Hidden Stroke Causes: How a CT-Based Plaque Scoring System (Carotid Plaque-RADS) Could Identify Dangerous Carotid Plaques in Patients with Unexplained Strokes

Table of Contents

• Key Points
• Understanding the Problem: What Is ESUS and How Can Carotid Plaques Cause Strokes?
• Study Methods: How the Research Was Conducted
• Key Findings: High-Risk Plaques Are More Common on the Stroke Side
• How These Results Compare with Previous Research
• Clinical Implications: What This Means for Patients
• Study Limitations: What This Research Could Not Prove
• Recommendations: What Patients Should Know and Do
• Frequently Asked Questions
• Source Information

Key Points

• In 94 ESUS patients, high-risk carotid plaques were about twice as common on the stroke side.
• CTA-based Plaque-RADS may identify dangerous plaques without additional scans or radiation exposure.
• Plaque features like thickness or ulceration matter even when carotid narrowing is under 50%.
• This study shows association, not causation; prospective research is still needed to confirm clinical impact.
• Patients should ask if structured plaque assessment can be applied to their existing CTA images.

Understanding the Problem: What Is ESUS and How Can Carotid Plaques Cause Strokes?

For roughly 1 in 6 people who have an ischemic stroke (a stroke caused by a blocked blood vessel), doctors cannot identify a clear cause even after a thorough evaluation. These cases are called embolic strokes of undetermined source, or ESUS. The "embolic" part means the stroke was likely caused by a clot or debris that traveled through the bloodstream to the brain, but the original source remains a mystery.

One possible hidden culprit is a build-up of plaque (fatty deposits) in the carotid arteries, the main arteries in the neck that carry blood to the brain. You might assume that only severely narrowed arteries cause trouble, but that is not always true.

Even plaques that cause only mild narrowing — less than 50% stenosis, or reduction in the artery's inner diameter — can rupture, form clots, and send debris to the brain. The danger lies less in the degree of narrowing and more in the plaque's specific features, its "vulnerability" to breaking apart.

To help doctors describe and compare these plaques, a scoring system called Carotid Plaque-RADS (Plaque Reporting and Data System) was recently introduced. It classifies plaques into 4 risk levels using imaging features such as plaque thickness, ulceration (a break or crater in the plaque surface), and intraplaque hemorrhage (IPH, bleeding inside the plaque).

However, the original Plaque-RADS system depends heavily on magnetic resonance imaging (MRI), which is expensive, time-consuming, and not available at every hospital. This limits its usefulness in everyday stroke care.

CTA (computed tomography angiography, a CT scan that uses injected contrast dye to create detailed images of blood vessels) is already widely used to evaluate stroke patients quickly. The researchers behind this study wanted to test whether a CTA-based version of Plaque-RADS could identify high-risk plaques in ESUS patients. If it works, it could be used in nearly any hospital that treats stroke patients, without needing specialized MRI equipment.

Study Methods: How the Research Was Conducted

The study was a retrospective observational cohort study, meaning the researchers looked back at medical records and images from patients who had already been treated. It took place at a single integrated health system, Penn Medicine, which includes three hospitals: the Hospital of the University of Pennsylvania, Penn Presbyterian Medical Center, and Pennsylvania Hospital. All three are part of a comprehensive stroke center network.

Patients were included if they were admitted between October 1, 2015, and April 1, 2017, with a diagnosis of acute ischemic stroke. To be eligible, patients had to be at least 18 years old, have a stroke affecting only one side of the front part of the brain (unilateral anterior circulation stroke), have undergone a neck CTA, and meet the formal criteria for ESUS as determined by a vascular neurologist (a doctor specializing in stroke and blood vessel disease of the brain).

There were several exclusion criteria (reasons patients were not included). The researchers excluded anyone with:

• Acute infarction (brain tissue damage) in more than one vascular territory (more than one blood supply region of the brain)
• Prior carotid revascularization (a previous procedure, such as stenting or surgery, to reopen a narrowed carotid artery)
• Occlusion (complete blockage) of either cervical internal carotid artery in the neck
• CTA performed more than 10 days after the stroke began

From 772 screened patients, 94 met all the inclusion criteria. This gave the researchers a total of 188 carotid arteries to examine (one on each side of the neck for each patient).

