Spiral CT of the lungs: optimal technique and resolution compared with conventional CT

The purpose of this study was to compare the resolution of spiral and conventional CT for imaging the lungs and to optimize spiral CT techniques for imaging lung parenchyma. A line-pair phantom, a dried lung specimen, and three human subjects were scanned. Spiral CT scans obtained on a Somatom PLUS-S CT scanner were compared with conventional CT scans. In addition, the influence of different collimation (1,3, 5, and 8 mm), pitch (1 and 2), display kernel (UltraHi, Hi, Standard, and Soft), and interpolation algorithms (360 degrees and 180 degrees linear interpolation) on spiral CT scans was evaluated. Scans obtained by using different techniques were compared for edge sharpness, contrast resolution, overall visibility of vessels, and noise. With the line bars parallel to the z-axis, a spatial resolution of 7.7 Ip/cm was obtained at all four levels of collimation used for both spiral and conventional CT scanning. Resolution in the z-axis (i.e., direction of table movement) on spiral CT scans decreased markedly when alignment of the line-pair phantom bars was changed to 45 degrees. Resolution along the z-axis decreased with increasing collimation, and this effect was more marked on spiral CT scans than on conventional CT scans. The relative decrease in resolution on spiral CT scans was visible on scans of human subjects only at narrow collimation (1 and 3 mm). Optimal visualization of structures in the lung parenchyma was obtained by using 3-mm collimation and the UltraHi display kernel. The choice of table speed (pitch 1 vs 2) and interpolation algorithm (360 degrees vs 180 degrees linear interpolation) did not significantly affect the image quality of scans of human subjects. CT scans of the lung obtained by using spiral and conventional modes at wide collimation (5 and 8 mm) show no differences in resolution, although, at narrow collimation (1 and 3 mm), curved structures such as walls of bullae and fissures are somewhat indistinct. On spiral CT scans of the lungs, narrow collimation and the UltraHi display kernel provide the best image resolution and have more effect on image quality than does choice of table speed or linear interpolation algorithm.