Radiation dose reduction with deep-learning image reconstruction for coronary computed tomography angiography

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Cardiovascular Imaging

Background:

• Coronary CT angiography (CCTA) is crucial for diagnosing coronary artery disease.
• Reducing radiation dose in CCTA is a significant clinical goal.
• Deep-learning image reconstruction (DLIR) shows promise for dose reduction without compromising image quality.

Purpose of the Study:

• To evaluate the effectiveness of DLIR in reducing radiation dose for CCTA.
• To assess the impact of DLIR on image noise, stenosis severity, plaque composition, and plaque volume quantification.
• To compare DLIR with traditional reconstruction methods at different radiation doses.

Main Methods:

• Prospective study involving 50 patients undergoing two CCTA scans: normal-dose (ND) with ASiR-V 100% and lower-dose (LD) with DLIR.
• Quantitative assessment of image noise (HU) and plaque volumes (mm³).
• Visual categorization of stenosis severity and plaque composition (calcified, non-calcified, mixed).

Main Results:

• DLIR achieved a 43% radiation dose reduction (1.4 mSv vs. 0.8 mSv) with no significant increase in image noise (28 HU vs. 27 HU).
• Excellent reliability was observed for stenosis severity (ICC=0.995) and plaque composition (ICC=0.974) between ND and LD scans.
• Bland-Altman analysis showed minimal mean difference (-0.8 mm³) in plaque volume quantification.

Conclusions:

• DLIR enables substantial radiation dose reduction in CCTA (over 40%) while preserving diagnostic image quality.
• The technique does not negatively affect image noise, stenosis assessment, plaque characterization, or volume measurements.
• DLIR represents a valuable tool for safer and more efficient CCTA examinations.