Investigation of an FFT-based correlation technique for verification of radiation treatment setup

Electronic portal imaging devices that capture an image of a patient's treatment field electronically by means of a computer data acquisition system operating in real time are becoming available. This paper reports on a study of a field verification correlation algorithm that can compare each treatment portal image to an image of the correct treatment field positioning. The algorithm requires no human intervention or analysis of the images but rather uses fast Fourier transforms to produce a correlation distribution. The position and amplitude of the correlation distribution maximum were tested as objective measures of translational and rotational differences of subject positions between pairs of images. The concept was tested by using a prototype algorithm to obtain the correlation distributions for images of an Alderson Rando head phantom. Images of the phantom setup with various errors were compared with an image of the phantom in the initial, correct treatment position. Translations, in-plane and out-of-plane rotations, and combinations of translations and rotations were studied. The algorithm accurately measured translations. The value of the correlation distribution maximum was found to be a reasonable candidate for an alignment parameter for which tolerable error thresholds might be established.