Towards task-based assessment of CT performance: system and object MTF across different reconstruction algorithms.

Area of Science:

• Medical Imaging Physics
• Radiological Sciences
• Image Quality Assessment

Background:

• Accurate assessment of computed tomography (CT) system resolution is crucial for diagnostic performance.
• Conventional methods for measuring modulation transfer function (MTF) may not fully represent real-world imaging tasks.
• Evaluating resolution across diverse reconstruction algorithms, radiation doses, and contrast levels is essential.

Purpose of the Study:

• To develop and validate a novel measurement method for evaluating the task-based modulation transfer function (MTF) of CT imaging systems.
• To assess the impact of different reconstruction algorithms (FBP, ASIR, MBIR), dose, and contrast on CT image resolution.
• To provide a more comprehensive understanding of CT system performance beyond conventional metrics.

Main Methods:

• Developed an algorithm to extract task-based MTF (MTF(Task)) from disk images of ACR phantom rod inserts.
• Validated the MTF(Task) method against the conventional wire technique.
• Measured and compared MTF(Task) for filtered back projection (FBP), adaptive statistical iterative (ASIR), and model-based iterative (MBIR) reconstruction algorithms across varying dose and contrast levels.

Main Results:

• The MTF(Task) method using inserts correlated well with the wire-based method for FBP reconstructions.
• For ASIR and MBIR, high-contrast MTF(Task) was comparable or superior to FBP, but low-contrast MTF(Task) was lower.
• Iterative reconstructions (ASIR, MBIR) showed MTF(Task) varying with noise (proportional to mA), unlike FBP; MBIR exhibited greater variability across dose and contrast.

Conclusions:

• The developed method enables task-based MTF assessment for CT systems using conventional and iterative reconstructions.
• Object-specific MTF is dependent on imaging parameters like dose and contrast, particularly for iterative algorithms.
• Iterative reconstructions offer noise benefits but present a trade-off in resolution that varies with imaging conditions, necessitating task-based evaluation.