pywsi¶

https://travis-ci.com/saketkc/pywsi.svg?token=GsuWFnsdqcXUSp8vzLip&branch=master

Free software: BSD license
Documentation: https://www.saket-choudhary.me/pywsi/

Features¶

See the Notebook.

Training InceptionV4 on Tumor/Normal patches¶

We currently rely on InceptionV4 model for training. It is one of the deepest and most sophesticated models available. Another model we would ideally like to explore is Inception-Resnet, but later.

Step 1. Create tissue masks¶

pywsi create-tissue-masks --indir /CAMELYON16/testing/images/ \
--level 5 --savedir /CAMELYON16/testing/tissue_masks

Step 2. Create annotation masks¶

pywsi create-annotation-masks --indir /CAMELYON16/testing/images/ \
--level 5 --savedir /CAMELYON16/testing/annotation_masks \
--jsondir /CAMELYON16/testing/lesion_annotations_json

Step 3A. Extract tumor patches¶

pywsi extract-tumor-patches --indir /CAMELYON16/testing/images/ \
--annmaskdir /CAMELYON16/testing/annotation_masks \
--tismaskdir /CAMELYON16/testing/tissue_masks \
--level 5 --savedir /CAMELYON16/testing/extracted_tumor_patches

Step 3B. Extract normal patches¶

pywsi extract-normal-patches --indir /CAMELYON16/training/normal \
--tismaskdir /CAMELYON16/training/tissue_masks --level 5 \
--savedir /CAMELYON16/training/extracted_normal_patches

Dataset download¶