Distributed Iterative Image Reconstruction for PET Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current positron emission tomography (PET) reconstruction methods, especially iterative techniques, are computationally expensive and time-consuming, particularly when using multiple longitudinal positions, and existing parallel processing solutions increase system cost and complexity.
Innovation Solution
A method and apparatus utilizing two processors to perform distinct portions of the reconstruction process in parallel, updating an object estimate iteratively, and distributing the reconstruction of projection data among multiple processors to accelerate convergence and reduce reconstruction time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If iterative reconstruction techniques are used to improve reconstruction quality, then manufacturing precision is improved, but loss of time increases
Solution Approach 1:
The reconstruction process is divided into multiple independent subsets of projection data. Each subset can be processed separately through multiple iterations, allowing parallel computation while maintaining the quality benefits of iterative reconstruction. The patent segments the projection data into subsets that can be independently reconstructed and then combined to form the final image.
Solution Approach 2:
The patent introduces a new dimension of parallel processing by processing multiple subsets simultaneously across multiple processors. This transforms the single-threaded iterative reconstruction into a multi-dimensional parallel computation scheme, where iterations occur across both the subset dimension and the processor dimension.
2Productivity
If parallel processing architecture is used to reduce reconstruction time, then productivity is improved, but device complexity increases
Solution Approach 1:
The system segments the projection data into multiple subsets that can be distributed to different processors. This segmentation enables parallel processing without requiring complex inter-processor communication infrastructure, as each processor independently handles its assigned subset through the iterative reconstruction process.
Solution Approach 2:
The patent employs universal reconstruction algorithms that can be executed on standard processors without requiring specialized hardware. The same iterative reconstruction algorithm is applied to each subset, allowing general-purpose processors to function as dedicated reconstruction units, thereby reducing system complexity while maintaining parallel processing capabilities.
3Productivity
If sequential processing of frames is used to reduce reconstruction time, then productivity is improved, but loss of time increases due to processor idle time
Solution Approach 1:
The patent ensures continuous utilization of processing resources by dividing projection data from multiple frames into subsets that are distributed across processors. While one processor is computing, others are preparing or receiving data, eliminating idle time and maintaining continuous productive action across the processing system.
Solution Approach 2:
The system performs preliminary distribution of projection data subsets to multiple processors before the actual iterative reconstruction begins. This preliminary action ensures that all processors are ready to compute simultaneously, eliminating startup delays and idle time that would occur with sequential processing.
Data Source
AI summary
A method and apparatus for performing an iterative image reconstruction uses two or more processors (130). The reconstruction task is distributed among the various processors (130). In one embodiment, the projection space data (300) is distributed among the processors (130). In another embodiment, the object space (200) is distributed among the processors (130).