ohbm / osr2020

Website for the Open Science Room at the OHBM 2020 meeting
https://ohbm.github.io/osr2020
Other
18 stars 6 forks source link

Open Workflows (Software/process demo): BrainSuite BIDS-App: Reproducible structural image processing pipelines with group level statistical analysis #46

Open jsheunis opened 4 years ago

jsheunis commented 4 years ago

BrainSuite BIDS-App: Reproducible structural image processing pipelines with group level statistical analysis

By Yeun Kim, University of California, Los Angeles

Abstract

BrainSuite is a collection of open-source software tools designed to perform processing and analysis of structural, diffusion, and resting functional magnetic resonance imaging data. We have developed BrainSuite BIDS-App [1], which follows the BIDS-App framework [3], to provide a portable, containerized version of BrainSuite's processing pipelines and statistical analysis tools to provide greater interoperability and reproducibility. In BrainSuite BIDS-App, there are two main processing pipelines for individual-level processing of: 1) T1-weighted (T1w) MRI and diffusion MRI (dMRI); and 2) resting fMRI data. We focus here on the structural analysis pipeline. For the group-level analysis stage, we have developed the BrainSuite Statistics Toolbox in R (bss-r), a set of statistics tools written in R that are designed to operate on the outputs of BrainSuite [2]. Briefly, the single-subject structural pipeline in the BrainSuite BIDSApp processes T1w and dMRI to perform: 1) cortical surface extraction from the T1w; 2) surface-constrained volume registration of the surface and volume to a labeled atlas; 3) distortion correction and coregistration of dMRI to the T1w; 4) generation of diffusion parameter maps in the T1w space. The outputs of the single-subject workflow are then processed using the BrainSuite BIDS-App group-level analysis, which incorporates bss-r. This provides functionality to perform ANOVA, t-tests, paired t-tests, linear regression, correlation tests, and mixed modeling of the BrainSuite outputs. These tests can be applied to surface cortical thickness measures, surface Jacobians (which measure the relative differences in local surface area), volumetric Jacobians (which measure the relative differences in local volume), DTI parametric maps, and regional anatomical ROI measures. Bss-r in the BrainSuite BIDS-App generates an RMarkdown page and a corresponding rendered HTML page that provide visualization of the results of the statistical tests that were performed. The bss-r report page displays a table of the subject demographic data used for the test, followed by the results, which are presented as numeric tables. For voxel-wise tests, image slices are also presented that show maps of correlations, p-values, adjusted p-values, t-values, and adjusted t-values, which are overlaid on the anatomical atlas image for reference. Bss-r identifies the largest connected clusters of t-values and shows images for each of these. The bss-r report can be used to share directly with collaborators or used as figures for publications. Our implementation of the BrainSuite workflows and BrainSuite statistical toolkit in a BIDS-App provides a powerful framework for neuroimaging researchers to perform automated image analysis with reproducible workflows.

[1] Kim Y, Wong J, Shattuck D (2018) BrainSuite BIDS-App: a Containerized Version of the BrainSuite Processing Pipelines. 23rd Annual Meeting of the Organization for Human Brain Mapping Singapore, June 2018. [2] Joshi S, Kim Y, Joshi A, Leahy R, Shattuck D (2017) Reproducible TBM and ROI Analyses Using the BrainSuite Statistics Toolbox (BSS). 23rd Annual Meeting of the Organization for Human Brain Mapping Vancouver, Canada, June 2017. [3] Gorgolewski, K J, et al. (2017). BIDS apps: Improving ease of use, accessibility, and reproducibility of neuroimaging data analysis methods. PLoS computational biology, 13(3), e1005209.

Useful Links

http://brainsuite.org/ http://brainsuite.org/bssr/

Tagging @yeunkim