CT Reconstruction Using Adaptive Resolution and Voxel Division
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional CT reconstruction methods are inefficient in terms of time, particularly when higher resolution is required for volume data, as they do not effectively adapt to the complexity of the object's shape, leading to unnecessary processing in areas with minimal shape changes.
Innovation Solution
A CT reconstruction method using filtered back projection that dynamically adjusts the resolution of projection images and voxel division based on the object's complexity, reducing the number of pixels and images, and provisionally dividing voxels only where significant changes occur, thereby optimizing processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If higher resolution is used for CT reconstruction, then measurement precision is improved, but reconstruction time increases significantly
Solution Approach 1:
The patent applies different resolution levels to different regions of the object based on shape complexity. Regions with simple shapes use lower resolution, while regions with complex shapes use higher resolution. This is achieved by calculating shape complexity metrics for different regions and dynamically adjusting the reconstruction resolution accordingly, thereby reducing overall reconstruction time while maintaining measurement precision where needed.
Solution Approach 2:
The patent dynamically adjusts the resolution level during the reconstruction process based on the detected shape complexity of the object. The system evaluates the object's geometry and adaptively modifies reconstruction parameters in real-time, transitioning from uniform high-resolution reconstruction to variable resolution reconstruction that optimizes the balance between precision and speed.
2Measurement precision
If uniform high resolution is applied to entire volume data, then measurement precision is improved, but processing efficiency deteriorates
Solution Approach 1:
The patent implements local quality optimization by dividing the volume data into multiple regions and applying different resolution levels to each region based on its shape complexity. Regions with simple geometric features are reconstructed at lower resolution, while regions with complex features receive higher resolution processing. This selective approach maintains measurement precision for complex regions while significantly improving overall reconstruction efficiency.
Solution Approach 2:
The patent segments the volume data into multiple regions based on shape complexity characteristics. By identifying and separating regions with different geometric complexities, the system can apply optimized reconstruction strategies to each segment independently, avoiding the computational overhead of uniform high-resolution processing across the entire volume.
3Manufacturing precision
If resolution is increased in all areas, then manufacturing precision is improved, but loss of time increases due to unnecessary processing
Solution Approach 1:
The patent applies local quality optimization by evaluating shape complexity in different regions of the workpiece and adjusting resolution accordingly. Regions with simple shapes that do not require high manufacturing precision are processed at lower resolution, while regions with complex geometries that demand high precision are processed at higher resolution. This targeted approach eliminates unnecessary processing time in simple regions while maintaining manufacturing precision where required.
Solution Approach 2:
The patent dynamically changes the resolution parameter during reconstruction based on local shape complexity analysis. By modifying the resolution parameter adaptively across different regions rather than maintaining a uniform high resolution, the system achieves manufacturing precision for complex features while reducing processing time for simpler features through parameter optimization.
Data Source
AI summary
Projection images of reduced resolution are generated by reducing resolution of filtered projection images and/or reducing the number of filtered projection images. Volume data of reduced resolution is generated by performing CT reconstruction using the projection images of reduced resolution. Each voxel of the volume data of reduced resolution is provisionally divided. The provisionally divided voxels are compared in voxel value before and after provisional division. If a difference in voxel value before and after the provisional division is greater than a threshold, the provisional division is determined to be valid, and division is further continued. If the difference in voxel value before and after the provisional division is less than or equal to the threshold, the provisional division is determined to be invalid and the voxel ends being divided.