Streak Artifact Correction in CT Slice Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for correcting streak artifacts in slice images, particularly those caused by objects with high X-ray density, are inadequate as they often misidentify anatomical structures like bones and require extensive annotated training data for learning-based methods, leading to errors and inefficiencies.
Innovation Solution
A computer-implemented method and device that processes initially reconstructed slice images using a variation algorithm to determine variation slice and projection images, allowing for the correction of streak artifacts by identifying and masking out pixels contributing to these artifacts, thereby preventing erroneous masking of anatomical structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If threshold value-based masking is applied to mask out metal objects in projection images, then streak artifacts are reduced, but anatomical structures like bones are erroneously masked out
Solution Approach 1:
The patent applies dynamic masking by iteratively updating the mask based on the difference between the current slice image and the reconstructed image from masked projection images. The mask is refined through multiple iterations, transitioning from a static threshold-based approach to a dynamic adaptive approach that progressively improves masking accuracy while preserving anatomical structures.
Solution Approach 2:
The patent implements a feedback mechanism where the masking process uses the difference image (between the original slice image and the reconstructed image from masked projections) to continuously refine and update the mask. This feedback loop allows the system to learn from each iteration and improve the accuracy of metal object identification without falsely masking anatomical structures.
2Measurement precision
If learning-based methods are used for projection-based metal masking, then masking accuracy improves, but extensive annotated training data is required
Solution Approach 1:
The patent enables the system to self-improve through iterative processing of the difference image without requiring external annotated training data. The masking algorithm automatically refines its own mask by using the feedback from reconstruction errors, making the system self-sufficient and eliminating the need for extensive manual annotation of training datasets.
Solution Approach 2:
The patent performs preliminary masking using a simple threshold-based approach before applying the iterative refinement process. This preliminary action provides an initial mask that guides the subsequent iterative improvement, allowing the system to start with a basic level of performance and progressively enhance accuracy without requiring pre-trained models or extensive training data preparation.
3Productivity
If simple threshold value techniques are used for masking, then processing speed is maintained, but false-positive and false-negative errors occur
Solution Approach 1:
The patent applies partial action by performing masking refinement only on regions where the difference image indicates discrepancies between the original and reconstructed images. Rather than processing the entire image at full resolution through multiple iterations, the method focuses computational effort on problematic regions, maintaining processing efficiency while improving masking reliability in critical areas.
Data Source
AI summary
A computer-implemented method is for the correction of streak artifacts in slice images. In an embodiment, the method includes: receiving at least one initially reconstructed slice image by a processor, the at least one initially reconstructed slice image being based on a plurality of initial projection images; determining at least one variation slice image, via the processor, using a variation algorithm, the at least one variation slice image being based on the at least one initially reconstructed slice image; determining at least one variation projection image based upon the at least one variation slice image; and determining at least one corrected slice image as a function of the at least one variation projection image.