3D Medical Image Marker Point Identification via Calibration Scale
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for automatically identifying marker points in three-dimensional medical images, such as the iterative closest point algorithm, suffer from low robustness, long calculation times, and a tendency to fall into local optimal solutions, making them unreliable for precise registration in surgical navigation systems.
Innovation Solution
A method involving a calibration scale with 4 to 12 marker points, where information about the scale is read, and DICOM images are processed through threshold segmentation and polygonal data extraction, followed by filtering and matching to calculate a rotation matrix and residual error, ensuring accurate identification of image marker points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the iterative closest point algorithm is used to identify marker points, then registration precision is improved, but calculation time increases and robustness deteriorates
Solution Approach 1:
The patent applies preliminary action by performing threshold segmentation and polygonal data extraction before marker point matching. This pre-processing step prepares the data in advance, creating a structured format that enables faster subsequent matching operations while maintaining precision.
Solution Approach 2:
The patent segments the marker point identification process into distinct stages: threshold segmentation of the DICOM image, extraction of polygonal data, filtering to obtain candidate marker points, and final matching. This segmentation allows each stage to be optimized independently, reducing overall calculation time while preserving accuracy.
2Measurement precision
If the iterative closest point algorithm is used to identify marker points, then registration precision is improved, but robustness deteriorates
Solution Approach 1:
The patent performs preliminary filtering and sorting of candidate marker points before the final matching step. This preliminary action removes obviously incorrect candidates and organizes the data structure, making the algorithm more robust against poor initial conditions and local optimal solutions.
Solution Approach 2:
The patent incorporates feedback mechanisms through iterative refinement of marker point matches. The algorithm uses the matching results to refine the transformation matrix, which then feeds back into the matching process to improve subsequent iterations, enhancing both robustness and precision.
3Productivity
If threshold segmentation and polygonal data extraction are performed, then marker point identification speed is improved, but processing complexity increases
Solution Approach 1:
The patent extracts only the essential features from the DICOM image through threshold segmentation and polygonal data extraction. By taking out only the relevant marker point information and discarding unnecessary data, the method simplifies subsequent processing while maintaining identification speed.
Solution Approach 2:
The patent replaces complex mechanical or manual marker point identification with automated image processing algorithms. The threshold segmentation and polygonal extraction processes automatically identify marker points without manual intervention, improving speed while the modular structure keeps processing complexity manageable.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a method for automatically identifying a marker point sequence in a three-dimensional medical image, including the following steps: 1) selecting a calibration scale, and reading information about the selected calibration scale; 2) reading a DICOM image that includes the calibration scale and marker points and is scanned by a CBCT or CT; 3) performing threshold segmentation on the obtained DICOM image, and performing extraction to generate polygonal data; 4) filtering the polygonal data according to the obtained information about the calibration scale to obtain image marker points in the DICOM image; 5) calculating the distance between each two image marker points in the image respectively; 6) performing a preliminary match on the obtained image marker points, and completing filtering and sorting of the image marker points to obtain scale marker points vector and the corresponding one-to-one image marker points vector; 7) calculating a rotation matrix between the scale marker points vector and the image marker points vector and calculating a residual error; and 8) determining whether the residual error is less than a target error; if NOT, then going to step 6) to perform filtering and sorting again; if YES, then using a result output in step 6) as a matching result. The present invention can quickly and accurately identify multiple image marker points.