Signal-Processing Approaches to Risk Assessment in Coronary Artery Disease

Cardiovascular disease has long been the leading cause of death in developed countries and it is rapidly gaining similar status in developing countries. Sudden heart attacks remain the primary cause of death in the United States. In the year 2004 it is estimated that more than 1.4 million Americans, and more than 19 million others worldwide, experienced a life-threatening heart attack; these unpredicted attacks account for the majority of the $280 billion burden of cardiovascular diseases. Coronary artery disease (CAD) currently claims over 12.9 million American patients. In addition, CAD is the leading cause of premature, permanent disability in the U.S. labor force and accounts for about 1/5 of all disability allowances by the Social Security Administration. 1 Dept. of Computer Science, University of Houston, TX 2 Dept. of Cardiology, University of Athens, Greece 3 Cardiovascular Research Foundation, New York, NY 4 Dept. of Mechanical Engineering, University of Houston, TX 5 Baylor College of Medicine, Houston, TX 6 Dept. of Cardiology, University of Aarhus, Denmark 7 Association for Eradication of Heart Attack, Houston, TX Signal-Processing Approaches to Risk Assessment in Coronary Artery Disease Ioannis A. Kakadiaris, Sean M. O’Malley, Manolis Vavuranakis, Stéphane Carlier, Ralph Metcalfe, Craig J. Hartley, Erling Falk, and Morteza Naghavi Cardiovascular disease has long been the leading cause of death in developed countries and it is rapidly gaining similar status in developing countries. Sudden heart attacks remain the primary cause of death in the United States. In the year 2004 it is estimated that more than 1.4 million Americans, and more than 19 million others worldwide, experienced a life-threatening heart attack; these unpredicted attacks account for the majority of the $280 billion burden of cardiovascular diseases. Coronary artery disease (CAD) currently claims over 12.9 million American patients. In addition, CAD is the leading cause of premature, permanent disability in the U.S. labor force and accounts for about 1/5 of all disability allowances by the Social Security Administration. CARDIOLOGY’S RECENT PARADIGM SHIFT Coronary artery disease occurs as a result of atherosclerosis, a condition in which fatty plaques build up on the walls of the coronary arteries. If these plaques rupture, thrombosis may occur and obstruct the flow of blood to the heart, thus causing a potentially fatal heart attack. Some plaques present a particularly high risk of rupture; these are defined as vulnerable plaques. In the past decade, the field of cardiology has witnessed a major paradigm shift. Previously, gradually-accumulating fatty deposits and the consequent narrowing of the coronary arteries were thought to be the culprits in acute cardiac events. Today, cardiovascular specialists know that heart attacks are caused by inflammation of the coronary arteries and thrombosis-related complications of vulnerable plaques. As a result, the discovery of vulnerable plaque has the definition of a vulnerable patient, i.e., an individual with a high likelihood of experiencing a heart attack in the near future [1]. CARDIOVASCULAR RISK SCREENING Atherogenesis, or the formation of plaque, starts at a very early age and takes place over many years. The classification of plaque formation is based on a timeline from the initial inflammation to the lipid-rich plaque that leads to myocardial infarction. There are 140 million Americans over the age of 35 who are at risk of a heart attack; among this population, those in the range of 45-75 years are at highest risk for loss of productive life years. About 50-60 million people in this category may not exhibit clinical evidence of the disease, and may or may not have risk factors (race, sex, family history, hyperlipidemia, hypertension, diabetes, etc.), but they are still at risk of a heart attack. Since approximately 50% of deaths relating to acute cardiac events occur in such individuals with no prior symptoms, there is an urgent need for a wide-area screening program. Toward this end, the Association for Eradication of Heart Attack (a non-profit organization based in Houston, Texas) has developed the Screening for Heart Attack Prevention and Education (SHAPE) program, which promotes screening based on a battery of increasingly invasive tests. Emerging diagnostic tests—unavailable even a few years ago—to assess patient vulnerability factors include novel genetic and serum biomarkers such as the inflammatory marker C-reactive protein (CRP); noninvasive imaging tests such as computed tomography (CT), ultrasonography (for which relevant risk-scoring methods have been developed), and magnetic resonance (MR) imaging; and interventional catheter-based tools that allow precise, localized assessment. All of these provide unprecedented opportunities for early detection and risk stratification for potentially vulnerable patients. But what effect in our