Reference: DOI: 10.1007/s10278-018-0154-z
dicom_header_extraction/extract_dicom_headers_w_generator_150K.py
dicom_header_extraction/normalize_selected_dcm_headers.py
caret_on_headers.R # most methods
caret_on_headers_nona.R # GLMNET
originally DICOMs were converted to 299x299 PNGs using convert_dicom_list_to_png.sh
script
Weight files are available here
-
scripts and config files:
image_classifiers/e5ce2d69b035975cb5336cec0da9a32a
-
weight file:
model-272-general-e5ce2d69b035975cb5336cec0da9a32a.hdf5
-
scripts and config files:
image_classifiers/e8e71fc090141d7c6fb334359152d295
-
weight file:
model-134-wire-e8e71fc090141d7c6fb334359152d295.hdf5
Scripts used to generate Fig. 1
combine_predictions_hdr_and_img.ipynb
visualize_predictions_hdr_and_img.ipynb
Scripts used to generate Supplementary Figures S1 & S2
calc_auroc_confidence_intervals.R
plot_auroc_difference_pvalue.ipynb