Event Abstract

Back to Event

Harnessing cloud computing for high capacity analysis of neuroimaging data from NDAR

Daniel Clark1, Christian Haselgrove2, David N. Kennedy2, Zhizhong Liu3, Michael Milham1, 4, Petros Petrosyan3, Carinna Torgerson3, John Van Horn3 and Cameron Craddock1, 4*
  • 1 Child Mind Institute, Center for the Developing Brain, United States
  • 2 University of Massachusetts Medical School, United States
  • 3 University of Southern California, United States
  • 4 Nathan S. Kline Institute for Psychiatric Research, United States

Introduction: The National Database for Autism Research (NDAR, http://ndar.nih.gov) and other NIH/NIMH data repositories are amassing and sharing thousands of neuroimaging datasets. With the availability of this deluge of data and the development of the NDAR infrastructure for its organization and storage, the bottleneck for applying discovery science to psychiatric neuroimaging has shifted to the computational challenges associated with data processing and analysis. Maximizing the potential of these data requires automated pipelines that can leverage high-performance computing (HPC) architectures to achieve high throughput computation without compromising on the quality of the results. A disadvantage of this approach is that it requires access to HPC systems that are not always available, particularly at smaller research institutions, or in developing countries. Cloud computing resources such as Amazon Web Services (AWS) Elastic Compute Cloud (EC2) offers a “pay as you go” model that might be an economical alternative to the large capital costs and maintenance burden of dedicated HPC infrastructures. Realizing this need, the developers of the Laboratory of Neuro Imaging (LONI) Pipeline, the Neuroimaging Informatics Tools and Resources Clearinghouse (NITRC) Computational Environment (CE) and the Configurable Pipeline for the Analysis of Connectomes (C-PAC) have implemented pipelines in EC2 that interface with NDAR. Each pipeline was used to perform a benchmark analysis of 2,000 structural images from the NDAR database to establish the feasibility of this approach. Methods: Each of three pipelines were installed into Amazon Machine Images (AMIs) and customized to perform structural preprocessing on NDAR data. The LONI Pipeline (http://pipeline.loni.usc.edu) was enhanced to permit direct access to NDAR collections for workflow-based data processing (Torgerson, in press). Workflows can be created from a combination of commonly available neuroimaging processing tools represented as Pipeline Modules. With respect the benchmark analysis, specifically developed Pipeline Modules captured the results from FreeSurfer and FSL FirstAll, updated the NDAR with the results and returned them back to the NDAR Amazon Cloud storage server. C-PAC (http://fcp-indi.github.io) is a configurable pipeline for performing comprehensive functional connectivity analyses that was extended to include the Advanced Normalization Tools (ANTs) cortical thickness methodology (Tustison, 2014) and to interface it with NDAR (https://github.com/FCP-INDI/ndar-dev). Results of this workflow include 3D volumes of cortical thickness and regional measures derived from the Desikan-Killiany-Tourville atlas (http://mindboggle.info/faq/labels.html). NITRC-CE (http://www.nitrc.org/projects/nitrc_es/) is an AMI that is pre-installed with popular neuroimaging tools. A series of scripts were developed for NITRC-CE to interact with NDAR, calculate a series of quality assessment measures on the data, perform structural imaging analysis using FreeSurfer and FSL FirstAll results, and to write the results back to NDAR (https://github.com/chaselgrove/ndar). Results: Speeds obtained for processing structural data in EC2 were consistent with those obtained for local multi-core processors. For example, using an EC2 instance with 4 processors and 15 GB of RAM (m3.xlarge), the C-PAC pipeline was able to complete the ANTS cortical thickness pipeline in 8.5 hours per subject, in comparison to 9 hours on a local workstation with 12 processors and 64 GB of RAM. EC2 processing cost $1.94 per image for on demand instances and an estimated $0.26 per image when using spot instances. Conclusions: Analyzing data using cloud computing is an affordable solution, with low hardware and software maintenance burdens; this can be beneficial for smaller laboratories and when data is already in the cloud. Further reductions in cost can be obtained using lower costs spot instances, which fluctuate in price and may get shut down if demand gets too high.

Acknowledgements

Pipeline development was generously funded by NDAR.

References

Torgerson, C.M., Quinn, C., Dinov, I.D., Liu, Z., Petrosayan, P., Kennedy, D.N., Toga, A.W., and Van Horn, J.D. (in press) “Interacting with the National Database for Autism Research (NDAR) via the LONI Pipeline Workflow Environment”, Brain Imaging and Behavior.

Tustison, N.J., Cook, P.A., Klein, A., Song, G., Das, S.R., Duda, J.T., Kandel B.M., van Strien, N., Stone J.R., Gee, J.C., Avants, B.B. (2014) “Large-scale evaluation of ANTs and FreeSurfer cortical thickness measurements.” Neuroimage. 2014 Oct 1;99:166-79. doi: 10.1016/j.neuroimage.2014.05.044. Epub 2014 May 29.

Keywords: Cloud computing, preprocessing, pipeline, AWS, cortical thickness

Conference: Neuroinformatics 2015, Cairns, Australia, 20 Aug - 22 Aug, 2015.

Presentation Type: Poster, to be considered for oral presentation

Topic: Neuroimaging

Citation: Clark D, Haselgrove C, Kennedy DN, Liu Z, Milham M, Petrosyan P, Torgerson C, Van Horn J and Craddock C (2015). Harnessing cloud computing for high capacity analysis of neuroimaging data from NDAR. Front. Neurosci. Conference Abstract: Neuroinformatics 2015. doi: 10.3389/conf.fnins.2015.91.00021

Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.

The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.

Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.

For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.

Received: 08 Apr 2015; Published Online: 05 Aug 2015.

* Correspondence: Dr. Cameron Craddock, Child Mind Institute, Center for the Developing Brain, New York, New York, United States, cameron.craddock@austin.utexas.edu