Temporal resolution and the evaluation of candidate algorithms for four‐dimensional CT

Publisher Website
Google Scholar
Abstract
The four‐dimensional computed tomography (“4D‐CT”) with area detector has been developed for dynamic volumetric imaging with large longitudinal coverage. In this paper one of the key technologies for 4D‐CT development is discussed: Image reconstruction algorithm with high temporal resolution. All of the cone‐beam algorithms investigated previously assume that the object is stationary. In this paper a new class of cone‐beam problem is addressed: a dynamic volumetric (4‐D) imaging. A continuously rotating circular (stationary couch) scanning is employed, and then, a generalized version of the well‐known Feldkamp algorithm with the following three steps is performed: (1) applying a weighting function (along the time axis) to projection data, (2) filtering the weighted data along the detector row direction, (3) cone‐beam backprojecting of the filtered data along the corresponding x‐ray path. The weighting function controls the time center, the temporal resolution, and the image quality. Four weighting functions developed for fan‐beam reconstruction were applied to the first step: (a) a constant weight fixed at 0.5 (FS‐FDK), (b) feathering both edges of the (time) window (OS‐FDK), (c) Parker's weight for a half‐scan (HF‐FDK), and (d) an extended Parker's weight, which allows us to use a larger range of projection data up to one rotation (NHS‐FDK). We evaluated them in terms of temporal resolution, image noise, and image quality. Also, the cause of the artifact has been investigated. The temporal resolution of NHF‐FDK equals that of HS‐FDK, which is half of the one rotation period. For the moving object, NHS‐FDK offers the best image quality. The images with FS‐FDK are degraded by streak artifacts; HS‐FDK provides poor image quality with good temporal resolution; and images by OS‐FDK are blurred due to insufficient temporal resolution. The cause of the artifact was found as an inconsistency of projection data due to object motion (in FS‐FDK) and lost 3‐D‐Radon data caused by applying Parker's weight (in HS‐FDK). A hand toy was employed for the preliminary evaluation of dynamic volumetric imaging with the real 256‐slice scanner. In an overall evaluation, NHS‐FDK provides the stable and the sufficient image quality both with moving and stationary objects.