Coronary plaque phenotyping with cardiac CTA: Separating the signal from the noise

In patients with stable coronary artery disease (CAD), a small percentage of atherosclerotic plaques can undergo vulnerable transformation over time, placing these patients at a higher risk of experiencing future myocardial infarction (MI). Coronary computed tomography angiography (CTA) based coronary imaging is emerging as a promising tool for identifying high-risk plaques as it can combine advanced imaging techniques to extract qualitative and volumetric features of plaque composition from radiographic images and apply machine learning algorithms to analyze these features. By analyzing a large number of radiomic features, ‘CT-radiomics' can provide a comprehensive assessment of plaque morphology and composition, including measuring diameter stenosis (DS%), total plaque volume (TPV), low-attenuated plaque volume (LAPV) and identifying the presence of adverse plaque characteristics (APC) such as low-attenuated plaque, positive remodeling, spotty calcification, and napkin ring sign [, , ]. In addition, combining CT-radiomics with computational fluid dynamics (CFD) can also enable hemodynamic assessment of lesions such as CT-fractional flow reserve (FFR_ct). This approach allows for the evaluation of functional aspects of plaques, which can help guide treatment decisions []. Furthermore, local plaque-specific indices such as wall shear stress (WSS) have been shown to play an important role in plaque transformation and can help assess for risk of MI []. Consequently, CT-based computational imaging can add to CT-radiomics for the early identification of high-risk plaques in patients with stable CAD, which may help guide interventions to reduce future MI and potentially improve outcomes (Fig. 1).

In this issue of Atherosclerosis, Yang and colleagues examined the 10-year prognostic implications of baseline CT-derived coronary hemodynamics and morphological characteristics in a cohort of patients (n = 103) with stable CAD, who were enrolled prospectively []. A vessel level analysis was performed on 78 vessels from 57 patients with a mean angiographic diameter stenosis of 50%. The mean invasive FFR was found to be 0.90, with nearly 90% of vessels demonstrating an FFR value greater than 0.80. Target vessel failure (TVF) occurred in 11 vessels (14.6%), defined as a composite of cardiovascular death, vessel-related MI, and clinically mandated vessel-related revascularization. The study revealed that vessels with lower FFR_ct and greater percent atheroma volume (PAV), were at a higher risk of TVF, even after adjusting for DS > 50% and the presence of at least two APCs or for age and other cardiovascular risk factors. In addition, when FFR_ct and PAV were combined with DS > 50% and APCs, the prognostic value of this combined model to predict 10-year TVF was significantly enhanced, as evidenced by an AUC of 0.82 (p < 0.01).

Among, 138 coronary lesions, the target lesion failure rate (TLF, a composite of cardiovascular deaths, vessel related-MI, and target-lesion revascularization) was 11.5% (15 lesions). Lesion level analysis revealed that a greater delta FFR_ct and a higher baseline WSS were associated with a greater risk of future TLF, even after adjusting for DS > 50% and APCs or for age and cardiovascular risk factors. Furthermore, a CT-derived model combining DS% > 50%, LAPV with WSS, and delta FFR_ct demonstrated the best 10-year prognostic value to predict future TLF (AUC 0.81, p < 0.01).

It is worth noting that previous studies examining the use of intracoronary imaging techniques for plaque phenotyping have shown limited success in predicting future adverse cardiac events []. However, a sub-analysis of the PROSPECT (Providing Regional Observations to Study Predictors of Events in the Coronary Tree) study revealed that combining low WSS with at least two high-risk plaque features, identified through invasive intravascular ultrasonography (such as thin cap fibroatheroma, minimum lumen area < 4mm², or plaque burden >70%), significantly improved the prognostic value for predicting the risk of future major cardiovascular events []. Another study conducted on patients with stable CAD, significant plaque burden, and hemodynamically significant lesions (mean DS% 49, median FFR 0.70) computed WSS using only angiographic images. This study showed that high WSS has additional prognostic value beyond invasive FFR in predicting vessel related MI at three years []. Furthermore, the EMERALD (Exploring the Mechanism of the Plaque Rupture in Acute Myocardial Infarction using Coronary CT Angiography and Computational Fluid Dynamics) study demonstrated that identifying plaques with both high-risk morphological features and adverse hemodynamic characteristics (such as high WSS and greater delta FFR_ct) using coronary CTA was more effective in helping clinicians identify lesions at a higher risk of acute coronary syndromes (ACS) during a follow-up period of approximately one year, compared to models that included only APCs or only plaque-hemodynamics characteristics []. However, the current study by Yang et al. is the first to demonstrate the potential of a comprehensive plaque assessment that incorporates CT-derived plaque morphological and hemodynamic features in identifying high-risk plaques and predicting future long-term cardiovascular events in patients with stable CAD who have significant plaque burden but non-functional lesions. Moreover, this study shows that using CTA to measure FFR_ct and percent atheroma volume can create a distinct digital signature for each coronary vessel, potentially assisting in identifying patients with a higher risk of future adverse cardiac events [].

