Statistical phase alignment of edge spread function for modulation transfer function measurement on computed tomography images

Purpose: This study aimed to develop a statistical approach for edge spread function (ESF) phase alignment to improve the accuracy of modulation transfer function (MTF) measurements at the edges of computed tomography (CT) images. Methods: A statistical approach to ESF phase alignment was initiated by collecting ESF samples with poor phase alignment. One ESF sample was selected as the reference ESF and the other as the treated ESF. The treated ESF was then shifted by 10-pixels in the right and left directions with a 1-pixel increment at each shift. The mean squared error (MSE) for each shift was calculated between the shifted and reference ESF, and the position with the minimum MSE indicated the best phase alignment between the two ESFs. All shifted ESFs were averaged and differentiated to obtain a single-line spread function (LSF). The MTF was generated by Fourier transformation of the LSF. The MTFs from the shifted ESF and the non-shifted MTF from images of the ACR CT, point-computational, CTDI, and anthropomorphic phantoms were compared. Results: The MTF curves obtained after the phase alignment showed higher and more consistent results than those obtained before the alignment. The MTF curves obtained after phase alignment were comparable to those obtained from a point computational phantom. Our method showed improved accuracy in measuring spatial resolution compared to those without the edge-shifting method. Conclusions: The results showed that a statistical approach for ESF phase alignment can overcome poor phase alignment and produce a more accurate MTF curve.