The CTA scans were acquired using a 4th-generation helical CT scanner. Key technical details included section thickness ranging from 0.625 to 1.5 mm, a matrix size of 512×512, and a field of view of 20–33 cm. Patients received 100 mL of iodinated contrast material (Isovue-370) injected intravenously through a 20-gauge catheter placed in a vein at the elbow (antecubital).

A neuroradiologist (a radiologist specializing in brain and neck imaging) reviewed each scan while blinded — meaning they did not know which side of the brain the stroke had occurred on, nor any of the patients' clinical details. The doctor measured the percent stenosis based on the North American Symptomatic Carotid Endarterectomy Trial (NASCET) criteria, measured the maximum plaque thickness (MPT, the thickest point of the plaque), and scored for the presence of plaque ulceration (a break in the plaque surface) across a 4 cm segment of the carotid bifurcation (the point where the common carotid artery splits into its two main branches).

To measure intraplaque hemorrhage (IPH, bleeding inside the plaque), the team used a semi-automated plaque quantification software called Elucid Bioimaging. A neuroradiologist checked the software's segmentations to confirm accuracy.

The original Plaque-RADS classification defines subtypes as follows: 1 means a normal vessel wall; 2 means maximum wall thickness (MWT) under 3 mm; 3 is divided into A (MWT greater than 3 mm without complex features), B (thin fibrous cap, the protective covering over the plaque), and C (ulceration); and 4 is divided into A (IPH), B (ruptured fibrous cap), and C (intraluminal thrombus, a blood clot inside the artery).

Because several of these features — especially fibrous cap status — are difficult to see on CTA, the researchers modified the classification using only the features CTA can reliably detect. Their modified CTA-based Plaque-RADS system was defined as:

1. Plaque-RADS 1: no plaque (normal vessel wall)
2. Plaque-RADS 2: plaque thickness less than 3 mm
3. Plaque-RADS 3: plaque thickness of 3 mm or more, OR an ulcerated plaque (regardless of thickness)
4. Plaque-RADS 4: plaque with intraplaque hemorrhage volume greater than 50 mm³, regardless of plaque thickness

The 50 mm³ threshold for IPH was chosen based on prior research showing that this threshold had the strongest association with cerebrovascular symptoms on the same side of the body. Plaque-RADS 3 and 4 were considered "high-risk" subtypes, while Plaque-RADS 1 and 2 were considered "low-risk."

For statistical analysis, the researchers used a mixed-effects logistic regression model. This approach tested whether high-risk plaques were more likely to be found on the side of the neck matching the stroke side (ipsilateral) versus the opposite side (contralateral). The model adjusted for age and sex and included random intercepts for subjects to account for natural variation between patients. Interaction terms were initially tested but removed from the final model because they were not statistically significant (p > 0.78).

Sensitivity analyses (additional tests to check whether results hold under different assumptions) were performed by re-classifying plaques using different IPH volume thresholds, including greater than 30 mm³, greater than 40 mm³, and a version with no IPH-based classification at all. Statistical significance was pre-specified at 0.05, and all analyses were performed using JMP version 17.0 software.

Key Findings: What the Researchers Discovered

The study included 94 patients. Of these, 55% were female, and the median age was 66 years. There were no significant differences in baseline demographics (age, sex, or other characteristics) across the different ipsilateral Plaque-RADS subtypes.

Here is the full distribution of CTA-based Plaque-RADS subtypes among all 188 carotid arteries, and broken down by stroke side:

• Plaque-RADS 1 (no plaque): 34 of 188 arteries overall (18.1%). On the stroke side: 14 of 94 (14.9%). On the opposite side: 20 of 94 (21.3%).
• Plaque-RADS 2 (plaque under 3 mm): 84 of 188 arteries overall (44.7%). On the stroke side: 40 of 94 (42.6%). On the opposite side: 44 of 94 (46.8%).
• Plaque-RADS 3 (plaque 3 mm or more, or ulcerated): 69 of 188 arteries overall (36.7%). On the stroke side: 39 of 94 (41.5%). On the opposite side: 30 of 94 (31.9%).
• Plaque-RADS 3 with ulceration specifically: 3 of 188 arteries overall (1.6%). On the stroke side: 1 of 94 (1.1%). On the opposite side: 2 of 94 (2.1%).
• Plaque-RADS 4 (IPH greater than 50 mm³): 1 of 188 arteries overall (0.5%). On the stroke side: 1 of 94 (1.1%). On the opposite side: 0 of 94 (0%).

