Deep Learning-based CT-FFR: A Gatekeeper to the Cath Lab

by | May 11, 2021 | All News, Blog

Coronary artery disease (CAD) remains the leading cause of death worldwide in both men and women, accounting for approximately 16% of the world’s total deaths in 2019 alone. Fortunately, with proper diagnosis and management, patients with CAD can be appropriately treated.

Computed tomography-derived fractional flow reserve (CT-FFR) has demonstrated the potential to improve the diagnosis of patients with CAD. DeepVessel FFR, Keya Medical’s deep learning-based CT-FFR software approved for clinical use in China and the European Union, combines the digital images created during a non-invasive coronary computed tomography angiography (CCTA) exam to estimate fractional flow reserve (FFR) using deep learning algorithms. Recent evidence and clinical trials have demonstrated that combining deep learning with non-invasive CT-FFR detection can help improve outcomes in patients with stable CAD as compared to alternative diagnostic tests.

What is Coronary Artery Disease?

Coronary artery disease occurs when the major blood vessels that supply blood to the heart become damaged or diseased, often due to plaque buildup. As plaque accumulates in the coronary arteries, less blood can flow through the arteries due to the vessel narrowing. The decreased blood flow to the heart can lead to chest pain, shortness of breath, or even a heart attack when left untreated.

Diagnosing Coronary Artery Disease in Symptomatic Patients

The path to diagnosing CAD in symptomatic patients presenting with chest pain is essential to improving outcomes and quality of care. Clinicians may recommend several different diagnostic tests to evaluate the function of the heart, each presenting potential benefits and limitations.

Single Photon Emission Computed Tomography

Single Photon Emission Computed Tomography (SPECT) is a non-invasive nuclear imaging test used for diagnosing and managing patients suspected of CAD. During the non-invasive test, a clinician will inject radioactive tracers through a vein in the patient’s arm to produce images of the heart. The test allows clinicians to see how blood flows through the arteries and veins to evaluate how well the heart is working.
Studies examining the effectiveness of SPECT have demonstrated that it offers a lower sensitivity for CAD than other diagnostic tests [1]. As a result, SPECT tests may lead to a higher proportion of CAD cases going undetected.

Exercise Stress Test

An Exercise Stress Test utilizes an electrocardiogram (EKG) machine to measure the function of the heart. During an exercise stress test, a patient may be asked to exercise on a treadmill or stationary bicycle while the speed is gradually increased to a faster pace. Throughout the test, the clinician will monitor the patient’s heart rate to see if the blood supply is reduced in the coronary arteries.

Exercise stress testing is commonly used for the detection of CAD in patients presenting with chest pain, and it offers a less expensive pathway to diagnosing CAD. However, according to research published in The International Journal of Clinical Practice, the diagnostic accuracy of exercise stress testing depends on factors including the age, gender, and clinical characteristics of the patient [2]. Additionally, exercise stress testing is more useful at excluding CAD rather than confirming that a patient has it.

Stress Echocardiogram

Stress Echocardiography (Stress Echo) is a diagnostic test in which a clinician uses ultrasound imaging to evaluate the function of the heart. During a stress echo exam, a patient may be asked to exercise on a treadmill or stationary bicycle while the speed is gradually increased, similar to an exercise stress test. During the exam, ultrasound waves create pictures of the heart that allow the clinician to assess the blood and oxygen flow during exercise.
Stress echocardiography is recommended for the diagnosis and assessment of patients presenting with chest pain in whom there is an intermediate probability of CAD. While stress echo is a commonly used non-invasive technique for imaging the heart, the technique has a relatively low sensitivity in recognizing multivessel disease in normal ventricles [3].

Invasive Coronary Angiogram

Invasive Coronary Angiography (ICA) is an invasive diagnostic procedure that is widely accepted as the standard for detecting CAD in patients with high clinical risk. The test involves injecting a contrast agent into the blood vessels of a patient through a catheter and taking X-ray pictures to detect blockages in the coronary arteries that are caused by plaque buildup. ICA is typically performed to identify treatable coronary artery lesions and determine the revascularization strategy prior to initiative definitive treatment.

According to a study published in the American Heart Journal, only 41% of patients referred for elective ICA are found to have obstructive CAD [4]. This exposes patients to greater risk as they undergo unnecessary invasive procedures while incurring additional costs.

Fractional Flow Reserve

Fractional Flow Reserve (FFR) provides clinicians with a tool to conduct functional analysis of the heart. It is widely accepted as the gold standard for assessing the functional significance of stenotic lesions and guiding coronary revascularization procedures [5].

FFR is performed by inserting a pressure guidewire into a patient’s groin or wrist and fishing it up to the heart to measure pressure, temperature, and blood flow to determine lesion severity. FFR is calculated by taking the ratio of blood flow in a coronary artery to the theoretical maximum flow in a normal artery. An FFR value higher than 0.8 is an indicator that the stenosis is not hemodynamically significant, while a value lower than 0.8 indicates that the stenosis is hemodynamically significant, and revascularization should be performed.