picture of vulnerable population would broad application of these tools have? It is estimated that screening with these techniques would identify 2-3 million additional individuals with lesions in their coronary arteries that are significant enough to require intervention. However, such a screening program would generate a huge influx of raw imaging data. For a single patient, a number of multidimensional imaging datasets are collected during the screening process, with increasing levels of storage and analysis complexity (2-D, 3-D, 2-D+time, 3-D+time, etc.). Applying these techniques to all those requiring diagnostic imaging, it is apparent that the amount of data that will be produced is overwhelming. Consequently, computational tools to assist in analyses of the pathological conditions that underlie sudden cardiac events are in high demand. INTRAVASCULAR ULTRASOUND-BASED IDENTIFICATION OF VULNERABLE PLAQUE Detection of the vulnerable plaques, particularly the rupture-prone vulnerable plaques (which are the culprit of many heart attacks and strokes) is one of the most active areas of research in both the cardiology and biomedical imaging communities. While there exist some non-invasive technologies (mentioned earlier) that can assess plaques morphologically, to the best of our knowledge no technology currently exists to reliably characterize plaques as vulnerable or non-vulnerable in a living human. To assist with this task, invasive technologies that are currently available, or under development, include visible-light angioscopy, near-infrared optical coherence tomography, thermography, spectroscopy, and intravascular ultrasound (IVUS). Among these technologies, IVUS is a particularly attractive option as it is wellknown, widely-available, has a long history of use, and is clinically-approved. Since the mid-1990’s, interest in IVUS signal processing has increased dramatically due to the emergence of frequency-domain techniques, which have enabled in vivo characterization of vessel and plaque tissue properties [2]. Yet in spite of its power, the major drawback of IVUS has been its inability to provide information about plaque vulnerability, i.e., an assessment of whether a plaque is stable and requires little or no treatment, or vulnerable and requires more aggressive intervention. To address this limitation, the presence of vasa vasorum (VV) microvessels (shown in Fig. 1a) could be exploited. While normally these vessels simply feed larger vessels, recent evidence suggests that VV proliferation is a strong marker of plaque inflammation, and a preceding or concomitant factor associated with plaque rupture and instability [3, 4]. As such, a technology capable of imaging VV in vivo would provide the clinician with a powerful indicator of plaque vulnerability. However, so far the VV could not be imaged in vivo; even utilizing a high-resolution intravascular technology such as IVUS, the small scale and echo-transparency of the VV typically renders them invisible. In what follows, we introduce a novel method that enables (for the first time) IVUS detection of atherosclerotic plaque inflammation based on quantification of VV density and perfusion. This method consists of a contrast-enhanced IVUS acquisition protocol and a series of signal/image-processing techniques to detect VV in the resulting contrast-enhanced sequences. DATA ACQUISITION To acquire the IVUS data, we have used both a rotating single-crystal 40 MHz scanner (Boston Scientific, Inc.; GalaxyTM) and a solid-state phased-array 20 MHz scanner (Volcano Therapeutics, Inc.; InvisionTM). For contrast we utilize OptisonTM: an ultrasound contrast agent composed of albumin microspheres filled with octafluoropropane gas. The IVUS catheter is advanced percutaneously, transluminally into the patient; and imaging is typically performed near a suspect region of the arterial wall. First, a baseline IVUS signal is recorded for 1-2 min (“pre-injection” period). Next, the contrast agent is injected, temporarily washing out the IVUS signal due to the echo-opacity of the lumen (“during-injection” period). Finally, IVUS signals are collected for 1-2 min (“post-injection” period), followed by the injection of normal saline to flush residual microbubbles. During these periods the IVUS catheter is not moved. SIGNAL PROCESSING Following acquisition, our VV detection and imaging method consists of three steps. First, the IVUS sequence is gated to eliminate gross motion artifacts. Next, the region-of-interest of the vessel is tracked with a hybrid rigid/elastic registration technique such that relative catheter/vessel motions can be eliminated and elastic wall deformations compensated for. Finally, enhancement detection is performed using difference-imaging and statistical techniques. The resulting enhancement is visualized and quantified. Step 1: Frame Gating – In stationary-catheter IVUS studies, maintaining a fixed catheter position with regard to an anatomic point of reference is impossible in practice due