Heart Smart: AI in the Cardiovascular Sector

“Irony is Fate’s most common figure of speech.”
      —Trevanian (Shibumi)

Life is full of ironies but perhaps one of the greatest (and arguably the most frustrating) is wish fulfillment bad timing. Think about it: The deepest mortal desires are often only realized when they are no longer wanted or desperately needed.

Alas, irony can be cruel at times.

But it also can be kind, delivering just the right opportunity to just the right person at exactly the right time—not in answer to an abandoned passion per se, but rather to meet an unrealized need.

Sharon Bruno was the heritor of irony’s kinder, gentler side not too long ago. Working in the healthcare field, Bruno has always felt obligated to be extra vigilant about her health, though she found additional motivation through friends and relatives with undiagnosed conditions and dark medical histories.

“It’s very scary because I’ve seen so many patients and friends and neighbors that we thought were healthy, eating right and exercising,” the registered nurse said in an online video, “but all of a sudden they had a heart attack. That could be me.”

It nearly was her. A silent killer—unbeknownst to Bruno—was lurking deep within the confines of her coronary arteries. 

As a nurse, Bruno was keenly aware of her family health history and genetic predisposition to cardiac disease. Knowing her father died at 55 from a massive stroke likely triggered by high blood pressure and high cholesterol, Bruno was adamant about monitoring her coronary calcium score, an assessment of coronary artery calcium deposits.

“So knowing that family history, I always wanted to be checked to make sure I’m okay. I had my calcium score test [done] three years in a row and it was zero,” Bruno recalled in the video. “So I was happy because according to the test, at the bottom, it actually says very low risk of a cardiac event. I was like, ‘This is great. I’m good.’”

Further testing proved otherwise, however. Given access through her employer to a more technologically advanced coronary plaque screening method, Bruno learned her true calcium deposit score and cardiac event risk.

“Thanks to my father, I got the genes,” she shared. “My score was not 0…”

Far from it: An artificial intelligence (AI)-enabled analysis developed by digital healthcare firm Cleerly showed moderate atherosclerotic plaque buildup in Bruno’s coronary arteries, with 348.1 mm³ total plaque and minimal stenosis. Her official diagnosis was stage 2 moderate plaque (with stages defined as normal, mild, moderate, and severe).

Zero to 348 in the blink of an eye (or in Bruno’s case, the beat of her heart).

Bruno’s high coronary calcium score (100 or less is desirable) was indicative of significant calcium deposits within her cardiovascular system; that buildup increased her risk of heart attack and coronary artery disease (CAD). 

Similar to a standard calcium score test, Cleerly’s analysis measures heart attack risk through a coronary computed tomography (CT) scan but accurately reveals the amount and type of plaque buildup. Its proprietary machine learning algorithms non-invasively gauge atherosclerosis (plaque), stenosis, and ischemia likelihood using coronary computed tomography angiography (CCTA) studies.

Cleerly’s machine-learning AI generates a 3D model of the coronary arteries that clinicians use to identify the lumen and vessel walls, locate and measure stenoses, and quantify and categorize plaque. The firm’s ISCHEMIA algorithm uses measurements based on invasive fractional flow reserve (FFR) data to determine the likelihood of vessel-level ischemia (restricted blood supply). 

Cleerly says its coronary calcium analyses are based on more than 10 million images from over 40,000 patients gathered over a 15-year period in multi-center clinical trials. Clinical studies have shown the company’s AI-powered coronary plaque assessment is comparable to invasive gold standards, including FFR, intravascular ultrasound, intracoronary near-infrared spectroscopy imaging, optical coherence tomography, and quantitative coronary angiography.

Data from six multi-center clinical trials also have validated Cleerly’s U.S. Food and Drug Administration-cleared AI solution as one of the most accurate approaches for identifying, quantifying, and characterizing CAD, according to the company.

“With the application of AI to CCTA, Cleerly LABS provides highly quantitative measures of atherosclerotic plaque burden, plaque morphology, and stenosis metrics, delivering high precision and performance compared to invasive ‘gold standards,’” Cleerly Founder/CEO James K. Min, M.D., said. “Several published studies report real-world examples of AI outperforming traditional diagnostic methods in cardiac imaging interpretation, as described by the respective study authors. For example, Griffin et al. (2023) reported that Cleerly’s AI-based analysis of coronary computed tomography angiography achieved a sensitivity of 94% and a specificity of 68% for detecting significant coronary artery stenosis, which is substantially higher than the 74% sensitivity and 43% specificity reported for traditional stress testing.”

That disparity helps explain the huge gap in Bruno’s test scores. Since learning her true plaque grade, however, the Fountain Life vice president has considerably narrowed the spread. By adopting a heart-healthy diet and daily exercise plan (running), Bruno has reduced her total plaque score by nearly 100 points (348 to 254.6), her non-calcified plaque by 56 points (216 to 159.8), and eliminated severe plaque.

“I was so, so happy hearing that. The Cleerly test gives you the information without any invasive procedures and you get pictures of your heart, it’s absolutely amazing,” she exclaimed. “…we don’t know exactly what’s happening inside our body. We don’t know what’s happening in our hearts. I had no symptoms of any kind. I didn’t have any indication that I’m having any problems or issues with my heart. I’m a nurse and I didn’t know what was going on until I got the CCTA Cleerly test done. Now I know and I’m making the changes and I know that it’s working. This is life-changing and life-saving.”

Bruno’s life is just one of the countless being changed—and more importantly saved—by the transformative capabilities of artificial intelligence. The technology has existed for decades, loitering on the periphery of mainstream adoption due to ongoing fears about its potential hazards. But AI is rapidly emerging as a standard component of care as its analytical prowess increasingly authenticates its proficiency in enhancing diagnostic accuracy and enabling earlier disease detection.

Technological advancements in AI are transforming cardiac imaging by providing better, more sophisticated tools for analyzing vast amounts of data and enhancing images. AI solutions are now commonplace in most imaging modalities, including CT, echocardiography, magnetic resonance imaging (MRI), and nuclear imaging, streamlining the clinical workflow and improving decision-making. 

Studies have found that AI algorithms can match or surpass human faculty in evaluating the coronary tree, matching symptoms with biomarkers, and predicting disease. AI’s ability to augment images and even generate new images of its own has revolutionized cardiac imaging.

“AI is already transforming cardiovascular diagnostics by increasing speed, accuracy, and efficiency across cardiac CT, MRI, and echocardiography. As workloads rise and staffing challenges persist, AI is seen as a vital support tool for clinicians and imaging departments alike,” noted Atul Gupta, M.D., chief medical officer of Diagnosis and Treatment at Philips N.V. “With cardiac CT usage expected to account for 14% of all CT exams by 2030, the pressure on imaging departments is growing fast. AI in MRI is also driving faster scans, sharper images, and automated reporting, making high-quality cardiac MR more accessible and efficient. The broader impact of AI for clinicians will be improved diagnostic confidence, reduced variability, and faster turnaround times while the impact for technologists includes more streamlined workflows, less set-up time, and reduced retake rates. And for patients, they will have lower radiation exposure, experience faster exams, and potentially receive earlier, more accurate diagnoses.”

Philips is delivering all those AI-driven benefits (and more) through such solutions as the CT 5300, built specifically for AI-based image reconstruction; the Spectral CT 7500, which allows for 15-minute quadruple rule-out scans; Precise Cardiac, an AI tool that helps compensate for cardiac motion; and MRI SmartSpeed software, a program that uses an AI reconstruction algorithm (Adaptive-CS-Net) to improve image resolution and shorten exam times compared to Philips scans without Compressed SENSE imaging. SmartSpeed leverages all the scan data to denoise the image as soon as possible in the reconstruction process. 

Philips added to its AI-powered cardiac toolkit earlier this year by introducing the Compact Ultrasound 5500CV, a portable system that reduces scan time by up to 50% while improving reproducibility. The portable ultrasound system is compatible with Philips’ X7-2t and X8-2t transesophageal echocardiography transducers. Leveraging the company’s xMatrix tranducers, the system supports xPlane Imaging, meaning the system can simultaneously capture two cross-sectional views using a single probe pass, thereby securing twice the clinical information in the same time as traditional 2D methods. Its AI-powered Auto Measure function speeds up routine two-dimensional and Doppler cardiac measurements by up to 50% while ensuring consistent results, according to Philips.

“AI is impacting earlier diagnosis and treatment of heart failure and this is where echocardiography and cardiac ultrasound are playing huge roles,” Dr. Gupta said. “By embedding AI into next-generation echocardiography platforms, clinicians can now detect signs of heart failure—such as ventricular strain—before symptoms ever appear. These AI tools automate and accelerate complex measurements like global longitudinal strain, eliminate variability between technicians, and make it possible to consistently track changes over time. This level of precision is not only improving diagnostic accuracy but also enabling earlier interventions, such as the introduction of cardioprotective therapies before permanent heart damage occurs.”

Diagnostic accuracy and earlier interventions are among the objectives of Philips’ collaboration with GPU-accelerated computing pioneer NVIDIA. The pair is teaming up to build an MRI foundational model powered by NVIDIA’s advanced AI computing platform. A large deep learning neural network trained on massive datasets, the foundational model will generate new applications aimed at enhancing MR image quality, accelerating scan times, and improving diagnostic workflow and accuracy across various clinical applications. 

