AI Meets CT Imaging: The Future of Atherosclerosis Diagnosis

Atherosclerosis is one of the leading causes of cardiovascular disease (CVD), which is also one of the reasons for the vast health issues among the elderly. This condition occurs when plaques made up of fatty substances, cholesterol, and other materials build up within the walls of arteries, resulting in restricted blood flow and increased risk of heart attacks and strokes. The traditional methods for detecting atherosclerosis have their limitations when assessing the rupturing of plaques, which causes sudden cardiac problems.

Recent technological advances in imaging, especially in computed tomography (CT), have made it easier to pick on Coronary Artery Disease (CAD). Moreover, the rise of artificial intelligence (AI) is transforming the way medical professionals analyze these CT scans. By using AI, especially deep learning models that are trained to analyze visuals, doctors can gain a clearer and more accurate picture of atherosclerotic plaques. This blog explores how AI is enhancing the detection, characterization, and risk prediction of atherosclerosis, and how these advancements may shape the future of cardiovascular healthcare.

Atherosclerosis and Coronary Artery Disease

Atherosclerosis develops when Low-Density Lipoprotein (LDL) cholesterol builds up in the inner lining of the arterial walls. Over time, this accumulation triggers an inflammatory response that leads to the formation of plaques. These plaques are made up of cholesterol, immune cells, and smooth muscle cells, and can gradually thicken the artery walls, causing narrowing or stenosis. However, it's not the narrowing of arteries alone that poses the biggest threat, rather it's the instability of certain plaques, which can rupture and cause blood clots, blocking blood flow to the heart or brain.

There are different types of plaques, including calcified plaques, non-calcified plaques, and low-attenuation plaques, which vary in their risk for rupture. Despite coronary angiography being a commonly used tool for diagnosing CAD, it primarily focuses on detecting the degree of narrowing (stenosis), which often misses the more dangerous non-calcified plaques. AI-based imaging technologies, such as CT angiography, can provide a more detailed and comprehensive way to assess both plaque size and its characteristics, which may offer a solution to overcome such limitations.

Current Imaging Techniques for Atherosclerosis

Coronary CT Angiography (CCTA). This imaging technique gives detailed pictures of the arteries, making it easy for physicians to identify blockages, measure the severity of stenosis, and estimate Coronary Artery Calcium (CAC) scores. The CAC score, depending on the CT scans, evaluates the amount of calcified plaque in the arteries, which in turn helps the medical professionals to span a patient’s risk for heart disease. But this method has its drawbacks. The CAC score only measures calcified plaques, which can be misleading because statin treatments may stabilize plaques by increasing calcification, without reducing the risk of plaque rupture.

On the other hand, CCTA can differentiate between calcified and non-calcified plaques, but it still has limitations. The earlier approach focuses mostly on stenosis severity, ignoring high-risk plaque features such as lipid-rich plaques or positive remodeling. These plaques are more dangerous as they are prone to rupture even without causing maximum narrowing of the artery. New imaging techniques, especially those aided by AI are designed to fill this gap by focusing on the plaque size, its type, composition, and risk to rupture.

Figure 1: Plaque Microstructures

AI in Cardiovascular Imaging: A Closer Look

Artificial Intelligence (AI) has emerged as a powerful tool in medical imaging, enabling more accurate and efficient interpretation of complex images. AI can be broadly divided into machine learning (ML) and deep learning (DL), with the latter being a more advanced subset that mimics human brain function. Deep learning algorithms, such as Convolutional Neural Networks (CNNs), are particularly well-trained for analyzing medical images, as they can detect patterns and features that are often difficult for human eyes to identify.

For cardiovascular imaging, AI tools can segment CT images, identify, and describe anatomical structures like the coronary arteries, plaques, and surrounding tissues. This mechanism speeds up the process and also ensures greater consistency and accuracy. They are also known to classify plaques according to their composition and type, which can be helpful for risk prediction.

