Neuroimaging study designs, computational analyses and data provenance using the LONI pipeline - PubMed (original) (raw)

. 2010 Sep 28;5(9):e13070.

doi: 10.1371/journal.pone.0013070.

Kamen Lozev, Petros Petrosyan, Zhizhong Liu, Paul Eggert, Jonathan Pierce, Alen Zamanyan, Shruthi Chakrapani, John Van Horn, D Stott Parker, Rico Magsipoc, Kelvin Leung, Boris Gutman, Roger Woods, Arthur Toga

Affiliations

Neuroimaging study designs, computational analyses and data provenance using the LONI pipeline

Ivo Dinov et al. PLoS One. 2010.

Abstract

Modern computational neuroscience employs diverse software tools and multidisciplinary expertise to analyze heterogeneous brain data. The classical problems of gathering meaningful data, fitting specific models, and discovering appropriate analysis and visualization tools give way to a new class of computational challenges--management of large and incongruous data, integration and interoperability of computational resources, and data provenance. We designed, implemented and validated a new paradigm for addressing these challenges in the neuroimaging field. Our solution is based on the LONI Pipeline environment [3], [4], a graphical workflow environment for constructing and executing complex data processing protocols. We developed study-design, database and visual language programming functionalities within the LONI Pipeline that enable the construction of complete, elaborate and robust graphical workflows for analyzing neuroimaging and other data. These workflows facilitate open sharing and communication of data and metadata, concrete processing protocols, result validation, and study replication among different investigators and research groups. The LONI Pipeline features include distributed grid-enabled infrastructure, virtualized execution environment, efficient integration, data provenance, validation and distribution of new computational tools, automated data format conversion, and an intuitive graphical user interface. We demonstrate the new LONI Pipeline features using large scale neuroimaging studies based on data from the International Consortium for Brain Mapping [5] and the Alzheimer's Disease Neuroimaging Initiative [6]. User guides, forums, instructions and downloads of the LONI Pipeline environment are available at http://pipeline.loni.ucla.edu.

PubMed Disclaimer

Conflict of interest statement

Competing Interests: The authors have declared that no competing interests exist.

Figures

Figure 1

Figure 1. Pipeline Study-Design Architecture.

Imaging and non-imaging meta-data for 104 subjects is used to stratify the entire population into 3 distinct cohorts – asymptomatic normal controls (NC), Alzheimer's disease (AD) patients, and Mild Cognitive Impairment (MCI) subjects. The nested inserts show the search and selection grouping criteria, cohort sizes and an instance of an XML meta-data file for one subject. The meta-data can be manually entered, automatically parsed from spreadsheets, databases or clinical charts, or fed in as results of the pipeline workflow calculations (derived data).

Figure 2

Figure 2. Examples of the Pipeline Database Plug-in Infrastructure using the Imaging Data Archive (IDA) database (left) and the XNAT database (right).

Secure user-authentication provides an appropriate data-access level. The user then selects a location for local/remote storage of the data for the computational Pipeline processing and a format for the data representation (e.g., study-design). The data-download progress monitor provides information about the status of the transfer. At the end a study-design, or a data-source module is constructed, which allows the stratification of the population into groups (Male/female, in this case).

Figure 3

Figure 3. Pipeline environment as a visual programming language.

Top panel (Figure 3.A) shows an example of a conditional flow of control (if-else), which splits a group of subjects into 3 cohorts, processes each cohort separately and finally maps statistical differences between the 3 groups. The bottom panel (Figure 3.B) demonstrates the Global Shape Analysis (GSA) pipeline workflow with efficient loop-group iterations – see the 3 loop-group modules on the top, one for each for the 3 population cohorts (AD, MCI and NC subjects).

Figure 4

Figure 4. The LONI tensor-based morphometry (TBM) pipeline workflow.

