Cleerly

Precision Plaque Analysis

Andrew Kuzemczak
April 08, 2025

In partnership with

Learn AI in 5 minutes a day

This is the easiest way for a busy person wanting to learn AI in as little time as possible:

Sign up for The Rundown AI newsletter
They send you 5-minute email updates on the latest AI news and how to use it
You learn how to become 2x more productive by leveraging AI

Hi friend,

Welcome back to Future Human! I hope you all enjoyed the look at digital physical therapy last week with Sword Health. We received some great feedback from that edition, plus the previous seven. I have been compiling a list for months now (well before Future Human launched) of the features I want to add to make the ideal healthtech newsletter. The feedback received has been a terrific tool to narrow my list to just the options which matter more to you. So look out for upcoming features:

TLDR bullets offering a summary of the key points at the beginning of each edition
LinkedIn newsletter option sending a partial version of the letter to those more active there (full version to remain here)
Job board with healthtech/medical jobs for those searching for their next gig

Here at Cornell, we are immersed in the renal unit. I am exhausted by electrolytes. Very interesting, but dense. If you ask the media, medical burnout tends to begin in the early years of residency. If you ask my classmates, they are burning right this moment. All is well though. Spring break is right around the corner. I look forward to writing next week’s Future Human newsletter from the beaches of the Outer Banks.

Alright, for our ninth edition of Future Human, we return with newsletter nepotism. In week #2, I wrote about EpiBone, a startup founded by the graduation speaker at my graduate school commencement. Well for this week, I am looking more to the present with a startup from a former professor at Weill Cornell. I have followed this company since before I received my medical school decisions. Admittedly, I got so deeply invested in it that it inspired me to come here to Cornell when making my final choice—and that remains one of my greatest decisions.

With that, let me ask you:

If medical AI flagged you as high-risk for a heart attack—even though you feel perfectly healthy—would you trust the algorithm enough to alter your habits? Or would you only accept the word of a human to change?

The Story

Let me start by answering that question myself. Although I value the opinion of a human physician first, data shows the AI may just have a point. 50% of individuals who suffer heart attacks do so without prior symptoms. We here in the US have many ideas for how to address this, but few have been able to comprehensively solve the problem. Every 40 seconds, someone in the US suffers a heart attack. For many, coronary artery disease—the deadliest type of cardiovascular disease and primary cause of heart attacks—was silently building beneath the surface undetected for years.¹^,²

This development of coronary artery disease has resulted in a crisis claiming lives every few seconds around the world, and costing the U.S. alone $400+ billion annually. It is simultaneously estimated that 80% of heart attacks are preventable—a sad but hopeful statistic. So why are we not doing more to detect the risk earlier?

This week’s startup believes they have the tool to do just that, so let me introduce you to Cleerly.

Cleerly was founded in 2017 by Dr. James K. Min. Dr. Min was a Professor of Radiology and Medicine at Weill Cornell Medical College and the Director of the Dalio Institute of Cardiovascular Imaging (currently doing research there, so I am basically his protégé). Dr. Min is a cardiologist who maintains a clinical focus on cardiovascular disease prevention and cardiovascular imaging. When he was working here at Weill Cornell, he introduced a preventative care model, based around Computed Tomography Angiography (CTA), that delivered impressive results for cardiac patients.

Cleerly is his attempt to replicate this impact at scale.

Put as simply as possible, Cleerly is using AI to analyze coronary CT angiography images to detect and quantify atherosclerosis (plaque buildup). The system helps non-invasively label and risk stratify each plaque in the artery, inform physicians on each plaque’s possibility of rupture to decide next steps, and prevent heart attacks before symptoms arise. The AI then generates a 3D model of the arteries, identifies their lumen and vessel walls, measures stenoses, and categorizes plaques.

Dr. Min left his clinical role to build this technology because the diagnostic ‘gold standard’ was still falling short for patients.

Nuclear stress tests are common, but offer little information on the plaque burden and type, resulting in false negatives and positives³
Cardiac catheterizations are accurate but invasive, so less ideal for routine checks⁴
CT scans (CTAs) are non-invasive, accurate, and lower cost, but manual interpretation is difficult⁵

CTA is the new frontier, and for Dr. Min, AI is the upgrade. With cardiovascular disease risk rising in an aging population with worse lifestyle decisions, the need for improved CTA technology is rapidly intensifying.

The Tech

If it was not clear above, after a patient undergoes a CTA, Cleerly’s AI analyzes the scan to offer deeper, plaque-specific insights—but first, what even is a CTA?

A cardiac CT angiography scan uses X-rays to take many images of your heart and blood vessels at various angles. A computer combines the images to create a 3D image of your heart.⁵ It is used to see narrowed arteries and blood vessels that supply blood to the heart or other parts of the body. One can also add on IV contrast to visualize the fluid movement through the vessels and the blockages along the way. So in sum, it can tell:

Heart structure
How well your heart pumps blood
Heart muscle scarring from heart attack
Fluid in the pericardial sac that covers the surface of the heart
Amount of plaque buildup and narrowing of your coronary arteries

So once the CTA is recorded, Cleerly’s AI jumps in. Their analyses are based on over 10 million images from over 40,000 patients gathered over a 15-year-period in multi-center clinical trials. They have shown their AI to be comparable to current, invasive diagnostic tools:⁶^,⁷

IVUS (Intravascular Ultrasound): catheter-based imaging technique that uses high-frequency sound waves to produce cross-sectional images of the inside of coronary arteries
- What it tells you:
  - Size of the artery
  - Plaque burden and composition
NIRS (Near-Infrared Spectroscopy): catheter-based method that uses infrared light to detect the chemical composition of plaques
- What it tells you:
  - Presence of lipid-core plaques
  - Risk of future cardiovascular events
FFR (Fractional Flow Reserve): pressure wire-based technique that measures the pressure difference across a coronary artery stenosis (vessel stiffening)
- What it tells you:
  - Whether stenosis is functionally significant

So what does the AI offer?

Cleerly user interface

Cleerly’s AI is obsessed with each individual plaque and will characterize the obstructions most at risk of dislodging and causing heart attack or stroke. It will measure the individual plaque size, type (lipid, calcium, etc.), and risk level. For the first time, a doctor can see a quantified score across the arteries and potentially intervene in specific areas to reduce the patient’s risk. Where as previously the doctor would have to sit with the scan and move through slice by slice (X-ray slices), now they receive a report after someone else (Cleerly AI) reviews it with the knowledge of 10 million previous image reviews.

Although the deep dive so far may present the platform as one single service, they offer multiple different tools trained on specific patient diagnoses and risks. Cleerly ISCHEMIA is an evaluation that uses nearly 40 measures of heart health to calculate a likelihood of coronary artery ischemia (inadequate blood supply) at a per-vessel level. Clearly Plaque Analysis is the core product described above, allowing AI to determine stenosis severity, plaque volume, and composition with the accuracy of a seasoned physician. Finally, Cleerly COMPARE assesses changes in plaques overtime for longitudinal analysis—a key to heart attack prevention.

Cleerly ISCHEMIA

Across my research, whether through publications, interviews, or other analyses, it seems Dr. Min’s goal for the technology is simple—become the new standard of care for heart disease prevention as mammograms are for breast cancer.

Prevention is the final frontier, and they are doubling down through their own research. In 2023, they announced their TRANSFORM trial—a randomized trial enrolling 7,500 patients who have pre-diabetes, type 2 diabetes, or metabolic syndrome but no symptoms of heart disease. Although we still await the data, they aim to prove that a personalized care strategy based on a Cleerly analysis is better than usual care based on traditional cardiovascular risk factors for primary prevention.¹

While we wait, however, some research already exists. I dove in and asked: can AI really support and predict CVD?

The most ‘groundbreaking’ among the papers was a 2023 Japanese multi-center study covering 133 atherosclerotic plaques from 47 patients who underwent coronary CT angiography and NIRS-IVUS. AI demonstrated excellent diagnostic accuracy in detecting significant low-density non-calcified plaques (LD-NCP). The method achieved an accuracy of 94%, sensitivity of 93%, and specificity of 94%. There was a strong correlation between the AI and intravascular ultrasound (IVUS), which as we mentioned above, is accurate but invasive.⁶

So there you have it—Cleerly uses AI to analyze CTAs and detect heart disease risk early. The team is building on a technology platform (artificial intelligence) that has already shown promise in the diagnostic arena with a scan type (CTA) that is increasing in frequency and improving in accessibility. It is a promising equation, for sure.

The Market

The computed tomography (CT) market was valued at $5.8 billion last year, growing at a CAGR of 7.1% to reach $10.7 billion by 2032. CT, magnetic resonance imaging (MRI), and ultrasound (US) always fight for the second place spot in the global hierarchy of medical imaging (X-ray reigns supreme). Within cardiology, however, given the reduced usefulness of 2D X-rays in heart evaluations, CT vies for the top spot. Usage of all three modalities is increasing, with the exception of CT in children (for fears of radiation exposure). As a note, when I say CTA (angiography), I am speaking of the cardiac specific CT that focuses on the blood vessels.⁸

The AI for CTA space is unsurprisingly smaller. For Cleerly, they find themselves perfectly bookended by the 18 year old ‘incumbent’ and the newer startup on the block.

Heartflow is the market leader for AI in coronary analysis. Founded in 2007, they were the first to bring AI together with cardiac imaging.⁹ They, however, are focused on producing 3D heart models from the scans to inform doctor’s to the locations of occlusions. While Cleerly has a similar end goal, they are working at the vessel level with individual plaque labels. Heartflow has raised approximately $883 million in funding across multiple rounds. Just last month, they raised $98 million through the sale and issuance of convertible notes. Their key innovation remains the ability to calculate FFR (Fractional Flow Reserve) from CT to measure how a blockage is impacting the blood flow. This, however, was matched by Cleerly in February of 2024 with the FDA clearance of Cleerly ISCHEMIA.

Elucid is the newest venture. With 510k FDA clearance in October of 2024, they released their PlaqueIQ software that…wait for it…turns coronary CT angiography images into 3D models that quantify and classify plaque morphology to improve predictions of heart attack and stroke risk. They are still early in their path. Elucid is currently working toward an FFR in CT indication like Cleerly and Heartflow have.¹⁰ Recall, this adds computational fluid dynamics to simulate blood flow through the coronary arteries.

Elucid last raised $80 million in a Series C in November of 2023, bringing total funding to $115 million. They remain focused on a similar end goal as Cleerly, with the labeling of plaque risk, but stand farther behind in the process.

So the market we face is small yet competitive, with each company fighting for every new functionality. Cleerly stands firmly as the second mover, after Heartflow, arguably putting them in one of the most advantageous market positions. Able to learn from the incumbent’s mistakes, while being pushed by the newer players, Dr. Min and his team have most forces on their side.

The Sick

We have written about heart disease burden before, but previously we focused on arrhythmias with Cairdac and heart failure with BiVACOR. Here we care just about the pipes, so lets look at coronary artery disease (CAD).

CAD impacts the larger arteries on the surface of the heart. It is the most common type of heart disease. In 2022, CAD caused 371,506 deaths in the United States.¹¹

Coronary arteries for your reference

About 1 in 20 adults aged 20 and older have CAD, but many do not know until they have crushing chest pain. This pain comes from the stenosis (narrowing) becoming so severe that blood flow is blocked and tissue dies. The blockage, as many of you know, is caused by deposits of cholesterol, forming plaques. This gives us a few different types of CAD:

Obstructive CAD means that the diameter of a large coronary artery is blocked by 50% or more
Nonobstructive CAD means it is blocked <50%.
Coronary microvascular disease means blood flow into the tiny arteries within the heart muscle is blocked (from damage to inner walls of blood vessels)

Preventative care is widely accepted as the best approach for CAD risk. Think healthier eating, exercise, reduced stress, and statins. But besides blood pressure monitoring, do we really have a technologically advanced and accessible way to check in on our cardiovascular status?

As CTA becomes more available and affordable with robust data defending its cost effectiveness long term, it may soon assume this contested position.¹² Data has shown in 2023 and 2024 that CTA was a cost-effective CAD diagnostic strategy when used alone or as part of a sequential system. Even in low-to-intermediate risk patients with stable chest pain, CTA provides a diagnostic option with improved long-term health outcomes when compared to functional testing. Add to this foundation the advent of AI, and we could have an affordable, accessible, and physician-light tool for diagnosis. The fewer physicians a diagnosis requires, the less it will be delayed by the bottleneck of physician shortages.

