Resource estimation in high performance medical image computing.

Area of Science:

• Medical Imaging
• Computational Science
• High-Performance Computing (HPC)

Background:

• Medical imaging analysis involves complex, multi-stage pipelines that are increasingly challenged by large datasets.
• High-performance computing (HPC) environments offer solutions for distributed computation but require accurate resource utilization modeling.
• Predicting computational resource needs (CPU, memory) for medical image processing algorithms is difficult due to variability.

Purpose of the Study:

• To address the challenge of inaccurate resource estimation in HPC for medical imaging analysis.
• To develop and implement a system for predicting computational resource requirements of medical image processing algorithms.
• To enable more efficient utilization of HPC resources by medical imaging researchers.

Main Methods:

• Implementation of a novel resource estimation system tailored for medical image processing.
• Development of predictive models to estimate CPU runtime and memory usage for computational tasks.
• Integration of the system within existing high-performance computing (HPC) pipeline environments.

Main Results:

• The developed system effectively estimates resource requirements for medical image processing tasks.
• Improved accuracy in resource prediction leads to reduced computational waste in HPC environments.
• Users can more efficiently schedule and execute complex medical imaging analyses on shared HPC resources.

Conclusions:

• Accurate resource estimation is a critical bottleneck for leveraging HPC in medical imaging.
• The implemented system successfully overcomes prediction difficulties, enhancing computational efficiency.
• This work facilitates broader and more effective use of HPC for advanced medical imaging studies.