Automated Fiducial Marker Tracking in X-Ray Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for detecting fiducial markers in x-ray images are inefficient due to the need for manual selection and assumptions about marker shapes, making fully automated detection challenging, especially in the presence of high-contrast features like bony anatomy and metallic objects.
Innovation Solution
The development of a system that generates high-quality, dynamic templates from a single CBCT scan, using filtered marker-enhanced images and iterative processes to create static and dynamic templates for accurate fiducial marker tracking, capable of handling motion and deformation, and incorporating breathing data for improved tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If template matching methods are used to detect fiducial markers, then marker detection can be performed, but detection accuracy deteriorates due to interference from high-contrast features like bony anatomy and metallic objects
Solution Approach 1:
The patent segments the detection process into multiple stages: initial marker detection, template generation from detected markers, and refined marker tracking. This segmentation allows the system to first identify markers without interference, then use those markers to create specialized templates that can distinguish markers from interfering high-contrast features like bone and metal objects in subsequent detection
Solution Approach 2:
The system performs preliminary detection of fiducial markers to generate templates before actual tracking begins. These pre-generated templates capture the specific appearance characteristics of the implanted markers, enabling the tracking algorithm to differentiate markers from interfering anatomical structures and improve detection accuracy during real-time tracking
2Extent of automation
If manual selection and shape assumptions are used in traditional detection methods, then some level of automation can be achieved, but full automation deteriorates due to the need for user intervention and shape assumptions
Solution Approach 1:
The system implements self-service by automatically detecting fiducial markers in the initial stage and using those detected markers to generate templates for subsequent tracking. This eliminates the need for manual template creation and shape assumptions, as the system generates its own detection parameters from the actual marker appearances observed in the imaging data
Solution Approach 2:
The system dynamically changes detection parameters based on actual marker observations. Instead of using fixed shape assumptions, the system adapts template parameters (such as marker size, shape, and appearance characteristics) based on the specific markers detected in the patient's anatomy, enabling fully automated detection without user intervention while maintaining operational simplicity
3Device complexity
If static templates are used for marker tracking, then computational complexity is reduced, but tracking accuracy deteriorates when markers undergo motion or deformation
Solution Approach 1:
The patent implements dynamic template updating where templates are initially generated from detected markers and then refined iteratively based on tracking results. This dynamic adaptation allows templates to accommodate marker motion and deformation while maintaining computational efficiency, as the system updates templates based on observed marker positions and appearances rather than using complex real-time models
Data Source
AI summary
Various embodiments of the present technology generally relate to identification of tumor location. More specifically, some embodiments of the present technology relate automated tracking of fiducial marker clusters in x-ray images for the real-time identification of tumor location and guidance of radiation therapy beams. Some embodiments use processed CBCT projection images, an automated routine of reconstruction, forward-projection, tracking, and stabilization generated static templates of the marker cluster at arbitrary viewing angles. Breathing data can be incorporated into some embodiments, resulting in dynamic templates dependent on both viewing angle and breathing motion. In some embodiments, marker clusters can be tracked using normalized cross correlations between templates (either static or dynamic) and CBCT projection images.