COMBINING HI-RESOLUTION SCAN MODE WITH DEEP LEARNING RECONSTRUCTION ALGORITHMS IN CARDIAC CT.

Area of Science:

• Medical Imaging
• Radiology
• Computational Imaging

Background:

• Computed tomography (CT) imaging is crucial for diagnosis.
• Advancements in CT technology, including high-resolution (Hi-res) scan modes and deep learning image reconstruction (DLIR), aim to improve diagnostic accuracy.
• Understanding the interplay between these technologies is essential for optimizing CT protocols.

Purpose of the Study:

• To evaluate the combined impact of Hi-res scan mode and DLIR on CT image quality.
• To assess the effects on spatial resolution, image noise, contrast-to-noise ratio (CNR), and radiation dose.
• To compare DLIR performance against traditional model-based iterative reconstruction (ASiR-V).

Main Methods:

• CT scans were performed on Catphan600® and Lungman phantoms using standard and Hi-res modes with standard and high-definition (HD) kernels.
• Images were reconstructed using Adaptive Statistical Iterative Reconstruction-Volume (ASiR-V) and three levels of DLIR.
• Quantitative assessments included spatial resolution, noise levels, CNR, and radiation dose measurements.

Main Results:

• Hi-res and HD kernels significantly improved spatial resolution.
• Image noise increased substantially with Hi-res & HD modes (Catphan600®: 69%, Lungman: 10-70%).
• DLIR reduced mean noise by 54% and improved CNR by 64%, without impacting spatial resolution. Radiation dose varied based on phantom size.

Conclusions:

• Hi-res scan mode enhances spatial resolution in CT imaging.
• DLIR effectively mitigates noise and improves CNR, offering superior image quality compared to ASiR-V.
• The combination of Hi-res and DLIR presents a promising approach for advanced CT applications, balancing resolution gains with noise management.