Super-Resolution Deep Learning Reconstruction for Coronary CT Angiography: Coronary Stenosis Assessment and CAD-RADS...

Area of Science:

• Radiology
• Artificial Intelligence
• Cardiovascular Imaging

Background:

• Coronary CT angiography (CCTA) plays a crucial role in assessing coronary artery stenosis.
• Current reconstruction algorithms may limit spatial resolution, impacting diagnostic accuracy.
• Super-resolution deep learning reconstruction (SR-DLR) is a novel technique with potential to enhance CCTA image quality.

Purpose of the Study:

• To compare the performance of SR-DLR against hybrid iterative reconstruction (HIR) for coronary stenosis assessment.
• To evaluate the impact of SR-DLR on Coronary Artery Disease Reporting and Data System (CAD-RADS) classification.
• To utilize invasive coronary angiography (ICA) as the reference standard for validation.

Main Methods:

• Prospective enrollment of 204 patients undergoing CCTA and ICA at 10 centers in China.
• Reconstruction of CCTA images using both HIR and SR-DLR algorithms.
• Quantification of percentage diameter stenosis (PDS) for various plaque types and determination of participant-level CAD-RADS categories.

Main Results:

• SR-DLR demonstrated superior performance in detecting significant coronary stenosis compared to HIR at both lesion (AUC, 0.97 vs 0.90) and participant levels (AUC, 0.90 vs 0.79).
• SR-DLR resulted in a 20% reclassification of participant-level CAD-RADS categories (41 of 204 patients).
• Median PDS for calcified plaques was significantly lower with SR-DLR (58%) versus HIR (63%) (P < .001).

Conclusions:

• SR-DLR significantly outperforms HIR in the assessment of coronary stenosis using CCTA.
• The enhanced resolution provided by SR-DLR impacts clinical decision-making, evidenced by CAD-RADS reclassification.
• SR-DLR holds promise for improving the diagnostic accuracy and clinical utility of CCTA.