Longitudinal image deblurring in spiral CT.

Abstract

To assess the feasibility of digital deconvolution techniques to improve longitudinal resolution of spiral computed tomography (CT) multiplanar reformations and evaluate how technical factors in deconvolution affect longitudinal resolution, noise, and edge ringing. Longitudinal line spread function (LSF) of the system was estimated from longitudinal reformations of transaxial spiral CT images of a step test phantom. By using the estimated LSF, longitudinal reformations of the phantom and three clinical spiral CT studies were deconvolved by the methods of Wiener filtering and constrained iterative deconvolution. Edge ringing and image noise were quantified for Wiener filtering and constrained iterative deconvolution. Longitudinal reformations were substantially deblurred and resolution improved after deconvolution. Anatomic boundaries in clinical images were more clearly delineated after restoration. The methods of Wiener deconvolution and constrained iterative deconvolution improved the sharpness of the phantom step boundary at the expense of increased edge ringing and image noise. In longitudinal spiral CT reformations, blurring along the longitudinal axis can be reduced by Wiener filtering or constrained iterative deconvolution.