Researchers have made a breakthrough in identifying patients at high risk of stroke. A recent study demonstrates how AI stroke prediction models significantly improve the detection of vulnerable carotid plaques. By combining traditional B-mode ultrasound with contrast-enhanced ultrasound (CEUS), deep learning algorithms offer a more accurate assessment of plaque stability. This innovation could transform stroke prevention in clinical settings.
The Power of AI Stroke Prediction
Standard ultrasound techniques often struggle to distinguish between stable and unstable plaques. Unstable plaques are dangerous because they can rupture and cause acute ischemic stroke (AIS). However, artificial intelligence now provides a solution. The research team developed an AI model using the ResNet34 architecture. This model analyzes ultrasound images to identify subtle features that human eyes might miss. Consequently, this technology helps clinicians prioritize high-risk patients for urgent intervention.
Superior Performance in Clinical Trials
The study retrospectively analyzed 923 patients and prospectively validated the findings across three other centers. Initially, researchers used simple logistic regression for prediction. The model using B-mode ultrasound alone achieved an AUC of 0.719. Adding CEUS data improved this score to 0.819. Remarkably, the AI-integrated model surpassed both, reaching an AUC of 0.917. This represents a substantial leap in diagnostic precision. Furthermore, the external test set confirmed the model’s reliability across different clinical settings.
Enhancing Diagnostic Consistency
Diagnostic accuracy often depends on the radiologist’s experience level. Fortunately, AI-based tools reduce this variability. These models provide objective data that supports clinical decision-making. Moreover, the integration of CEUS allows for better visualization of plaque neovascularization. Neovascularization is a key indicator of plaque vulnerability and future stroke risk. Therefore, combining advanced imaging with deep learning creates a robust early-warning system for stroke prevention.
Frequently Asked Questions
Q1: How does AI improve carotid ultrasound?
AI identifies microscopic patterns in plaque texture that indicate high stroke risk. This allows for a more detailed analysis than traditional manual assessment.
Q2: Is CEUS necessary for the AI model?
Yes, combining CEUS with AI provides the highest predictive value for vulnerable plaques. It visualizes blood flow within the plaque itself.
Q3: Can this model be used in routine check-ups?
The model’s high accuracy makes it a potentially valuable tool for screening asymptomatic patients. It helps identify those who require proactive medical management.
References
- Jin ZY et al. Features of B-mode ultrasound and contrast-enhanced ultrasound of carotid plaque based on deep learning enhance the prediction of vulnerable plaques associated with acute ischemic stroke. Eur Radiol. 2026 Feb 21. doi: 10.1007/s00330-026-12376-z. PMID: 41721846.
- Skandha SS et al. Artificial Intelligence in Symptomatic Carotid Plaque Detection: A Narrative Review. Diagnostics. 2023. doi: 10.3390/diagnostics13071234.
- Mantella LE et al. The utility of ultrasound and computed tomography in the assessment of carotid artery plaque vulnerability. Frontiers in Cardiovascular Medicine. 2022. doi: 10.3389/fcvm.2021.783111.