One of the advantages of AI is its ability to analyze large datasets. Unlike traditional methods that rely on the manual interpretation of a small set of images, AI models can process vast amounts of visual data, making it possible to predict plaque related factors including location and inflammation.

AI in Coronary Plaque Characterization

Coronary plaque classification is important because different types of plaques pose varying levels of risk. For example, calcified plaques tend to be more stable and less likely to rupture, while non-calcified plaques, particularly low-attenuation plaques, are prone to rupture and cause heart attacks.

AI-based models have been shown to excel in differentiating between these various plaque types, using image features that may not be immediately obvious. For example, AI can identify the "napkin-ring sign," a hallmark of high-risk plaques, or "positive remodeling," which indicates that a plaque is growing outward rather than narrowing the artery. The AI systems can classify these plaques in a way that traditional techniques cannot.This precise detection allows doctors to more accurately assess a patient's risk for acute coronary syndrome (ACS). AI tools can also predict the progression of plaques over time, which helps in determining which patients may benefit from early intervention, such as statin therapy or other procedures.

AI in Assessing Coronary Inflammation

Coronary inflammation is a key symptom of atherosclerosis and plaque rupture. The inflammatory response triggered by LDL accumulation leads to changes in the Perivascular Adipose Tissue (PVAT) around the coronary arteries. AI has made it possible to assess this inflammation using CT imaging. One key factor is the Fat Attenuation Index (FAI), which measures the density of PVAT and can diagnose the inflammatory processes in the arteries.

Studies have shown that elevated FAI values are related with increased risk for cardiovascular diseases. AI can automatically calculate the FAI from CCTA images, offering a valuable tool for identifying coronary inflammation early, even before a significant plaque buildup occurs. This provides a more personalized approach to CVD risk and can help with more targeted interventions. For example, anti-inflammatory treatments, such as statins or biological therapies, may be advised based on biomarkers like FAI.

AI for Risk Prediction: The Future of Cardiovascular Health

AI models can integrate imaging biomarkers, such as plaque type and FAI, with traditional risk factors like age, sex, hypertension, and cholesterol levels. One example of this integration is the CaRi-Heart Risk score, a model that combines imaging data with clinical risk factors to predict a patient's eight-year risk of fatal cardiac events.

By incorporating these AI-driven insights into patient assessments, doctors can move to a more personalized treatment approach. Large international cohorts, such as the Oxford Risk Factor and Non-invasive Imaging (ORFAN) study, are helping to validate these AI models against real-world clinical outcomes. Over time, AI might even predict plaque vulnerability and determine the best course of treatment for each patient.

Figure 2: Flow Diagram illustrating how AI based Diagnosis fits into the larger workflow

Challenges and Limitations

Despite the promising developments, there are several challenges to the widespread adoption of AI in cardiovascular imaging. One major issue is the lack of standardization in plaque definitions and segmentation methods. Different research studies may use varying reference points to classify plaques, making it difficult to compare results across studies and implement a standard clinical approach.

Secondly, AI models are still often viewed as “black boxes,” meaning it’s difficult for doctors to understand how the AI arrives at its conclusions. For AI models to gain the trust of clinicians, they must be transparent, interpretable, and backed by clinical evidence.

Lastly, there are concerns regarding the quality of data used to train such models. CT images can vary widely in quality depending on factors such as scan resolution and slice thickness. AI models trained on inconsistent data may produce unreliable results, limiting clinical adoption.

Future Implications

AI has the potential to completely transform the way we detect, analyze, and treat atherosclerosis. By improving the accuracy of coronary plaque characterization and providing new insights into coronary inflammation, AI can offer a more reliable and personalized approach to cardiovascular risk assessment. While challenges remain in standardizing imaging methods and ensuring data quality, the continued development of AI in cardiovascular imaging is seen to help in the prevention and management of atherosclerosis.

As AI-driven tools are adopted to clinical practice, they will enable doctors to detect high-risk plaques earlier, predict cardiovascular events more accurately, and provide better treatment plans for patients.