Artificial Intelligence for Detecting Aortic Arch Calcification on Chest Radiographs: A Systematic Review

Area of Science:

• Radiology
• Artificial Intelligence
• Cardiovascular Disease

Background:

• Aortic arch calcification (AAC) on chest radiographs (CXRs) is a significant cardiovascular event predictor.
• AAC is frequently overlooked in routine radiological assessments.
• Artificial intelligence (AI) offers potential for automated AAC detection.

Purpose of the Study:

• To systematically review the diagnostic accuracy of AI models for detecting AAC on CXRs.
• To assess the clinical implementation potential of AI-driven AAC detection.

Main Methods:

• Systematic review following PRISMA 2020 guidelines.
• Searched Embase, PubMed, Scopus, and Web of Science (Jan 2020-Oct 2025).
• Included retrospective studies using Convolutional Neural Networks (CNNs) on large datasets; bias assessed with QUADAS-2.

Main Results:

• Three studies (2022-2024) involving ~2.7 million images were analyzed.
• AI models showed high diagnostic discrimination (AUROC 0.81-0.99), with performance varying in external cohorts.
• Significant sensitivity-specificity trade-offs were observed; GRADE certainty was low due to heterogeneity and lack of cross-sectional imaging standards.

Conclusions:

• Deep learning models show promise for reliable AAC detection on CXRs, enabling opportunistic cardiovascular risk stratification.
• Heterogeneity in AI model architectures and validation hinders broad comparability.
• Standardized annotation and external validation are crucial for clinical generalizability of AI tools.