Parallel LOR Data Processing for PET Image Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current nuclear medicine diagnostic systems face challenges in achieving high-speed calculations due to the large amount of data processing required for list mode data, particularly in parallel calculations, which often lead to memory competition issues and inefficiencies when using GPUs or other SIMD mechanisms.
Innovation Solution
A data processing method that performs forward projection and back projection calculations in parallel for every Line Of Response (LOR) and image space area, respectively, allowing for versatile application across different arithmetic mechanisms and preventing memory competition, thereby enhancing calculation speed and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If parallel calculations are performed using GPU for list mode data processing, then calculation speed is improved, but memory competition occurs and device complexity increases
Solution Approach 1:
The patent implements a universal parallel calculation framework that can operate on both GPU and CPU architectures through a single unified algorithm design. The forward projection and back projection calculations are structured to be architecture-independent, allowing the same codebase to function on different computing platforms without requiring GPU-specific mechanisms or additional complexity.
Solution Approach 2:
The patent segments the image reconstruction process into distinct forward projection and back projection calculation modules, each handled by separate processing units. This segmentation allows independent optimization of each module and enables flexible deployment on different hardware architectures, avoiding the need for complex GPU-specific parallelization strategies.
2Productivity
If MIMD type parallel calculations are used for list mode data processing, then processing capability is improved, but memory area and processor requirements increase
Solution Approach 1:
The patent changes the parallelization approach from MIMD to a more efficient model where multiple data items are processed through coordinated operations. By adjusting the parallel calculation parameters and data access patterns, the system achieves high processing capability while reducing the total number of processors and memory resources required compared to traditional MIMD approaches.
Solution Approach 2:
The patent merges multiple calculation operations into unified processing steps where forward projection and back projection are performed through coordinated memory access patterns. This merging reduces the number of separate processor tasks and minimizes memory competition, achieving high productivity with fewer resources.
3Manufacturing precision
If list mode data processing is performed with iterative approximation, then image reconstruction quality is improved, but calculation amount increases significantly
Solution Approach 1:
The patent performs preliminary organization of list mode data into structured formats before entering the iterative approximation process. By pre-processing and organizing the data structure, the system reduces the computational burden during iterative calculations, maintaining high image reconstruction quality while decreasing the total calculation amount required.
Solution Approach 2:
The patent implements partial iterative approximation steps that achieve sufficient image reconstruction quality without completing the full iterative process. By using a limited number of iterations or stopping criteria, the system balances reconstruction accuracy with computational efficiency, avoiding excessive calculation while maintaining acceptable image quality.
Data Source
AI summary
A PET apparatus and a processing method according in this invention carry out arithmetic processes in parallel, in steps S4 (forward projection) and S5 (back projection), for every LOR, on list data (created from event data obtained by detecting gamma rays), and can therefore be applied to other arithmetic mechanisms and have versatility. Since parallel processing is carried out on the list data and the parallel processing is carried out in each of steps S4 (forward projection) and S5 (back projection), a competition for memory can be prevented to realize an improvement in speed. As a result, high versatility and an improvement in speed of the calculations can be attained.