3574
MicroQIT - A Computational Framework for Population Microstructure Imaging1Laboratory of Neuro Imaging, Stevens Neuroimaging and Informatics Institute, Keck School of Medicine of USC, Los Angeles, CA, United States
Synopsis
Microstructure imaging provides a quantitative tool for characterizing neural tissue with diffusion MRI and expanding our understanding of how the brain changes in health and disease. However, robust computational tools are needed for population imaging studies, so we developed a computational framework (MicroQIT) using the Quantitative Imaging Toolkit to meet this need by providing regional summaries and spatially normalized microstructure parameter maps. It supports a variety of ways to extract microstructure information from multi-shell diffusion MRI, leverages grid computing environments, and is available for use by the research community for future studies.
Introduction
Microstructure imaging constitutes a wide variety of techniques for estimating quantitative parameters that describe brain tissue using diffusion MRI1. These techniques can expand our understanding of how the brain changes in health and disease; however, robust computational tools are needed for conducting microstructure imaging studies and evaluating the efficacy of microstructure models2. While a variety of software packages exist for estimating microstructure parameters from individual scans3,4 there remains a need for integrated workflows for processing large datasets from population imaging studies. To address this challenge, we developed a computational framework (MicroQIT) using the Quantitative Imaging Toolkit5 and provide it for use by the research community. We describe the design of the framework, present results from evaluative studies of its reliability and sensitivity to age associations in neurodevelopment, and discuss its capabilities and prospects.Design
The primary design goal of MicroQIT is to automate the microstructure imaging studies that involve large populations. It can be used from the command line or from the LONI Pipeline6 to take diffusion-weighted MRIs as input and produce spatially normalized microstructure parameter maps and statistical summaries from region-based analysis. A variety of parameters are included, such as diffusion tensor imaging with and without free water elimination (DTI, fwDTI)7, neurite orientation dispersion and density imaging (NODDI)8, white matter tract integrity (WMTI)9, and multi-compartment microscopic diffusion imaging (MCSMT)10.
MicroQIT uses a template-based approach to spatially normalize data and statistically aggregate parameters through ROI analysis11. To provide accurate alignment within white matter12, it uses the tensor-based registration algorithm provided by DTI-TK13. We created population averaged diffusion tensor image based on the IIT template14,15 for coregistration and added ROIs from the Johns Hopkins University (JHU) and Desikan-Killiany (DK) white matter atlases. The JHU ROIs were aligned using FSL fnirt16, and the DK ROIs were found by averaging Freesurfer segmentations17 in the HCP subjects.
MicroQIT is implemented using GNU Make. Make provides a declarative interface for choosing which type of analysis to conduct, and it facilitates parallel computing and data reuse. Unlike previous neuroimaging applications using Make18, we extend it to include data provenance regarding the workflow, commands, and compute environment in each run. We also include extensions to support grid computing, which enables parallel processing of subjects when a cluster is available and parallel model fitting when a particularly expensive optimization is needed.
Evaluation
We evaluated our framework with two experiments with the following typically used microstructure parameters: DTI fractional anisotropy (FA), DTI mean diffusivity (MD), fwDTI FA, fwDTI MD, NODDI neurite density index (NDI), NODDI orientation dispersion index (ODI), MCSMT neurite density, and WMTI axonal water fraction (AWF). First we conducted a scan-rescan analysis to assess reproducibility. We used 44 pairs of test-retest in vivo human scans with 1.25 mm isotropic voxels and b=0,1000,2000,3000 s/mm2 from the Human Connectome Project (HCP)19. We measured the coefficient of variation (CV) and intra-class correlation (ICC) of each ROI in several commonly used microstructure parameters. Figure 1 shows excellent reliability with mean CV < 2.5% and ICC > 0.85 cross all metrics. Second, we investigated the relative sensitivity of these parameters to age-associated changes in white matter neurodevelopment. We used 133 in vivo scans with 2.0 mm isotropic voxels and b=0,1000,3000 s/mm2 from the CMIND dataset. We fit cubic polynomials to the ROI averaged parameters and compare the results from the region with the highest R2. Figure 2 shows the relative sensitivity of the parameters, where the highest performing methods achieved an R2 > 0.8.
Discussion and Conclusions
MicroQIT provides a robust computational framework for analyzing microstructure parameters across a large population. The results of our evaluation demonstrates high reliability provided in a variety of parameters and high sensitivity to typical changes in white matter through neurodevelopment. The design of MicroQIT can be easily extended to use new microstructure models that can be useful for population studies, and the present implementation may help to understand the changes in the brain in future studies. MicroQIT is available for download online20,21.Acknowledgements
This work was supported by NIH grant P41EB015922.
