| COMMENTARIES |
Cardiac CT in the Assessment of Non-Calcified Plaque: Progress in the Quest for a Non-Invasive Definition of the Vulnerable Atheroma
Harald Brodoefel, Department of Diagnostic Radiology, Eberhard-Karls-University, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany, Email: h.brodoefel@t-online.de
The majority of myocardial infarctions result from rupture of vulnerable atheroma. The latter is characterized by a distinct histology, notably a large lipid core covered by a thin fibrous cap [1]. Due to the phenomenon of vascular remodeling significant luminal stenosis may be absent in more than half of these plaques which might thus be invisible to two-dimensional invasive coronary angiography (ICA) [2]. Intravascular-ultrasound (IVUS) is the current gold standard for the detection and quantification of non-obstructive plaque. Only recently, a new technique has been designed that relies on spectrum analysis of IVUS-derived radiofrequency and, thereby, differentiates the major components of coronary plaque [3,4]. This so-called IVUS virtual histology (IVUS-VH) holds the promise of identifying non-obstructive but vulnerable atheroma and may serve as a tool for risk stratification or serial monitoring of therapy. However, the method is time-consuming, costly, and ultimately restricted to a small number of highly specialized centers. Also, IVUS is limited to proximal segments; the plaque burden of the entire coronary tree cannot be assessed. Together, these drawbacks have fuelled the quest for alternative techniques to detect, quantify, and characterize the vulnerable plaque.
In recent years, multi-slice computed tomography (CT) has seen a revolutionary technical advance. By means of its excellent negative predictive value it has evolved into an accepted alternative to ICA in patients with intermediate risk of CAD and clinical presentation that mandates the evaluation of possible underlying CAD [5]. While initial applications were restricted to the detection or coronary calcium, visualization of the lumen and stenosis detection have by now become the major focus and the driving force for further technical refinement. However, with ongoing increases of spatial and temporal resolution, applications of CT beyond lumenography are finally appreciated. Since CT may assess the complete coronary tree and visualize not only the vessel lumen but also the wall and perivascular fat it seems ideally suited to the analysis of the entire atherosclerotic disease process, most notably the non-obstructive and non-calcified vulnerable plaque.
The general feasibility of CT plaque detection, quantification, and compositional characterization has been long been established by 16- or 64-slice technology [6-9]. Achenbach et al. reported a sensitivity and specificity for the detection of any non-calcified plaque of 78% or 87%, respectively [10]. Plaque volume was shown by Leber et al. to closely correlate (r = 0.83) to IVUS [11]. Most importantly, attenuation in Houndsfield Units (HU) of fibrous and lipid plaque components were proved to be significantly different [6,12,13]. Accordingly, a recent study has found a higher prevalence of plaques with a low CT attenuation of 30 < HU in lesions associated with acute coronary syndromes when compared to stable coronary atheroma [14].
However, up to now quantification and characterization of non-calcified plaque by CT have remained challenging for many reasons. Temporal and spatial resolutions constitute a fundamental technical restriction that limits accuracy and reproducibility of plaque analysis. Moreover, until now, CT attenuation of non-calcified plaque has only retrospectively been assessed from manual measurements in atheroma. The latter were just visually correlated to hypo- and hyperechoic areas in conventional grey-scale IVUS. The potential to prospectively use CT attenuation for plaque analysis has never been exploited, meaning that plaques or plaque compartments had to be segmented by visual estimate and manual editing. However, such approach is associated with an unacceptable inter-observer variability of up to 37% for plaque volume [11].
Only recently, dedicated software has become available that relies on CT attenuation and provides automated segmentation of non-calcified plaque through definition of a set of HU ranges for lumen, wall, and plaque components. This prospective use of CT attenuation is largely independent of observer and allows for volumetric quantification of both the entire plaque burden and the plaque components.
At the same time, the release of the latest dual-source CT (DSCT) system has introduced a remarkable improvement on temporal resolution (83 ms). In concert with the
When applying these thresholds in a blinded prospective plaque analysis, no systematic biases were found for plaque volume or percentages of plaque composition. However, while correlation to IVUS-VH was r = 0.83 for plaque volume, it turned out to be insignificant for percentages of plaque composition. With intra-class correlation coefficients being 0.90 for plaque volume and 0.81, 0.94, or 0.98 for percentages of lipid, fibrous, or fatty components, respectively, inter-observer reproducibility proved to be excellent.
The good reproducibility of our measurements is particularly promising and reflects the elimination of manual reader-dependent steps in the HU-based analysis approach. The method, thus, seems to overcome the primary drawback of manual segmentation and provides the key prerequisite for larger studies on non-invasive plaque characterization.
Non-significant correlation of percentage plaque composition to IVUS-VH seems to be a serious limitation but has to be put into perspective. Since we studied a very small number of plaques and all turned out to have a very similar composition in IVUS-VH, further investigation is mandatory to assess the accuracy of CT in demonstrating clinically relevant differences in plaque morphology. While IVUS-VH has a much better spatial resolution and may better dissect the complex anatomy of atheroma, it is more susceptible to artifacts resulting from acoustic shadowing of calcifications. Notably, in a recent comparison between in vivo IVUS-VH and histology, similar observations have been highlighted and correlation of any non-calcified plaque component, likewise, turned out to be insignificant [16]. Ultimately, some inter-method discrepancy appears to be unavoidable. Indeed, as long as future studies succeed in providing a link between CT-based compositional data and the vulnerability of plaque, weak correlation to IVUS-VH does not limit the attractiveness of CT plaque characterization. In my opinion, the latter does have the potential for an independent, non-invasive definition of the vulnerable atheroma.
However, on the way towards possible clinical routine a number of challenges remain. The most critical is spatial resolution of current scanners which needs further and substantial improvement to provide greater robustness in detection and characterization of non-obstructive, non-calcified plaque. Reproducibility has to be confirmed not only between observers of the same study but also between successive CT studies of the same patient. With intra-luminal contrast being variable between different CT examinations, optimal HU thresholds will change and algorithms need to be developed to compensate for inter-study variability. In case we might want to not only do risk stratification but also follow plaque stabilization in patients receiving therapy radiation, dose needs must be reduced to justify long-term studies. Finally, clinical indications need to be defined. At this point, the prognostic value of soft plaque analysis is not entirely clear. Use of cardiac CT solely for the purpose of risk stratification will only be clinically indicated once a significant benefit over traditional methods of risk stratification has been demonstrated.
Ultimately it seems that cardiac CT is slowly coming closer to playing its full potential as a comprehensive non-invasive imaging modality. Thanks to both its combined visualization of coronary artery lumen and wall as well as its capacity to characterize plaque by means of CT attenuation, the method offers far more than mere lumenography. At this stage, the feasibility of reproducible plaque characterization has been demonstrated. Further studies are needed to finally define the vulnerable plaque in a non-invasive manner or assess the CT criteria for risk stratification or plaque stabilization.
References