Comparative evaluation of uncertainty in MTF estimation in computed tomography using periodic bar patterns: analysis...

Objective.To compare the uncertainties in the estimation of the Modulation Transfer Function (MTF) in computed tomography (CT) using periodic bar patterns with two different methods: the variance method (VM) and the harmonic method (HM).Approach.Synthetic CT images of bar phantoms were generated using two reconstruction kernels (Soft and Detail) from a commercial CT scanner. Both white and correlated noise scenarios were simulated. The MTF was calculated using VM and HM across 1000 noise realizations per condition. Theoretical models for uncertainty estimation were developed for the white noise case and validated against simulation results. Whereas white noise simulations served to validate the theoretical model, the conclusions were drawn from the correlated noise simulations, which represent actual CT behavior.Main results.At low and intermediate spatial frequencies, both VM and HM yield low uncertainty in MTF estimation, even under high noise levels. However, significant differences arise at high frequencies. For correlated noise-typical in CT imaging-HM benefits from noise energy reduction at higher frequencies, leading to improved precision. In contrast, VM shows increasing uncertainty in this range. A positive bias is also observed in VM results at high frequencies. These behaviors are consistent with the theoretical predictions for white noise in which the increase in uncertainty at high frequencies was due to an inverse dependence on the MTF value.Significance.The HM offers superior accuracy for MTF estimation at high frequencies in CT imaging with correlated noise. Its robustness against frequency-dependent noise behavior makes it a preferable choice for precise spatial resolution characterization using periodic bar phantoms.