3D-2D Medical Image Registration with Self-Calibrating Geometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional 3D-2D image registration methods require a known and fixed geometric relationship between the x-ray source and detector, limiting their application to imaging systems with unconstrained source-detector geometry, such as mobile x-ray radiography systems where the x-ray source can be freely positioned.
Innovation Solution
A method of 3D-2D registration that solves for image transformation parameters and source-detector geometry parameters, allowing the x-ray source to be freely positioned relative to the detector, by using image similarity metrics and optimization algorithms to determine the best match between 3D and 2D images without assuming a fixed geometric relationship.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional 3D-2D registration methods are used with a known fixed geometric relationship between x-ray source and detector, then the registration can be performed with six degrees of freedom, but the method cannot be applied to imaging systems with unconstrained source-detector geometry such as mobile x-ray radiography systems
Solution Approach 1:
The patent extends the registration parameters from six degrees of freedom (position and orientation of the 3D volume) to nine degrees of freedom by adding three parameters for source-detector geometry (source-to-detector distance and angles). This parameter expansion allows the method to adapt to mobile imaging systems where the geometric relationship between source and detector is not fixed, eliminating the need for separate geometric calibration while maintaining registration accuracy.
2Reliability
If the x-ray source is rigidly related to the detector, then conventional 3D-2D registration methods work well, but the system cannot accommodate freely positionable x-ray sources
Solution Approach 1:
The patent introduces three additional parameters to the conventional six-degree-of-freedom registration model, creating a nine-degree-of-freedom system. These additional parameters describe the geometric relationship between the x-ray source and detector (source-to-detector distance and angular orientation). By optimizing all nine parameters simultaneously using an iterative algorithm, the method maintains registration accuracy while accommodating any source-detector configuration, including freely positionable mobile x-ray sources.
3Measurement precision
If geometric calibration is performed using a calibration phantom, then the relative position between x-ray source and detector is known, but this limits general application to systems with fixed geometry
Solution Approach 1:
The patent extracts the source-detector geometry parameters from the separate geometric calibration process and integrates them directly into the 3D-2D registration optimization. Instead of determining source-detector geometry through a distinct calibration step using a phantom, the method treats these geometric parameters as unknowns to be solved alongside the image transformation parameters. This integration eliminates the need for geometric calibration while maintaining measurement precision through the optimization process.
Solution Approach 2:
The registration algorithm performs self-calibration by simultaneously optimizing for both image transformation parameters and source-detector geometry parameters. The system uses the image similarity metric (such as mutual information or gradient-based measures) between the 2D projection and the rendered 3D volume to automatically determine the correct geometric relationship, eliminating the need for external calibration phantoms or manual geometric measurement.
Data Source
AI summary
A method of 3D-2D registration for medical imaging includes the following steps: providing a first input interface for acquiring a three-dimensional image; providing a second input interface for acquiring a fixed two-dimensional image using an imaging system that includes a source and a detector and that has an unknown source-detector geometry; initializing image transformation parameters and source-detector geometry parameters; generating a reconstructed two-dimensional image from the three-dimensional image using the image transformation parameters and the source-detector geometry parameters; determining an image similarity metric between the fixed two-dimensional image and the reconstructed two-dimensional image; and updating the image transformation parameters and the source-detector geometry parameters using the image similarity metric, and a corresponding non-transitory computer-readable medium and apparatus.