Computed Tomography Advancements in Plaque Analysis: From Histology to Comprehensive Plaque Burden Assessment

ABSTRACT

Advancements in coronary computed tomography angiography (CCTA) facilitated the transition from traditional histological approaches to comprehensive plaque burden assessment. Recent updates in the European Society of Cardiology (ESC) guidelines emphasize CCTA's role in managing chronic coronary syndrome by enabling detailed monitoring of atherosclerotic plaque progression. Limitations of conventional CCTA, such as spatial resolution challenges in accurately characterizing plaque components like thin‐cap fibroatheromas and necrotic lipid‐rich cores, are addressed with photon‐counting detector CT (PCD‐CT) technology. PCD‐CT offers enhanced spatial resolution and spectral imaging, improving the detection and characterization of high‐risk plaque features while reducing artifacts. The integration of artificial intelligence (AI) in plaque analysis enhances diagnostic accuracy through automated plaque characterization and radiomics. These technological advancements support a comprehensive approach to plaque assessment, incorporating hemodynamic evaluations, morphological metrics, and AI‐driven analysis, thereby enabling personalized patient care and improved prediction of acute clinical events.

CCTA in coronary disease: from morphological features to advanced imaging biomarkers redefining non‐invasive imaging through comprehensive plaque burden assessment, enabling personalized therapeutic strategies

Abbreviations

ACS

acute coronary syndrome

AI

artificial intelligence

CAC

coronary artery calcium

CAD

coronary artery disease

CCTA

coronary computed tomography angiography

CNR

contrast to noise ratio

DECT

dual‐energy CT

ESC

European Society of Cardiology

FFR‐CT

fractional flow reserve—CT

HR

hazard ratio

IVUS

intravascular ultrasound

MACE

major adverse cardiac events

OCT

optical coherence tomography

PCD‐CT

photon‐counting detector CT

UHR

ultra‐high resolution

VMI

virtual monoenergetic images

VNC

virtual non‐contrast

WSS

wall shear stress

1. History of Plaque Analysis

Atherosclerosis begins with the accumulation of cholesterol‐rich lipids in the arterial wall and progresses through inflammation, exhibiting increasingly complex patterns of evolution and varied manifestations among individuals [1, 2]. Factors such as intra‐plaque hemorrhage, neovascularization, and recurrent cycles of plaque healing and rupture can contribute to the development of acute coronary syndromes (ACS) [3]. The ability of Coronary Computed Tomography Angiography (CCTA) to assess and predict the rupture or erosion of atherosclerotic plaques, referred to as “virtual histology,” has been validated against histopathology as well as invasive imaging modalities [3, 4]. Most ACS result from the sudden thrombosis of plaques with distinct morphological features [4]. Among these, thin‐cap fibroatheroma and necrotic lipid‐rich core lesions have been demonstrated to be the precursors of plaque rupture [5]. Therefore, discrimination between lipid‐rich and fibrous plaque components is central for ACS prevention [6].

However, CCTA characterization of coronary atheromas can be affected by the limited spatial resolution of conventional CT [7], as well as several technical factors, including plaque size, image noise, contrast and spatial resolution, and reconstruction parameters [8, 9, 10, 11, 12, 13].

Moreover, around two‐thirds of ruptured plaques in patients experiencing sudden cardiac death contain microcalcifications, which are difficult to detect solely by CCTA, especially when they are small (<1 mm), due to spatial resolution issues [14]. Along with spatial and contrast resolution related drawbacks, several technical acquisition and reconstruction aspects may influence the assessment of coronary plaques using CCTA [15]. Coronary lumen attenuation measured by CCTA significantly influences the measured plaque attenuation, with higher lumen attenuation leading to higher plaque attenuation [16]. Moreover, partial volume effects and interpolation can alter attenuation profiles within the soft tissue range. This is particularly critical when calcifications in the plaque and contrast material in the lumen are adjacent to the region of measurement, making absolute plaque attenuation values from different studies difficult to compare. Moreover, attenuation values in CCTA have been largely demonstrated to depend on kernels and iterative reconstruction level, leading to inter and intra‐subject variability and making it challenging to define absolute attenuation ranges corresponding to specific plaque characteristics [9, 10].

1.1. CT Advances in Plaque Analysis

To overcome the limitations of conventional CT scanners in stenosis assessment and plaque composition analysis, advancements have progressed in two main directions: improving spatial resolution and enhancing spectral analysis.

