Limited-data image registration for radiotherapy positioning and verification.

Abstract

One benefit to having on-line CT imaging integrated into a radiotherapy system is that images can be collected with the patient in the treatment position. These images can then be automatically registered to planning images for improved positioning and verification. However, many such on-line imaging systems have a limited field of view (LFOV) that could potentially impair registration. Thus, the viability of automatic registration was investigated in the context of collecting on-line LFOV and also limited-slice CT images for radiotherapy. Mutual information and two new voxel-based registration algorithms were tested to align LFOV and limited-slice prostate and breast images given known displacements. Success rates were tallied for different field-of-view sizes, slice distributions, and initial displacements. Most of the automatic registration algorithms tested were useful for solving these LFOV and limited-slice problems. Registration of LFOV images was generally successful, especially for fields of view of at least half the patient's size. For limited-slice images, success was more closely correlated to the slice spacing than to the number of slices used, with sparse slice spacing being preferable. Mutual information and other automatic registration algorithms have been identified as useful methods for registering LFOV and limited-slice radiotherapy images with planning CT images.