Virtual X-Ray Image Generation for Structure Pose Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining the position of a structure within the body, such as a tumor or bone, are not efficient or reliable, especially when requiring precise positioning relative to an external reference point for treatment, as they often rely on time-consuming 3D data generation and may not account for structural changes during procedures like radiation therapy.
Innovation Solution
A method utilizing pre-generated 3D data from CT or MRI scans, combined with multiple 2D X-ray images from known viewing directions, generates virtual X-ray images for comparison with actual images to calculate a cumulative similarity measure, allowing for precise determination of structure position relative to a reference coordinate system, while minimizing the influence of noise and unwanted structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If 3D data are generated using CT or MRI before treatment, then the position determination can be performed quickly using pre-generated data, but the time required for data acquisition increases and the structure position may change during the procedure
Solution Approach 1:
The patent generates 3D data (CT or MRI scans) in advance before the treatment procedure, so that when position determination is needed, the data is already available. This preliminary action reduces the time required for position determination during treatment, though it may result in the structure position changing between data acquisition and actual treatment.
2Measurement precision
If multiple 2D X-ray images are used for position determination, then the reliability and precision of structure positioning is improved, but the complexity of data processing and device operation increases
Solution Approach 1:
The patent divides the 3D data into multiple 2D image slices at different depths. Each 2D slice is then registered with corresponding actual X-ray images independently. This segmentation allows the system to process complex 3D positioning problems through simpler 2D comparisons, improving precision while managing computational complexity.
Solution Approach 2:
The patent transforms the 3D positioning problem into a series of 2D image registration problems by slicing the 3D data. This dimensionality change simplifies the comparison process between virtual and actual images, as 2D image matching is computationally more tractable than direct 3D registration, while still achieving accurate 3D position determination through aggregation of 2D results.
3Measurement precision
If image registration is performed iteratively adjusting 3D data position, then the positioning accuracy is improved, but the time required for position determination increases
Solution Approach 1:
The patent performs preliminary alignment by matching anatomical landmarks or reference points between the 3D data and actual X-ray images before iterative optimization. This preliminary action provides a good initial estimate, reducing the number of iterative adjustments needed and thus decreasing registration time while maintaining accuracy.
Solution Approach 2:
The patent may perform registration on a subset of key anatomical features or critical regions rather than processing the entire 3D volume with full iterative registration. This partial action approach achieves sufficient positioning accuracy for clinical purposes while significantly reducing computational time compared to exhaustive registration of all image data.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Method for determining the location of a structure in a body comprising the following steps: - Providing 3D data representing a three-dimensional image of at least the part of the body containing the structure; - Providing at least two 2D image datasets, each representing an X-ray image of at least the part of the body containing the structure from a known viewpoint; - Generating virtual X-ray images from the 3D data for a virtual location of the 3D data, whereby for each viewpoint for which a 2D image dataset has been provided, a corresponding virtual X-ray image is generated; - Pairwise comparing the virtual X-ray images with the corresponding 2D image datasets and generating a cumulative similarity measure (Fcumm) from all comparisons for the virtual location of the 3D data; - Repeating the steps of generating virtual X-ray images.Pairwise comparison and generation of a cumulative similarity measure for different virtual positions of the 3D data and determination of the position of the structure in the body from the virtual position of the 3D data that leads to the highest cumulative similarity measure.