A New Graph Node Classification Benchmark: Learning Structure from Histology Cell Graphs
Vanea, Claudia, Campbell, Jonathan, Dodi, Omri, Salumäe, Liis, Meir, Karen, Hochner-Celnikier, Drorith, Hochner, Hagit, Laisk, Triin, Ernst, Linda M., Lindgren, Cecilia M., Nellåker, Christoffer
–arXiv.org Artificial Intelligence
We introduce a new benchmark dataset, Placenta, for node classification in an underexplored domain: predicting microanatomical tissue structures from cell graphs in placenta histology whole slide images. This problem is uniquely challenging for graph learning for a few reasons. Cell graphs are large (>1 million nodes per image), node features are varied (64-dimensions of 11 types of cells), class labels are imbalanced (9 classes ranging from 0.21% of the data to 40.0%), and cellular communities cluster into heterogeneously distributed tissues of widely varying sizes (from 11 nodes to 44,671 nodes for a single structure). Here, we release a dataset consisting of two cell graphs from two placenta histology images totalling 2,395,747 nodes, 799,745 of which have ground truth labels. We present inductive benchmark results for 7 scalable models and show how the unique qualities of cell graphs can help drive the development of novel graph neural network architectures.
arXiv.org Artificial Intelligence
Nov-11-2022
- Country:
- North America > United States
- Washington > King County
- Seattle (0.04)
- Illinois > Cook County
- Chicago (0.04)
- Washington > King County
- Europe
- United Kingdom > England
- Oxfordshire > Oxford (0.05)
- Estonia > Tartu County
- Tartu (0.04)
- United Kingdom > England
- Asia > Middle East
- Israel > Jerusalem District > Jerusalem (0.04)
- North America > United States
- Genre:
- Research Report (0.85)
- Industry:
- Technology: