By Andréanne Proulx
*Source: https://researchfeatures.com/2016/06/20/decoding-language-brain/
I am a Master's student in Psychology currently enrolled at the University of Montreal. My main focus is in genomic imagery which consists of investigating the effect of genetic mutations on fonctional and structural brain phenotypes. More specifically, I have been interested in resing-state functional connectivity measures in carrier populations.
This project's main objective will be to learn machine learning and visualisation tools available in Python. As for the scientific purpose, it will be to evaluate the performance of different linear and non-linear classification algorithms (SVM, decision tree, random forest, etc.) at a task which consists of predicting autism diagnosis from the functional connectivity data. I will also try out cross validation and different hyperparameters in order to maximise the classification performance. A second objective for this project will be to integrate as much of the tools (Github notably) we were introduced to in the previous week allowing therefore science to be as reproducible as possible.
This project will rely on the following tools:
- Jupyter Notebook
- Libraries: Scikit-learn, Nilearn, Seaborn, Matplotlib, Pyplot
- Github
For the purpose of this project, I will be using an the preprocessed open source database ABIDE (see Preprocessed Connectomes Project), available through Nilearn. This data contains structural, functional and phenotypic data of 539 individuals with autism and 573 typical controls.
- Two jupyter notebook containing the code for the visualization and the machine learning
- Requirements.txt
- Readme file
- Python script (team repository)
- Presentation slides (team repository)
- Interactive plot (README file)
- Python librairies: Nilearn (specialized library for neuroimaging), Matplotlib (static plotting), Plotly (interactive plotting), Sklearn (Machine learning)
- Jupyter Notebook
- Git & Github : Track files and building a repository
- Python librairies: Nilearn, Matplotlib, Plotly, Sklearn (dimensionality reduction, test-train split, gridsearch, cross-validation methods, evaluate performance)
- Jupyter Notebook
- Github: Push, pull, fork, merge, pull requests, issues, etc. Also learned team management and organization.
- Git: Track Jupyter Notebook
- Venv: Make requirement.txt file
Part of the ABIDE TEAM. See team repository at: https://github.com/orgs/brainhack-school2020/teams/abide-team.
Visualizations: https://plotly.com/~anproulx/2/data-visualization/
This repository contains the two Jupyter Notebooks: the first notebook contains the code related to the data exploration and visualization (and week 3 deliverable), whereas the second notebook contains the code for the machine learning part of this project. To install all necessary requirements to rerun this analysis, see the requirements.txt file included in the repository.
- Choosing appropriate estimators with scikit-learn https://scikit-learn.org/stable/tutorial/machine_learning_map/
- Benchmarking functional connectome-based predictive models for resting-state fMRI (for classification estimator inspiration) https://hal.inria.fr/hal-01824205
- Getting the data using nilearn fetch https://nilearn.github.io/modules/generated/nilearn.datasets.fetch_abide_pcp.html
- Python Data Science Handbook GitHub and website https://github.com/jakevdp/PythonDataScienceHandbook
A special thank you to all the 2020 BrainHack School Organizers for making this event an amazing learning experience!
- Pierre Bellec (Mentor)
- Désirée Lussier-Lévesque (Mentor)
- Alexa Pichet-Binette (Mentor)
- Sebastian Urchs (Mentor)
- Jean-Baptiste Poline
- Tristan Glatard
- Benjamin de Leener
- Elizabeth DuPre
- Ross Markello
- Peer Herholz
- Samuel Guay
- Valerie Hayot-Sasson
- Karim Jerbi
- Greg Kiar
- Jake Vogel
- Agâh Karakuzu