COINSTAC

COINSTAC performs decentralized, privacy-preserving analysis of neuroimaging and brain imaging data across multiple sites to enable aggregated statistical inference without sharing raw data.


Key Features:

  • Decentralized Framework: Executes analyses locally on each site's machines using their own datasets to avoid centralizing raw neuroimaging data.
  • Privacy Preservation: Reduces transfer of sensitive neuroimaging data by keeping subject-level data on-site and exchanging only derived results.
  • Synchronization and Aggregation: Synchronizes local analysis outputs to a cloud-based platform for aggregate analysis across contributors.
  • Algorithm Adaptation for Decentralization: Implements decentralized algorithms adapted to operate in a distributed setting, enabling complex computations such as regression statistics calculation.
  • Pipeline Specifications: Defines pipeline specifications to standardize decentralized analysis workflows and result aggregation.

Scientific Applications:

  • Large-scale brain imaging studies: Enables multi-site aggregation of neuroimaging results for population-level analyses without pooling raw imaging files.
  • Collaborative multi-site research projects: Supports cross-institutional studies that require privacy preservation while combining analytic results.
  • Advanced statistical analyses: Facilitates distributed computation of complex statistics, including regression analyses, across heterogeneous datasets.

Methodology:

Local computation on site machines followed by synchronization of derived results to a cloud-based platform for aggregate analysis, using decentralized algorithms (including implementations for regression statistics calculation).

Topics

Details

License:
MIT
Tool Type:
desktop application
Programming Languages:
JavaScript
Added:
8/17/2018
Last Updated:
12/10/2018

Operations

Publications

Ming J, Verner E, Sarwate A, Kelly R, Reed C, Kahleck T, Silva R, Panta S, Turner J, Plis S, Calhoun V. COINSTAC: Decentralizing the future of brain imaging analysis. F1000Research. 2017;6:1512. doi:10.12688/f1000research.12353.1. PMID:29123643. PMCID:PMC5657031.

Funding: - National Institutes of Health: 1R01DA040487, P20GM103472/5P20RR021938, R01EB005846 - National Science Foundation: 1539067, 1631819