Coronary Stent Evaluation by CTA: Image Quality Comparison Between Super-Resolution Deep Learning Reconstruction and...

Area of Science:

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

Background:

• Coronary artery stent assessment on CT angiography (CTA) is crucial for patient management.
• Traditional reconstruction algorithms may limit image quality, impacting stent evaluation.
• Super-resolution deep learning reconstruction (SR-DLR) offers potential for improved image sharpness.

Purpose of the Study:

• To compare the performance of SR-DLR against traditional reconstruction algorithms for coronary stent evaluation.
• To assess image quality metrics and diagnostic confidence using different reconstruction techniques.

Main Methods:

• Retrospective analysis of coronary CTA scans from patients with coronary artery stents.
• Reconstruction of images using hybrid iterative reconstruction (HIR), model-based iterative reconstruction (MBIR), normal-resolution deep learning reconstruction (NR-DLR), and SR-DLR.
• Quantitative and qualitative image quality assessments, including artifact analysis, noise levels, lumen diameter, strut sharpness, contrast-to-noise ratio (CNR), and diagnostic confidence.
• Calculation of stent assessability rates for stents ≤ 3.0 mm in diameter.

Main Results:

• SR-DLR demonstrated significantly lower blooming artifacts, reduced stent-induced attenuation, and lower image noise compared to other methods.
• SR-DLR yielded higher in-stent lumen diameter, improved stent strut sharpness, and superior CNR.
• Qualitative assessments showed significantly higher scores for SR-DLR in image sharpness, noise texture, and delineation of stent components.
• Stent assessability rates for small-diameter stents (≤ 3.0 mm) were substantially higher with SR-DLR (86.5%-89.2%) compared to HIR (35.1%-43.2%), MBIR (59.5%-62.2%), and NR-DLR (62.2%-64.9%).

Conclusions:

• SR-DLR provides superior image quality for coronary stent assessment compared to HIR, MBIR, and NR-DLR.
• The algorithm enhances delineation of stent struts and lumens, reduces artifacts, and improves image sharpness.
• SR-DLR holds significant clinical potential for improving the assessment of coronary stents, particularly those with small diameters, on 320-row scanners.