AI-derived carotid plaque vulnerability from B-mode ultrasound for cerebrovascular risk stratification: a multicenter cohort study

Q: What are the key findings of this study?

An AI-derived high-risk carotid plaque vulnerability score from routine B-mode ultrasound independently predicted future cerebrovascular events beyond conventional clinical risk factors.

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: High cardiovascular risk with carotid atherosclerotic plaques (n=6618). Intervention: AI-derived carotid plaque vulnerability score vs. Low-risk AI score. Primary outcome: Ischemic cerebrovascular events (ischemic stroke or transient ischemic attack) (HR 2.09, 95% CI 1.69-2.58, p=<0.001).

Q: What does this research mean for the field?

An artificial intelligence framework can accurately assess carotid plaque vulnerability from routine B-mode ultrasound, outperforming human readers and independently predicting ischemic cerebrovascular events. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Q: What question did this study set out to answer?

To develop and validate a cascaded multitask AI framework for automated carotid plaque detection, feature characterization, and vulnerability assessment using routine B-mode ultrasound to predict cerebrovascular events.

Qi Xu; Huanhuan Ding; Lin Zhou; Xiaoli Ji; Xiang Zhang; Zengqiao Lin; Zenan Lu; Lina Ye; Zhongshan Gou; Wenbing Jiang; Zhihan Yan

doi:10.1038/s41746-026-03201-4

September 8, 2026npj Digital MedicineOpen Access

AI-derived carotid plaque vulnerability from B-mode ultrasound for cerebrovascular risk stratification: a multicenter cohort study

View Full Paper

Ask AI

Bookmark

Share

Key result

AI-derived carotid plaque vulnerability score linked to ~109% higher cerebrovascular event risk beyond clinical factors.

HR 2.09
95% CI 1.69-2.58
P<0.001
n=6,618

Why the study?

Carotid B-mode ultrasound vulnerability evaluation remains subjective, and its prognostic value is uncertain.

Does an AI-derived carotid plaque vulnerability score from routine B-mode ultrasound improve cerebrovascular risk stratification and predict future ischemic events in high-risk patients?

Key Points

To develop and validate a cascaded multitask AI framework for automated carotid plaque detection, feature characterization, and vulnerability assessment using routine B-mode ultrasound to predict cerebrovascular events.
A multicenter cohort of 6618 participants at high cardiovascular risk with 38,090 B-mode ultrasound images from four centers was split into a training cohort (3229 participants; 15,897 images), an internal test cohort (1468 participants; 7093 images), and an external test cohort (1921 participants; 15,100 images).
Plaque presence and five B-mode features were evaluated against expert B-mode consensus, while vulnerability classification was evaluated against multimodal expert consensus.
Prognostic performance was evaluated longitudinally in 2174 participants over a median follow-up of 37 months.
The AI model achieved an AUC of 0.95 (95% CI, 0.94–0.95) internally and 0.96 externally for plaque detection, and 0.90 (95% CI, 0.89–0.91) internally and 0.88 (95% CI, 0.87–0.89) externally for vulnerability prediction, outperforming six independent readers (all DeLong P < 0.001).
During follow-up, 277 ischemic cerebrovascular events occurred, including 206 ischemic strokes and 71 transient ischemic attacks.
AI-defined high-risk status was independently associated with future cerebrovascular events (adjusted HR, 2.09; 95% CI, 1.69–2.58; P < 0.001), improving the concordance index from 0.769 to 0.795.

Study Design

Type

Cohort (n=6,618)

Multicenter

Yes

Structured PICO

Does an AI-derived carotid plaque vulnerability score from routine B-mode ultrasound improve cerebrovascular risk stratification and predict future ischemic events in high-risk patients?

P

Population

6,618 participants at high cardiovascular risk undergoing routine carotid B-mode ultrasound, with a subset of 2,174 participants followed longitudinally for a median of 37 months.

E

Exposure

Cascaded multitask AI framework for plaque detection, characterization of five B-mode features, and vulnerability assessment from routine B-mode ultrasound.