The collaboration ideally will induce a solution that can seamlessly be integrated into existing MR workflows. Philips will build on NVIDIA’s VISTA-3D, a specialized interactive foundation model for 3D medical imaging and MAISI, a state-of-the-art 3D model designed to generate high-quality synthetic images with or without anatomical annotations. Together, the technologies will create a domain-specific solution tailored to MR imaging’s unique challenges.

“Accuracy is the key word here. There are three major ways we’re enhancing accuracy; one is enhancing the [image] resolution. The second is by automating the ability to segment organs or segment parts of the heart so then (three) you can provide in real time some of these quantitative assessments,” explained David Niewolny, NVIDIA’s business development director for healthcare/medical. “In electrocardiography for example, there’s the ability to do things like automated image acquisition. AI algorithms can now acquire electrocardiographic images while reducing significant variability and improving the quality. We’re really moving in this world now where we’re taking AI from the static kind of images to much more dynamic real-time insights that give clinicians the opportunity to personalize care. With that, as a foundational accelerated computing platform company, NVIDIA becomes the center of that transition. We’re moving from not just interpreting the cardiac images to help predict and prevent disease. NVIDIA is really building the platform that’s accelerating this entire ecosystem.”

Joining Philips on that platform are Fujifilm, GE HealthCare, and Siemens Healthineers, all of which are striving to streamline their medical imaging workflows and/or improve picture quality. Case in point: Fujifilm leverages NVIDIA GPUs in its Cardio StillShot software, which integrates with the company’s whole-body X-ray CT system SCENARIA View for precise cardiac imaging at any heart rate. 

Siemens Healthineers, by contrast, uses NVIDIA’s MONAI Deploy—part of the open-source R&D platform MONAI—to accelerate and streamline the integration of AI workflows for medical imaging into clinical practice. GE HealthCare, meanwhile, is leveraging NVIDIA’s expertise to develop ultrasound and X-ray applications using its partner’s new Isaac for healthcare medical device simulation platform.

GE HealthCare’s AI-powered cardiac care repertoire stretches well beyond the NVIDIA platform, though. Its CardIQ Suite, for example, is an integrated workflow for seamlessly reviewing coronary calcium scoring and CCTA data. The suite features a fully automated calcium scoring algorithm that quickly identifies calcium burden and location, providing both total and per territory scores within seconds, and enables clinicians to visualize and estimate heart fat volume. 

The company’s AltiX AI.i edition of Mac-Lab, CardioLab, and ComboLab editions aim to elevate workflow in the cardiac catheterization lab, augment interoperability, and enhance precision care for various cardiac procedures.

“The National Institute for Health and Care Excellence recommends CCTA as the first-line investigation for patients with chest pain due to suspected CAD, highlighting its importance in improving diagnostic certainty. To assist clinicians in this effort and the use of this technology, GE HealthCare is proud to offer CardIQ Suite,” stated Chad Rowland, general manager, Global Premium CT and Photon Counting, at GE HealthCare. “…readers can immediately proceed to the CCTA read using advanced 2D and automated 3D processing tools as well as enjoy automated coronary segmentation and tracking AI algorithms to significantly reduce the need for manual intervention, enhancing efficiency with ready-to-read multi-planar images. Altogether, CardIQ Suite aims to drive greater efficiency and streamline workflows.”

Those same objectives underpin Revolution Vibe, a new CT system GE HealthCare premiered at the American College of Cardiology Scientific Session & Expo in March. The system incorporates the company’s Unlimited One-Beat Cardiac Imaging, ECG-less Cardiac TrueFidelity DL, SnapShot Freeze 2, and Effortless Workflow’s AI-powered solutions to deliver fast, accurate diagnoses and more efficient workflows. 

The Unlimited One-Beat Cardiac Imaging solution reconstructs clear CCTA images at low dose, improving access for patients with complex conditions, while the TrueFidelity DL deep learning imaging technology and the SnapShot Freeze 2 algorithm bolster GE Healthcare’s motion-free CCTA imaging capabilities. 

Revolution Vibe’s Effortless Workflow feature leverages AI to automatically adjust protocols and patient positioning when necessary. This one-step decision tree workflow can save radiologists up to four minutes per scan and reduce the patient preparation process by up to five minutes per scan, according to the company.

“GE HealthCare’s Effortless Cardiac Workflow is a great example of how AI can streamline and expedite cardiac imaging,” Rowland said. “It optimizes CT systems for cardiac scans, leveraging AI to automatically select protocols and position the patient—optimizing scanning time and making it easy to use for every user, even junior or inexperienced technologists.”

