Clinica
Poster No:
1920
Submission Type:
Abstract Submission
Authors:
Alexandre Routier1, Arnaud Marcoux1, Mauricio Diaz Melo2, Jorge Samper-González1, Adam Wild1, Alexis Guyot1, Junhao WEN1, Elina Thibeau--Sutre1, Simona Bottani1, Stanley Durrleman1, Ninon Burgos1, Olivier Colliot1
Institutions:
1ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria, Paris, France, 2Inria Paris, SED, Paris, France
First Author:
Alexandre Routier
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
Co-Author(s):
Arnaud Marcoux
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
Jorge Samper-González
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
Elina Thibeau--Sutre
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
Simona Bottani
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
Stanley Durrleman
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
Olivier Colliot
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
ARAMIS Lab, ICM, Inserm U1127, CNRS UMR 7225, Sorbonne University, Inria
Paris, France
Introduction:
We present new advances made to Clinica (www.clinica.run), an open source software platform for clinical neuroscience studies. Neuroimaging studies are challenging since they involve several data analysis steps such as image preprocessing, extraction of image-derived features or statistical analysis. The development of machine learning methods for neuroimaging also involves most of these steps. The objective of Clinica is to automate the processing and statistical analysis of neuroimaging data and ease the development of machine learning approaches.
New functionalities have been integrated to Clinica to enable the longitudinal analysis of T1w MRI and PET data, and the development of deep learning classification approaches. Other advances aim to consolidate the platform.
New functionalities have been integrated to Clinica to enable the longitudinal analysis of T1w MRI and PET data, and the development of deep learning classification approaches. Other advances aim to consolidate the platform.
Methods:
The core of Clinica is written in Python and mainly relies on Nipype (Gorgolewski et al., 2011) to create pipelines. These pipelines involve the combination of different software packages widely used in the neuroimaging community: FreeSurfer (Fischl, 2012), FSL (Jenkinson et al., 2012), SPM (Frackowiak et al., 1997), ANTs (Avants et al., 2014), MRtrix3 (Tournier et al., 2012), and PETPVC (Thomas et al., 2016). Features extracted with the different pipelines can be used as inputs to statistical or machine learning analysis. The input neuroimaging data are expected to follow the BIDS data structure (Gorgolewski et al., 2016).
New functionalities of Clinica can be divided into three main parts.
1) Clinica now comprises pipelines that are specifically dedicated to the processing of longitudinal data sets: t1-freesurfer-longitudinal and pet-surface-surface. The longitudinal FreeSurfer stream (Reuter et al., 2012) processes a series of T1w MR images acquired at different time points for the same subject to increase the accuracy of volume and thickness estimates. The resulting pial and white surfaces can then be used for the projection of the PET signal using the methodology described in (Marcoux et al., 2018). Finally, longitudinal surface-based data can be plugged to the statistics-surface pipeline, which relies on SurfStat (Worsley et al., 2009).
2) Previous releases provided integration between outputs of Clinica and machine learning algorithms from scikit-learn (Pedregosa et al., 2011). We now provide pipelines that preprocess neuroimaging data for deep learning classification: the t1-linear mainly performs affine registration of T1w images to a standard space using ANTs (Avants et al., 2014) and dl-prepare-dl prepares raw or processed data for integration with PyTorch (e.g. extraction of patches/slices).
3) Other functionalities were also developed or enhanced:
creation of Docker and Singularity images to automate the installation of software dependencies and ease the execution of pipelines;
update of the adni-2-bids converter that now includes ADNI3 and new modalities (fMRI, new PET tracers);
development of nifd-2-bids that converts the NIFD dataset (http://4rtni-ftldni.ini.usc.edu), which contains data of patients with frontotemporal lobar degeneration, to BIDS;
update of the dwi-preprocessing pipeline that now uses the FSL eddy tool (Andersson et al., 2016a 2016b);
development of the statistics-volume pipeline that enables statistical analysis on voxel-based features using the general linear model through the wrapping of SPM functions (Friston et al, 2007).
Pipelines available in Clinica are listed in Figure 1.
New functionalities of Clinica can be divided into three main parts.
