Dr Thomas Close1,2, Dr Phillip Ward3,4, Mr Francesco Sforazzini1, Dr Zhaolin Chen1,5, Prof Gary Egan1,3,4
1Monash Biomedical Imaging (MBI), Monash University, Melbourne, Australia,
2National Imaging Facility (NIF), Melbourne, Australia,
3Centre for Integrative Brain Function (CIBF), Melbourne, Australia,
4Monash Institute of Cognitive and Clinical Neurosciences (MCCIN), Monash University, Melbourne, Australia,
5Department of Electrical and Computer Systems Engineering, Monash University, Melbourne, Australia
Magnetic Resonance Imaging (MRI) is a flexible imaging modality, which can be manipulated to produce different soft tissue contrasts in order to detect a range of biomarkers. A large ecosystem of tools has been developed for MR image analysis. A typical image analysis workflow requires the application of many tools, particularly for workflows that combine contrasts and/or modalities. The complexity of these workflows, coupled with the trend in neuroimaging research towards larger cohort sizes, makes ad hoc implementations unwieldy. Monash Biomedical Imaging (MBI) has developed a workflow architecture for analysis of large neuroimaging studies, which is integrated with an imaging data repository.
Data acquired using MBI’s clinical research MRI and PET scanners are stored in an XNAT repository in accordance with requirements for a trusted data repository service. These requirements provide guarantees around continuity of access, and stipulate QC procedures and required metadata. Pipelines that check the integrity of uploaded imaging sessions and which analyse the QC data are automatically triggered using the XNAT container service.
The workflow architecture stores intermediate outputs and provenance data in the repository for reuse by subsequent analyses. Intermediate outputs are automatically detected before constructing workflow graphs. Workflows are run on MASSIVE and their results are uploaded to the repository.
Future work is planned to improve the interconnection between MBI’s XNAT repository and MASSIVE, including enabling users to launch preconfigured workflows from the XNAT UI. Such workflows could be triggered on ingest of imaging data into the repository, and thereby fully automate the image analyses.
Tom Close received his PhD on advanced methods in diffusion MRI tractography from the University of Melbourne. After completing his PhD he took up a post-doctoral position at the Okinawa Institute of Science and Technology (OIST), Okinawa, Japan, during which he developed standardised model description languages for computational neuroscience. In 2015, he returned to Melbourne as the senior informatics officer at Monash Biomedical Imaging (MBI), where he has migrated MBI’s imaging archive to the XNAT platform, and is developing integrated analysis pipelines for the automatic analysis of MRI and PET data.