A structure-based method for on-line matching of portal images for an optimal patient set-up in radiotherapy

Abstract

In radiotherapy, portal images are used to ensure a correct patient position during every radiation session. A reliable on-line verification is of clinical interest to interrupt the radiation in time in case the patient is not at the right position. A great problem for successful image registration is the poor image quality of portal images. They are corrupted by noise and of very low contrast. A method directly based on the grey levels is not sufficient. Therefore a structure-based method was developed which is almost insensitive to distrubances (air bubbles, noise, slowly varying grey levels). In most cases the selection of a region of interest (ROI) can be omitted. Besides the automatical segmentation of the radiation field, only the structures relevant for matching the anatomy are enhanced by using a bandpass filter. It is possible to detect the maximum correlation between different image modalities reliably (simulator image, digitally reconstructed radiograph, portal image). By using Fast Fourier Transformation (FFT), the calculation time is smaller than five seconds, which enables a clinical on-line verification. We have matched 1139 pairs of images of different modalities and various regions of the body (pelvis, nasopharyngeal space, head, lung). The success rate is greater than 95%.