Artificial intelligence diagnostic performance in image-based vulnerable carotid plaque detection: a systematic...

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Medicine
• Cardiovascular Disease Diagnostics

Background:

• Carotid artery atherosclerosis is a major cause of ischemic stroke due to embolic events.
• Accurate identification of unstable carotid plaques is crucial for stroke prevention strategies.
• Artificial intelligence (AI) offers potential for improved plaque risk stratification.

Purpose of the Study:

• To evaluate the diagnostic performance of AI algorithms in differentiating unstable from stable carotid plaques using medical imaging.
• To synthesize evidence on AI's accuracy in carotid plaque characterization for stroke risk assessment.

Main Methods:

• A systematic review and meta-analysis of studies using AI for unstable carotid plaque identification from medical images.
• Searches conducted across major databases (Medline, Embase, Web of Science, IEEE, PubMed, Cochrane Library) up to June 2023.
• Extracted diagnostic accuracy metrics (sensitivity, specificity, AUC) and assessed risk of bias using QUADAS-AI.

Main Results:

• Meta-analysis of 14 studies (from 31 reviewed) showed pooled sensitivity of 91%, specificity of 84%, and AUC of 0.94.
• Significant heterogeneity (I² > 90%) and limited external validation (1 study) were noted, affecting generalizability.
• AI performance varied by sample size, AI type (ML/DL), segmentation method, and publication year, despite observed publication bias.

Conclusions:

• AI algorithms exhibit promising diagnostic performance for identifying unstable carotid plaques.
• Future research must prioritize rigorous external validation and generalizability of AI models.
• Enhancing AI explainability is essential for successful clinical translation in stroke prevention.