The key comparison: 42.6% of plaques on the stroke side were high-risk (Plaque-RADS 3 or 4), compared with 31.9% of plaques on the opposite side. In plain terms, about 43 in 100 patients had a high-risk plaque on the stroke side, versus about 32 in 100 on the other side.

This difference was statistically significant. In the mixed-effects logistic model adjusting for age and sex, plaques on the stroke side were significantly associated with high-risk Plaque-RADS subtypes, with an adjusted odds ratio of 2.10 (95% confidence interval 1.20–3.71, p=0.01).

What does an odds ratio of 2.10 mean in everyday language? Patients were about twice as likely to have a high-risk plaque on the same side as their stroke. The 95% confidence interval (1.20–3.71) tells us the researchers are 95% confident that the true effect falls somewhere between a 20% increase and a 3.7-fold increase — and since the entire range is above 1, the association is robust, not likely to be due to chance. The p-value of 0.01 means there is only a 1% probability that this result occurred by random luck.

The findings held up in all sensitivity analyses. The significant association persisted regardless of IPH volume threshold — whether the researchers used 30 mm³ or 40 mm³ as the cutoff, the results remained significant (p < 0.005 for both). Even when the IPH-based classification was removed entirely (using only plaque thickness and ulceration to define high-risk), the association remained significant (p=0.01). This consistency strengthens confidence in the main finding.

How These Results Compare with Previous Research

This study is not the first to test Plaque-RADS in stroke patients, but it is one of the first to apply it using CTA specifically. Two prior studies are particularly relevant.

The first study examined 61 patients with carotid stenosis of 50% or greater (moderate to severe narrowing) using MRI. It found a higher proportion of high-risk Plaque-RADS subtypes in symptomatic carotid arteries (those linked to stroke symptoms) compared with asymptomatic arteries (p < 0.001). However, because the study relied on MRI, its findings were limited in how broadly they could be applied — MRI is not universally available for emergency stroke imaging.

The second study was a population-based retrospective cohort of 1,378 patients with less than 70% carotid stenosis. Using a combination of ultrasound, CT, and MRI to apply Plaque-RADS, the researchers found that a score below 3 versus 3 or above predicted disease-free versus recurrence-free survival rates (p < 0.001). In other words, patients with lower-risk plaques had better outcomes over time.

This new study extends that work by focusing specifically on ESUS patients with less than 50% stenosis and using only CTA, the imaging modality most commonly available in stroke centers. One important point of discussion is the 3 mm maximum wall thickness threshold that separates Plaque-RADS 2 from Plaque-RADS 3. This cutoff aligns with prior studies of mild-to-moderate stenosis showing that a wall thickness of 3 mm or more is more common in carotid arteries on the same side as a stroke. That consistency suggests the 3 mm threshold is clinically meaningful for mildly narrowed plaques.

The researchers also discussed the role of plaque ulceration, which is classified as Plaque-RADS 3c in the original system. Here the scientific evidence is mixed. In the prospective Plaque At Risk (PARISK) study, ulcerated plaques with less than 70% stenosis showed no significant association with recurrent stroke on the same side over an average follow-up of 5 years. By contrast, a retrospective study of 152 ESUS patients found that plaque ulceration was more common in carotid arteries on the stroke side.

Detecting ulcerations is technically challenging. The sensitivity of CTA for spotting ulcerations depends on the slice thickness of the scan, and tiny calcifications in the artery wall can mimic the appearance of an ulcer. The researchers note that further work is needed to refine how plaque features are detected and classified by CTA in mildly narrowed arteries.

Clinical Implications: What This Means for Patients

The main message for patients is hopeful: a simple, widely available CT test may help solve a frustrating medical mystery.

