Iterative Image Reconstruction Regularization Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Iterative image reconstruction algorithms with regularization in CT scanners face challenges in achieving consistent image quality across different patient sizes, anatomies, and tube currents due to the dataset-dependent nature of the regularization parameter β, which requires manual adjustment for each patient.
Innovation Solution
The regularization term β is scaled dynamically based on the number of detected photons in the projection data, varying its strength in different image regions to ensure consistent image quality, with a scaling factor algorithm generating a value that adjusts β in real-time during the reconstruction process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a fixed regularization term β is used in iterative reconstruction, then the algorithm is simple to implement, but image quality becomes inconsistent across different patient sizes, anatomies, and tube currents
Solution Approach 1:
The patent applies dynamics by transforming the fixed regularization term β into a dynamic, data-dependent regularization term β(y). The regularization parameter is adjusted automatically based on the projection data y, specifically using metrics like the standard deviation of the projection data or the number of detected photons. This allows the reconstruction algorithm to adapt the regularization strength according to the specific dataset being processed, thereby maintaining consistent image quality across different patient sizes, anatomies, and tube currents without manual intervention.
Solution Approach 2:
The patent implements parameter changes by modifying the regularization term β from a constant value to a variable parameter β(y) that changes based on the input data. Specific implementations include setting β proportional to the standard deviation of the projection data, or β proportional to the average number of detected photons. This parameter adaptation ensures that the regularization strength matches the noise characteristics of each specific dataset, resolving the contradiction between implementation simplicity and image quality consistency.
2Manufacturing precision
If the regularization parameter β is manually adjusted for each patient, then image quality consistency can be achieved, but the operation time and complexity increase significantly
Solution Approach 1:
The patent applies self-service by enabling the reconstruction algorithm to automatically determine the appropriate regularization parameter β without requiring manual input from the operator. The system computes β automatically based on characteristics of the projection data y, such as the standard deviation or photon count statistics. This self-determination mechanism eliminates the time-consuming manual adjustment process while maintaining consistent image quality across different patients and scan conditions.
Solution Approach 2:
The patent implements feedback by using the projection data y to inform the selection of the regularization parameter β. The algorithm analyzes characteristics of the input data (such as noise levels or photon statistics) and uses this feedback to automatically adjust β to the appropriate value for that specific dataset. This closed-loop approach ensures optimal image quality without requiring manual intervention or additional time for parameter tuning.
3Object-affected harmful factors
If stronger regularization is applied to reduce noise, then image noise is reduced, but image detail and resolution may be lost
Solution Approach 1:
The patent applies local quality by making the regularization strength spatially and data-dependently variable rather than uniformly applied. Through the data-dependent parameter β(y), the regularization effect is automatically stronger in regions with higher noise (lower photon counts) and weaker in regions with better signal quality. This localized adaptation preserves image details in high-signal regions while effectively suppressing noise in low-signal regions, resolving the contradiction between noise reduction and detail preservation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method includes scaling a regularization term of an update algorithm of an iterative reconstruction algorithm with regularization with a scaling value. The scaling value is variable in at least one dimension, thereby varying the regularization of the iterative reconstruction in the least one dimension. The method further includes iteratively reconstructing an image based at least on the update algorithm, the varying scaled regularization term, and projection data.