Connectome-based machine learning models are vulnerable to subtle data manipulations: v1.0.0

Code for the manuscript "Connectome-based machine learning models are vulnerable to subtle data manipulations."

For the most updated version, please see https://github.com/mattrosenblatt7/trust_connectomes.

Please see the README file for details about scripts and running the code. Essentially, the "minimal_code" folder contains code for which you can demonstrate enhancement attacks in your own data. The "paper_experiments" folder contains more detailed information about the experiments run in the manuscript. 

The specific data used in this study cannot be shared, but all four datasets used in this study are open-source: ABCD (NIMH Data Archive, https://nda.nih.gov/abcd), HCP (ConnectomeDB database, https://db.humanconnectome.org), PNC (dbGaP Study, accession code: phs000607.v3.p2, https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000607.v3.p2), and SLIM (INDI, http://fcon_1000.projects.nitrc.org/indi/retro/southwestuni_qiu_index.html). Data collection was approved by the relevant ethics review board for each of the four datasets.