Improving image quality with super-resolution deep-learning-based reconstruction in coronary CT angiography

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Cardiovascular Imaging

Background:

• Coronary CT angiography (CCTA) is crucial for diagnosing coronary artery disease.
• Image quality in CCTA is often limited by noise and artifacts, impacting diagnostic accuracy.
• Advanced reconstruction algorithms are needed to overcome these limitations.

Purpose of the Study:

• To evaluate the effectiveness of super-resolution deep-learning-based reconstruction (SR-DLR) in improving CCTA image quality.
• To compare SR-DLR with traditional reconstruction methods like hybrid iterative reconstruction (HIR), model-based iterative reconstruction (MBIR), and normal-resolution deep-learning-based reconstruction (NR-DLR).

Main Methods:

• Retrospective analysis of CCTA data from 41 patients using a 320-row scanner.
• Reconstruction of images using HIR, MBIR, NR-DLR, and SR-DLR algorithms.
• Quantitative assessment of image noise, contrast-to-noise ratio (CNR), and blooming artifacts.
• Subjective evaluation of image sharpness, noise, texture, edge smoothness, and delineation of cardiac structures.
• Task-based image quality assessment using a physical phantom to evaluate object detectability.

Main Results:

• SR-DLR demonstrated significantly lower image noise and blooming artifacts compared to HIR, MBIR, and NR-DLR.
• SR-DLR achieved a higher contrast-to-noise ratio (CNR) than all other methods (p < 0.001).
• Subjective image quality scores were highest for SR-DLR across all criteria (p < 0.001).
• Phantom studies showed SR-DLR provided superior spatial resolution, noise characteristics, and object detectability.

Conclusions:

• SR-DLR significantly enhances both objective and subjective image quality in CCTA.
• The algorithm improves spatial resolution, noise properties, and object detectability, aiding in accurate coronary artery disease assessment.
• SR-DLR offers faster reconstruction times than MBIR and has the potential to become a new standard for CCTA on 320-row CT scanners.