Enterprise AI Analysis

The Transformative Potential of AI-Supported Tools

AI-supported tools are revolutionizing medical image analysis, offering unprecedented opportunities to improve patient care and accelerate scientific discovery. These tools provide reliable and fast quantification and characterization of atherosclerotic plaque, significantly boosting efficiency and precision for clinicians and researchers.

Automated systems using advanced deep learning methods can identify and label coronary branches, segment vessel borders, and subclassify atherosclerotic plaque based on attenuation values or other characteristics. These AI-powered analyses have demonstrated high accuracy and agreement when compared with intravascular ultrasonography (IVUS), though large-scale head-to-head comparisons between different AI tools are still pending.

Crucially, AI allows for the generation of age-adjusted and gender-adjusted percentile curves for atherosclerosis imaging biomarkers like total plaque volume (TPV), enabling highly individualized risk stratification and tailored treatment strategies.

Shifting Paradigms in Atherosclerosis Management

Traditional approaches to coronary atherosclerosis, such as "treat to target" LDL-C reduction, risk-factor-based strategies, or a "fire-and-forget" model, have contributed to global mortality reduction but have limitations. They often fail to address residual cardiovascular risk, overlook plaque burden and morphology, and may result in overtreatment or undertreatment due to static, fixed-dose therapies.

This Consensus Statement proposes a paradigm shift towards integrating individualized treatment recommendations informed by CCTA-derived, AI-supported evaluation of atherosclerotic plaque. This new approach leverages percentile-based risk stratification and patient-tailored treatment, combining the strengths of current strategies while mitigating their weaknesses.

Consensus Generation Methodology

To develop explicit, evidence-based treatment recommendations based on population percentiles of AI-supported atherosclerotic plaque volume, the Quantitative Cardiovascular Imaging (QCI) Study Group assembled 35 experts from diverse fields, including cardiologists, radiologists, computer scientists, biomedical engineers, general practitioners, and epidemiologists.

The consensus was achieved through a rigorous three-round Delphi process, conducted after the third international QCI Study Group meeting in September 2024. Participants responded to a set of 12 questions (including binary, ranking, and quantitative questions) designed to derive expert opinion on the feasibility and best way to translate CCTA plaque imaging into medical treatment recommendations. This iterative process facilitated the exchange and convergence of expert opinions to reach a collective understanding and consensus.

Key Biomarkers & Clinical Parameters for Individualized Treatment

CCTA commonly differentiates between calcified and non-calcified plaque (NCP) components, each with unique clinical implications. While calcified plaque (CPV) is widely used for risk stratification, particularly in asymptomatic patients, its utility for individual treatment recommendations is limited because it often overestimates risk in the elderly, and its volume can increase with lipid-lowering therapies despite reduced MACE risk. Vulnerable plaques are primarily inflammatory and non-calcified.

Non-calcified plaque (NCPV) is a critical component for prognosis, especially low-attenuation NCP. It is a valuable prognostic marker and can indicate metabolically active CAD, particularly in younger patients. However, its role in individual treatment recommendations is still under investigation, and it is considered the plaque subtype best suited for monitoring treatment response.

Total Plaque Volume (TPV), which combines CPV and NCPV, is identified as the most meaningful biomarker for initiating or escalating medical treatment due to its superior sensitivity and prognostic value. However, TPV alone is not ideal for monitoring drug response as it doesn't differentiate between changes in NCPV and CPV.

High-risk plaque features (low-attenuation plaque, napkin-ring sign, positive remodeling, spotty calcification) are strong risk modulators. While their individual predictive power is high, their low prevalence and assessment reliability limit their standalone use for guiding treatment initiation, making them better suited for adjusting existing treatment intensity.

AI in Atherosclerosis Imaging: Applications and Reliability

AI and ML are transforming cardiovascular risk assessment by enabling automated quantification of imaging biomarkers like CAC scoring, CAD-RADS scoring, and various plaque types. Fully automated deep learning methods accelerate quantification and save reader time, demonstrating concordance with IVUS and expert human readers.

The reliability of AI tools for atherosclerotic plaque quantification is excellent, showing high agreement with IVUS (TPV r=0.91, NCPV r=0.87, CPV r=0.91) and good interscan reliability. However, limitations exist, including wide limits of agreement between scans/rescans and considerable variability across different software vendors, which impacts reclassification of patients near risk thresholds.

Individualized Recommendations for Medical Treatment

The QCI Study Group recommends a treatment paradigm based on age-adjusted and gender-adjusted percentile curves of Total Plaque Volume (TPV). The presence of any atherosclerotic plaque detected by CCTA should lead to a recommendation of pharmacological treatment (standard intensity).

For escalation to high-intensity treatment, the consensus recommends the 70th percentile of total plaque volume (TPV) adjusted for age and gender as the threshold. Additionally, clinical risk factors such as smoking, positive family history of CVD, and high-risk plaque features (low-attenuation plaque, positive remodeling) serve as crucial modulators for escalating treatment intensity.

Pharmacological management primarily involves statins, with options for other lipid-lowering agents (ezetimibe, PCSK9 inhibitors) for very high-risk patients or those unable to achieve LDL-C goals. Lifestyle modifications and management of other risk factors (blood pressure, glucose) remain essential.

Limitations and Future Directions

While a consensus on treatment percentiles was reached, the role of Non-Calcified Plaque (NCP) in individualized treatment recommendations and monitoring treatment response requires further determination through future randomized trials. Defining an optimal treatment threshold for small atherosclerotic plaques also remains a challenge.

Future advancements in AI technologies and CT image reconstruction, including photon-counting CT, are expected to significantly enhance the imaging and characterization of atherosclerotic plaque. These innovations promise to unlock the potential for even more precise individualized treatment, but will necessitate robust standardization efforts for CT acquisition and analysis, and further development of AI methods to handle large, high-dimensional data.

Concluding the Future of AI in Cardiovascular Imaging

AI-supported tools provide reliable and accurate assessments of atherosclerotic plaque burden and morphology, significantly refining risk stratification and clinical decision-making. The QCI Study Group identified age-adjusted and gender-adjusted percentiles of Total Plaque Volume (TPV) as the most meaningful parameter for initiating and escalating medical treatment.

Specifically, the presence of any atherosclerotic plaque should trigger pharmacological treatment, with the 70th percentile of TPV warranting high-intensity treatment. Clinical risk factors and high-risk plaque features modulate these decisions. This paradigm shift, integrating advanced imaging biomarkers and AI-driven analysis, offers a path towards truly individualized cardiovascular care. Future large-scale randomized controlled trials are essential to validate these recommendations and ultimately improve patient outcomes globally.