Computation of FFR CT from Cardiac CT: Does it improve diagnostic accuracy? Jonathon Leipsic MD FRCPC FSCCT Assistant Professor of Radiology and Cardiology Vice Chairman of Research Dept Radiology UBC President Canadian Chapter Society of Cardiovascular CT Disclosures: Research support (NHLBI; QNRF; GE Healthcare; Philips Medical, Vital Images); Medical Advisory Board (GE Healthcare); Equity Interest (TC3 Cardiovascular Core Laboratories; Cedars-Sinai Medical Center)
Coronary CT Angiography for Diagnosis of Obstructive CAD Test Sensitivity Specificity Exercise ECG treadmill 1 68% 77% Exercise Echo treadmill 2 86% 81% Dobutamine Echo 2 ~85% ~85% Exercise nuclear treadmill 3 87% 73% Pharmacologic nuclear 3 89% 75% Coronary CTA 4 95% 83% 1. ACC/AHA 2002 Guideline Update for Exercise Testing 2. ACC/AHA/ASE 2003 Guideline Update for the Application of Echocardiography 3. ACC/AHA/ASNC Guidelines for the Clinical Use of Cardiac Radionuclide Imaging 4. ACCURACY study
Decisions of Revascularization Based Upon Angiography Alone COURAGE Trial (D/MI/CVA) BARI 2D Trial (D/MI/CVA) 2,287 individuals with angiographically obstructive CAD and ischemia assigned to PCI or medical tx and followed for 4.6 years 2,368 diabetic patients assigned to revascularization or medical tx and followed for 5 years Source: Boden et al. NEJM 2007; BARI 2D Study Group, NEJM 2009
log Hazard Ratio Ischemia-Guided Revascularization Therapeutic Benefit Based Upon MPS Findings Myocardial Perfusion Imaging Revascularization vs Medical Rx 0 1 2 3 4 5 6 10,627 patients No CAD 1.9 yr cardiac death Medical Rx Revasc *p<0.001 0 12.5% 25% 32.5% 50% Source: Hachamovitch, et al Circulation 2003 % Myocardium Ischemic
Assessment of Lesion-specific Ischemia by FFR is the Only Method To Improve Event-free Survival Source: Tonino et al. NEJM 2009;360:213-24; Pijls et al. J Am Coll Cardiol 2010
Can CCTA diagnose ischemia? CCTA has a high sensitivity and high NPV for diagnosis of obstructive CAD However, CCTA does not define the hemodynamic significance of coronary lesions Only one of these patients has hemodynamically significant CAD. Which one is it? >50% diameter stenosis >50% diameter stenosis
Reference Gold Standard for Ischemia: Fractional Flow Reserve (FFR) FFR at the time of invasive coronary angiography (ICA) is the only method for specific determination of the hemodynamic significance of coronary artery lesions (lesion-specific ischemia) FFR = Ratio of maximal myocardial blood flow through a diseased artery to the blood flow in the hypothetical case that this artery is normal Values <0.80 or <0.75 considered diagnostic of lesion-specific ischemia Source: Pijls NH et al. J Am Coll Cardiol. 2007; Pijls NH et al. J. Am. Coll. Cardiol. 2010
CCTA Stenosis Demonstrates an Unreliable Relationship to Lesion-Specific Ischemia CCTA correlates favorably with angiographic estimate of coronary stenosis but unreliably estimates hemodynamic significance of lesions as determined by FFR FFR 0.89 Source: Meijboom et al. J Am Coll Cardiol 2008 75% false positives
Computational Fluid Dynamics Computational fluid dynamics (CFD) quantifies fluid pressure and velocity, based on physical laws of mass conservation and momentum balance CFD for Patient-Specific Models of Coronary Arteries 1. Numerical method approximates governing equations 2. Obtain solution for velocity / pressure at finite (but very large) number of points 3. Simultaneously solve millions of non-linear equations and repeating process for thousands of time intervals within cardiac cycle Images courtesy of Prof. Charbel Farhat, Dept. of Aeronautics & Astronautics, Stanford University
FFR can now be derived from CCTA Detailed maps of pressure and velocity can be obtained computationally Input Data Geometry - obtained from design specifications Fluid properties viscosity and density of air Equations of fluid flow solved on supercomputers Computed velocity, pressure, lift and drag Geometry represented as a mesh Equations of fluid flow solved for each point on this mesh P = P atm Atmosphe ric pressure Velocity of air relative to wing Boundary Conditions Velocity of incoming air relative to wing Atmospheric pressure Source: Courtesy of Charley Taylor, Stanford University
Simulating coronary blood flow uses similar principles but is even more complicated Patient-specific geometry Heart-Vessel Interactions Complex fluid Properties Inputs Accurate coronary geometric models including branching structure and pathology Physiologic models personalized using minimal measured data Boundary Conditions Heart/vascular interaction Aortic impedance Time-varying coronary resistance related to intramyocardial pressure Models to simulate hyperemia Flow demand related to myocardial mass, cardiac work, etc Microcirculatory resistance depends on complex vessel structure Physiology changes due to administration of drugs to induce hyperemia Numerical Methods Anisotropic, adaptive, boundary layer mesh generation to reduce computation time Tight coupling between heart model and aorta/coronary model High performance parallel incompressible flow solver
