arcana 0.2

Archive-centric analysis workflow architecture based on NiPype

ARchive Centric ANAlysis (Arcana) is Python package for “archive-centric” analysis of study groups (e.g. NeuroImaging studies)

Arcana interacts closely with an archive, storing intermediate outputs, along with the parameters used to derive them, for reuse by subsequent analyses. Archives can either be XNAT repositories or (http://xnat.org) local directories organised by subject and visit, and a BIDS module (http://bids.neuroimaging.io/) is planned as future work.

Analysis workflows are constructed and executed using the NiPype package, and can either be run locally or submitted to high HPC facilities using NiPype’s execution plugins. For a requested analysis output, Arcana determines the required processing steps by querying the archive to check for missing intermediate outputs before constructing the workflow graph. When running in an environment with the modules package installed, Arcana manages the loading and unloading of software modules per pipeline node.

Design

Arcana is designed with an object-oriented philosophy, with the acquired and derived data sets along with the analysis pipelines used to derive the derived data sets encapsulated within “Study” classes.

The Arcana package itself only provides the abstract Study and MultiStudy base classes, which are designed to be sub-classed by more specific classes representing the analysis that can be performed on different types of data (e.g. FmriStudy, PetStudy). These specific classes can then be sub-classed further into classes that are specific to the a particular study, and integrate the complete workflow from preprocessing to statistic analysis.

Installation

Arcana can be installed using pip:

$ pip install arcana