Use a convolutional neural network, MobileNetV2, to identify the activated B-cell-like (ABC) and germinal centre B-cell-like (GCB) classes of diffuse large B-cell lymphoma (DLBCL) from immunohistochemistry-stained histopathology slides.
The software currently gives a classification accuracy of around 67% and overfits to the training set, but I think this may be an issue with the image preprocessing.
Developed during a first-year PhD rotation project, supervised by Professor David R Westhead at the Leeds Institute for Data Analytics, University of Leeds.
- Linux
apt-get:
- python >= 3.6.9
- openslide-tools
pip-install:
- openslide-python
- opencv-python
- pandas
- xlrd
- numpy
- tensorflow-gpu >= 2.1.0
- Set directories and constants:
python parameters.pyandpython dl_parameters.py - Generate images:
python main.py <gene_name>wheregene_nameisABCorGCB - Populate datasets and train a model:
python learning.py