Concurrent Sinogram Reconstruction for PET Imaging Memory Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current PET scanner reconstruction techniques face challenges with memory requirements and reconstruction speed, especially in whole-body scans, due to the need for concurrent data acquisition and processing, which is not feasible with 3D iterative algorithms.
Innovation Solution
The method involves loading an acquired sinogram of projection data into memory and reconstructing sinogram subsets of a prespecified axial length sequentially into successive regions of a common entire object image, allowing for reduced memory usage and concurrent reconstruction during data acquisition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all data frames are combined and reconstructed as one whole body at once, then image quality and statistics are improved, but memory requirements increase and reconstruction speed decreases
Solution Approach 1:
The patent divides the whole body sinogram into multiple axial subsets (e.g., head, chest, abdomen, pelvis regions) and reconstructs each subset separately using 3D iterative algorithms. This segmentation allows each region to be processed with appropriate memory allocation while maintaining overall image quality, resolving the contradiction between comprehensive reconstruction and memory requirements.
2Productivity
If concurrent reconstruction is performed while acquiring data, then patient throughput increases, but 3D iterative reconstruction algorithms cannot process all data frames simultaneously
Solution Approach 1:
The patent segments the acquisition process into multiple bed frames with overlapping fields of view, allowing concurrent reconstruction of different axial regions while data is being acquired. Each region can be reconstructed independently using 3D iterative algorithms, enabling parallel processing that improves patient throughput without requiring simultaneous processing of all data frames.
Solution Approach 2:
The patent performs preliminary reconstruction of earlier bed frames while subsequent data frames are being acquired. This preliminary action allows the system to start processing data immediately upon acquisition completion, reducing total scan time and improving patient throughput while maintaining the benefits of 3D iterative reconstruction.
3Measurement precision
If multiple bed frames with overlapping regions are used, then edge statistics are improved, but reconstruction time and computational complexity increase
Solution Approach 1:
The patent segments the reconstruction process by processing each bed frame's overlapping regions independently using 3D iterative algorithms. By reconstructing each segment separately and then combining the results, the system achieves improved edge statistics without requiring simultaneous processing of all frames, thus reducing total reconstruction time.
Data Source
AI summary
In an imaging system (10), a short axial length 4D sinograms are loaded one at a time from a data memory (40). A portion of an image memory (44) that corresponds to a currently reconstructed sinogram subset (1112), is initialized. If a part of the object is already reconstructed, an iterative reconstruction is performed in which the previously reconstructed image (m1) is iteratively improved by using the data from the currently reconstructed overlapping image (m2) to converge on the final image.