Role of Artificial Intelligence in Detecting and Classifying Aortic Dissection: Where Are We? A Systematic Review and...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Cardiovascular Diseases

Background:

• Aortic dissection (AD) is a life-threatening condition requiring rapid and accurate diagnosis.
• Computed tomography (CT) is a primary imaging modality for AD detection.
• Artificial intelligence (AI) offers potential for automated analysis of medical images.

Purpose of the Study:

• To systematically review and meta-analyze the diagnostic performance of AI models for AD detection and classification using CT images.
• To assess the quality and risk of bias in studies evaluating AI for AD diagnosis.

Main Methods:

• Systematic search of PubMed, Web of Science, Embase, and Medline (January 2010 - October 2023).
• Inclusion of all primary studies evaluating AI for AD detection on CT.
• Quality assessment using METRICS and CLAIM checklists; risk of bias using QUADAS-2.
• Univariate and bivariate meta-analyses to estimate sensitivity, specificity, and AUC.

Main Results:

• Thirteen studies were included, mostly using contrast-enhanced CT (CECT).
• AI demonstrated high pooled sensitivity (94%) and specificity (88%) for AD detection.
• Overall model performance was good (AUC, 0.97), with strong performance for both CECT and noncontrast CT (NCCT) subgroups.
• Only three studies classified AD by Stanford criteria.
• Most included studies were of low quality with high risk of bias.

Conclusions:

• AI models can feasibly detect aortic dissection on CT images with high accuracy.
• Current AI diagnostic performance does not yet translate to proven clinical applicability.
• Further high-quality research is needed to establish clinical utility.