1) Clinica now comprises pipelines that are specifically dedicated to the processing of longitudinal data sets: t1-freesurfer-longitudinal and pet-surface-surface. The longitudinal FreeSurfer stream (Reuter et al., 2012) processes a series of T1w MR images acquired at different time points for the same subject to increase the accuracy of volume and thickness estimates. The resulting pial and white surfaces can then be used for the projection of the PET signal using the methodology described in (Marcoux et al., 2018). Finally, longitudinal surface-based data can be plugged to the statistics-surface pipeline, which relies on SurfStat (Worsley et al., 2009).
2) Previous releases provided integration between outputs of Clinica and machine learning algorithms from scikit-learn (Pedregosa et al., 2011). We now provide pipelines that preprocess neuroimaging data for deep learning classification: the t1-linear mainly performs affine registration of T1w images to a standard space using ANTs (Avants et al., 2014) and dl-prepare-dl prepares raw or processed data for integration with PyTorch (e.g. extraction of patches/slices).
3) Other functionalities were also developed or enhanced:
creation of Docker and Singularity images to automate the installation of software dependencies and ease the execution of pipelines;
update of the adni-2-bids converter that now includes ADNI3 and new modalities (fMRI, new PET tracers);
development of nifd-2-bids that converts the NIFD dataset (http://4rtni-ftldni.ini.usc.edu), which contains data of patients with frontotemporal lobar degeneration, to BIDS;
update of the dwi-preprocessing pipeline that now uses the FSL eddy tool (Andersson et al., 2016a 2016b);
development of the statistics-volume pipeline that enables statistical analysis on voxel-based features using the general linear model through the wrapping of SPM functions (Friston et al, 2007).
Pipelines available in Clinica are listed in Figure 1.
Results:
The new functionalities of Clinica aim to answer the needs of its target audience. Neuroscientists and clinicians conducting clinical neuroscience studies will benefit from longitudinal pipelines, and researchers developing advanced machine learning algorithms will benefit from new pipelines that ease the application of deep learning approaches to neuroimaging data.
Conclusions:
Clinica (www.clinica.run) is an open source software platform that aims to make clinical research studies easier and to pursue the community effort of reproducibility.
Neuroinformatics and Data Sharing:
Workflows 1
Novel Imaging Acquisition Methods:
Multi-Modal Imaging 2
Keywords:
Data Organization
Machine Learning
Positron Emission Tomography (PET)
STRUCTURAL MRI
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
Other - Reproducibility
1|2Indicates the priority used for review
My abstract is being submitted as a Software Demonstration.
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Was any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using author date format
Andersson, J. L. R., and Sotiropoulos, S. N. (2016b). An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging. Neuroimage 125, 1063–1078.
Avants, B. B.., et al. (2014). The Insight ToolKit image registration framework. Front Neuroinform 8.
Fischl, B. (2012). FreeSurfer. NeuroImage 62, 774–781.
Frackowiak, R. S. J., et al. (1997). Human Brain Function. Academic Press USA
Friston, K., et al. (2007). Statistical Parametric Mapping. Elsevier
Gorgolewski, K., et al. (2011). Nipype: A Flexible, Lightweight and Extensible Neuroimaging Data Processing Framework in Python. Front. Neuroinform. 5.
Gorgolewski, K., et al. (2016). The brain imaging data structure, a format for organizing and describing outputs of neuroimaging experiments. Scientific Data 3, 160044.
Jenkinson, M., et al. (2012). FSL. Neuroimage 62, 782–790. doi:10.1016/j.neuroimage.2011.09.015.
Pedregosa, F., et al. (2011). Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research 12, 2825−2830.
Marcoux, A., et al. (2018). An Automated Pipeline for the Analysis of PET Data on the Cortical Surface. Frontiers in Neuroinformatics 12.
Reuter, M., et al. (2012). Within-subject template estimation for unbiased longitudinal image analysis. NeuroImage 61, 1402–1418.
Thomas, B. A., et al. (2016). PETPVC: a toolbox for performing partial volume correction techniques in positron emission tomography. Physics in Medicine and Biology 61, 7975–7993.
Tournier, J.-D., et al. (2012). MRtrix: Diffusion tractography in crossing fiber regions. Int. J. Imaging Syst. Technol. 22, 53–66.
Worsley, K., et al. (2009). SurfStat: A Matlab toolbox for the statistical analysis of univariate and multivariate surface and volumetric data using linear mixed effects models and random field theory. NeuroImage 47, S102.