class: center, middle ## RAMP #### Prediction of autism using functional and structural MRI ##### Guillaume Lemaitre ###### CDS Pitching Day 2017 .affiliations[   ] --- ### Project overview * Collaboration with the Institut Pateur ### Problem statement <center>Classification of autism spectrum disorder (ASD) vs. control subjects using brain MRI imaging</center> ### Dataset * Type of data: resting-state functional MRI (R-fMRI) and structural MRI * ABIDE I: 539 ASD / 573 controls -> 17 sites * ABIDE II: 521 ASD / 593 controls -> 19 sites * Robert Debre Hospital: around 1000 subjects --- ### R-f MRI dataset * 4D dataset -> 3D (volume) + T (time) .center[<iframe width="600" height="300" src="//www.youtube.com/embed/l7iWbfvOycs?rel=0" frameborder="0" allowfullscreen></iframe>] * Time -> change of oxygenation -> neuronal activation * Analysis of the functional connectivity --- ### R-f MRI dataset #### Connectome classification pipeline <center></center> 1. ROIs definition 2. Time-series extraction 3. Matrix estimation 4. Classification --- ### R-f MRI dataset #### Opening the box <center></center> #### Powered by nilearn and scikit-learn <center></center> --- ### Structural MRI <center></center> * Volume, area, cortical thickness, mean curvature ... ### Ground-truth * Autism status * WISCC4 IQ * Social responsiveness --- ### RAMP preparation * .green[DONE] - Problem definition --- ### RAMP preparation * .green[DONE] - Problem definition * .green[DONE] - Feature extraction R-fMRI <center></center> --- ### RAMP preparation * .green[DONE] - Problem definition * .green[DONE] - Feature extraction R-fMRI * .green[DONE] - Feature extraction structural MRI --- ### RAMP preparation * .green[DONE] - Problem definition * .green[DONE] - Feature extraction R-fMRI * .green[DONE] - Feature extraction structural MRI * .yellow[ONGOING] - Quality check of the dataset <center></center> --- ### RAMP preparation * .green[DONE] - Problem definition * .green[DONE] - Feature extraction R-fMRI * .green[DONE] - Feature extraction structural MRI * .yellow[ONGOING] - Quality check of the dataset * .red[TODO] - Design of RAMP starting kit (classification pipeline) * .red[TODO] - Organize brainhack --- ### Contributions #### Development * scikit-learn: http://scikit-learn.org/ * RAMP: https://github.com/paris-saclay-cds/ramp-workflow * RAMP kits and preparation: https://github.com/paris-saclay-cds/ramp-pasteur-scraping #### Usage * nilearn: http://nilearn.github.io/ * pandas: https://pandas.pydata.org/ <center>  </center>