Image Quality Variation with Gantry Rotation Time and Reconstruction Algorithm in Ultra-high-resolution CT

Area of Science:

• Medical Imaging
• Radiology
• Computed Tomography

Background:

• Ultra-high-resolution CT (U-HRCT) may use longer gantry rotation times to maintain image quality with small focal spots.
• Different CT reconstruction algorithms (DLR, MBIR, FBP) may be affected differently by varying gantry rotation times.

Purpose of the Study:

• To evaluate the impact of gantry rotation time on image quality for deep learning reconstruction (DLR), model-based iterative reconstruction (MBIR), and filtered back projection (FBP) in U-HRCT.

Main Methods:

• A phantom was scanned using U-HRCT at various dose levels and gantry rotation times (0.5s, 0.75s, 1.0s).
• Images were reconstructed using DLR, MBIR, and FBP algorithms.
• Image quality was assessed by analyzing noise characteristics (noise power spectrum) and high-contrast resolution.

Main Results:

• MBIR and FBP demonstrated consistent image quality across all tested gantry rotation times.
• DLR showed significantly reduced high-contrast resolution at a 1.0s rotation time compared to 0.5-0.75s.
• At 1.0s, DLR exhibited a shift in the noise power spectrum towards lower frequencies, indicating degraded noise texture.

Conclusions:

• Deep learning reconstruction (DLR) provides superior image quality at gantry rotation times of 0.5-0.75s.
• At 1.0s, DLR demonstrates a loss of resolution and altered noise texture, potentially due to limitations in processing underrepresented data distributions from its training set.
• Optimizing diagnostic performance requires tailoring U-HRCT scan parameters to the specific reconstruction algorithm used.