A new technique based on a rho-theta coordinate system for determining differences in patient position between portal and simulator images is presented. Unlike the conventional point matching method, which requires the fiducial points to be labeled in pairs before the registration, the rho-theta technique avoids this manual procedure. It accomplishes the treatment verification in two major steps; image alignment and field displacement analysis. For the same number of fiducial points in the simulator and portal images, it first finds the corresponding paired points if the points are not distributed symmetrically about their centroid. This is followed by alignment of these paired points using the least squares matching method to find the optimal two-dimensional rigid body transformation parameters (shift, rotation, and scaling factor). The transformation parameters are then used to transform the portal field edge into the simulator image, so that the portal field can be compared with the prescribed field on the simulator image. A number of parameters were explored to describe the field displacement errors, including treatment field size, under/over irradiated size, the shift in center of gravity of the field, the field edge shift, and rotation of the field. The rho-theta technique as implemented is both fast and accurate. Experiments on the registration of radiological phantom portal images acquired with an on-line portal imaging system mounted on a linear accelerator indicate an accuracy on the order of 1 mm in detecting the shift of the field's center of gravity and approximately 1 degree in detecting the field rotation. The results of a clinical trial are also presented.(ABSTRACT TRUNCATED AT 250 WORDS)