Practical approaches to approximating MTF and NPS in CT with an example application to task-based observer studies.

Area of Science:

• Medical Imaging Physics
• Radiological Sciences
• Image Quality Assessment

Background:

• Accurate characterization of image quality in computed tomography (CT) is crucial for diagnostic performance.
• Modulation Transfer Function (MTF) and Noise Power Spectrum (NPS) are key metrics for quantifying CT image quality.
• Determining MTF and NPS typically requires extensive data acquisition and analysis.

Purpose of the Study:

• To investigate two novel methods for approximating CT MTF and NPS.
• To evaluate these approximations across a range of scan parameters and reconstruction kernels.
• To assess the utility of these methods for limited image acquisitions.

Main Methods:

• A linear systems approach using filtered backprojection (FBP) kernels and derived MTF filter functions.
• An empirical fitted model for approximating MTF and NPS.
• Both methods incorporate a scaling function for mAs and kV variations and compare figure of merits (FOM) like d' and AUC.

Main Results:

• Both approximation methods demonstrated low Root Mean Square Error (RMSE) compared to determined NPS values (max 0.05 of peak).
• The empirical model method yielded superior RMSE for FOM: d' (≤0.02) and AUC (≤0.001).
• The kernel ratio method also showed acceptable RMSE for FOM: d' (≤0.1) and AUC (≤0.01).

Conclusions:

• The proposed methods offer a convenient approach for approximating MTF and NPS in CT.
• These approximations are suitable for applications such as mathematical observer studies.
• Both methods require minimal direct NPS determinations to model MTF/NPS for diverse scan settings.