CT Image Reconstruction Using Dual-Algorithm Blending for Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
CT imaging systems face challenges in generating clear images of dynamic organs like the heart due to motion-related artifacts caused by inconsistent projection data acquisition, leading to increased noise in reconstructed images.
Innovation Solution
A method that decomposes the image object into regions based on temporal behavior, using a combination of full and short scan projection data to reduce noise-related artifacts, by adaptively blending image volumes with different temporal resolution and applying filters to isolate and remove noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the image data is acquired as quickly as possible to minimize motion artifacts, then temporal resolution is improved, but fewer projections are acquired resulting in increased noise in the final image
Solution Approach 1:
The patent segments the image reconstruction process into two distinct stages: a first reconstruction using all available projection data to minimize noise, and a second reconstruction using a limited angular range to maximize temporal resolution. The final image is then formed by combining these two reconstructions, allowing each to optimize for its respective strength.
Solution Approach 2:
The patent changes the reconstruction parameters between two algorithms: the first algorithm uses a full angular range with all projection data for noise reduction, while the second algorithm uses a limited angular range for temporal resolution. By adjusting these parameters and blending the results, the system resolves the contradiction between speed and precision.
2Measurement precision
If more projection data is acquired to reduce noise, then image quality is improved, but the acquisition time increases causing motion-related artifacts
Solution Approach 1:
The patent divides the projection data into two sets: one used for a first reconstruction with full angular coverage to minimize noise, and another used for a second reconstruction with limited angular coverage to maintain temporal resolution. This segmentation allows the system to use more data when needed without increasing overall acquisition time.
Solution Approach 2:
The patent applies partial action by using only the necessary subset of projection data for the second reconstruction (limited angular range), while the first reconstruction uses the full dataset. This partial use of data for specific purposes allows optimization without the penalty of full data acquisition time.
3Device complexity
If a single reconstruction algorithm is used, then the process is simple, but it cannot simultaneously optimize both temporal resolution and noise reduction
Solution Approach 1:
The patent merges the results of two different reconstruction algorithms into a final combined image. The first algorithm optimizes for noise reduction using all projection data, while the second optimizes for temporal resolution using limited angular data. By combining these two reconstructions, the system achieves both goals simultaneously.
Solution Approach 2:
The patent changes the parameters of the reconstruction algorithm between two passes: the first uses full angular range and all projection data, while the second uses limited angular range. This parameter change allows the system to optimize for different qualities in different passes and then combine the results.
Data Source
AI summary
A method for reconstructing an image of an object includes acquiring a set of measured projection data, reconstructing the measured projection data using a first algorithm to generate a first reconstructed image dataset, reconstructing the measured projection data using a second algorithm to generate a second reconstructed image dataset, the second algorithm being utilized to improve the temporal resolution of the second reconstructed image dataset, and generating a final image dataset using both the first and second image datasets.