C

Comparator

Expert B-mode consensus (for plaque presence and features), multimodal expert consensus (for vulnerability), six independent B-mode readers, and a clinical risk factor model (for prognostic evaluation).

O

Outcome

Diagnostic performance (AUC, AP, accuracy) for plaque detection and vulnerability classification, and prognostic association with ischemic cerebrovascular events (ischemic stroke and transient ischemic attack) over a median follow-up of 37 months.composite

Main Result

Hazard Ratio: 2.09 (95% CI 1.69–2.58)

p-value: p=<0.001

An AI framework applied to routine carotid B-mode ultrasound accurately identifies plaque vulnerability and independently predicts future ischemic cerebrovascular events, improving risk stratification beyond conventional clinical factors.

Limitations

Study population was derived from high cardiovascular risk individuals within a single country, which may limit generalizability.
Consensus labels included B-mode assessment, which may introduce incorporation bias when evaluating a B-mode-based model.
Prognostic analysis was retrospective, and residual confounding cannot be excluded.
Event laterality could not be consistently adjudicated across centers, so overall ischemic events were evaluated rather than ipsilateral events.
The patient-level AI-derived vulnerability score and its fixed 0.5 threshold require further external validation.
Study population was derived from high cardiovascular risk individuals within a single country, which may limit generalizability
Consensus labels included B-mode assessment, which may introduce incorporation bias
Prognostic analysis was retrospective, and residual confounding cannot be excluded
Event laterality could not be consistently adjudicated across centers
Patient-level AI-derived vulnerability score and its fixed 0.5 threshold require further external validation

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

May aid cerebrovascular risk stratification on routine ultrasound; leaves open prospective validation before clinical adoption.

Abstract

Routine carotid B-mode ultrasound is widely used for plaque assessment, but vulnerability evaluation remains subjective, and its prognostic value is uncertain. We developed and validated a cascaded multitask AI framework for plaque detection, characterization of five B-mode features, and vulnerability assessment. The multicenter cohort included 6618 participants at high cardiovascular risk and 38,090 images from four centers. Plaque presence and B-mode features were evaluated against expert B-mode consensus, whereas vulnerability was evaluated against multimodal expert consensus. Internal and external AUCs were 0.95 and 0.96 for plaque detection, 0.85–0.94 and 0.81–0.92 for feature characterization, and 0.90 and 0.88 for vulnerability classification. The model outperformed six independent B-mode readers and showed greater net benefit in decision-curve analysis. Prognostic validation included 2174 participants; 277 ischemic cerebrovascular events occurred over a median of 37 months. AI-defined high-risk status remained independently associated with events after adjustment for clinical risk factors (hazard ratio, 2.09; 95% confidence interval, 1.69–2.58), and adding the AI score improved the C-index from 0.769 to 0.795. These findings support standardized plaque vulnerability assessment and clinically meaningful cerebrovascular risk stratification from routine B-mode ultrasound.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Xu et al. (2026) conducted a cohort in High cardiovascular risk with carotid atherosclerotic plaques (n=6,618). AI-derived carotid plaque vulnerability score vs. Low-risk AI score was evaluated on Ischemic cerebrovascular events (ischemic stroke or transient ischemic attack) (HR 2.09, 95% CI 1.69-2.58, p=<0.001). An AI-derived high-risk carotid plaque vulnerability score from routine B-mode ultrasound independently predicted future cerebrovascular events beyond conventional clinical risk factors.

synapsesocial.com/papers/6aa27b0c58559d80afc74051 https://doi.org/10.1038/s41746-026-03201-4

View Full Paper

Ask AI

Bookmark

Share

AI-derived carotid plaque vulnerability from B-mode ultrasound for cerebrovascular risk stratification: a multicenter cohort study

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

AI-derived carotid plaque vulnerability from B-mode ultrasound for cerebrovascular risk stratification: a multicenter cohort study

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading