Scalable analysis of Big pathology image data cohorts using efficient methods and high-performance computing...

Area of Science:

• Digital pathology
• Computational oncology
• Bioinformatics

Background:

• Addressing challenges of large data sizes and high computational demands in large-scale investigative pathology and oncology studies.
• Developing core capabilities for researchers and clinical investigators to evaluate analytical pipelines.
• Quantifying sensitivity and variability of results in complex biological studies.

Purpose of the Study:

• To present a suite of tools and methods for efficient analysis of large-scale pathology and oncology data.
• To enable researchers to evaluate multiple analytical pipelines and their results.
• To manage high computational demands and large data sizes inherent in modern research.

Main Methods:

• Utilizing state-of-the-art parallel machines for high-performance computing.
• Implementing efficient content-based image retrieval (CBIR) algorithms.
• Employing hierarchical analysis for rapid detection and retrieval of similar image patches.
• Performing consensus clustering on large datasets using shared memory systems.

Main Results:

• Demonstrated efficient CBIR algorithms for large microscopy image analysis.
• Showcased high-performance computing capabilities for handling extensive datasets (e.g., 500,000 data points).
• Validated the effectiveness of the tools in meeting challenges of clinically relevant pathology applications.

Conclusions:

• Efficient CBIR and high-performance computing are crucial for analyzing large microscopy images in pathology.
• The developed technologies empower researchers to effectively utilize information from digitized microscopy specimens.
• These advancements facilitate more effective research and clinical investigation in oncology and pathology.