Machine Learning classifier on fMRI data

Team contributors: Mikkel Schöttner

Summary

This project will use machine learning on the preprocessed ABIDE data set to predict autism spectrum disorder (ASD) from resting state functional magnetic resonance imaging (rs-fMRI) data.

Because of the large size of the data set, the main analysis will be run on a high performance computing (HPC) cluster.

Project definition

Background

Several studies have found lower functional connectivity in the default mode network in people with ASD (Anderson, 2014). Based on these findings, rs-fMRI data have been used to predict autism by training a classifier on the multi-site ABIDE data set (Nielsen et al., 2013). This project aims to replicate these findings.

Tools

The project will rely on the following technologies:

• nilearn
• scikit-learn
• plotly
• HPC/compute Canada

Data

The Autism Brain Imaging Data Exchange contains resting state fMRI data from several sites of 539 individuals with autism spectrum disorder (ASD) and 573 neurotypical controls. It has been preprocessed using several different preprocessing pipelines. More information on that can be found here. The data set is included in the Nilearn data sets. It can be downloaded using nilearn.datasets.fetch_abide_pcp. See here for all parameter specifications.

Deliverables

At the end of this project, we will have:

• a Jupyter notebook that contains an interactive graph/dashboard created using plotly
• a Jupyter notebook that includes preparation of the data, visualization, explanations etc.
• a Python script that contains the main analysis

Results

Week 3 deliverable: data visualization

The ABIDE dataset contains data from several different sites. One caveat of training a classifier on this data is that the sites have age distributions that differ quite much from each other. The interactive plot linked below shows the age distribution over different sites, split by autism/control in box plots. You can toggle the two groups on and off. You can also hover with the mouse over each boxplot to see the median, min, max and quartiles.

Link to the plot

Tools I learned during this project

• machine learning
• using HPC clusters

Literature

Anderson, J. S., Patel, V. B., Preedy, V. R., & Martin, C. R. (2014). Cortical underconnectivity hypothesis in autism: evidence from functional connectivity MRI. Comprehensive guide to autism, 1457, 1471.

Nielsen, J. A., Zielinski, B. A., Fletcher, P. T., Alexander, A. L., Lange, N., Bigler, E. D., ... & Anderson, J. S. (2013). Multisite functional connectivity MRI classification of autism: ABIDE results. Frontiers in human neuroscience, 7, 599.