megFingerprinting

I received an NSERC-CREATE graduate fellowship in 2017 (find the details here). As part of this fellowship, I completed a 16 week internship at the Montreal Neurological Institute. During this time, I worked with OMEGA, the Open MEG Archive. The projected I worked on during this internship aims to identify people based on resting state brain connectivity patterns using both Deep Learning and linear methods

GitHub Repository

You can find it here

Status

On going - We have an accuracy of ~95% using linear methods (correlations) and dimensionality reduction techniques (PCA; see figure below). We are currently running follow up analysis (i.e. is it really the dynamic connectivity that helps us identify people or is it a by-product of source modelling?)

fingerprinting

Supervisors

Dr Bratislav Misic and Dr Sylvain Baillet

Skills developed

  • Deep learning using Keras (Python library)
  • Bash programming language
  • Big data (> 150 MEG datasets)
  • Linear modelling and dimensionality reduction techniques using NumPy, SciPy, and Scikit-Learn (Python libraries)
  • Advanced Signal Processing, specifically focused for MEG data using Brainstorm (MATLAB environment)
  • Basic structural MRI processing (using FreeSurfer)