Interactive Volume Exploration of Petascale Microscopy Data Streams Using a Visualization-Driven Virtual Memory...

Area of Science:

• Scientific Visualization
• Electron Microscopy Data Analysis
• Petascale Computing

Background:

• Visualizing large-scale, high-resolution datasets from electron microscopy presents significant computational challenges.
• Existing systems often struggle with the scale and continuous nature of petascale imaging data.

Purpose of the Study:

• To develop and present the first volume visualization system capable of handling petascale volumes from continuous electron microscopy image streams.
• To design an architecture that scales to dense, anisotropic petascale volumes efficiently.

Main Methods:

• A scalable multi-resolution virtual memory architecture is employed, decoupling representation construction from data acquisition.
• The system utilizes a visualization-driven approach, restricting computations to visible volume data and processing missing data on-demand.
• Ray-casting accesses 3D data blocks constructed from 2D image tiles only when needed (out-of-core processing).

Main Results:

• The system demonstrates scalability to petascale volumes, handling dense and anisotropic data.
• It efficiently processes continuous streams of high-resolution electron microscopy images.
• Performance is evaluated against previous state-of-the-art systems, showing significant improvements.

Conclusions:

• The developed system offers a novel and effective solution for visualizing petascale electron microscopy data.
• Its unique architecture enables efficient handling of large, continuous datasets, advancing scientific discovery in fields like neuroscience.