Improvement of Spatial Resolution on Coronary CT Angiography by Using Super-Resolution Deep Learning Reconstruction

Area of Science:

• Medical Imaging
• Radiology
• Artificial Intelligence in Medicine

Background:

• Coronary CT angiography (CCTA) is crucial for diagnosing coronary artery disease.
• Image quality in CCTA impacts diagnostic accuracy, particularly for detecting subtle plaques.
• Traditional iterative reconstruction (IR) methods have limitations in noise reduction and sharpness.

Purpose of the Study:

• To compare the image quality of CCTA reconstructed using super-resolution deep learning reconstruction (SR-DLR) versus hybrid iterative reconstruction (IR).
• To evaluate noise levels, contrast-to-noise ratio (CNR), edge rise slope (ERS), and plaque detectability between the two reconstruction methods.

Main Methods:

• Retrospective analysis of 100 CCTA patient datasets acquired on a 320-detector-row CT scanner.
• Images were reconstructed using both hybrid IR and SR-DLR techniques.
• Quantitative analysis included image noise, CNR, and ERS; qualitative assessment involved overall image quality and plaque detectability scoring.

Main Results:

• SR-DLR yielded significantly lower image noise (15.6 vs 22.9 HU) and higher CNR (32.4 vs 20.4) compared to hybrid IR.
• SR-DLR showed a steeper ERS (300.0 vs 198.2 HU/mm), indicating sharper coronary artery margins.
• Overall image quality scores were superior for SR-DLR (3.6 vs 3.1), with improved detection of plaques < 50% stenosis.

Conclusions:

• SR-DLR significantly outperforms hybrid IR in CCTA image quality.
• Key improvements include reduced noise, enhanced margin sharpness, and superior plaque detectability.
• SR-DLR holds promise for improving the diagnostic performance of CCTA.