References
[1] Ferizi, U., Schneider, T., Panagiotaki, E., Nedjati‐Gilani, G., Zhang, H., Wheeler‐Kingshott, C. A., & Alexander, D. C. (2014). A ranking of diffusion MRI compartment models with in vivo human brain data. Magnetic resonance in medicine, 72(6), 1785-1792.
[2] Alexander, D. C., Dyrby, T. B., Nilsson, M., & Zhang, H. (2017). Imaging brain microstructure with diffusion MRI: practicality and applications. NMR in Biomedicine, e3841.
[3] Fick, R., Wassermann, D., & Deriche, R. (2018, June). Mipy: An Open-Source Framework to improve reproducibility in Brain Microstructure Imaging. In OHBM 2018-Human Brain Mapping (pp. 1-4).
[4] Alimi, A., Fick, R., Wassermann, D., & Deriche, R. (2018, September). Dmipy, a Diffusion Microstructure Imaging toolbox in Python to improve research reproducibility. In Computational Diffusion MRI 2018.
[5] Cabeen, R. P., Laidlaw, D. H., Toga, A. W., (2018). Quantitative Imaging Toolkit: Software for Interactive 3D Visualization, Processing, and Analysis ofNeuroimaging Datasets. ISMRM 2018, Abstract 2854
[6] Dinov, I., Van Horn, J., Lozev, K., Magsipoc, R., Petrosyan, P., Liu, Z., ... & Toga, A. W. (2009). Efficient, distributed and interactive neuroimaging data analysis using the LONI pipeline. Frontiers in neuroinformatics, 3, 22.
[7] Hoy, A. R., Koay, C. G., Kecskemeti, S. R., & Alexander, A. L. (2014). Optimization of a free water elimination two-compartment model for diffusion tensor imaging. NeuroImage, 103, 323-333.
[8] Zhang, H., Schneider, T., Wheeler-Kingshott, C. A., & Alexander, D. C. (2012). NODDI: practical in vivo neurite orientation dispersion and density imaging of the human brain. Neuroimage, 61(4), 1000-1016.
[9] Fieremans, E., Benitez, A., Jensen, J. H., Falangola, M. F., Tabesh, A., Deardorff, R. L., ... & Helpern, J. A. (2013). Novel white matter tract integrity metrics sensitive to Alzheimer disease progression. American Journal of Neuroradiology.
[10] Kaden, E., Kelm, N. D., Carson, R. P., Does, M. D., & Alexander, D. C. (2016). Multi-compartment microscopic diffusion imaging. NeuroImage, 139, 346- 359.
[11] Cabeen, R. P., Bastin, M. E., & Laidlaw, D. H. (2017). A Comparative evaluation of voxel-based spatial mapping in diffusion tensor imaging. Neuroimage, 146, 100-112.
[12] Wang, Y., Gupta, A., Liu, Z., Zhang, H., Escolar, M. L., Gilmore, J. H., ... & Styner, M. (2011). DTI registration in atlas based fiber analysis of infantile Krabbe disease. Neuroimage, 55(4), 1577-1586.
[13] Zhang, H., Yushkevich, P. A., Alexander, D. C., & Gee, J. C. (2006). Deformable registration of diffusion tensor MR images with explicit orientation optimization. Medical image analysis, 10(5), 764-785.
[14] Zhang, S., Peng, H., Dawe, R. J., & Arfanakis, K. (2011). Enhanced ICBM diffusion tensor template of the human brain. Neuroimage, 54(2), 974-984.
[15] Zhang, S., & Arfanakis, K. (2018). Evaluation of standardized and study-specific diffusion tensor imaging templates of the adult human brain: Template characteristics, spatial normalization accuracy, and detection of small inter-group FA differences. NeuroImage, 172, 40-50.
[16] Jenkinson, M., Beckmann, C. F., Behrens, T. E., Woolrich, M. W., & Smith, S. M. (2012). Fsl. Neuroimage, 62(2), 782-790.
[17] Fischl, B. (2012). FreeSurfer. Neuroimage, 62(2), 774-781.
[18] Askren, M. K., McAllister-Day, T. K., Koh, N., Mestre, Z., Dines, J. N., Korman, B. A., ... & Rane, S. D. (2016). Using make for reproducible and parallel neuroimaging workflow and quality-assurance. Frontiers in neuroinformatics, 10, 2.
[19] Sotiropoulos, S. N., Jbabdi, S., Xu, J., Andersson, J. L., Moeller, S., Auerbach, E. J., ... & Feinberg, D. A. (2013). Advances in diffusion MRI acquisition and processing in the Human Connectome Project. Neuroimage, 80, 125-143.
[20] http://cabeen.io/
[21] http://resource.loni.usc.edu/resources/downloads/Figures
