/hnc-outcome-analysis

Primary LanguagePython

Head and Neck cancer treatment outcome analysis

All experiments were run by submitting to the slurm scheduling system on Orion (The High Computing Clusters of NMBU), which requires an slurm script (.sh), a container (we used Singularity *.sif) and the code file (python *.py).

Singularity definition file

Singularity.ray

Manually build the container using

singularity build --fakeroot Singularity.ray deoxys.sif

3D EfficientNet

Check the new_layer.py and customize_obj.py files, together with the architecture & scripts folders.

Tradition ML experiments

See outcome_traditional.py and outcome_traditional_radiomics.py

Neural net configuration

See config/clinical*, config/radiomics*, and config/tabular* folders

CNN configuration

See config/outcome_img* folders

Bootstrap sampling results

See outcome_plot.py

Interpretability

To generate vargrad, see interpretability.py & interpretability.sh To analyze results, see interpret_analyze.py & interpret_analyze.sh

Figures generation

See outcome_vargrad_plot.py