For improving cardiovascular health outcomes, the combination of AI with various types of CT exams is generating evidence of their potential usefulness for radiologists, cardiologists, and patients.
With help from AI, CT imaging techniques such as coronary CT angiography (CCTA), stress myocardial CT perfusion imaging, and CT-derived fractional flow reserve (FFR) analysis can play a key role in reducing the use of invasive coronary angiography, helping to prevent major adverse cardiovascular events (MACE), and influencing therapy selection for people at high risk of cardiovascular complications, according to heart and cardiovascular medicine researchers.
Certain applications are progressing faster than others, however.
For improving cardiovascular health outcomes, the combination of AI with various types of CT exams is generating evidence of their potential usefulness for radiologists, cardiologists, and patients.
With help from AI, CT imaging techniques such as coronary CT angiography (CCTA), stress myocardial CT perfusion imaging, and CT-derived fractional flow reserve (FFR) analysis can play a key role in reducing the use of invasive coronary angiography, helping to prevent major adverse cardiovascular events (MACE), and influencing therapy selection for people at high risk of cardiovascular complications, according to heart and cardiovascular medicine researchers.
Certain applications are progressing faster than others, however.
Uncovering MACE risks
With many factors underlying MACE, radiologists and data scientists have been mining various regions of interest on cardiac CT images to bring potentially useful predictive and prognostic tools to the forefront. Researchers from Cedars-Sinai Medical Center in Los Angeles are among those monitoring the clinical and investigational role of AI in advanced cardiac CT imaging.
Damini Dey, PhD, a scientist and professor at Cedars-Sinai, has developed and validated AI-assisted coronary plaque quantification on CCTA exams, as noted in a Cedars-Sinai news feature. Automated coronary plaque quantification would shorten the process from at least 30 minutes to just five or six seconds, according to Dey's research.
In addition, Cedars-Sinai is studying automated epicardial adipose tissue volume quantification to predict MACE, and they are demonstrating how opportunistic screening as part of routine chest CT scans can identify patients at high risk.
A study designed and implemented by investigators at Cedars-Sinai found that AI can accurately evaluate cardiovascular risk during routine noncontrast-enhanced chest CT scans. More than 15 million CT exams are performed in the U.S. each year -- and many of these are underutilized or understudied, according to the experts at Cedars-Sinai.
Two AI models evaluated data on coronary calcium and heart muscle chamber sizes from nearly 30,000 patient imaging records. The researchers were able to determine that those measures are a better indicator of cardiac risk than a radiologist’s identification of abnormalities.
“These results are likely practice-changing for many patients because this technology can accurately identify cardiovascular risk without the use of invasive tests or contrast dye that some patients cannot receive,” said Piotr Slomka, PhD, for a Cedars-Sinai blog published April 23. Slomka is director of Innovation in Imaging at Cedars-Sinai, a professor of medicine in the Division of Artificial Intelligence in Medicine, and senior author of the study.
Dey, who is the director of the quantitative imaging analysis program at the Biomedical Imaging Research Institute, added for AuntMinnie.com, "AI is pervasive. It's used in image acquisition, reconstruction, improvement of image quality, and then segmentation and quantification, and this is where there's the most efficient application."
"But obviously there's a need for validation and testing in your own data," she said.
Changes in fat tissue
AI technology can detect the level of inflammation of the heart arteries by spotting changes in the fat tissue around arteries that are not visible to the human eye, according to Kenneth Chan, a clinical research fellow at the University of Oxford's Acute Multidisciplinary Imaging and Interventional Centre.
A British Heart Foundation-funded study published May 29 in The Lancet co-authored by Chan used AI to analyze CCTA scans and clinical data from over 40,000 people at eight hospitals in the U.K.
"Many thousands of patients who undergo a cardiac CT scan do not have significant narrowings of the heart arteries," added Prof. Charalambos Antoniades, British Heart Foundation chair of cardiovascular medicine at the University of Oxford and lead author of The Lancet article. "These scans are often stored in hospital radiology systems for years without further use, but we could extract more information from these scans to guide preventative treatments."
The Oxford Risk Factors and Non-invasive imaging (ORFAN) pilot study in four National Health Service (NHS) hospitals involved integrating AI technology, specifically, an AI-enhanced cardiac risk prediction algorithm, which integrates Fat Attenuation Index (FAI) score, coronary plaque metrics, and clinical risk factors. Researchers provided the resulting risk scores to doctors for 744 patients. Among the takeaways: In up to 45% of cases, doctors altered their patients' treatment plans as a result.
The FAI score is one way to capture inflammatory risk beyond the current clinical risk stratification and CCTA interpretation, particularly among patients without obstructive [coronary artery disease], according to Chan, Antoniades, and colleagues. Higher FAI may be linked to a higher inflammatory burden, and such prognostic algorithms could be used as an alternative to traditional risk factor-based risk calculators, the researchers noted.
Inflammation in autoimmune disorders
Similarly, researchers such as Brittany Weber, MD, PhD, of Brigham and Women's Hospital in Boston are also assessing the use of AI in detecting and measuring inflammation in patients with autoimmune disorders.
