Influence of retrospective sorting on image quality in respiratory correlated computed tomography.

Abstract

To evaluate the influence of retrospective sorting on image quality in four-dimensional respiratory correlated CT. Twelve patients with intrapulmonary tumors were examined using a 24-slice CT-scanner in helical mode. Images were reconstructed after retrospective sorting based on five algorithms: amplitude-based sorting with definition of peak-exhalation and peak-inhalation separately/locally for all breathing cycles (LAS) and globally for the time of image acquisition (GAS). Drifts of the breathing signal were corrected in dc-GAS. In phase-based (PS) and cycle-based (CS) algorithm the projections were sorted relative to time. Motion artifacts were scored by a radiologist. The tumor volumes were measured using automatic image segmentation. Averaged over all breathing phases, LAS and PS achieved significantly improved image quality and lowest tumor volume variability compared to GAS, dc-GAS and CS. Imaging redundancy of 5s was not sufficient for GAS and dc-GAS: missing corresponding amplitude positions in one or several breathing cycles resulted in incomplete reconstruction of peak-ventilation images in 11/12 and 10/12 patients with GAS and dc-GAS, respectively. Limiting the analysis to mid-ventilation phases showed GAS and dc-GAS as the most reliable algorithms. LAS and PS are suggested as a compromise between image quality and radiation dose.