PyHRF: A Python Library for the Analysis of fMRI Data Based on Local Estimation of Hemodynamic Response Function

Transcript

1. A Python Library for the Analysis
   of fMRI Data Based on Local
   Estimation of Hemodynamic
   Response Function
   J. Arias, P. Ciuciu, M. Dojat, F. Forbes, A. Frau-Pascual,
   T. Perret, J. M. Warnking
   

   View Slide

2. 1.
   Introduction
   “Begin at the beginning”, the King said, very gravely,
   “and go on till you come to the end; then stop.”
   -- Lewis Carroll, Alice in Wonderland
   2
   

   View Slide

3. × The current fMRI methods (e.g., SPM) estimates
   the neural activity using a canonical HRF for the
   whole brain.
   × There are local variations of the HRF in brain
   regions of patients with anomalies (e.g., Stenosis).
   Conjecture: Local variations of the HRF in patients
   with anomalies cause a wrong estimation of the neural
   activity.
   3
   Motivation
   

   View Slide

4. × Separate estimation of the response function
   (HRF)
   × A Bayesian framework allows to account for prior
   knowledge
   × A parcellation is used to deal with the HRF
   estimation regionally.
   JDE is solved with Variational Expectation
   Maximization (VEM).
   4
   Joint Detection-Estimation
   (JDE)
   

   View Slide

5. “Never send a human to do a machine’s job”
   -- Agent Smith, The Matrix
   5
   2.
   PyHRF
   

   View Slide

6. 6
   PyHRF
   × Open source software
   × Pythonic
   × Robust scientific libraries
   × e.g., scipy, numpy, scikit-learn, nilearn, nipype ...
   × Implementation of several models
   × MCMC-BOLD, JDE-BOLD, JDE-ASL …
   × Parallel computation
   × Documentation, Notebooks, Docker image
   http://www.pyhrf.org https://github.com/pyhrf
   

   View Slide

7. 7
   PyHRF
   

   View Slide

8. 8
   PyHRF
   http://www.neurovault.org/images/307
   Posterior Probability Map (PPM)
   Z-Scores
   Right Hand
   Gorgolewski, K. J., Storkey, A., Bastin, M. E., Whittle, I. R., Wardlaw, J. M., & Pernet, C. R. (2013). A test-retest fMRI dataset for
   motor, language and spatial attention functions. GigaScience, 2(1), 6. https://doi.org/10.1186/2047-217X-2-6
   

   View Slide

9. × Improvement of the PyHRF API for further
   developments
   × Bring PyHRF to clinicians (GUI, CLI).
   × Pythonic framework from acquisition to
   visualization
   × Interaction SPM PyHRF
   Show the variations of the HRF response in brain
   regions of patients with anomalies (e.g., Stenosis).
   9
   Semavi Project
   

   View Slide

10. Semavi Project
    Stimuli &
    Acquisition
    Pre-
    processing
    Analysis
    10
    PyHRF
    

    View Slide

11. 11
    Semavi Project
    

    View Slide

12. Semavi Project
    12
    I
    f
    

    View Slide

13. Alice: “How long is forever?”
    White Rabbit: “Sometimes, just one second.
    -- Lewis Carroll, Alice in Wonderland
    13
    3.
    Concluding
    Remarks
    15 minutes
    

    View Slide

14. × PyHRF allows to estimate the HRF locally
    × Open source software
    × Multiplatform (Conda, Docker)
    × Extensible framework for the fMRI workflow
    Now we have a powerful tool for non-experts that
    will allow to test our hypothesis and new ideas.
    14
    Summary
    

    View Slide

15. × Continue to improve PyHRF
    × Internships
    × Use of PyHRF for the analysis of small animals
    × rats
    × Calibration of some parameters of the model to be
    able to compare our results with SPM
    × ROC analysis
    Show the variations of the HRF response in brain
    regions of patients with anomalies (e.g., Stenosis).
    15
    Future Work
    

    View Slide

16. Thanks!
    Any questions?
    

    View Slide