Medical conditions such as autoimmune disease are among the common causes of heart inflammation which can occur suddenly, or progress slowly, and may have severe symptoms or almost no symptoms, according to the National Heart, Lung, and Blood Institute (NHLBI).
An associate physician in cardiovascular medicine and cardiovascular imaging at Brigham and Women’s Hospital in Boston, Weber has penned much on the emerging specialty of cardio-rheumatology and, specifically, on the "cardinal signs" of inflammation.
"In the cardiovascular system, the impact of inflammation ranges broadly from the initiation or promotion of atherosclerosis to vasculitis, myocarditis, pericarditis, and even heart failure," Weber noted for a 2023 editorial in the Journal of the American Heart Association.
Middle-aged women are often affected by heart inflammation, according to Weber, and CT imaging plus AI can play a role in both measuring the signs of potential cardiovascular complications and evaluating the impact of new therapies. An AI-calculated FAI score can detect and measure vascular inflammation on CCTA exams.
Systemic inflammatory syndromes contribute to unique elevated atherosclerotic risk and incident cardiovascular disease, according to Weber. In addition, heart vascular inflammation can affect the lining of the heart or valves, the heart muscle, or the tissue around the heart.
"Equally important to treating the underlying inflammation process (e.g., recurrent pericarditis, myocarditis, vasculitis) is the recognition of the increased risk of cardiovascular disease, particularly atherosclerotic disease, among patients with systemic inflammatory disorders," Weber and colleagues added. And, "as this field continues to grow, it could become a separate program or at least allow the programmatic structure for complex medical decision-making across multiple different specialties with a footprint in cardiology and rheumatology. It is important for the cardiologist to recognize that our standard risk calculators do not adequately capture risk in these individuals. Improved risk stratification is needed and is the ongoing area of multiple research efforts and an area where cardiovascular imaging tools may help to advance the field."
According to Weber et al, the expanding role of cardiovascular imaging biomarkers as a surrogate of cardiovascular risk and outcomes is well-suited. Her research paper, The Frequency, Prevalence, And Outcomes Of Incidentally Detected Coronary Artery Calcium Using Artificial Intelligence Analysis Among Patients With Immune Mediated Inflammatory Disease, has been named a finalist for the Young Investigator Awards at the 2024 Society of Cardiovascular CT (SCCT) annual scientific meeting in July.
AI adoption patterns
When an AI tool is cleared by the U.S. Food and Drug Administration (FDA), that product then undergoes further validation and clinical studies or trials, and physicians can reach a consensus on what can be reported from AI-enabled imaging results, Dey of Cedars-Sinai told AuntMinnie, adding that confidence of the end user is key to the adoption of AI tools in cardiac CT.
An analysis of AI billing and coding trends published in September 2023 in NEJM AI offered an early glimpse into what type of cardiac CT AI tools physicians have been billing for. As of June 2023, AI devices for cardiac CT applications were used in two of the top three highest number of Current Procedural Terminology (CPT) claims, according to Kevin Wu, PhD, and Eric Wu, PhD, of Stanford University and colleagues who used U.S. insurance claims data to study the clinical adoption of medical AI devices in the U.S.
However, the researchers reported that overall utilization of medical AI products was still limited and focused on a few leading procedures. The most used product at that time was a Class 2 AI-based medical device software that supports the risk assessment and functional evaluation of coronary artery disease and relies on CCTA with FFR-CT calculations. The product provides anatomic data, plaque identification, and characterization, as well as calculations of FFR-CT, coronary physiological simulation computed from simulated pressure, velocity, and blood flow information obtained from a 3D computer model generated from static coronary CT images.
In a breakdown, Wu and colleagues reported that the coronary artery disease FFR-CT product (reimbursed at around $1,000 per procedure) topped the chart for the most claims at 67,306, followed by a coronary atherosclerosis tool (reimbursed around $700) at only 4,450 claims -- fractions of total billings but still representing cardiac CT AI products to watch.
Regardless, Wu and colleagues noted that the commercialization of FDA-approved AI products is still nascent but growing.
The researchers also observed that the presence of academic medical centers is a significant factor in the adoption of medical AI, and they acknowledged that the addition of AI may require significant changes to the clinical workflow as many others have also pointed out.
A 'one-stop shop'
There has been intense attention on AI for its potential to improve cardiovascular disease prevention, detection, diagnosis, and treatment, according to the American Heart Association, which in February called for the development of best practices to advance the use of AI in cardiovascular care.
Toward that end, CT imaging is becoming a “one-stop shop" for imaging of the vasculature, with CCTA the noninvasive imaging modality of choice for the noninvasive examination of the coronary arteries, noted Antoniades in 2021.
Work at Brigham and Women's Hospital, Cedars-Sinai, NHS hospitals, and elsewhere helps raise awareness and increase understanding of the emerging roles of AI-derived measures for prediction, prognostication, and monitoring of underlying cardiovascular risks -- in cardiac CT and other imaging modalities as well. Eventually, these AI tools could reduce hospital length of stay and improve quality of life and health for people at high risk for cardiovascular events.