1.2. Spatial Resolution Improvements in Plaque Characterization

Photon‐counting detector CT (PCD‐CT) represents an advanced technology offering significantly enhanced spatial resolution by directly converting x‐ray photons into electrical signals using semiconductor materials like CdTe or CZT [17]. Additional benefits include improved geometric dose efficiency and a higher contrast‐to‐noise ratio (CNR) for lower‐energy x‐rays, which is especially advantageous in CCTA imaging [18]. Improved spatial resolution and CNR with PCD‐CT have been proven to provide additional benefits in coronary arteries luminal assessment [19], especially when dealing with stenosis quantification in calcified plaques, leading to a significant reduction in blooming and partial volume artifacts and a more accurate stenosis assessment [20, 21]. When compared with conventional CCTA, ultra‐high resolution (UHR) PCD‐CT results in more accurate stenosis measurements in calcific coronary plaques and leads to a substantial reclassification rate in around 50% of patients [22] (Figure 1). Beyond improving the precision of stenosis measurements, UHR PCD‐CT significantly enhances advanced plaque analysis: a recent study showed that UHR acquisition leads to a reduction in total plaque volume in approximately one‐third of patients compared with EID‐CT (total plaque volume: 723.5 mm³ [IQR, 500.6–1184.5 mm³] vs. 1084.7 mm³ [IQR, 710.7–1609.8 mm³], respectively; p < 0.001), while also improving intra‐ and inter‐reader reproducibility [23].

FIGURE 1.

A case of high calcium load over the right coronary artery (RCA) in 3D reconstruction (left image). CT image of the RCA acquired with ultra‐high resolution (UHR) mode Photon counting detector (PCD‐CT) (center image) and with energy integrating detector (EID)‐CT (right image), underscoring the added value of PCD‐CT in stenosis grading in highly calcific plaques. Ao root indicates aortic root; Asc Ao, ascending aorta; D, diagonal branch artery; LAD, left anterior descending artery; LCx, left circumflex artery; MO, marginal branch artery; RCA, right coronary artery.

1.3. Spectral Analysis for Compositional Assessment

Several studies have explored the added value of dual‐energy CT (DECT) in the diagnostic performance of CCTA in the evaluation of coronary artery disease (CAD) [24]. Symons et al. [25] showed that low energy noise‐optimized Virtual Monoenergetic Images (VMI) 40–70 keV can improve the CNR of both calcified and non‐calcified coronary plaques when compared with conventional images. Different VMI levels can be used to improve image quality and stenosis assessment depending on plaque composition [26].

PCD‐CT technology provides intrinsic spectral information by generating signal pulses proportional to the absorbed x‐ray energy. This innovation offers an additional approach to overcome artifacts from highly calcified plaques [27]. A novel calcium removal reconstruction algorithm (PureLumen) has been demonstrated to be able to eliminate only the calcified components of coronary plaques, to overcome the risk of stenosis overestimation due to blooming artifacts [28]. PCD‐CT offers the possibility to directly and accurately identify high‐risk plaque components, like thin‐cap fibroatheroma and spotty calcifications, which are still a challenge with conventional CT technology [29]. In an in vitro study, PCD‐CT provided superior detectability for both 0.5 mm thick non‐calcific (AUC = 95% vs. 75%) and lipid‐rich plaques (AUC = 85% vs. 77.5%) in comparison with conventional energy integrating detectors [30]. Thanks to spectral imaging, PCD‐CT can improve plaque composition characterization via material decomposition algorithms; differences in contrast agent concentration and their spectrum have been demonstrated to be able to identify different plaque components [31]. The accuracy of PCD‐CT to quantify vulnerable plaque features has been proved ex vivo, comparing CT features with histology: no significant differences were found between histological results and PCD‐CT measurements [32, 33]. An example of plaque analysis is shown in Figure  2 .

FIGURE 2.

A 50‐year‐old male patient with arterial hypertension referring to the cardiologist for an episode of retrosternal pain during jogging. A mainly calcific plaque determining minimal vessel stenosis is shown in mid‐left anterior descending artery (left image, red arrow). Advanced plaque analysis is performed (central image, longitudinal, and axial view) with inner and outer vessel wall identification and plaque segmentation. Plaque analysis results are shown (right image), showing a total plaque volume of 8 mm³ and confirming mainly calcific plaque (dense calcium volume = 5.5 mm³, 67% of total plaque volume). Ao root indicates aortic root; LAD, left anterior descending artery.

CAC scoring remains a key tool in cardiovascular risk stratification, with a class‐IIa recommendation in patients with a borderline/intermediate risk [34]. Both DECT and PCD‐CT allow for CAC score quantification on virtual non‐contrast (VNC) images, reducing radiation exposure, with PCD‐CT offering a substantial increase in spatial resolution and accurate results [35, 36].

1.4. Shift Paradigm: From Histology to “Plaque Burden”

The so‐called “CCTA high‐risk plaque features” include positive vessel remodeling, the napkin ring sign, the presence of low‐attenuation plaques, and spotty calcification. All these four features detection with CCTA have been validated against optimal computed tomography (OCT) and have been associated with an increased risk of major adverse cardiac events (MACE) [35, 37, 38]. However, in the acute setting, their predictive value for ACS remains controversial, with studies yielding mixed results [39, 40].

