Deep Source

This is the Github repository for the Deep Source project, created during Neurohackademy 2019.

Goal

• Simulate MEG source data with deep sources (e.g. hippocampal activity) and estimate which source reconstruction method performs better when reconstructing them.
• For a detailed overview of what we are going to achieve, see our project-outline

Contributors

*(in alphabetic order)

• Azeez Adebimpe
• Ryan Timms
• Martina G. Vilas

Next steps:

1. • Tidy-up jupyter notebooks
2. • Incorporate and compare other source reconstruction methods (other than MNE and Beamformer)
3. • Translate HMM source reconstruction methods into Python
4. • Quantitatively compare the source reconstruction methods using the following algorithms:
     • Crosstalk-to-Signal Ratio (CSR)
     • Neural Activity Index (NAI) --> translate them to Python
     • Point-Spread Functions --> translate them to Python
5. • Simulate the ground truth data to vary in the following aspects:
     • Depth of the sources
     • Correlation of the sources
     • Closeness of the sources
     • The SNR
       ... and compare the performance of different source reconstruction methods.