CT Image Reconstruction Kernel Frequency Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
CT images often suffer from aliasing artifacts due to mismatched convolution kernel cutoff frequencies and image cutoff frequencies, leading to poor image quality and potential misdiagnosis.
Innovation Solution
The method involves determining the image cutoff frequency based on the CT image matrix size and imaging field of view, adjusting the convolution kernel cutoff frequency if necessary, and truncating it using a window function to ensure it matches the image cutoff frequency, thereby eliminating aliasing artifacts through a convolution back-projection algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a preset convolution kernel with fixed cutoff frequency is used for CT image reconstruction, then the reconstruction process is simple and fast, but aliasing artifacts occur when the convolution kernel cutoff frequency does not match the image cutoff frequency, degrading image quality
Solution Approach 1:
The patent makes the convolution kernel dynamically adjustable by allowing its cutoff frequency to be modified based on the specific imaging conditions (FOV, matrix size). The system dynamically selects or adjusts the convolution kernel parameters to match the current reconstruction requirements, transforming a static kernel into an adaptive one that prevents aliasing artifacts while maintaining efficient reconstruction.
Solution Approach 2:
The patent changes the parameters of the convolution kernel (specifically the cutoff frequency) to match the image cutoff frequency calculated from the FOV and matrix size. By adjusting this key parameter, the system ensures optimal matching between the convolution kernel characteristics and the reconstruction parameters, eliminating aliasing artifacts while preserving reconstruction efficiency.
2Measurement precision
If the convolution kernel cutoff frequency is increased to preserve high-frequency details, then image detail resolution improves, but aliasing artifacts increase and image quality deteriorates
Solution Approach 1:
The patent optimizes the convolution kernel cutoff frequency parameter by calculating the image cutoff frequency from the FOV and matrix size, then adjusting the kernel parameters to match. This parameter matching ensures that high-frequency details are preserved up to the appropriate limit without introducing aliasing artifacts, achieving optimal balance between detail resolution and artifact suppression.
Solution Approach 2:
The system implements a feedback mechanism where the image cutoff frequency is calculated based on reconstruction parameters (FOV, matrix), and this calculated value feeds back to adjust the convolution kernel cutoff frequency. This closed-loop parameter adjustment ensures that the kernel always operates at the optimal frequency match, preventing aliasing while preserving necessary detail information.
3Manufacturing precision
If different convolution kernels are used for different imaging conditions to optimize image quality, then image quality improves, but the complexity of selecting and managing multiple kernels increases
Solution Approach 1:
Instead of managing multiple distinct convolution kernels, the patent uses a single kernel whose parameters (cutoff frequency) are dynamically adjusted based on the imaging conditions. This parameter-based approach simplifies kernel management while still achieving optimal image quality for different FOV and matrix size combinations, as the same kernel adapts to different conditions through parameter modification.
Solution Approach 2:
The patent creates a universal convolution kernel that can function across different imaging conditions by making its parameters adjustable. Rather than requiring separate kernels for different scenarios, this single multi-functional kernel adapts to various FOV and matrix size combinations, reducing the number of kernels that need to be stored and managed while maintaining optimal performance for each condition.
Data Source
AI summary
Methods, devices, and apparatus, including computer programs encoded on a computer storage medium for reconstructing an image are provided. An example method includes: acquiring Computed Tomography (CT) projection data of a subject, determining an image cutoff frequency Fimg according to a CT image matrix size for reconstructing a CT image and an imaging field of view size, obtaining a convolution kernel and a cutoff frequency Fker of the convolution kernel, using the convolution kernel as a reconstruction convolution kernel when Fker<Fimg; adjusting Fker to a product of a preset value k and Fimg and truncating the convolution kernel to obtain a portion of the convolution kernel having a cutoff frequency no more than the adjusted Fker as the reconstruction convolution kernel when Fker≥Fimg, and reconstruct the CT image with the CT projection data and the reconstruction convolution kernel by using a convolution back-projection algorithm.