The effects of out-of-plane rotations on two dimensional portal image registration in conformal radiotherapy of the prostate

Purpose: Rotations of the patient out of the image plane can significantly degrade the accuracy of two-dimensional (2D) image registration. This study determines the magnitude of the geometric errors introduced by 2D image registration as a result of out-of-plane rotations, and analyzes the dosimetric effects of these errors. Methods and Materials: The magnitude of the errors introduced by 2D registration were determined by comparing orthogonal view portal images of a rotated phantom to simulator reference images of the same phantom without rotation. Dosimetric effects were calculated for three-dimensional (3D) conformal prostate treatments by applying the registration errors to patient treatment plans, The calculations were performed using a modified version of the dose calculation software used in our Cancer Center for 3D treatment planning based on computed tomography (CT), A method to detect out-of-plane rotations, specific to pelvic treatments, is introduced that uses the relative displacement of the centers of gravity of the acetabula in lateral images. Results: The inherent uncertainty in the registration algorithm was 0.6 +/- 0.5 mm in translation and 0.7 +/- 0.8 degrees in rotation within the image plane, For a 2 degrees out-of-plane rotation, the errors increase to 2.3 +/- 1.0 mm and 1.2 +/- 1.1 degrees, In some clinically realizable treatment scenarios it was observed that the errors introduced by the registration procedure could result in an overdosing of the rectal wall. The method to detect out-of-plane rotations was found to have an accuracy of better than 1 degrees for rotations of less than 10 degrees. Conclusions: The errors introduced to the patient position by 2D image registration have dosimetrically significant consequences for out-of-plane rotations of 2 degrees or more, However when used in conjunction with the method to detect out-of-plane rotations, 2D registration software was found to cause insignificant dose errors and, thus, become a more reliable and accurate clinical tool.