This information details the QuNex ecosystem and documents how to efficiently execute QuNex workflows. The guide sections are organized according to the natural expected data progression, from on-boarding to analyses, which are provided as links on the Wiki landing page.
What is QuNex?#
Analyses of neuroimaging data involves an extensive set of steps, from initial conversion of DICOM images generated by the scanner to generating statistical images of the results of whole-brain analyses or preparing figures of resting-state functional connectivity graphs. QuNex was developed with the mission of enabling efficient analysis of both small studies on a local server (e.g. involving 20 participants), as well as large-scale datasets with >1000s of participants in a high-performance computing environment.
Originally, QuNex was developed to achieve the specific analytical needs of its developers and was thus designed to evolve and integrate novel and improved methods. It is not a standalone tool, but rather a 'toolbox' suite that allows integration with state-of-the-art neuroimaging software developed by the neuroimaging community. In that sense, while QuNex offers its own analytic capacity, it also serves as a 'nexus' ecosystem to facilitate unification of neuroimaging 'best practices' for rapid and agile deployment.
To achieve this goal, QuNex follows core design principles:
1. QuNex is command-line interface (CLI) driven
By design, QuNex does not feature a point-and-click graphical user interface (GUI). This design choice encourages users to rely on command-line interface (CLI) skills that robustly enable:
Documentation, i.e. there is a complete record of what exactly was done;
Replication, i.e. workflows can be repeated with different variations, adjustments and/or corrections;
Parallelization, i.e. deployment across high-performance computing clusters;
Flexibility, i.e. CLI-based design model facilitates implementation of varied analytic workflows;
Extensibility, i.e. CLI-based design model enables rapid development of new features;
Training, i.e. learning CLI-based frameworks ultimately drives education of 'power user' neuroimagers.
2. QuNex is purpose-built for agile biomarker data science
A core design principle for QuNex was the capacity to rapidly and flexibly iterate over distinct analytic choices, either on the processing or analysis workflow side. For instance, QuNex features a robust data hierarchy specification, but is algorithmically flexible, and thus enables rapid deployment and testing of new neuroimaging data science features.
3. QuNex is platform agnostic and inter-operable
A sustaining design choice for QuNex distribution was a Container framework, both for High-Performance computing (via Singularity) and secure cloud platforms (via Docker).
4. QuNex architecture is modular
QuNex was built with a vision towards sustained growth by the neuroimaging community. To that end, the source repository architecture is built in extendable and modular fashion. Thus, future expansion of QuNex capabilities can be achieved without direct interference of core functionality.