FFR CT is determined from typically-acquired CCTA 1. No additional image acquisition 2. No excess radiation 3. No modification to imaging protocols (prospective or retrospective gating) 4. No administration of adenosine or other medications 3D FFR CT map computed FFR CT = 0.72 (can select any point on model)
DISCOVER-FLOW: Diagnosis of ischemia-causing stenoses obtained via non-invasive fractional flow reserve Objective: To determine the diagnostic performance of non-invasive FFR CT, as compared to invasively measured FFR Study design Prospective multicenter study 1. Seoul National University Hospital, Seoul, Korea 2. Pauls Stradins University, Riga, Latvia 3. Inje University Paik Hospital, Koyang, Korea 4. Stanford University, Stanford, CA, USA 5. New York Presbyterian Hospital, NY, USA Sample size calculation Assumed diagnostic accuracy of CCTA stenosis for FFR-ischemia = 49% 2 150 vessels required to detect an improvement in diagnostic accuracy of FFR CT by 25% Type I error: 0.05, statistical power: 90% Source: Koo et al. J Am Coll Cardiol 2011; 2 Meijboom et al. J Am Coll Cardiol 2008
Diagnostic Performance of FFR CT and CCTA 25% Source: Koo et al. J Am Coll Cardiol 2011
Diagnostic Performance of FFR CT and CCTA ROC curve analysis Per-Vessel Per-Patient 1 1 FFR-CT FFR-CT Sensitivity CT alone Sensitivity CT alone Area Under the Curve Area Under the Curve FFR CT = 0.90 CCTA = 0.75 P=0.001 FFR CT = 0.92 CCTA = 0.70 P=0.0001 0 1 - Specificity 0 1 0 1 - Specificity 0 1 Source: Koo et al. J Am Coll Cardiol 2011
Examples DISCOVER-FLOW CCTA FFR CT Invasive angiography FFR FFR 0.74 >50% diameter stenosis FFR CT 0.74 ischemia >50% diameter stenosis FFR 0.74 ischemia FFR 0.85 >50% diameter stenosis FFR CT 0.85 no ischemia >50% diameter stenosis FFR 0.84 no ischemia
What about intermediate stenoses? 50-70% 50-70% Source: Tonino PA et al. J Am Coll Cardiol
Moderate vs. Severe Stenosis: Moving Beyond a Binary Approach Stenosis Severity 0-24% 25-49% 50-69% >70% 100% Cheng et al. JACC CV Imaging (Jul 2008)
Diagnostic Performance of FFRCT for 30-69% Stenoses by CCTA FFR CT Accuracy Sensitivity Specificity PPV NPV FFR CT 0.80 83.8 64.9 88.5 60.5 90.0 CCTA 43.1 85.2 31.2 25.2 88.9 100 80 83.8 85.2 88.5 90.0 88.9 60 43.1 64.9 60.5 40 20 31.2 25.2 0 Accuracy Sensitivity Specificity PPV NPV Leipsic et al. RSNA 2011
30-70% QCA Confirmed Stenosis FFR CT True Positive False Positive True Negative False Negative FFR CT 0.80 8 2 33 3 CCTA 10 25 10 1 40 33 25 20 0 8 10 10 2 3 1 True Positive False Positive True Negative FalseNegative
The DeFACTO Trial (Determination of Fractional Flow Reserve by Anatomic Computed Tomographic AngiOgraphy) 17-center international study of 252 patients undergoing CCTA and invasive FFR to evaluate the diagnostic performance of FFR CT Per-patient endpoint Enrollment complete LBCT ESC 2012 FFR 0.85 >50% diameter stenosis FFR CT 0.85 no ischemia >50% diameter stenosis FFR 0.84 no ischemia Source: http://clinicaltrials.gov/ct2/show/nct01233518?term=defacto&rank=1
JAMA. 2012;308(12):doi:10.1001/2012.jama.11274
Per-patient Diagnostic Performance of FFR CT 0.80 and CT 50% vs. FFR 0.80 For the Intention to Diagnose Sample FFR CT <0.80 CT >50% Estimate %, (95% CI) Subjects in group Estimate %, (95% CI) Subjects in group Sensitivity 90 (84-95) 129 84 (77-90) 129 Specificity 54 (46-83) 123 42 (34-51) 123 PPV 67 (60-74) 172 61 (53-67) 180 NPV 84 (74-90) 80 72 (61-81) 72 Accuracy 73 (67-78) 252 64 (58-70) 252
Areas under the receiver operator operating characteristics curve on a (A) per-patient and (B) per-vessel level for ischemia by FFR CT 0.80 and CT stenosis 50% as compared to invasive FFR
Conclusions FFR CT 1. Demonstrates high diagnostic performance compared to invasive measured FFR with 6 fold reduction in false positives for intermediate lesiosn as compared to CCTA when adequate image quality is achieved 2. >40% improvement in diagnostic accuracy compared to CCTA 3. Robust discrimination of ischemia-causing lesions FFR CT is a new computational method that may be considered for the physiologic assessment of intermediate stenoses identified by CCTA Careful adherence to best practice to ensure optimal image quality is highly valuable to ensure the most accurate results