Gated cardiac CT in infants: What can we expect from deep learning image reconstruction algorithm?

Area of Science:

• Medical Imaging
• Radiology
• Artificial Intelligence in Medicine

Background:

• Electrocardiogram (ECG)-gated cardiac computed tomography (CT) is a standard imaging modality for infants with congenital heart disease (CHD).
• Deep Learning Image Reconstruction (DLIR) presents a promising approach to enhance image quality and reduce radiation exposure in cardiac CT.

Purpose of the Study:

• To evaluate the potential radiation dose reduction achievable with DLIR in pediatric cardiac CT using an anthropomorphic phantom.
• To compare the image quality and diagnostic performance of DLIR against traditional iterative reconstruction methods.

Main Methods:

• An anthropomorphic pediatric phantom underwent ECG-gated cardiac CT scans at varying radiation dose levels.
• Images were reconstructed using both Adaptive Statistical Iterative Reconstruction (ASIR-V) and DLIR algorithms.
• Image quality was assessed by evaluating the detectability of high-contrast vessels and the spatial resolution for vessel discrimination.

Main Results:

• DLIR demonstrated a 2.4% improvement in vessel detectability and a 20.9% enhancement in discrimination performance compared to ASIR-V.
• To achieve comparable detection levels, DLIR enabled a dose reduction of up to 64% relative to ASIR-V.

Conclusions:

• DLIR offers significant potential for dose reduction in pediatric cardiac CT.
• DLIR effectively preserves or enhances image quality, providing a valuable alternative to conventional reconstruction techniques like ASIR-V.