plaque progression and rupture.Image-based biomechanical plaque models have been developed and used to quantify plaque mechanical conditions and seek their linkage to plaque progression and vulnerability development activities.Based on recent advances in imaging and modeling,this paper attempts to provide a brief review on plaque research,including histological classification,image preparation,biomechanical modeling and analysis methods including medical imaging techniques represented by intravascular ultrasound(IVUS)and optical coherence tomography(OCT),computational modeling and their applications in plaque progression and vulnerability analyses and predictions.The clinical application and future development direction are also briefly described.We focus more on human coronary plaque modeling and mainly included results from our group for illustration purpose.We apologize in advance for our limitations.

Machine Learning Model Comparison for Automatic Segmentation of Intracoronary Optical Coherence Tomography and Plaque Cap Thickness Quantification

Optical coherence tomography(OCT)is a new intravascular imaging technique with high resolution and could provide accurate morphological information for plaques in coronary arteries.However,its segmentation is still commonly performed manually by experts which is time-consuming.The aim of this study was to develop automatic techniques to characterize plaque components and quantify plaque cap thickness using 3 machine learning methods including convolutional neural network(CNN)with U-Net architecture,CNN with Fully convolutional DenseNet(FC-DenseNet)architecture and support vector machine(SVM).In vivo OCT and intravascular ultrasound(IVUS)images were acquired from two patients at Emory University with informed consent obtained.Eighteen OCT image slices which included lipid core and with acceptable image quality were selected for our study.Manual segmentation from imaging experts was used as the gold standard for model training and validation.Since OCT has limited penetration,virtual histology IVUS was combined with OCT data to improve reliability.A 3-fold cross-validation method was used for model training and validation.The overall tissue classification accuracy for the 18 slices studied(total classification database sample size was 8580096 pixels)was 96.36%and 92.72%for U-Net and FC-DenseNet,respectively.The best average prediction accuracy for lipid was 91.29%based on SVM,compared to 82.84%and 78.91%from U-Net and FC-DenseNet,respectively.The overall average accuracy(Acc)differentiating lipid and fibrous tissue were 95.58%,92.33%and 81.84%for U-Net,FC-DenseNet and SVM,respectively.The average errors of U-Net,FC-DenseNet and SVM from the 18 slices for cap thickness quantification were 8.83%,10.71%and 15.85%.The average relative errors of minimum cap thickness from 18 slices of U-Net,FC-DenseNet and SVM were 17.46%,13.06%and 22.20%,respectively.To conclude,CNN-based segmentation methods can better characterize plaque compositions and quantify plaque cap thickness on OCT images and are more likely to be used in the clinical arena.Large-scale studies are needed to further develop the methods and validate our findings.

Preface:Innovations and Current Trends in Computational Cardiovascular Modeling and Beyond:Molecular,Cellular,Tissue and Organ Biomechanics with Clinical Applications

Innovative computational modeling and methods have been developed and widely used in cardiovascular research and beyond.In this special issue,leading experts in their respective areas were invited to present their novel computational models,numerical methods with new and ground-breaking biomedical and clinical applications in ventrical valve mechanics,arteries,aneurysm,medical devices and treatment techniques,as well as microscale studies at cell and molecule levels,and others.Thirteen papers were selected to form this special issue which has 3 parts:Part I included 5 cardiovascular modeling papers.Part II has 4 papers where numerical methods were developed to design,analyze,and improve various medical devices.Part III selected 4 papers with wide applications including eye,ear,endocytosis,and liver.A brief introduction on each paper is given below.