Instructions to setup and run HistoQC on H&E slides datasets

Clone HistoQC

git clone https://github.com/choosehappy/HistoQC

Then edit HistoQC/environment.devenv.yml and remove line 17 completely, which contains a string like this ... set HISTOQC_DEVEL = os.environ.get('HISTOQC_DEVEL', False).

Create a Conda environment called histoqc by running

conda env create --file environment.devenv.yml
conda activate histoqc

Then install the full repo as a Python package in editable mode:

pip install -e .

Install Blobfuse

The following steps are useful when the datasets are located on the cloud. The section can be skipped when the datasets are available locally on the machine.

Get blobfuse to access slides datasets:

sudo apt-get install blobfuse

Create a file blobfuse.cfg with

accountName <your_storage_account>
accountKey <redacted>
containerName <container_name>

blobfuse /<your_storage_account> \
            --tmp-path=/tmp/blobfusetmp \
            --config-file=blobfuse.cfg \
            -o attr_timeout=240 \
            -o entry_timeout=240 \
            -o negative_timeout=120 \
            -o allow_other  # important if you want other users to see that mount too

Test run HistoQC

Run a test run as follows (on only a small set of images to start with). The example below works with .svs slides, can be changed for other slide formats.

python -m histoqc /<your_storage_account>/<slides_dataset_folder>/00*/**.svs -n 8 -c <path_to_config_file> -o <target_folder>

Argument slides_dataset_folder refers to the folder where the slides are stored. Argument -n specifies the number of parallel processes. The argument -o specifies the output folder where histoqc results will be stored (results.tsv, error.log, and intermediate result folders). The argument path_to_config_file refers to the histoqc config file used for pre-processing the slides. The original configs in histoqc/configs can be used, such as config_v2.1.ini for H&E slides and config_ihc.ini for IHC slides.**

**For best results with H&E images, we commented out lines 4 and 16 in config_v2.1.ini as shown below (pen marking and blur detection modules respectively).

; ClassificationModule.byExampleWithFeatures:pen_markings
; BlurDetectionModule.identifyBlurryRegions