METHODOLOGIC CONSIDERATIONS IN MULTIPLANAR REFORMATTING OF TRANSAXIAL CT IMAGES

Abstract

The ability to reformat conventional transaxial images obtained by computed tomography (CT) into coronal and saggital sections is now an option available on several commercial CT scanners. We evaluated two important technical variables that can greatly influence the quality of the reformatted image: (1) slice thickness of the transaxial sections, and (2) slice thickness of the reformatted image. Using two 3rd generation scanners (GE 8800 and Siemens Somatom II), studies were performed on phantoms designed or modified to evaluate high contrast (13%) and low contrast (3%) objects. Clinical studies of the spine and a bladder tumor were also evluated. Contiguous transaxial images of slice thickness 10, 5, and 1.5 mm were separately obtained, and the multiplanar images were reformatted into images from 1 to 20 pixels in thickness Analysis of the multiplanar reformatted images showed that spatial resolution of high contrast objects were best using the narrowest (1.5mm) collimation for the transaxial images. Geometric distortion of spherical test objects along the z axis was minimized by use of the 1.5 mm collimation. Edge definition of low contrast objects was degraded by the noise present on image reformatted from 1.5 mm thick slices, but this disadvantage could be compensated by reformatting the images into slices several pixels thick. Reformatted images too thick in size resulted in distortion or lack of detection of small objects. Selection of technical variables for multiplanar reformatting of CT images is a complex problem, ultimately dependent on the expected size and relative density of the object being evaluated. Our study demonstrates the value of 5 and 1.5 mm collimation of the transaxial slices, and shows that the increased noise resulting from narrow collimation can be overcome by judicious use of thick pixel reformation rather than increased radiation dosage.