Coronary artery calcium quantification: comparison between filtered-back projection, hybrid iterative reconstruction,...

Area of Science:

• Radiology
• Medical Imaging
• Cardiovascular Disease

Background:

• Current coronary artery calcium (CAC) quantification protocols require updates to align with advanced imaging technologies.
• Evaluating novel image reconstruction techniques is crucial for improving diagnostic accuracy in cardiovascular imaging.

Purpose of the Study:

• To investigate the impact of filtered-back projection (FBP), hybrid iterative reconstruction (IR), and deep learning reconstruction (DLR) on CAC quantification.
• To compare the performance of different reconstruction methods in both in vitro and in vivo settings for accurate CAC scoring.

Main Methods:

• An in vitro study utilized a phantom and bone samples, with volumes measured by water displacement.
• An in vivo study involved 100 patients undergoing CAC scoring using 120 kVp tube voltage and 3 mm slice thickness.
• Images were reconstructed using FBP, hybrid IR, and three DLR levels: mild (DLRmild), standard (DLRstd), and strong (DLRstr).

Main Results:

• In vitro calcium volume measurements showed no significant differences across all reconstruction methods (P=.949).
• In vivo studies revealed significantly lower image noise with DLRstr (P<.001).
• No significant differences in in vivo calcium volume (P=.987) or Agatston scores (P=.991) were observed among the reconstruction methods. DLR and hybrid IR showed high agreement with FBP.

Conclusions:

• Deep learning reconstruction, particularly the strong setting (DLRstr), demonstrates minimal bias in Agatston score agreement.
• DLRstr is recommended for accurate CAC quantification, offering potential for improved image quality and diagnostic precision.