Skip to content

Predicting risky behavior in structural brain volume using the UK Biobank

Notifications You must be signed in to change notification settings

brainhack-school2020/hannahkiesow_RiskyBrain

Repository files navigation

Project for BrainHack School 2020

Team Contributors: Hannah Kiesow & Elise Douard

Still brainstorming!

Introduction

Hi! I am a second year PhD student studying computational neuroscience, working in the lab of Dr. Danilo Bzdok at McGill University. My research interests are in social cognition, and I am extremely fascinated with social interaction. My lab focuses on big data in neuroscience using data-driven methods. For my current projects, I mainly use Bayesian probabilistic modelling to analyze my data, using the UK BioBank, one of the world's largest biomedical databases. However I am starting to branch into machine learning and am super eager to apply ML methods in neuroimaging. My experience so far has been working mostly with structural MRI images, but I would love to branch out and work with other neuroimaging modalities.



Looking forward to collaborating with you!
twitter: @hannahmaykiesow

Project Definition

Goals:

  • I would ideally like to focus on using machine learning techniques. I would definitely like to focus on getting intuitions behind different ML techniques, and perhaps try out different ensemble methods for prediction

  • I also would like to get to know the nilearn API more in depth, especially working with the different atlases for feature engineering.

  • Lastly, I think visualization is an extremely important skill to invest in. That's why I want to explore other visualization tools using python such as bokeh or ptit prince.

Progress overview

  • So far, I am learning Canonical Correlation Analysis (CCA) so that I can first create my features using two atlases, the Social Brain Atlas and the Harvard Oxford Atlas

Tools

look at this pretty raincloud figure

  • Nilearn (working with different atlases)
  • Machine Learning (sklearn)
  • Visualization (ptit prince, seaborn, bokeh)

Data

  • UK Biobank data (~10,000 person release)

Deliverables

At the end of this project, we will have:

  • a better understanding and intuition of several machine learning algorithms suitable for neuroimaging
  • a deeper understanding of feature engineering via atlases
  • hopefully made a pretty figure :D

Tools I learned during this project

  • Canonical Correlation Analysis

Results

Conclusion and acknowledgement