GPU Management for CT Sinogram Processing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current computed tomography systems face performance bottlenecks in processing large sets of sinogram data, leading to inefficiencies in processing speed and idle time for graphics processing units due to single-threaded operations and minimal memory allocation strategies.

Innovation Solution

Implementing a multi-threaded approach with separate processing and output threads for graphics processing units, allowing concurrent data processing and storage operations, and dynamic memory allocation to optimize resource utilization and reduce bottlenecks.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Power

If additional graphics processing units are added to increase processing power, then the processing capability for large sinogram datasets is improved, but the system complexity and cost increase

Engineering Contradiction:

Improveprocessing powerVSAvoidsystem complexity

Core Design Contradiction:

PowerVSDevice complexity

Solution Approach 1:

The patent segments the processing workflow into distinct functional threads: processing threads that handle sinogram data reconstruction and output threads that manage storage operations. This segmentation allows independent optimization of each thread and enables better resource utilization without adding hardware complexity

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent implements preliminary action by having output threads prepare storage operations in advance and maintain ready queues for processed data. This allows graphics processing units to continue processing without waiting for storage operations, maximizing their utilization without requiring additional hardware

Inventive Principle:

Principle #10Preliminary action

2Speed

If graphics processing units continuously process sinogram data, then processing speed is improved, but idle time occurs when waiting for storage operations to complete

Engineering Contradiction:

Improveprocessing speedVSAvoididle time

Core Design Contradiction:

SpeedVSLoss of time

Solution Approach 1:

The patent ensures continuity of useful action by implementing a multi-threaded architecture where processing threads continuously receive and process sinogram data while output threads simultaneously handle storage operations. The ready queue mechanism ensures that processed data is immediately available for storage without blocking the processing pipeline, eliminating idle time for graphics processing units

Inventive Principle:

Principle #20Continuity of useful action

Solution Approach 2:

The patent introduces an intermediary ready queue that buffers processed graphics data between the processing threads and output threads. This intermediary decouples the processing and storage operations, allowing them to proceed independently at different speeds without causing idle time for either operation

Inventive Principle:

Principle #24Intermediary (Mediator)

3Manufacturing precision

If large sets of sinogram data are processed to achieve high resolution images, then image quality is improved, but processing time increases significantly

Engineering Contradiction:

Improveimage qualityVSAvoidprocessing time

Core Design Contradiction:

Manufacturing precisionVSDuration of action of moving object

Solution Approach 1:

The patent segments large sinogram datasets into smaller manageable blocks that can be processed in parallel by multiple processing threads. Each thread handles a specific block independently, allowing high-resolution reconstruction without requiring the entire dataset to be processed sequentially, thus reducing total processing time while maintaining image quality

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent implements partial action by processing sinogram data in incremental blocks rather than requiring complete datasets before reconstruction. This allows intermediate results to be generated and stored progressively, enabling high-resolution imaging with reduced processing time by avoiding the need to wait for complete data acquisition

Inventive Principle:

Principle #16Partial or excessive action

4Ease of operation

If single-threaded operations are used to simplify system management, then ease of operation is improved, but resource utilization efficiency deteriorates

Engineering Contradiction:

Improvesystem managementVSAvoidresource utilization efficiency

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The patent implements self-service through automatic thread management and dynamic resource allocation. The system automatically creates and manages processing and output threads based on workload characteristics, and dynamically adjusts the number of threads and their priorities without requiring manual intervention. This maintains ease of operation while achieving high resource utilization through automated multi-threading

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS10062135B2Graphics processing unit management system for computed tomography

Publication Date: 2018.08.28 NATIONAL TECHNOLOGY & ENGINEERING SOLUTIONS OF SANDIA LLC

AI summary

A method and apparatus for managing operation of graphics processing units in a computer system. A processing thread in the computer system controls processing of a set of sinogram data by a graphics processing unit to form a set of graphics data. An output thread in the computer system controls writing of the set of graphics data to a storage system. The graphics processing unit is available to perform processing of another set of sonogram data while the set of graphics data is written to the storage system such that the increased availability of a plurality of graphics processing units is present.