Currently, when a patient has an ESUS — a stroke with no clear cause — doctors often struggle to decide on the best long-term treatment to prevent another stroke. Identifying a high-risk carotid plaque on the stroke side could change the conversation. It might point toward more aggressive management of that plaque, whether through medications, lifestyle changes, or closer monitoring.

The CTA-based Plaque-RADS system has practical advantages. Most stroke patients already receive a CTA as part of their emergency workup, so applying this scoring system would not require additional scans, contrast doses, or radiation exposure. It would simply add a structured way of interpreting images that are already being obtained.

This matters because the original Plaque-RADS system, which relies on MRI-based features like fibrous cap status and intraluminal thrombus, is harder to implement in routine practice. A CTA-based system could make Plaque-RADS accessible to a much broader range of hospitals and clinics, allowing consistent, standardized reporting of plaque risk across centers.

The researchers emphasize that plaque features, not just the degree of narrowing, determine stroke risk. Even plaques that narrow the artery by less than 50% — which many doctors might consider low-risk — can be dangerous if they have high-risk characteristics like thickness above 3 mm or bleeding within the plaque.

Study Limitations: What This Research Could Not Prove

The researchers were open about the limitations of their work, and patients should understand what this study cannot tell us.

First, the number of cases with specific complex features was small. Only 3 plaques had ulceration and only 1 had IPH above the 50 mm³ threshold. This means the study had limited power to draw firm conclusions about these specific subtypes, even though the overall high-risk versus low-risk comparison was statistically significant.

Second, this was a retrospective observational study, not a randomized trial. It can show an association between high-risk plaques and stroke side, but it cannot prove that those plaques actually caused the strokes. To establish causation, a prospective study (one that follows patients forward in time) would be needed to evaluate whether patients with high-risk CTA-based Plaque-RADS scores have worse outcomes or higher recurrence rates.

Third, inter-observer agreement (the degree to which different doctors looking at the same images agree on the score) has not yet been tested for this modified CTA-based system. During the development of the original Carotid Plaque-RADS system, agreement across different imaging modalities (ultrasound, CT, MRI) was measured with a kappa of 0.866, which indicates strong agreement. But the modified CTA-based version needs its own reproducibility assessment before it can be confidently used in clinical practice.

Fourth, the study used a commercialized software (Elucid Bioimaging) to measure IPH volumes. IPH is typically identified on MRI, but the researchers note that certain CT features — such as attenuation characteristics below 25 Hounsfield Units (a scale measuring how dense a tissue appears on CT) or the presence of a "rim sign" (a bright edge around the plaque) — are associated with IPH and may allow detection without specialized software. These methods are still under investigation.

Fifth, the original Plaque-RADS system classifies IPH based on its presence or absence, not on volume thresholds. The modified CTA-based system uses a 50 mm³ cutoff, and future work will need to investigate how IPH volume relates to plaque vulnerability and stroke risk.

Finally, the study was conducted at a single health system, and the patient population may not represent all ESUS patients everywhere. Validation in other centers and populations is an important next step.

Recommendations: What Patients Should Know and Do

If you or a loved one has had an embolic stroke of undetermined source, this research offers several practical takeaways.

First, know that your CTA images may contain important clues about plaque risk that go beyond simple measurement of narrowing. Plaque thickness, ulceration, and signs of intraplaque hemorrhage can all matter, even when the artery looks only mildly narrowed.

Second, consider asking your neurologist or stroke specialist whether a structured plaque assessment, such as a Plaque-RADS scoring system, was or could be applied to your imaging. Not every center uses this system yet, but awareness is growing.

Third, understand that high-risk plaque features may influence treatment decisions. While this study does not prove that treating these plaques prevents future strokes, it provides a foundation for future research that may eventually guide more personalized therapy.

Fourth, continue to manage the standard stroke risk factors — blood pressure, cholesterol, diabetes, smoking cessation, and physical activity — since these remain the bedrock of stroke prevention regardless of plaque scoring.