The EMERALD trial emphasized the role of the hemodynamic assessment of the lesions and identified the best prediction model for future ACS culprit lesions as a combination of stenosis severity, adverse plaque characteristics, and hemodynamic parameters [41]. Moreover, Bon‐Kwon Koo et al. employed an AI‐based quantitative analysis (AI‐QPCHA) to enhance the ability of CCTA to identify culprit lesions of ACS: in this patient cohort, hemodynamic parameters outperformed all other plaque characteristics [42]. Indeed, the concept of “plaque vulnerability” extends beyond morphological risk features, with hemodynamic parameters such as wall shear stress (WSS) and Fractional Flow reserve (FFR)‐CT playing a central role in triggering events [43, 44]. In a recent sub‐analysis of the SCOT‐HEART trial, the all‐vessels low‐attenuation plaque burden was the strongest predictor of outcome (hazard ratio [HR], 1.60 [95% CI, 1.10–2.34] per doubling; p = 0.014), in comparison with the degree of the stenosis and cardiovascular risk scores [45]. These findings have further fueled the growing interest in overall plaque burden and its association with an increased risk of MACE. Building on this, Tzimas et al. analyzed over 10 000 CCTA scans using an AI‐based tool to create gender‐ and age‐stratified percentile nomograms for atherosclerotic plaque volumes, with the aim of improving the integration of CCTA findings into clinical decision‐making [46] (Figure  3 )

FIGURE 3.

Artificial intelligence‐enabled quantitative coronary plaque analysis showing total plaque burden of 333 mm³ in a 63‐year‐old male patient with 76% non‐calcified atherosclerosis. According to the age‐ and sex‐stratified population percentile nomograms for atherosclerotic plaque at the 54° percentile (Heart Flow, Mountain View, CA, USA). LAD indicates left anterior descending artery; LCx, left circumflex artery; RCA, right coronary artery.

1.5. Artificial Intelligence (AI) in Advanced Coronary Plaque Analysis

The rapid advancements in AI across all radiological fields hold significant potential to revolutionize advanced coronary plaque quantification and analysis [47]. AI application has already been extensively explored to facilitate the Coronary Artery Calcium (CAC) Score, a robust imaging‐based instrument for cardiovascular risk stratification and treatment [48]. Different machine‐learning and deep‐learning models have been proposed for automated CAC score quantification, showing excellent correlation with semi‐automated software currently used in clinical practice [49, 50, 51]. Different AI models have been proposed to speed and facilitate plaque characterization. AI models have been developed for plaque volume quantification and discrimination of calcified and non‐calcified components, demonstrating strong agreement with intravascular ultrasound (IVUS) [52]. Moreover, a promising automated function to evaluate low‐attenuation plaque excluding voxels adjacent to the outer vessel wall of coronaries has been proposed and compared to IVUS, showing an improvement in lipid‐rich component discrimination, further enhancing cardiovascular risk stratification [53].

Various approaches have also been explored to enhance the extraction of plaque‐related information from CCTA and radiomics has emerged as a supplementary technique to expand the characterization of abnormalities using radiological images [54]. Park et al. [55] developed a new deep‐learning model to accurately detect plaque erosions on CCTA images, allowing for aggressive patient‐tailored therapy to spare invasive procedures. Integration of qualitative plaque features with functional information about stenosis assessment is nowadays possible thanks to AI models: Ihdayhid et al. developed an automated unsupervised deep learning model to evaluate stenosis severity as well as high‐risk plaque features with good diagnostic performance based on CCTA data, further enhancing its clinical translation [56].

Radiomics may furthermore help in overcoming the intrinsic CT limitations for the identification of vulnerable plaques. Kolossvary et al. compared the diagnostic performance of a radiomics‐based machine learning model with visual and histogram‐based assessment ex vivo and found machine learning model outperformed visual assessment (AUC = 0.73 vs. 0.65, 95, p = 0.04) [57]. On the basis of these promising results, AI can be used to integrate extensive quantitative imaging data with clinical parameters to enhance prediction and diagnosis. Tesche et al. reported that integrating coronary CTA plaque features with clinical data into a machine‐learning model enhances risk stratification for MACE (AUC 0.96) compared to traditional methods [58]. Furthermore, in the SCOT‐HEART trial, AI models incorporating radiomics features predicted cardiac risk based on coronary perivascular adipose tissue in patients with MACE within 5 years and matched controls [59].

2. Conclusions

CCTA, with its high negative predictive value and ability to identify and monitor histopathological plaque features over time, has strengthened its position as the first‐line diagnostic modality for CAD. Advances in technology are set to enhance its role even further, enabling the integration of hemodynamic evaluations with stenosis assessments and morphological metrics. This evolution toward a “comprehensive plaque assessment” holds the potential to develop personalized scores for predicting acute clinical events, paving the way for more precise and individualized patient care.

Conflicts of Interest

The authors declare no conflicts of interest.