Spinal Cord Injury Application

What is this codebase about?

It contains the code for predicting ASIA classification after a spinal cord injury.

Where is it used?

The work herein has been published and the entirety of the research discussion can be found at: https://www.eneuro.org/content/10/1/ENEURO.0149-22.2022.

How is the codebase structured?

Path	What does it contain?
~/	Everything (including this file)
~/src/	All code
~/src/ml/	All ML code
~/src/ml/notebooks/	ML ad hoc research code
~/src/ml/modelling/	ML training, testing, and feature importance code
~/src/ml/modelling/pickles/	ML model training outputs
~/src/ml/modelling/plots/	ML feature importance graphs
~/src/ml/data/	ML data handling and transformation code
~/src/ml/data/csvs/	Raw dataset
~/src/ml/data/docs/	Raw dataset documentation
~/src/ml/data/utils/	Commonly used utilities for data handling and transformation

Requirements to run machine learning

• Windows 10
• Git
• Anaconda or Miniconda (https://www.anaconda.com/)
• NSCISC Dataset (https://www.nscisc.uab.edu/Research/NSCISC_DatabasePublicUse)

Steps to setup machine learning (Anaconda Terminal)

1. Clone repository: git clone https://github.com/kapoor1992/spinal_cord_injury_recovery.git
2. Navigate to the shared directory: cd src/ml
3. Create the conda environment: conda env create --file environment_windows.yml
4. Activate the environment: conda activate sci

Steps to run machine learning (Anaconda Terminal)

1. Copy f1_public.csv from NSCISC into src/ml/data/csvs/
2. Activate conda environment: conda activate sci
3. Navigate to the modelling directory: cd src/ml/modelling
4. Run model training (flags are optional): python model_runs.py --interpret --drop-patients-who-worsened
5. If --interpret was used, verify that importance.png can be seen in src/ml/modelling/plots/