The researchers also outline their own roadmap for the future. They plan to expand this dataset retrospectively and to prospectively image patients using ultra-high spatial resolution photon-counting CTA, a newer technology that may be able to detect fibrous cap status and ulcerations or fissures that standard CTA cannot resolve. They also call for a prospective study to evaluate patient outcomes and for formal inter-observer agreement testing of the modified CTA-based system.

In their conclusion, the authors state that Plaque-RADS could be an impactful tool in clinical practice, allowing for risk stratification of vulnerable plaque phenotypes in ESUS. This study represents a meaningful step toward making that tool practical and widely available.

Frequently Asked Questions

What is an embolic stroke of undetermined source (ESUS)?

ESUS is an ischemic stroke where doctors cannot find a clear cause even after a thorough evaluation. The stroke likely comes from a clot or debris traveling to the brain, but the original source remains unknown. About 1 in 6 ischemic stroke patients have ESUS.

What is a carotid plaque and why can it cause a stroke even without severe narrowing?

A carotid plaque is a fatty deposit in the neck arteries. Even if it narrows the artery by less than 50%, it can rupture, form clots, and send debris to the brain. The danger depends more on plaque features like thickness, ulceration, or bleeding inside the plaque.

Did this study find that high-risk plaques are more common on the stroke side?

Yes. Among 94 ESUS patients, high-risk plaques appeared in 42.6% of arteries on the stroke side versus 31.9% on the opposite side. After adjusting for age and sex, patients were about twice as likely to have a high-risk plaque on the same side as their stroke.

Does this study prove that high-risk plaques cause ESUS strokes?

No. This was a retrospective observational study, so it can show an association but cannot prove causation. To establish that these plaques actually cause strokes, researchers need prospective studies that follow patients over time and evaluate whether high-risk plaques lead to worse outcomes or more recurrent strokes.

What are the limitations of this study that patients should know about?

The study had few cases of ulceration (3) and intraplaque hemorrhage (1), so conclusions about these specific features are limited. Inter-observer agreement for this CTA-based system has not been tested. It was also a single health system study, so results may not apply to all ESUS patients everywhere.

After an embolic stroke of undetermined source (ESUS), should I get a second opinion to check whether my CT angiogram shows a high-risk carotid plaque that might have caused my stroke?

After an ESUS stroke, a second opinion focused on detailed review of your CT angiogram may be worthwhile. In patients with unexplained strokes, high-risk carotid plaque features—such as plaque thickness of 3 mm or more, ulceration, or intraplaque hemorrhage—have been found significantly more often on the side of the stroke (42.6% versus 31.9%; adjusted odds ratio 2.10). Standard reports often measure only the degree of narrowing, so these features can be missed. Your existing CTA images can be re-examined with a structured plaque scoring system. Diagnostic Detectives Network provides independent expert second opinions.

Source Information

Original article: "Prevalence of High-risk CTA-based Carotid Plaque-RADS Subtypes In Patients with Embolic Stroke of Undetermined Source."

Authors: Jae W Song, MD, MS (Department of Radiology, University of Pennsylvania, Philadelphia, PA); Huy Q Phi, BS (Drexel University College of Medicine); Manisha Koneru, MD, MS (Cooper Medical School of Rowan University); Quy Cao, PhD (Department of Biostatistics, University of Pennsylvania); Jeremy Rubin, MS (Department of Biostatistics, University of Pennsylvania); Yu Sakai, MD; Lamya Ibrahim, MD; Sonya E Zhou, MD; John H Woo, MD; Scott E Kasner, MD (Department of Neurology, University of Pennsylvania); Luca Saba, MD (Department of Radiology, University of Cagliari, Cagliari, Italy); and Brett L Cucchiara, MD (Department of Neurology, University of Pennsylvania).

Publication details: Published in Stroke, 2025 March, volume 56, issue 3, pages 737–740. DOI: 10.1161/STROKEAHA.124.048305. Published in final edited form as a Stroke author manuscript, available in PMC 2026 March 01.

This patient-friendly article is based on peer-reviewed research. The study was approved by the local institutional review board and reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Funding was provided by the American Heart Association (grant 938082) and the National Science Foundation Graduate Research Fellowship (DGE-2236662). The authors report no conflicts of interest.