Use of volume-rendered images in registration of nuclear medicine studies

Abstract

A simple operator-guided alignment technique based on volume-rendered images was developed to register tomographic nuclear medicine studies, For each of 2 three-dimensional data sets to be registered, volume-rendered images were generated in 3 orthogonal projections (x,y,z) using the method of maximum-activity projection. Registration was achieved as follows: a) One of the rendering orientations (e.g. x) was chosen for manipulation; b) The two-dimensional rendering was translated and rotated under operator control to achieve the best alignment as determined by visual assessment; c) This rotation and translation was then applied to the underlying three-dimensional data set, with updating of the rendered images in each of the orthogonal projections; d) Another orientation was chosen, and the process repeated. Since manipulation was performed on the small two-dimensional rendered image, feedback was instantaneous. To aid in the visual alignment, difference images and flicker images (toggling between the two data sets) were displayed. Accuracy was assessed by analysis of separate clinical data sets acquired without patient movement. After arbitrary rotation and translation of one of the two data sets, the 2 data sets were registered. Mean registration error was 0.36 pixels, corresponding to a 2.44 mm registration error. Thus, accurate registration can be achieved in under 10 minutes using this simple technique.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>