The Acuson Sequoia ultrasound system is Siemen’s Healthineers’ prime example of AI-driven cardiac imaging streamlining. Unveiled last May, the system’s AI-powered cardiology features include AI Measure, which reduces routine echocardiography exam times by automatically collecting (via 120 artificial intelligence-enabled calculations) detailed test-related measurements. Similarly, the 2D HeartAI tool leverages AI to improve exam efficiency and workflow during cardiac strain imaging, which measures myocardium deformation, delivering auto view detection and auto contour placement with or without an electrocardiogram. In addition, 2D HeartAI—with or without contrast—streamlines the diagnostic capabilities of ejection fraction evaluation and cardiac strain analysis, thus reducing unnecessary supplemental exams.

The Acuson Sequoia’s cardiac software also enables clinicians to generate a stress echo wall motion scoring report. These scores are used to assess cardiac function at various stress levels, reducing unneeded follow-up and supporting presurgical cardiac evaluation. 

“Increasingly, AI algorithms are being integrated on the scanners for performing tasks such as image reconstruction and image quality enhancement,” noted Puneet Sharma, vice president of AI and Digital Innovations at Siemens Healthineers. “There’s been interesting AI innovations for chest CT exams typically used for lung cancer screening. For example, a non-contrasted chest CT exam may not be ideal for assessing a cardiac condition, but AI can detect and quantify findings from that non-cardiac exam, such as coronary artery calcium and aortic disease. AI is making echocardiograms more quantitative by providing automated cardiac strain and ejection fraction assessment. With such AI-driven tools, we can more effectively help in detecting the risk of heart failure and other heart disease.” 

Siemens Healthineers has further improved its heart failure/heart disease risk detection competency by partnering with like-minded comrades, namely semiconductor manufacturer Intel and MRI solutions provider HeartVista.

The company is using second-generation Intel Xeon Scalable processors with Intel Deep Learning Boost to perform semantic segmentation of the heart’s left and right ventricles, with possible extension to all four chambers. The model input is a stack of beating heart MRI images; the output identifies and color codes heart regions or structures. This process aims to automate the manual segmentation process, accelerating result times.

Siemens Healthineer’s partnership with HeartVista, meanwhile, brings MRI scan sequences into one automated exam protocol, combining built-in and native sequences from the two companies. HeartVista’s AI-guided image acquisition software shortens cardiac MRI exam times, simplifies workflows, and improves image consistency, according to the firm. It is FDA-cleared for use with Siemens Healthineers MRI scanners.

Siemens Healthineers is not alone in its pursuit of AI-enabled cardiac imaging partnerships, though. Viz.ai, developer of AI-powered disease detection and intelligent care solutions, has collaborated with several companies to improve patient care and streamline clinical workflows.

The company teamed up with Bristol Myers Squibb to deploy an AI algorithm and provider workflow software that would identify and triage patients needed further testing for hypertrophic cardiomyopathy (thickened heart muscle). Trained on a diverse dataset of more than 831,000 ECGs from over 300,000 patients, Viz HCM leverages a multilayer convolutional neural network to find underlying signals of HCM.

“Our AI algorithm for hypertrophic cardiomyopathy (HCM) uses ECG data to identify suspected HCM with high accuracy, often flagging patients who otherwise would go undetected,” explained Meghana Valluri, manager, Product Management, at San Francisco-based Viz.ai. “In clinical studies the AI model improved detection rates and interpretability when calibrated for diverse patient populations.”

Viz.ai’s alliance with ultrasound imaging algorithm developer iCardioAI aims to advance the rapid detection of aortic stenosis and facilitate patient triage via iCardio.ai’s ultrasound interpretation technology. Similarly, Viz.ai has incorporated Cleerly’s AI algorithms for evaluating CCTA images into its Viz.ai One platform, which now includes more than 48 clinical AI modules. Key innovations include last year’s introduction of the Viz ICH Plus algorithm for automated volumetric assessment, cutting-edge 3D CT angiography functionality, as well as launching Viz Connect for cryptogenic stroke and Viz ACS for acute coronary syndrome.

 “AI excels at uncovering subtle patterns that may be missed by the human eye, such as early signs of structural heart disease or low-grade perfusion abnormalities. For heart failure, AI-enhanced echocardiography or ECG-linked tools can detect reduced ejection fraction or hypertrophic changes long before clinical symptoms manifest. These capabilities support earlier diagnosis and treatment, ultimately improving long-term outcomes,” Valluri said. “While AI alone may not always outperform human experts, the combination of AI plus providers has shown to significantly improve sensitivity. For example, in identifying incidental findings like pulmonary embolism on cardiac CT scans. This collaborative approach helps ensure that critical findings are not missed, ultimately driving better patient outcomes.”

And earlier intervention.

Behold the future of cardiac care, powered by transformative AI. The best is yet to come.