If you still are not sure of the value of AI in diagnosis, consider that the American College of Radiology has already compiled dozens of FDA-approved AI algorithms for radiological use. Their studies in cardiac radiology have shown the potential for AI to help improve efficiency, time to diagnosis, prognostic stratification, and identification of ischemic lesions.¹³ Put simply, applying AI to diagnosis, especially in cardiology can:

Reduce image analysis time and time to diagnosis
Catch patients without symptoms of disease that may benefit from therapy
Stratify the risk of major cardiovascular events accurately

We have a globally leading cause of death, a diagnostic mechanism that is becoming more affordable and accessible but could use some help, and now the AI functionality that could help it achieve the global reach, efficiency, and affordability it needs. Sounds like countless lives saved.

The Economy

One in three US adults received care for a cardiovascular risk factor or condition in 2020. That frequency does not come cheap. Inflation-adjusted health care costs of cardiovascular risk factors are projected to triple between 2020 and 2050, from $400 billion to $1.34 trillion.

As we have said many times before, when people are sick and being treated, they are also not working and contributing to society as they would like. Productivity losses from cardiovascular conditions are projected to increase by 54% between 2020 and 2050, from $234 billion to $361 billion. Stroke is projected to account for the largest absolute increase in costs.¹⁴

With cardiovascular disease (CVD) as the leading cause of death globally, it should be no surprise that the direct medical costs tower all other disease states at >$400 billion. How much it leads, though, should shock you. Cancer is the second leading cause of death in the US., only slightly behind CVD in number of deaths. Even then, cost of cancer care is projected to reach only $250 billion by 2030. The heart firmly holds the lead.

So what do we do then?

Treatment for a heart attack can total $20,000 for a fatal episode and $15,000 for a nonfatal one. CT angiography, however, costs $300-$600. With the introduction of Cleerly’s AI, we can save even more with fewer physician hours spent assessing the results post-scan.

Of course, I am not blind to the concern of overuse and over diagnosis, which can drive up costs and diminish economic savings. Medicine is notorious for lauding new systems which improve access to diagnosis and then overusing it, over diagnosing many who do not need treatment, and spending more money in the end. That said, with the improved efficiency with physicians not needing to assess from scratch every CTA, and the AI offering more accurate insights to reduce over diagnosis, I believe Cleerly offers a clear (haha) path forward to improved economics for CVD prevention, diagnosis, and treatment.

The economic data for responsible AI use in medicine is clear:¹⁵

Reduction in hospitalization: fewer emergency care visits, surgeries, and admissions
- AI can identify high-risk plaques earlier than traditional methods, allowing for proactive management and reducing the likelihood of heart attacks or strokes (duhhh, its the focus of these last 2,500 words)
Efficient resource allocation: targeted care for high-risk patients and better use of diagnostic resources
- AI's ability to stratify patients based on their true risk levels allows healthcare providers to focus resources on those most at risk
Reduced procedure costs: lower reliance on invasive diagnostic tests and surgeries
- Traditional diagnosis methods involve invasive procedures like catheterization or angiograms, which are costly, carry risks, and often have long recovery times.
Improved long-term outcomes: fewer long-term complications like heart attacks, strokes, and heart failure.
- By improving the accuracy of diagnosis through AI, patients are more likely to receive appropriate care early, preventing future complications
Operational cost savings: automation of image analysis to reduce labor and staffing costs
- AI will not replace a human touch, but it can certainly take on the hours of screen time physicians accumulate assessing their patient scans

My Thoughts

I thoroughly enjoyed exploring a software-oriented venture in the physiological space I am most excited about, cardiology. The work of Dr. Min and his team, beginning years back in these halls of Weill Cornell, is inspiring to say the least. It represents a new standard of diagnosis that will not just impact the health of millions globally, but simultaneously make the lives of thousands of physicians much easier.

Most agree that shifting from reactive treatment to proactive prevention is the way of the future. Cleerly and its AI platform are laying the groundwork for this future where heart attacks are caught well before they occur.

In a world where cardiovascular disease continues to lead in both mortality and economic burden, Cleerly’s AI-driven CTA analysis represents more than just an innovation—it is a blueprint for the future of preventative medicine: scalable, precise, efficient, and life-saving.

To more lives saved,

Andrew

I always appreciate feedback, questions, and conversation. Feel free to reach out on LinkedIn @andrewkuzemczak.