In computer vision and image analysis, image registration between 2D projections and a 3D image that achieves high accuracy and near real-time computation is challenging. In this paper, we propose a novel method that can rapidly detect an object’s 3D rigid motion or deformation from a 2D projection image or a small set thereof. The method is called CLARET (Correction via Limited-Angle Residues in External Beam Therapy) and consists of two stages: registration preceded by shape space and regression learning. In the registration stage, linear operators are used to iteratively estimate the motion/deformation parameters based on the current intensity residue between the target projec-tion(s) and the digitally reconstructed radiograph(s) (DRRs) of the estimated 3D image. The method determines the linear operators via a two-step learning process. First, it builds a low-order parametric model of the image region’s motion/deformation shape space from its prior 3D images. Second, using learning-time samples produced from the 3D images, it formulates the relationships between the model parameters and the co-varying 2D projection intensity residues by multi-scale linear regressions. The calculated multi-scale regression matrices yield the coarse-to-fine linear operators used in estimating the model parameters from the 2D projection intensity residues in the registration. The method’s application to Image-guided Radiation Therapy (IGRT) requires only a few seconds and yields good results in localizing a tumor under rigid motion in the head and neck and under respiratory deformation in the lung, using one treatment-time imaging 2D projection or a small set thereof.
Purpose: To evaluate traditional 3D/3D and 3D/2D rigid registration strategies for tomosynthesis images obtained from the Nanotube Stationary Tomosynthesis (NST) geometry. Method and Materials: NST is a multi‐source kV imager which is mounted on a linear accelerator gantry. The multiple sources allow imaging without gantry motion before and concurrent with radiation treatment. Due to the nature of the reconstructed images it is not immediately clear how to register tomosynthesis images to planning CTs. Depending on the amount of angular sampling in the geometry better performance can be achieved with 3D/3D registration as is the case with cone‐beam CT or 3D/2D registration as is the case with portal imaging. Tomosynthesis images contain angular sampling somewhere in between these two extremes. The question remains whether NST images should be considered a set of 2D projections or a 3D volume for the purpose of rigid registration. Simulated NST images were used to evaluate treatment time rigid registration for patient setup. Two GPU‐accelerated planning CT to tomosynthesis rigid registration methods were considered characterized by the domain in which the similarity metric is computed. Results: Simulated data sets suggest that evaluation of the similarity metric in projection space reduces mean target registration error (mTRE) and increases speed over reconstruction space methods. A rate limiting step of the 3D/3D method is the requirement for repeated iterative reconstructions. Conclusion: We have demonstrated that 3D/2D methods are faster and result in decreased mTRE over 3D/3D methods for rigid registration of NST images. It is suggested that 3D/2D methods will be faster in cases where a significant number of reconstructions must be generated. Conflict of Interest: This work is partially funded by a grant from Siemens Medical.
We have designed and built a stationary digital breast tomosynthesis (DBT) system containing a carbon nanotube based field emission x-ray source array to examine the possibility of obtaining a reduced scan time and improved image quality compared to conventional DBT systems. There are 25 individually addressable x-ray sources in our linear source array that are evenly angularly spaced to cover an angle of 48°. The sources are turned on sequentially during imaging and there is no motion of either the source or the detector. We present here an iterative reconstruction method based on a modified Ordered-Subset Convex (MOSC) algorithm that was employed for the reconstruction of images from the new DBT system. Using this algorithm based on a maximum-likelihood model, we reconstruct on non-cubic voxels for increased computational efficiency resulting in high in-plane resolution in the images. We have applied the reconstruction technique on simulated and phantom data from the system. Even without the use of the subsets, the reconstruction of an experimental 9-beam system with 960x768 pixels took less than 6 minutes (10 iterations). The projection images of a simulated mammography accreditation phantom were reconstructed using MOSC and a Simultaneous Algebraic Reconstruction technique (SART) and the results from the comparison between the two algorithms allow us to conclude that the MOSC is capable of delivering excellent image quality when used in tomosynthesis image reconstruction.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.