It is essential to consider the limitations of the study conducted by Yang and colleagues. Despite enrolling 103 patients initially, the final analysis only included 57 patients, which restricts the statistical power of the study. In addition, the low incidence of adverse events made it challenging to confirm the relationship between plaque-specific morphological and hemodynamic features with adverse cardiac events, even after adjusting for confounding variables. Furthermore, while the authors used appropriate statistical methodologies to account for multiplicity that arises from analyzing greater than one vessel and lesion from the same patient, the absence of a derivation cohort limited the ability to perform robust sensitivity analyses to verify the stability of their findings. It is worth mentioning that an increase in WSS was associated with a greater risk of 10-year target lesion failure in a cohort of patients with predominantly functionally negative lesions. Low and oscillatory WSS is associated with early atherosclerosis, plaque progression, and the development of non-stenotic thin cap fibroatheroma []. Atherosclerotic plaques in vascular beds exposed to multiple cardiovascular risk factors such as diabetes and hyperlipidemia evolve into stenotic lesions, demonstrating high WSS at the throat of the lesions and low WSS in the downstream segments, which creates an inflammatory microenvironment that triggers biological pathways. These pathways cause endothelial cells to produce plasmin, attract macrophages to release metalloproteinases that degrade collagen and induce apoptosis of smooth muscle cells, ultimately leading to an increase in necrotic core and decreased fibrous cap stability, which are precursors of future acute coronary syndromes []. It is important to mention that the lesion-level analysis in this study only included those lesions with at least 30% DS. Hence, the cohort of lesions included in the analysis had considerable plaque burden. Indeed, after dividing the WSS data into tertiles, it was noted that lesions with WSS >19.4 Pa accounted for a substantial proportion (34.6%) of the target lesion failures, whereas lesions with WSS <11.9 Pa showed a TLF rate of 10.1%. Interestingly, lesions demonstrating more physiologic WSS between 11.9 and 19.4 Pa exhibited a lower TLV rate of only 3.2%, indicating a U-shaped relationship between WSS and target vessel failure rates over a period of 10 years, emphasizing the intricate interplay between plaque morphology and local hemodynamic forces. It is also noteworthy that, unlike prior studies that used a much lower value to define high WSS [], the current study computed WSS based on the predictable response of the microcirculation to the maximal hyperemic state []. Hence, the authors reported much higher WSS values that were derived from hyperemic simulations for FFR_ct. As such, there is a need for standardization in reporting these values to ensure consistency and reliability of results across different studies and clinical settings. Moreover, artifacts arising from coronary calcification can also hinder accurate CT-phenotyping, and every effort should be made to improve the quality of image acquisitions. Lastly, reducing the number of mathematical assumptions used to compute hemodynamic indices can help to fully exploit the potential of CT-based imaging in this area. Overall, continued research and development in CT-based imaging for hemodynamic measurements are necessary to improve the accuracy and reliability of this technique for clinical use. Nonetheless, the authors should be commended for evaluating the combined prognostic value of CT-enabled histological, volumetric, and hemodynamic markers of plaque vulnerability. Indeed, LAPV has been demonstrated to offer incremental prognostic information beyond traditional clinical and imaging risk factors in predicting future ACS [,]. Adverse plaque characteristics, such as spotty calcification and napkin ring signs, are now widely recognized as features of high-risk atherosclerotic plaque [,]. However, to fully realize the prognostic potential of radiomics, larger studies with more comprehensive datasets that leverage imaging-based machine learning algorithms and incorporate additional markers of plaque vulnerability such as pericoronary adipose tissue attenuation and axial plaque stress in their prediction models, should be undertaken (Table 1) [,]. These advancements may enable us to tap into the remarkable potential of cardiac CTA for early identification of high-risk patients, allowing for targeted intervention.

In conclusion, the combination of CT-radiomics and computational imaging offers a more comprehensive evaluation of plaque characteristics by analyzing both morphological and hemodynamic features. This approach has the potential to identify high-risk patients who may benefit from proactive treatment. With ongoing research and development, CT-radiomics, in conjunction with computational imaging techniques, is likely to become an essential tool for managing stable CAD patients. Although the study by Yang and colleagues is a positive development, larger studies are necessary to distinguish the true signal from the noise in this promising field of cardiovascular medicine.