Full model-based iterative reconstruction (MBIR) in abdominal CT increases objective image quality, but decreases...

Area of Science:

• Medical Imaging
• Radiology
• Computed Tomography (CT)

Background:

• Iterative reconstruction (IR) algorithms offer potential improvements over filtered back projection (FBP) in CT imaging.
• Model-based iterative reconstruction (MBIR) represents an advanced form of IR, aiming for superior image quality.

Purpose of the Study:

• To evaluate and compare the image quality and radiologist acceptance of a full MBIR algorithm against a hybrid IR (HIR) and FBP algorithm.
• To assess the impact of different dose levels on image quality metrics for each reconstruction algorithm.

Main Methods:

• CT acquisitions were performed on a 320 detector-row scanner across seven dose levels.
• Images were reconstructed using FBP, full HIR, and full MBIR algorithms.
• Objective assessments included sensitometry, spatial resolution, image texture, and low-contrast detectability. Subjective analysis involved radiologist questionnaires on image quality and diagnostic confidence.

Main Results:

• Full MBIR demonstrated significantly higher contrast-to-noise ratio and improved spatial resolution at high frequencies compared to FBP and HIR.
• Full MBIR achieved better low-contrast detectability at low dose levels, with increased confidence in identifying inserts.
• A notable change in image texture with full MBIR was observed, leading 80% of radiologists to rate its general appearance and texture as worse than HIR, and 50% reporting lower diagnostic confidence.

Conclusions:

• Full MBIR significantly enhances image quality and low-contrast detectability compared to FBP and HIR.
• The altered image texture of full MBIR negatively impacts radiologist acceptance and diagnostic confidence.
• Increased awareness of MBIR's texture characteristics and its inherent quality improvements may enhance clinical adoption.