AI-driven assessment of over-scanning in chest CT: A systematic review and meta-analysis.

INTRODUCTION

Scan range is crucial for CT acquisitions. However, irrelevant over-scanning in CT is common and contributes to a significant radiation dose. This review explores the role of artificial intelligence (AI) in addressing manual over-scanning in chest CT imaging.

METHODS

A systematic search of peer-reviewed publications was conducted between December 2015 and March 2025 in Embase, Scopus, Ovid, EBSCOhost, and PubMed. Two reviewers and an academic lecturer independently reviewed the articles to ensure adherence to inclusion criteria. The quality of the included studies was assessed using CLAIM and QUADAS-2 tools. Summary estimates on over-scanning at the upper and lower boundaries of the scan range in chest CT were derived using meta-analysis.

RESULTS

Five studies employed AI algorithms to assess manual over-scanning in chest CT using either 2D topograms or 3D axial images at low and standard doses. These models accurately determine the extent of over-scanning, demonstrating strong agreement with radiologist evaluations. All included studies revealed significant variation in over-scanning at the superior (13.5 mm) and inferior (30.2 mm) boundaries of the scan range (p < 0.001), with approximately two-thirds of the total over-scanning (43.2 mm) occurring at the inferior level (abdomen).

CONCLUSIONS

Integrating AI tools into the over-scanning evaluation process may optimise chest CT imaging protocols and enhance patient safety by reducing over-scanning and radiation dose through real-time monitoring and retrospective analysis.