HNSC-classifier: an accuracy tool for head and neck cancer detection in digitized whole-slide histology using deep learning

an accuracy tools for head and neck cancer detection and stage inferred in digitized whole-slide histology using deep learning

The HNSC-classifier scheme and Deep learning framework:

options

Dependents

pandas==1.4.3
pillow==8.4.0
matplotlib==3.5.2
scipy==1.8.0
numpy==1.22.3
openslide-python
fastai==2.7.9
histolab==0.5.1

Installation:

1. install system dependency:

HNSC-classifier has one system-wide dependency: OpenSlide.

You should first download and install it from https://openslide.org/download/ according to your operating system.

2. install HNSC-classifier

$pip install HNSC-classifier

Usage:

$ HNSC-calssifier - i TCGA-BB-4223-01A-01-BS1.7d09ad3d-016e-461a-a053-f9434945073b.svs -c learn.pkl

Example (test in linux OS: Ubuntu 20.4, python 3.9)

Download test data

The test Whole slide image download form TCGA TCGA-BB-4223-01A-01-BS1.7d09ad3d-016e-461a-a053-f9434945073b.svs.

Download deep learning model

If you can not clink the hyperlink to obtain test data and DP model, you can download test data from ftp://23.105.208.65

Run HNSC-classifier in virtualenv

1. install virtualenv

$ pip install virtualenv

2. Create virtual environment

$ virtualenv ven

3. Activate environment

$ source ven/bin/activate

4. install HNSC-classifier

$pip install HNSC-classifier

5. validate installation

$HNSC-classifier -h

HNSC-classifier for cancer detected.

$ HNSC-calssifier - i TCGA-BB-4223-01A-01-BS1.7d09ad3d-016e-461a-a053-f9434945073b.svs -c learn.pkl

HNSC-classifier for stage detected.

$ HNSC-calssifier - i TCGA-BB-4223-01A-01-BS1.7d09ad3d-016e-461a-a053-f9434945073b.svs -s learn_S.pkl

HNSC-classifier for TNM Staging System detected.

$ HNSC-calssifier - i TCGA-BB-4223-01A-01-BS1.7d09ad3d-016e-461a-a053-f9434945073b.svs -t learn_T.pkl -m learn_M.pkl -n learn_N.pkl

Output

Extract_tiles/  
                tile_0_level0_1499-7466-1723-7690.png
                tile_1_level0_1499-7690-1723-7914.png
                tile_2_level0_1499-8810-1723-9034.png
                tile_3_level0_1499-9034-1723-9258.png
                ...
cancer_heatmap.png
stage_heatmap.png
TNM_system_M_heatmap.png
TNM_system_N_heatmap.png
TNM_system_T_heatmap.png
summary.png
summary.csv

• Extract_tiles: the tiles extract from WSI.
• cancer_heatmap.png: cancer detected result.
• stage_heatmap.png: stage detected result.
• TNM_system_M_heatmap.png: TNM stage system (M) detected result.
• TNM_system_N_heatmap.png: TNM stage system (N) detected result.
• TNM_system_T_heatmap.png: TNM stage system (T) detected result.
• summary.png: the summary of extracted and predicted tiles info.
• summary.csv: the summary of extracted and predicted tiles info.