GNN-based-diag

A repository for the 02-740 Bioimage Informatics course project on GNN-based cancer detection and classification.

WIP: https://drive.google.com/file/d/18EfaqRfm8PwYN7Vg5b_r0sqqOF8FZt5o/view?usp=sharing
Pipeline:
- Images --> Node/Edge list --> Graphical Embedding (Pytorch Geometric/Node2Vec) -> GNN/Sk-learn Classifiers

Initial Results:

We use RGB values on truncated portions of this dataset (https://www.kaggle.com/andrewmvd/malignant-lymphoma-classification) as node labels.
Euclidean pairwise distances are used as edge labels.
Graph is fully connected.
All of the above are good targets for improvement.
120 epochs with a baseline GCN:
120 epochs with a resnet-style skip connection GCN:

Name		Name	Last commit message	Last commit date
Latest commit History 96 Commits
data		data
resnet_baseline		resnet_baseline
.DS_Store		.DS_Store
.gitignore		.gitignore
Bioimage_Graph_Classification_Pipeline.ipynb		Bioimage_Graph_Classification_Pipeline.ipynb
Bioimage_Graph_Proc.ipynb		Bioimage_Graph_Proc.ipynb
LymphomaClassificationUsingResNet.ipynb		LymphomaClassificationUsingResNet.ipynb
LymphomaSVM.ipynb		LymphomaSVM.ipynb
README.md		README.md
SegmentinSimpsUpdated2.ipynb		SegmentinSimpsUpdated2.ipynb
SegmentinSimpsUpdated3.ipynb		SegmentinSimpsUpdated3.ipynb
SegmentinSimpsUpdated4.ipynb		SegmentinSimpsUpdated4.ipynb
data_DS_DS_node_labels.txt		data_DS_DS_node_labels.txt
data_utils.py		data_utils.py
graph.png		graph.png
install_requirements.sh		install_requirements.sh
main.py		main.py
modules.py		modules.py
plot_runs.ipynb		plot_runs.ipynb
run1.png		run1.png
run1.txt		run1.txt
run2.png		run2.png
run2.txt		run2.txt
testing.ipynb		testing.ipynb
visualizations.py		visualizations.py