Iterative Image Reconstruction Using Optimization-Transfer Algorithm
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical image reconstruction algorithms, particularly in PET and X-ray CT imaging, are computationally intensive and slow due to their statistical nature, necessitating improved methods for faster convergence and reduced computational resources to achieve better image quality at lower radiation doses and real-time feedback.
Innovation Solution
The implementation of an optimization-transfer algorithm using a quadratic surrogate function with curvature, calculated using an inverse-background image, combined with ordered subsets and Nesterov acceleration methods, for iterative image reconstruction, which accelerates convergence and reduces computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If statistical image reconstruction algorithms are used to improve image quality at reduced radiation doses, then manufacturing precision is improved, but productivity deteriorates due to computationally intensive processing
Solution Approach 1:
The patent applies ordered subsets decomposition by dividing the complete set of projection data into multiple subsets. Each iteration processes only one subset rather than all data, significantly reducing computational workload per iteration while maintaining convergence to the optimal solution. This segmentation enables faster reconstruction without sacrificing image quality.
Solution Approach 2:
The patent implements periodic action through iterative reconstruction where the algorithm cycles through multiple passes over the data. Each iteration refines the image estimate periodically, allowing the system to achieve high image quality through repeated refinement rather than requiring excessive computational resources in a single pass.
2Measurement precision
If conventional iterative reconstruction methods are used to achieve accurate image reconstruction, then measurement precision is improved, but loss of time increases due to slow convergence
Solution Approach 1:
The patent applies preliminary action by using Nesterov acceleration techniques that incorporate momentum from previous iterations to predict and accelerate convergence toward the solution. This preliminary momentum building allows the algorithm to reach accurate reconstruction faster than conventional methods that process each iteration independently.
Solution Approach 2:
The patent changes parameters by dynamically adjusting acceleration factors and subset sizes during the reconstruction process. These parameter modifications allow the algorithm to optimize convergence speed while maintaining reconstruction accuracy, adapting to the specific characteristics of the data being processed.
3Measurement precision
If high radiation doses are used to improve signal-to-noise ratio in measured signals, then measurement precision is improved, but object-affected harmful factors increase due to radiation exposure
Solution Approach 1:
The patent replaces the mechanical approach of increasing radiation dose to improve signal quality with a computational approach. By using advanced iterative reconstruction algorithms with ordered subsets and Nesterov acceleration, the system achieves high measurement precision from low-dose data through sophisticated signal processing rather than brute-force signal enhancement.
Data Source
AI summary
An embodiment provides a medical image processing apparatus that comprises circuitry. The circuitry obtains detection data representing detection events of radiation at a plurality of detector elements. The circuitry reconstructs an image by iteratively using an optimization-transfer algorithm to the detection data. The optimization-transfer algorithm uses a quadratic surrogate function that includes a curvature. The curvature is calculated using an inverse-background image.