X-ray CT Reconstruction Using ATRACT with Intensity Gradient Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computed tomography (CT) with volume of interest (VOI) collimation leads to data truncation, resulting in inconsistent image reconstruction and artifacts, especially when the volume of interest is not centered, due to removal of high-frequency spikes which affects global intensity gradients.
Innovation Solution
The ATRACT algorithm is extended with correction steps that include determining a fit parameter for the projection data characteristics at image region boundaries and applying an additive correction using a scaling factor to reintroduce lost information, effectively reducing artifacts by compensating for intensity gradient and bias-like artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If high-frequency spikes are removed after Laplace filtering to prevent cupping/capping artifacts, then truncation robustness is improved, but intensity gradient information is lost causing artifacts in non-centered cases
Solution Approach 1:
The patent applies feedback by detecting intensity gradient artifacts in the reconstructed image and using this information to adjust the filtering process. The algorithm calculates intensity gradients in the reconstructed image, identifies regions with artifacts, and feeds this information back to modify the residual filtering parameters, thereby compensating for the information loss from spike removal.
Solution Approach 2:
The patent changes filtering parameters dynamically based on the detected artifact severity. By adjusting the residual filtering strength or range in response to detected intensity gradient artifacts, the system adapts the processing parameters to recover lost information while maintaining truncation robustness.
2Object-affected harmful factors
If collimators are used to restrict X-ray field to volume of interest, then radiation dose is reduced, but data truncation occurs leading to inconsistent reconstruction
Solution Approach 1:
The patent segments the filtering process into local and nonlocal components. The local Laplace filtering handles high-frequency noise and spikes, while the nonlocal residual filtering recovers low-frequency intensity gradient information. This segmentation allows the system to handle truncated data from collimated views by processing different frequency components separately and combining them to produce consistent reconstructions.
Solution Approach 2:
The patent introduces an intermediary intensity gradient compensation step between the Laplace filtering and final reconstruction. This intermediary process detects and corrects intensity gradient artifacts caused by data truncation, acting as a mediator that reconciles the conflicting requirements of spike removal and gradient preservation.
3Productivity
If ATRACT algorithm is applied to non-centered volumes of interest, then reconstruction speed is maintained, but intensity gradient artifacts occur due to information loss
Solution Approach 1:
The patent applies preliminary intensity gradient analysis before final reconstruction. By detecting and compensating for intensity gradient artifacts in an intermediate step before completing the reconstruction, the algorithm prevents quality degradation while maintaining the efficient ATRACT workflow and reconstruction speed.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly improves image quality by reducing artifacts in non-centered cases, achieving high-quality three-dimensional image datasets that are largely free of intensity gradient artifacts, as confirmed by experiments with phantom and clinical data.
Implementation Method 1
The collimator thus reduces the X-ray dose significantly by using collimation to block (or at least severely attenuate) X-ray radiation in regions outside a predetermined volume of interest
Implementation Method 2
The computed tomography uses ionizing radiation as the X-ray technology
Data Source
AI summary
A method for reconstructing a three-dimensional image dataset from two-dimensional projection images includes recording the projection images using a collimator downstream of an X-ray source. A local Laplace filter is initially applied to the projection data of the projection images during the reconstruction using filtered back-projection. After this, high-frequency spikes arising in the Laplace-filtered projection data at boundaries to the image region are removed by a spike filter, and a global residual filter is applied. A fit parameter describing a fit function approximating the projection data characteristic in the uncorrected projection images inside the image region is determined based on at least marginal values of the projection data present at the boundaries. Next, following on from the use of the residual filter, an additive correction of the residual-filtered projection data in the image region is performed with the fit function scaled by a scaling factor.