Demonstration of unintentional misuse of FC identifiability to achieve superb prediction results
Figure: Identifiability using cosine similarity from FC and derivatives.
Figure: Unintentionally exploiting identifiability to get 20-25% better results.
We provide the OpenNeuro dataset ds004144 of 33 female Fibromyalgia patients and 33 female healthy controls.
Due to NIH/UKB data policy, we are not able to provide UKB, PNC, or BSNIP data. If you are an approved researcher or have any questions, please contact us using my email at the bottom.
You need to install Jupyter Notebook and some python dependencies. You can do so by running the following command:
pip install -r requirements.txt
You may need to add --break-system-packages
to the command above or run in a virtual python environment.
Open Jupyter Notebook in the notebooks directory. Run all cells to reproduce Fibromyalgia dataset results. For example, run the commands:
cd notebooks
jupyter notebook
Load the Ident4Fibromyalgia notebook and execute each of the cells in order.
My email: aorlichenko@tulane.edu
Personal website: https://aorliche.github.io
Lab website: http://www2.tulane.edu/~wyp/Home.html
A preprint is available on arXiv: https://doi.org/10.48550/arXiv.2308.01451