"Super-Resolution" Deep Learning Image Reconstruction in Dynamic Myocardial Perfusion: A Prospective Evaluation of...

Area of Science:

• Cardiovascular Imaging
• Medical Physics
• Artificial Intelligence in Medicine

Background:

• Dynamic myocardial computed tomography (CT) perfusion is crucial for assessing coronary artery disease.
• Traditional image reconstruction methods like filtered-back projection (FBP) and hybrid iterative reconstruction (IR) can be limited by image noise and suboptimal signal-to-noise ratio (SNR).
• Deep learning reconstruction (DLR) techniques offer potential improvements in image quality for CT applications.

Purpose of the Study:

• To evaluate the impact of Super-Resolution Deep Learning Reconstruction (SR-DLR) on image quality and myocardial hemodynamic parameters in dynamic myocardial CT perfusion.
• To compare SR-DLR against conventional filtered-back projection (FBP), hybrid iterative reconstruction (IR), and normal-resolution deep learning reconstruction (NR-DLR).

Main Methods:

• A prospective single-center study included 25 patients undergoing dynamic myocardial CT perfusion.
• Image noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were quantitatively assessed.
• Overall image quality, lesion visibility, myocardial blood flow (MBF), and coronary flow reserve (CFR) were analyzed across four reconstruction methods: FBP, hybrid IR, NR-DLR, and SR-DLR.

Main Results:

• SR-DLR demonstrated significantly lower image noise and higher SNR and CNR compared to FBP, hybrid IR, and NR-DLR in both rest and stress imaging (P < 0.001).
• Qualitative analysis revealed SR-DLR achieved the highest overall image quality and lesion visibility, outperforming FBP and comparable to hybrid IR and NR-DLR.
• No statistically significant differences in coronary flow reserve (CFR) were observed among the reconstruction methods.

Conclusions:

• Super-Resolution Deep Learning Reconstruction (SR-DLR) significantly enhances image quality in dynamic myocardial CT perfusion by reducing noise and improving SNR and CNR.
• SR-DLR maintains accurate hemodynamic quantification, comparable to other reconstruction methods.
• Both SR-DLR and normal-resolution deep learning reconstruction (NR-DLR) offer substantial improvements in image quality for myocardial CT perfusion imaging.