A visi-coil fiducial markers detection method on kV projection images during prostate radiation therapy

Here an accurate automated method based on fast marching approach for detecting visi-coil fiducial markers by using kV imaging is presented. Visi-coil segmentation in each kV image is performed on one marker at a time via fast marching method which requires user input of the endpoints. Initially, a 3D model was extracted by thresholding CT images and is used here as reference. The reference 3D model is re-projected onto the kV images. The new 2D re-projection points provide information about the direction of the markers and zones within which each marker endpoint may lie. This information is employed to detect the start and end point of each visi-coil and remove user input in the fast marching method. A 3D reconstruction of the markers is computed from the segmentation results. The proposed automated visi-coil detection algorithm can increase accuracy of marker location on noisy images and when obscured by boney tissue.