Cause

There are some reasons that you should perform Feature Extraction stage with a python script named "feature_extraction.py".

1. Incompatible Dependency

If you running iPython notebook ("kidney_radiomics.ipynb") with only python version 3.7 and install pyradiomics, you could perform Feature Extraction stage with iPython notebook.

However, sklearn version 1.1+, which could be only compatible with only python 3.8+, only support n_features_to_select='auto' option in sklearn.feature_selection.SequentialFeatureSelector.

In other words, Feature Extraction stage and Feature Selection 2 & Model Selection stage could not be run in a single iPython notebook with a single environment.

2. Multiprocessing

It's because iPython notebook do not support multiprocess which could dramatically speed up feature extraction process.

Performing Feature Extraction stage with a single process require too long time to extract radiomics features.

Thus, it is highly recommend to perform Feature Extraction with a python script named "feature_extraction.py" before running code blocks in iPython notebook ("kidney_radiomics.ipynb").

Environment (anaconda)

For Feature Extraction stage, use "feature_extracion.yml" which based on python == 3.7 and contain pyradiomics python module.

For running iPython notebook ("kidney_radiomics.ipynb"), use "kidney_radiomics.yml" which based on python >= 3.8 and contain scikit-learn >= 1.1.0.

Usage

1. Performing Feature Extraction with a stand alone python script

$python3 feature_extraction.py --n_process 4 --param_config /Users/wangheehwan/Desktop/kidney_radiomics/pipeline/params/Params2.yaml --nifti_Dir /Users/wangheehwan/Desktop/kidney_radiomics/NifTi --mask_Dir /Users/wangheehwan/Desktop/kidney_radiomics/Segmentation --save_Dir /Users/wangheehwan/Desktop/kidney_radiomics/pipeline/result

2. Running code blocks in iPython notebook while skipping Feature Extraction stage

running code blocks in the following procedure:

• import modules and functions
• perform Split Dataset stage. In this stage, loading csv files resulted from Feature Extraction stage and meta data files
• perform machine learning experiments
• visualize the result