ComPRePS 2.0: Enabling Massive-Scale Distributed Computing on High-Performance Computing Cluster for...

Area of Science:

• Computational pathology
• Digital pathology
• Health informatics

Background:

• The digitization of histological data into whole-slide images (WSI) has driven advancements in computational pathology.
• Large-scale, high-resolution image processing is essential for analyzing complex biomedical data, especially private patient information.
• Existing computational pathology tools face limitations in scalability and security for massive datasets.

Purpose of the Study:

• To develop an advanced computational pathology suite (ComPRePS 2.0) addressing the scalability and security limitations of its predecessor.
• To leverage high-performance computing clusters (HPCCs) for efficient processing of gigapixel WSI.
• To enhance AI-driven analysis of histopathological slides for clinical tasks and research.

Main Methods:

• Integration of the Computational Renal Pathology Suite (ComPRePS 2.0) with the University of Florida's HiperGator HPCC.
• Utilization of on-demand CPU, GPU, and memory resources.
• Implementation of Apptainer-based containerization and parallel file system access for distributed computing.

Main Results:

• ComPRePS 2.0 demonstrated a 15x performance improvement compared to ComPRePS 1.0.
• Achieved unprecedented scalability and enhanced security for processing large-dimensional WSIs.
• Successfully processed 920 large-dimensional WSIs, generating crucial data for kidney disease research.

Conclusions:

• ComPRePS 2.0 provides a scalable and secure platform for automated histopathological slide analysis.
• The integration with HPCCs significantly boosts computational power for digital pathology workloads.
• This advancement facilitates large-scale data analysis, benefiting researchers, pathologists, and students in understanding kidney disease progression.