Edge detection of the radiation field in double exposure portal images using a curve propagation algorithm

Abstract

An accurate detection of the radiation field in 3D-CRT is crucial to conformal radiotherapy. Automated techniques to detect the field on double exposure portal images have previously focused on thresholding techniques. In this paper, we present a new approach based on a curve propagation technique (the fast marching method) which proves to be more effective at detecting the radiation field than its thresholding counterpart. Comparison both in terms of computational times and effectiveness of the detection are presented: on complex images (non-homogeneous intensity level inside the radiation field, very gradual variation of the intensity level at the boundaries of the radiation field) our results show that the fast marching method is more easily automated, and converges quickly to the boundaries of the radiation field. The computational comparison is made between the fast marching method and an improved implementation of the automatic thresholding technique which uses a binary search to detect an optimal threshold. The execution time of the fast marching technique is on average 5 times faster for our representative set of images.