Iterative Reconstruction System Concurrent Projection Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image reconstruction methods in medical imaging, such as computed tomography, face challenges with high processing time and bandwidth requirements due to the need to perform back-projection and forward-projection on entire volumes, which limits system performance and scalability.
Innovation Solution
The approach involves breaking the image volume into smaller sections or tiles, allowing concurrent back-projection and forward-projection on each section independently, reducing the need to move large volumes of data and enabling simultaneous processing, thereby decreasing bandwidth requirements and improving reconstruction time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire image volume is processed as a single data partition for back-projection and forward-projection, then the reconstruction accuracy is maintained, but the bandwidth requirements and processing time increase significantly
Solution Approach 1:
The patent divides the image volume into multiple smaller partitions or tiles that can be processed independently and concurrently. Each partition contains a subset of the total voxels, allowing parallel back-projection and forward-projection operations across multiple processing units without requiring the entire volume to be loaded into memory simultaneously, thus reducing bandwidth requirements and processing time while maintaining overall reconstruction accuracy through coordinated processing of all partitions.
2Reliability
If the entire image volume is processed as a single data partition, then complete image data is available for reconstruction, but the system bandwidth and memory requirements increase
Solution Approach 1:
The patent segments the image volume into multiple partitions that can be processed in parallel. Each partition contains a subset of voxels that are back-projected and forward-projected independently. This segmentation reduces the amount of data that needs to be transferred across the system bandwidth at any given time, while the coordinated processing of all partitions ensures that the complete image data is reconstructed with full reliability.
Solution Approach 2:
The patent introduces a new dimension of parallel processing by dividing the volume processing task across multiple processing units simultaneously working on different partitions. This dimensional expansion from sequential whole-volume processing to parallel partitioned processing reduces the temporal and bandwidth constraints while maintaining data completeness through the aggregation of results from all partitions.
3Ease of manufacture
If sequential back-projection and forward-projection are performed on the entire volume, then the processing is straightforward to implement, but the reconstruction performance is constrained
Solution Approach 1:
The patent divides the volume into multiple partitions that can be processed in parallel, significantly improving reconstruction performance by utilizing multiple processing units simultaneously. While the implementation becomes more complex than sequential processing, the patent manages this complexity through systematic partitioning strategies and coordinated data flow management, achieving a favorable balance between implementation complexity and performance improvement.
Solution Approach 2:
The patent merges multiple parallel processing streams working on different partitions into a unified reconstruction output. By combining the results from concurrent back-projection and forward-projection operations on multiple partitions, the system achieves high-performance reconstruction that maintains the coordinated simplicity of unified processing while leveraging parallel computational power.
4Productivity
If concurrent back-projection and forward-projection are performed on section subsets, then processing time and bandwidth usage are reduced, but the coordination complexity increases
Solution Approach 1:
The patent segments the processing task into independent section subsets that can be concurrently processed. Each section undergoes back-projection and forward-projection independently, reducing the coordination overhead compared to processing the entire volume as a single unit. The segmentation strategy minimizes inter-dependencies between processing units while ensuring that all sections are eventually combined to form the complete reconstructed image.
Data Source
AI summary
A method, a non-transitory computer-readable storage medium, and an image processing apparatus are provided for performing iterative reconstruction to generate a medical image. The method includes generating, by circuitry of the image processing apparatus, a first image data set by separately back projecting subsets of a first view data set. Each of the subsets of the first view data set corresponds to one of a plurality of different non-overlapping sections of the medical image to be reconstructed. The method further includes generating, by the circuitry, a second view data set by separately forward projecting subsets of the first image data set. Each of the subsets of the first image data set corresponds to one of the sections of the medical image to be reconstructed. Further, the step of generating the second view data set starts before the step of generating the first image data set is completed.