Machine learning model to investigate lung cancer response after SBRT (radiotherapy treatment).
We investigated the clinical and radiomics data regarding lung cancer response after SBRT for the following predictions:
- survival
- local relapse
- remote relapse
We also investigated feature removal, data-preprocessing and prediction timeframe.
Authors: Camille Invernizzi, Pierre-Louis Benveniste
Create a new environment
conda create -n venv_lung_response python=3.9
Activate it
conda activate venv_lung_response
Then install all required libraries
pip install -r requirements.txt
The code is divided in two folders:
- data_preprocessing: here we investigate the data for data preprocessing, dataset merging, dataset statistics and feature elimination.
- model training: here we investigate the training of model prediction for survival, local relapse and final relapse.
NB: the investigations are detailed in the issues.
After doing the steps in installation section (section 1) and downloading the model from the release, you can run an inference using the file predict_3year_survival.py. Use the following command:
python predict_3year_survival.py --model-path PATH/TO/MODEL --sex X --BMI X --score_charlson X --smoke_cessation X --dose_tot X --BED_10 X --MeanIntensity X --IntensitySkewness X --IntensityKurtosis X --AreaUnderCurveCIVH X --RootMeanSquareIntensity X --IntensityHistogramMean X --IntensityHistogramVariance X --NGTDM_Strength X