Super-resolution deep learning reconstruction achieves low-dose high-resolution imaging in pediatric cardiac computed...

Area of Science:

• Medical Imaging
• Radiology
• Pediatric Cardiology

Background:

• Cardiac computed tomography (CT) is crucial in pediatric cardiology.
• Minimizing radiation exposure is a key challenge in pediatric CT.
• Super-resolution deep learning reconstruction (SR-DLR) offers a potential advancement over conventional deep learning reconstruction (C-DLR).

Purpose of the Study:

• To evaluate SR-DLR's potential for further radiation dose reduction in pediatric cardiac CT.
• To compare SR-DLR with C-DLR in terms of image quality and radiation dose.

Main Methods:

• Phantom studies assessed noise power spectrum and spatial resolution.
• Clinical study compared nine C-DLR (conventional dose) and 11 SR-DLR (lower dose) images.
• Evaluation metrics included noise, contrast-to-noise ratio, edge rise distance, and edge rise slope.

Main Results:

• SR-DLR (250 mA) showed equivalent noise to C-DLR (800 mA) with superior spatial resolution in phantom studies.
• SR-DLR significantly improved spatial resolution metrics (edge rise distance/slope) in clinical images (P<0.001).
• No significant differences in noise, contrast, or spatial resolution were found between SR-DLR (4.25±0.54 mGy) and C-DLR (13.3±6.43 mGy).

Conclusions:

• SR-DLR demonstrates non-inferior image quality compared to C-DLR, even at reduced radiation doses.
• SR-DLR holds significant potential for further radiation dose reduction in pediatric cardiac CT.
• Diagnostic image quality can be maintained while lowering radiation exposure in pediatric cardiac imaging.