2D-3D Image Registration Using Simulated X-Ray Projections
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image registration methods for bronchoscopy, such as 2D-3D registration, face challenges in fluoroscopic guidance due to low image contrast in X-ray images, where soft tissue structures like lesions and airways are difficult to visualize, and existing methods rely on bony features that exhibit duplicative patterns, making real-time data fusion infeasible.
Innovation Solution
An instrument-based 2D/3D registration system that segments tubular structures from CT images and projects them into 2D space, using Digital Radiological Reconstruction to simulate X-ray fluoroscope images, and registers the segmented airways with the endoscope displayed in fluoroscopic images, enabling effective fluoroscopic guidance by fusing anatomic structures and real-time instrument images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If feature-based registration methods are used, then computational speed is improved, but registration accuracy deteriorates due to insufficient distinctive features in X-ray images
Solution Approach 1:
The patent introduces a simulated X-ray image as an intermediary representation of the 3D CT data. This simulated image serves as a mediator that can be directly compared with the actual 2D X-ray fluoroscopy image, enabling intensity-based registration without relying on extracted features. The simulation process transforms 3D anatomical structures into a 2D projection that matches the viewing geometry of the fluoroscopy, creating a common reference frame for accurate alignment.
Solution Approach 2:
The patent replaces the mechanical feature extraction and matching system with an intensity-based comparison system. Instead of detecting edges, corners, or specific anatomical landmarks and matching them between modalities, the system compares the overall intensity patterns of the simulated 3D projection with the actual 2D X-ray image. This substitution leverages the full intensity information available in both images, improving registration accuracy while maintaining computational feasibility through efficient simulation algorithms.
2Measurement precision
If intensity-based registration methods are used, then registration accuracy is improved, but computational overhead increases making real-time data fusion infeasible
Solution Approach 1:
The patent performs preliminary actions by pre-segmenting the 3D CT data to identify relevant anatomical structures (such as blood vessels, bones, or organs) before the registration process. This segmentation is done once on the high-resolution 3D data, and the segmented structures are then used to generate the simulated 2D projection. By preparing this information in advance, the actual real-time registration only requires comparing intensity patterns rather than processing raw 3D volumes, significantly reducing computational overhead during the procedure.
Solution Approach 2:
The patent extracts only the essential intensity information from the 3D CT data by projecting it into a 2D simulated X-ray image that matches the fluoroscopy geometry. This extraction process removes unnecessary 3D spatial information and retains only the relevant 2D intensity patterns that can be directly compared with the actual X-ray image. By taking out only the necessary information for registration, the system reduces computational complexity while maintaining registration accuracy.
3Difficulty of detecting and measuring
If bony features are used for registration, then distinctive features are available, but registration accuracy deteriorates due to duplicative patterns in space
Solution Approach 1:
The patent merges multiple types of anatomical information into a unified intensity-based comparison. Instead of relying solely on bony features, the simulated 3D projection incorporates intensity patterns from all segmented structures (bones, soft tissues, blood vessels, airways) in their correct spatial relationships. This merging creates a more comprehensive and distinctive intensity pattern that reflects the unique anatomical configuration of each patient, eliminating the ambiguity caused by duplicative bony patterns alone.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This method allows for precise localization of lesions and their spatial relationship to the endoscope, facilitating real-time lesion visualization and effective fluoroscopic guidance in bronchoscopy procedures, even in areas with limited visible features.
Implementation Method 1
projecting the three-dimensional image of the organ into two-dimensional space to provide a projected image
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A system and method for registering three-dimensional images with two-dimensional intra-operative images includes segmenting (24) a tubular structured organ in a three- dimensional image of the organ, and projecting (26) the three-dimensional image of the organ into two-dimensional space to provide a projected image. A medical instrument depicted in a two-dimensional image of the medical instrument is segmented (28). A similarity score is computed between the projected image and a shape of the medical instrument depicted in the two-dimensional image to determine a best match. The projected image is registered (30) to the two-dimensional image based on the best match.