Abstract—Atherosclerosis is one of the main causes of cardiovascular disease, accounting for more than one third of all deaths in the United States, there is a growing need to develop non-invasive techniques to assess the severity of atherosclerotic plaque burden. Recent research has suggested that not the size of the atherosclerotic plaque but rather its composition is indicative for plaque rupture as the underlying event of stroke and acute coronary syndrome. With its excellent soft-tissue contrast, magnetic resonance imaging (MRI) is a favored modality for examining plaque composition. In an ex- vivo study, aimed to show the feasibility of quantifying the components of carotid atherosclerotic plaques in-vivo, we acquired multi-contrast MRI images of 13 freshly excised endarterectomy tissues with commercially available MRI sequences and a human surface coil. Feature space analysis (FSA) was utilized in four representative tissues to determine the total relative abundance of calcific, lipidic, fibrotic, thrombotic and normal components as well as in consecutive 2 mm sections across the carotid bifurcation in each tissue. Excellent qualitative agreement between the FSA results and the results obtained from histological methods was observed. This study demonstrates the feasibility of combining MRI with FSA to quantify carotid atherosclerotic plaques in-vivo. I. INTRODUCTION The prevalence for cardiovascular disease (CVD) in the United States in 2003 amounted to 34.2 % of the total population. This corresponds to health care costs of $403.1 billion for 2006. Of all the deaths suffered from CVD, 53.5 % are attributed to coronary heart disease and 17 % to Stroke with atherosclerosis as its underlying cause [1]. While the mechanisms of atherogenesis remain uncertain, the response-to-injury theory is widely accepted. Circulating monocytes infiltrate the injured vessel wall resulting in subendothelial accumulation of cholesterol-engorged macrophages (‘foam cells’) leading to so-called fatty-streak lesions. While not clinically significant, these fatty-streak lesions are precursors of more advanced lesions containing a lipid-rich necrotic core and smooth muscle cells (SMC). These SMC together with the extracelluar matrix typically Manuscript received April 3, 2006. This work was supported in part by the National Institutes of Health under Grants RO1-HL63090 and T32- HL02812. C. Karmonik is a Research Scientist with The Methodist Hospital, Houston, TX 77030 USA (phone: 713-441-1583; fax: 713-790-6474; ckarmonik@ tmh..tmc.edu.) P. Basto was a Summer Medical And Research Training (SMART) student with Baylor College of Medicine, Houston, TX 77030 USA. J. D. Morrisett is Professor of Medicine and Biochemistry at Baylor College of Medicine, Houston, TX 77030 USA ( morriset@bcm.tmc.edu). form a thin fibrous cap enclosing the necrotic core. Complexity of plaque composition can increase further, involving calcification, ulceration at the luminal surface, and hemorrhaging from small blood vessels penetrating the lesion. Although atherosclerotic plaques are able to grow and thereby gradually occlude the vessel, the more serious clinical event is the acute occlusion due to the formation of thrombus leading to myocardial infarction (coronary arteries) or stroke (carotid arteries). Thrombus formation is usually associated with rupture or erosion of the fibrous cap covering the lesion. Recent results indicate that it is not the size of the plaque but rather its composition that is an indicator of the severity of the disease and risk for rupture [2]. A noninvasive method for characterizing plaque composition would therefore be highly beneficial for assessing clinical risk. Among all non-invasive medical imaging modalities, magnetic resonance imaging (MRI) is one of the best suited to examine the composition of atherosclerotic plaque tissue because of its excellent soft-tissue contrast originating from the local molecular environment of the proton spins. Multi- contrast approaches in MRI have been utilized both in vivo and ex vivo to identify features in atherosclerotic plaque tissue. Lipidic, calcific, fibrotic and thrombotic tissue components have been observed to exhibit different MRI image intensity using various MRI acquisition techniques [3] – [8]. It would be highly advantageous to combine the information contained in MRI images acquired with different parameters into one single image for visualization and quantitative analysis of the plaque composition. Feature space analysis (FSA) or cluster analysis is a well established technique that partitions a set of objects into relatively homogeneous subsets based on inter-object similarities [9]. In an ex vivo study of excised coronary atherosclerotic plaque tissue, multi-contrast MRI images (proton-density weighted (PDW), T2-weighted (T2W), intermediate weighting) were segmented using FSA [10]. An ex vivo MRI study at 9.4 T has described quantifying mineralization content in atherosclerotic tissue using k-means clustering and employing PDW, T2W, T1-weighted (T1W) and diffusion weighted MRI. [11]. The classification based on k-means cluster analysis of 9.4T MRI images (T1W, T2W and PDW) from coronary atherosclerotic plaque tissue showed very good agreement with histopathology images for all AHA classification plaque types [12]. Quantification of Carotid Atherosclerotic Plaque Components using Feature Space Analysis and Magnetic Resonance Imaging Christof Karmonik, Pamela Basto, and Joel D. Morrisett Proceedings of the 28th IEEE EMBS Annual International Conference New York City, USA, Aug 30-Sept 3, 2006 FrC01.4 1-4244-0033-3/06/$20.00 ©2006 IEEE. 3102