The Open Connectome Project Data Cluster: Scalable Analysis and Vision for High-Throughput Neuroscience

Area of Science:

• Neuroscience
• Computer Science
• Bioinformatics

Background:

• High-throughput brain imaging generates massive datasets requiring specialized infrastructure.
• Mapping neural connectivity (connectomes) is crucial for understanding brain function.
• Existing systems often struggle with the scale and complexity of neuroimaging data.

Purpose of the Study:

• To present a scalable database cluster for spatial analysis and annotation of brain imaging data.
• To facilitate the construction of neural connectivity maps (connectomes) using parallel computer vision algorithms.
• To provide open-science data sets and services for the neuroscience community.

Main Methods:

• Utilized a NoSQL scale-out architecture for data distribution via spatial index partitioning.
• Implemented distinct I/O systems for reads (parallel disk arrays) and writes (solid-state storage) to optimize throughput.
• Employed RESTful Web services for programming interfaces to enhance scalability and usability.

Main Results:

• Demonstrated a scalable database cluster effective for 3-D electron microscopy, time-series, and multi-channel data.
• Achieved high throughput and avoided I/O interference through optimized data distribution and storage.
• Validated the system's performance and the effectiveness of spatial data organization.

Conclusions:

• The developed database cluster provides a scalable and efficient solution for high-throughput brain imaging data analysis.
• The system architecture supports the parallel processing required for connectomics research.
• Publicly available services and data at openconnecto.me promote open science in neuroscience.