Medical Imaging Compute Resource Pooling for Hardware Cost Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical imaging systems face high costs and low utilization due to dedicated, expensive computing resources that are underutilized, and software designs are inflexible and hardwired to specific hardware configurations, making it difficult to share resources across multiple imaging instruments.
Innovation Solution
A compute processing service that transparently discovers and utilizes available hardware, prioritizes and parallelizes computations, allows for resource sharing across multiple imaging instruments, and enables algorithms to be moved between systems without modification, using a dataflow model to execute computations on a network of CPUs, GPUs, and other accelerators.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If dedicated computing resources are used for each imaging instrument, then real-time image processing performance is guaranteed, but hardware cost increases and resource utilization decreases
Solution Approach 1:
The patent merges computing resources from multiple imaging instruments into a shared pool that can be dynamically allocated to different instruments as needed. Instead of each instrument having dedicated GPUs and CPUs, the system combines these resources into a common infrastructure that serves multiple modalities (CT, MRI, ultrasound, etc.), thereby reducing total hardware cost while maintaining performance through resource sharing.
Solution Approach 2:
The computing resources in the patent are designed to be universal and multi-functional, capable of serving multiple imaging modalities and applications. The same GPU cluster can process data from CT scanners, MRI machines, and ultrasound systems interchangeably, eliminating the need for modality-specific hardware and improving overall resource utilization.
2Reliability
If software is hardwired to specific hardware configurations, then performance requirements are met, but adaptability to different systems and evolution of designs is restricted
Solution Approach 1:
The patent introduces a software intermediary layer (runtime environment, abstraction layer) between the imaging applications and the underlying hardware. This intermediary handles hardware-specific details, allowing applications to run on different hardware configurations without modification. The abstraction layer translates generic computation requests into hardware-specific operations, enabling both performance optimization and hardware independence.
Solution Approach 2:
The system employs dynamic resource allocation and runtime adaptation mechanisms that allow software to adjust to different hardware configurations at runtime. Instead of being statically bound to specific hardware, the software can dynamically discover available resources, allocate them appropriately, and adapt its execution strategy based on the actual hardware present, enabling easy system evolution and deployment across different platforms.
3Productivity
If expensive dedicated hardware is deployed, then high performance is achieved, but the cost of equipment increases
Solution Approach 1:
The patent combines computing resources from multiple imaging instruments into a shared infrastructure, reducing the total amount of expensive hardware needed. By merging GPU clusters, CPU resources, and storage systems into a common pool that serves multiple modalities, the system achieves high processing performance while lowering overall hardware investment and operational costs.
Solution Approach 2:
The system implements self-service resource allocation where the computing infrastructure automatically manages and distributes resources to different imaging modalities based on demand. The runtime environment monitors resource usage, performs load balancing, and allocates computing power dynamically without requiring manual intervention or dedicated hardware for each modality, thereby optimizing cost-efficiency while maintaining high productivity.
Data Source
AI summary
A control method for controlling data processing acquired from medical imaging modalities by using multiple data processors connected to multiple medical imaging modalities via a network. The method includes obtaining image information for imaging to be performed with an imaging modality from the multiple imaging modalities. The method also includes obtaining load information of the multiple data processors before the imaging is completed. Allocating, based on graph information generated based on the obtained load information, at least a part of the multiple data processors to processing of data acquired in imaging based on the imaging information. The control method may conclude by performing processing of the acquired data with the allocated data processing resource.