Development and validation of an AI system for AAS CTA diagnosis and mapping

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Radiology

Background:

• Acute aortic syndrome (AAS), including aortic dissection (AD), intramural hematoma (IMH), and penetrating aortic ulcer (PAU), requires rapid diagnosis.
• Current manual interpretation of computed tomography angiography (CTA) for AAS is time-consuming and prone to variability.
• Automated diagnostic tools are needed to improve efficiency and accuracy in time-critical cardiovascular emergencies.

Purpose of the Study:

• To develop and externally validate an AI-powered decision support system (AAS-DSS) for AAS.
• To enable automated 17-zone aortic segmentation and slice-level AAS subtype classification using CTA.
• To assess the performance of AAS-DSS compared to radiologists and evaluate its generalizability.

Main Methods:

• The AAS-DSS was developed using nnUNet v.2, TotalSegmentor, and ResNet-18 on 586 CTA scans.
• External validation was performed on 198 multi-institutional CTA cases.
• The system performed automated 17-zone aortic segmentation and AAS subtype classification.

Main Results:

• AAS-DSS achieved excellent aortic segmentation and high AAS classification accuracy.
• The AI system demonstrated performance non-inferior to senior radiologists and outperformed junior radiologists.
• Interpretation time was significantly reduced, and strong cross-institutional generalizability was confirmed.

Conclusions:

• The AI-powered AAS-DSS effectively supports the diagnosis of acute aortic syndrome.
• The system accelerates diagnosis, potentially improving patient management, especially in resource-limited settings.
• AAS-DSS shows promise for standardized and efficient interpretation of AAS on CTA.