Key result
AI-driven aortic CT analysis accurately detects dissection with ~0.97 median AUC across 118 studies.
Why the study?
Manual segmentation and geometric analysis of CT and CTA for aortic aneurysm and dissection are time-consuming and operator-dependent, while evidence on artificial intelligence applications across segmentation, measurement, and detection tasks remains fragmented.
Does artificial intelligence-driven CT/CTA analysis accurately perform segmentation, measurement, and detection in thoracic and abdominal aortic disease?
Systematic Review
Yes
Does artificial intelligence-driven CT/CTA analysis accurately perform segmentation, measurement, and detection in thoracic and abdominal aortic disease?
Answers are grounded in this paper and cite their sources.
AI-driven CT/CTA analysis shows high technical accuracy for aortic segmentation, diameter measurement, and dissection detection, though widespread clinical integration requires prospective, multicenter validation due to high risk of bias in current studies.
Background/Objectives: Computed tomography (CT) and CT angiography (CTA) are central to assessing aortic aneurysm and dissection, but manual segmentation and geometric analysis remain time-consuming and operator-dependent, while evidence on artificial intelligence (AI) is fragmented. This task-based systematic review synthesised AI applications to CT/CTA of the thoracic and abdominal aorta by primary task: segmentation, measurement, or detection and classification. Methods: PubMed/MEDLINE and Europe PMC were searched from 2015 to September 2026 following PRISMA 2020; the protocol was registered retrospectively (INPLASY2026110004). Segmentation and measurement studies were appraised with CLAIM and detection studies with QUADAS-2/QUADAS-AI, and results were synthesised without meta-analysis. Results: Of 118 included studies (42 segmentation, 38 measurement, 38 detection), half were published since 2024. Median Dice similarity coefficient was 0.90 or higher for the outer wall, lumen and true and false lumina, but 0.86 for intraluminal thrombus and 0.63 for false lumen thrombus. Automated maximum diameter showed a median absolute bias of 0.26 mm (95% limits of agreement about ±5 mm). Detection of aortic dissection reached a median AUC of 0.96–0.97, mostly in enriched samples. External validation was reported in 14%, 34% and 55% of segmentation, measurement and detection studies, 33 of 38 detection studies were at high risk of bias, and only two systems were evaluated prospectively. Conclusions: AI supports technical automation of aortic CT/CTA segmentation and measurement and is progressing toward multistructure geometric phenotyping and clinical decision support, but independent multicentre validation, consecutive cohorts and prospective reader-in-the-loop studies are required before clinical integration. Detection/classification tools require independent diagnostic and workflow validation.
No takes yet. Share an insight, caveat, or question.
Paparo et al. (2026) conducted a systematic review in Aortic disease (abdominal aortic aneurysm, thoracic aortic aneurysm, aortic dissection, acute aortic syndrome). Artificial intelligence (deep learning) CT/CTA analysis vs. Manual or expert-adjudicated segmentation, measurement, or clinical diagnosis was evaluated on Segmentation overlap (Dice similarity coefficient), maximum diameter measurement bias, and aortic dissection detection accuracy (AUC). AI-driven aortic CT analysis achieved median DSC >=0.90 for lumen/wall segmentation, 0.26 mm median diameter bias, and 0.96-0.97 median AUC for dissection detection across 118 studies.
Synapse has enriched 10 related papers. Consider them for comparative context: