Fourier Based Patient Posititioning for Radiotherapy Treatment

Abstract

Accurate patient positioning during each fraction of a radiotherapy treatment is crucial to ensure accurate treatment of the tumor while preserving surrounding healthy tissues. The precision of the treatment can be increased with portal imaging devices which provide real time digital images during treatment. Detecting discrepancies in the patient setup can be derived from a comparison of images obtained during the planning phase (reference images) and images taken during each fraction of the treatment (portal images). As the megavoltage images used in treatment have low contrast and resolution, limited contrast adaptive histogram equalization is used to enhance image quality. A Fourier-Mellin transform is used to recover the translation and rotation parameters of the portal images in comparison to a reference image. In previous works, this approach was tested by registering a reference image against a scaled and rotated version of itself. In this paper, we applied the method to two experimental data sets containing phantom images taken on a radiotheraputic simulator and linear accelerator (LINAC). The average error rates obtained for different rotation angles were 0.76 and 2.8 degrees on simulator and LINAC images, respectively. Translational errors were 1.07 and 1.57 pixels in the x and y direction for simulator images and 1.77 and 1.55 pixels on the x, and y axis for LINAC images.