Multiple randomized trials have demonstrated the diagnostic value of FFR. The Fractional Flow Reserve Versus Angiography for Multi-vessel Evaluation (F.A.M.E.) trial illustrated that FFR-guided intervention could improve patient outcomes. In this trial, FFR-guided intervention resulted in a 33% reduction in MACE when compared to invasive coronary angiography [6]. However, current guidelines recommend assessing myocardial ischemia in stable patients with CAD through non-invasive functional testing before considering invasive coronary angiography or conducting myocardial revascularization [5,7].

Coronary Computed Tomography Angiography

Coronary computed tomography angiography is a non-invasive imaging test used to examine the arteries that supply blood to the heart. During a CCTA procedure, contrast material is injected into a vein in a patient’s arm and a computed tomography scanner is used to take pictures of the heart vessels from outside the body. This allows clinicians to obtain 3D images of the heart and vessels that can be used to observe, measure, and characterize plaque buildup in the coronary arteries and determine if there is any vessel narrowing.

Recent clinical trials suggest that CCTA should be used as a first-line diagnostic test for evaluating patients presenting with stable chest pain. The Scottish Computed Tomography of the Heart (SCOT-HEART) trial and the Prospective Multicenter Imaging Study for Evaluation of Chest Pain (PROMISE) trial have both shown that CCTA can lead to more appropriate use of ICA while providing better information than functional testing in patients who have stable chest pain with a low burden of obstructive CAD [8,9]. Additionally, the Randomized Evaluation of Patients with Stable Angina Comparing Utilization of Noninvasive Examinations (RESCUE) trial demonstrated similar outcomes between patients who underwent CCTA and patients who underwent SPECT but demonstrated that CCTA was a better predictor of major adverse cardiovascular events (MACE) and revascularization [10].

In response to this mounting body of evidence supporting the usage of CCTA, professional societies including the National Institute for Health and Care Excellence (NICE), European Society of Cardiology (ESC), American College of Cardiology (ACC), and the Society of Cardiovascular Computed Tomography (SCCT) have published new guidelines and recommendations on the first-line usage of CCTA in clinical practice. These guidelines recommend that clinicians adopt a CCTA-first strategy when diagnosing CAD in patients with stable chest pain in whom obstructive CAD cannot be excluded by clinical assessment alone.

DeepVessel FFR: Deep Learning-based CT-FFR

Combining CT-FFR with deep learning technology offers an alternative, non-invasive way of diagnosing CAD that can overcome the potential limitations of standalone diagnostic tests. DeepVessel FFR performs a non-invasive physiological functional assessment of the coronary arteries and accurately predict FFR values based on CCTA digital images. The software uses deep learning technology to construct 3D models of the coronary arteries and estimate FFR values non-invasively. As compared to traditional FFR techniques, DeepVessel FFR provides clinicians with a more efficient and affordable diagnostic tool to identify potential narrowing of the coronary arteries that may impact blood flow to the heart.

DeepVessel FFR software displayed on desktop

Figure 1: Deep learning-based CT-FFR software, DeepVessel FFR, generates a 3D model of the coronary artery tree and computes non-invasive FFR values from CCTA DICOM images.

Benefits of Deep Learning-based CT-FFR

Multiple studies have investigated the performance of deep learning-based CT-FFR compared to invasive diagnostic tests. A 2019 study published in the Journal of Geriatric Cardiology evaluated the performance of DeepVessel FFR in calculating FFR values from CCTA images using deep learning. The single-center prospective study included 63 patients and the performance of the software was assessed using invasive FFR as the reference standard.

According to the study findings, DeepVessel FFR correlated with invasive FFR, demonstrating that deep learning-based CT-FFR techniques can allow for efficient, non-invasive assessment of the functional significance of coronary stenosis [11].

Additional studies have demonstrated that deep learning-based CT-FFR can help to reduce unnecessary invasive procedures performed in the cardiac catheterization laboratory.

Researchers at Fuwai Hospital, a leading medical center in Beijing specializing in the treatment of complex cardiovascular diseases, conducted a retrospective study to evaluate deep learning-based CT-FFR in detecting hemodynamic changes of stenosis. The study included 73 patients suspected of CAD who received CCTA followed by invasive FFR within 90 days.

Diagnostic performance of DeepVessel FFR as compared to ICA and CCTA

Figure 2: Per-vessel diagnostic performance of DeepVessel FFR as compared to CCTA and invasive coronary angiography alone.

In the study, the performance of CT-FFR exceeded that of CCTA at both the patient and vessel levels while also offering improvements in specificity and positive predictive value without losing sensitivity and negative predictive value. Additionally, the study findings demonstrated that CT-FFR predicted clinical outcomes of patients, including MACE and re-hospitalization. As depicted in Figure 2, DeepVessel FFR achieved an accuracy of 92%, a sensitivity of 94%, and a specificity of 90%, exceeding that of both CCTA and ICA [13].

A separate study conducted by researchers at the First Affiliated Hospital of Jinan University published in European Radiology investigated the use of deep learning-based CT-FFR as an alternative to invasive coronary angiography in the selection of patients for coronary intervention. A total of 243 patients with symptomatic CAD were included in the study.

The study results showed that up to 72% of invasive diagnostic procedures could have been avoided by using a DL-CTFFR value greater than 0.8 as a cut-off for intervention without increasing MACE [14]. The study also showed that revascularization for symptomatic CAD patients guided by the combination of CCTA and DL-CTFFR was associated with a MACE rate like that when guided by coronary angiography in a 2-year follow-up investigation.

Enabling Better Diagnosis & Management of Patients with Coronary Artery Disease

These studies illustrate the value that deep learning-based CT-FFR can add when diagnosing and managing patients with CAD. Additional research is needed to help further demonstrate the potential of deep learning-based CT-FFR software in improving patient outcomes.

Collaborate with Us

At Keya Medical, we are committed to working closely with clinicians to improve the diagnosis and treatment of cardiovascular disease. To learn more about available collaboration opportunities, please contact us.
Contact Us
References
[1] Parker MW, Iskandar A, Limone B, Perugini A, Kim H, Jones C, Calamari B, Coleman CI, Heller GV. Diagnostic accuracy of cardiac positron emission tomography versus single photon emission computed tomography for coronary artery disease: a bivariate meta-analysis. Circ Cardiovasc Imaging. 2012 Nov;5(6):700-7. doi: 10.1161/CIRCIMAGING.112.978270. Epub 2012 Oct 10. PMID: 23051888.
[2] Banerjee A, Newman D, Van den Bruel A, & Heneghan C. Diagnostic accuracy of exercise stress testing for coronary artery disease: A systematic review and meta‐analysis of prospective studies. International Journal of Clinical Practice. 2021;(66):477-492. https://doi.org/10.1111/j.1742-1241.2012.02900.x
[3] Marwick TH. Stress echocardiography. Heart. 2003;89(1):113-118. doi:10.1136/heart.89.1.113
[4] Patel MR, Dai D, Hernandez AF, et al. Prevalence and predictors of nonobstructive coronary artery disease identified with coronary angiography in contemporary clinical practice. Am Heart F, 167 (2014), pp. 846-852.e2

[5] Montalescot G, Sechtem U, Achenbach S, et al. 2013 ESC guidelines on the management of stable coronary artery disease: the Task Force on the management of stable coronary artery disease of the European Society of Cardiology. Eur Heart J 2013;34:2949–3003.

[6] Tonino PA, De Bruyne B, Pijls NH et al. FAME Study Investigators. Fractional flow reserve versus angiography for guiding percutaneous coronary intervention. N Engl J Med 2009;360:213-24.
[7] Neumann, F, Sousa-Uva, M, Ahlsson, A, et al. 2018 ESC/EACTS Guidelines on myocardial revascularization. Eur Heart J 2018;40:87-165.
[8] Williams MC, Hunter A, Shah ASV, et al; SCOT-HEART Investigators. Use of Coronary Computed Tomographic Angiography to Guide Management of Patients With Coronary Disease. J Am Coll Cardiol. 2016 Apr 19;67(15):1759-1768. doi: 10.1016/j.jacc.2016.02.026. PMID: 27081014; PMCID: PMC4829708.
[9] Douglas PS, Hoffmann U, Patel MR, et al; PROMISE Investigators. Outcomes of Anatomical versus Functional Testing for Coronary Artery Disease. N Engl J Med 2015; 372:1291-1300.
[10] Stillman AE, Gatsonis C, Lima JAC, et al; the RESCUE Investigators. Coronary computed tomography angiography compared with single photon emission computed tomography myocardial perfusion imaging as a guide to optimal medical therapy in patients present with stable angina: The RESCUE trial. Journal of the American Heart Association 2020;9:e017993.
[11] Wang ZQ, Zhou YJ, Zhao YX, et al. Diagnostic accuracy of a deep learning approach to calculate FFR from coronary CT angiography. J Geriatr Cardiol. 2019 Jan;16(1):42-48. doi: 10.11909/j.issn.1671-5411.2019.01.010. PMID: 30800150; PMCID: PMC6379239.
[12] Nørgaard BL, Leipsic J, Gaur S, et al. Diagnostic performance of noninvasive fractional flow reserve derived from coronary computed tomography angiography in suspected coronary artery disease: the NXT trial (Analysis of coronary blood flow using CT angiography: Next steps). Journal of the American College of Cardiology, 2014, 63(12):1145-1155.
[13] Li Y, Qiu H, Hou Z, et al. Additional value of deep learning computed tomographic angiography-based fractional flow reserve in detecting coronary stenosis and predicting outcomes. Acta Radiologica (Stockholm, Sweden: 1987), 2021: 284185120983977
[14] Liu X, Mo X, Zhang H, et al. A 2-year investigation of the impact of the computed tomography-derived fractional flow reserve calculated using a deep learning algorithm on routine decision-making for coronary artery disease management. Eur Radiol. 2021 Feb 25. doi: 10.1007/s00330-021-07771-7. Epub ahead of print. PMID: 33630159.