This pipeline demonstrates the interoperability of several independently-developed computational neuroscience tools, management of grid-distributed jobs (different subjects and independent operations are executed in parallel), and the interactive process-monitoring framework for exploring the state of the entire execution workflow, as well as each individual module and input case.

Figure 5

Figure 5. The global shape analysis (GSA) pipeline workflow.

This workflow illustrates the protocol for construction of study-designs, automated ROI parsing, volumetric and shape measure calculations and between cohorts statistical analysis. 134 ADNI subjects are used in this study representing three independent cohorts classified by their Mini Mental State Exam (MMSE) scores - 18 Alzheimer's disease (AD) patients, 49 Mild Cognitive Impairment (MCI) subjects and 61 asymptomatic normals (NC). The workflow computes 6 global shape measures for each of 56 automatically extracted regions of interest (ROIs) for all subjects . Then, between-cohort statistics of these shape measures are calculated for each region of interest. The output of this workflow includes 18 3D scenes (3 possible group comparisons and 6 different shape measures). Each 3D scene contains only the ROI models of the regions where of the pair of cohorts showed statistically significant global shape differences. In this study-design, there are a total of 18 3D scene outputs reflecting the 3 possible pairs of group analyses (contrasts) comparing two cohorts (AD-NC, NC-MCI, AD-MCI) and the 6 different shape measures (mean-curvature, surface area, volume, shape-index, curvedness and fractal dimension). This workflow completed in about 46 hours on a small 56-node cluster and included a total of 3,209 jobs. The insert image only shows the 3D scene result for the ROIs which had statistically significant difference in mean-curvature between AD and NC subjects. For this comparison and shape measure, the resulting ROIs and their (labels) included right insular cortex (102), right middle orbitofrontal gyrus (30), right postcentral gyrus (42), left cingulated gyrus (121), left gyrus rectus (33) and left postcentral gyrus (41).

Figure 6

Figure 6. Pipeline meta-algorithm developments.

(Top Panel) This figure shows the LONI Pipeline implementation of the image registration meta-algorithm (IRMA). The four different registration algorithms AIR Linear, AIR Warp, FLIRT and MINC Tracc are present as individual nodes in this pipeline. Parameter sets specifying altogether 186 different runs for these algorithms are stored as editable lists within the LONI Pipeline environment. (Bottom Panel) Parallel-coordinates plot showing the rank-transformed metrics for one input image. Eleven distance metrics were employed to evaluate all 186 alignment instances. The results of the IRMA analysis illustrate that the FLIRT family performed consistently better than the other alignment methods for this input volume. The EDI (Entropy of Difference of Intensities) and the Woo (Woods' coefficient) metrics disagree somewhat with the other 9 metrics.

References

    1. Jack C, Bernstein M, Fox N, Thompson P, Alexander G, et al. The Alzheimer's disease neuroimaging initiative (ADNI): MRI methods. Journal of Magnetic Resonance Imaging. 2008;27:685–691. -PMC -PubMed
    1. Tu Z, Narr KL, Dinov I, Dollar P, Thompson PM, et al. Brain Anatomical Structure Segmentation by Hybrid Discriminative/Generative Models. IEEE Transactions on Medical Imaging. 2008;27:495–508. -PMC -PubMed
    1. Rex DE, Ma JQ, Toga AW. The LONI Pipeline Processing Environment. Neuroimage. 2003;19:1033–1048. -PubMed
    1. Dinov I, Van Horn J, Lozev K, Magsipoc R, Petrosyan P, et al. Efficient, Distributed and Interactive Neuroimaging Data Analysis using the LONI Pipeline. Frontiers of Neuroinformatics. 2009;3:1–10. -PMC -PubMed
    1. Mazziotta JC, Toga AW, Evans A, Fox P, Lancaster J. A probabilistic atlas of the human brain: theory and rationale for its development. The International Consortium for Brain Mapping (ICBM). Neuroimage. 1995;2:89–101. -PubMed

Publication types

MeSH terms

Grants and funding

LinkOut - more resources