Iterative Image Reconstruction Regularization Factor Automation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The manual determination of the regularization factor β in iterative image reconstruction and de-noising algorithms for computed tomography (CT) is time-intensive and dependent on specific image and dataset characteristics, requiring multiple iterations to achieve a suitable final image.
Innovation Solution
An automated method to determine the regularization factor β based on an image quality metric and a predetermined regularization level, allowing the regularization factor to be updated iteratively until a desired quality metric is reached, reducing the need for manual trial and error and minimizing computational time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual determination of regularization factor β is performed through multiple iterations, then image quality is improved, but time consumption increases significantly
Solution Approach 1:
The system performs self-adjustment by automatically determining the regularization factor β through an algorithm that evaluates image quality metrics and iteratively optimizes the parameter without requiring manual intervention. The processor automatically updates β based on the relationship between the initial noisy image and the reconstructed image, eliminating the need for operators to manually test multiple values.
Solution Approach 2:
The invention dynamically changes the regularization factor β based on image quality metrics and the relationship between initial and reconstructed images. The system adjusts β iteratively during the reconstruction process, transitioning from a static manually-determined parameter to a dynamic automatically-optimized parameter that adapts to the specific imaging conditions and noise characteristics.
2Adaptability or versatility
If the regularization factor β is manually determined for each image and dataset, then the algorithm adapts to specific characteristics, but the process becomes computational and time intensive
Solution Approach 1:
The system automatically adapts to each image and dataset by computing the optimal regularization factor β through an embedded algorithm that analyzes the specific characteristics of the input data. The processor performs self-adjustment by evaluating image quality metrics and updating β iteratively, eliminating the need for manual determination while maintaining adaptability to different imaging conditions.
Solution Approach 2:
The system performs preliminary automatic determination of the regularization factor β before final image reconstruction is completed. By pre-computing the optimal β value through iterative evaluation of image quality metrics, the system prepares the optimal parameter in advance, avoiding time-consuming manual adjustments during the reconstruction process.
3Measurement precision
If multiple iterations with different β values are run in parallel or series, then the desired β is identified, but computational resources are heavily consumed
Solution Approach 1:
The system performs self-optimization by automatically determining the optimal regularization factor β through a single integrated iterative process that evaluates image quality metrics and updates β dynamically. This eliminates the need to run multiple separate iterations with different β values in parallel or series, reducing computational resource consumption while maintaining accuracy.
Solution Approach 2:
The system implements a feedback mechanism where the regularization factor β is continuously updated based on the evaluation of image quality metrics from the reconstructed image. The processor uses this feedback to iteratively adjust β, converging to the optimal value without requiring multiple independent trial iterations, thereby reducing computational resource usage while achieving accurate determination.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A processing component (122) processes images based on an iterative reconstruction algorithm with regularization and/or de-noising algorithm. The processing component includes a set point determiner (224) that determines a quality set point (216) between predetermined lower and upper quality bounds (226) based on a quality variable (228) indicative of an image quality of interest. The processing component further includes a comparator (214) that compares, each processing iteration, a quality metric of a current generated image with the quality set point and generates a difference value indicative of a difference between the quality metric and the quality set point. The processing component further includes a regularization factor updater (220) that generates an updated regularization factor for a next processing iteration based on a current value (222) of the regularization factor and at least the quality metric in response to the difference value indicating that the quality metric is outside of a predetermined range about the quality set point.