CrowdGleason

This repo contains associated information with the publication "The CrowdGleason dataset: Learning the Gleason grade from crowds and experts" (https://doi.org/10.1016/j.cmpb.2024.108472?). We include the related citation, the dataset access at Zenodo, and the code to reproduce the experiments.

Citation

@article{LOPEZPEREZ2024108472,
title = {The CrowdGleason dataset: Learning the Gleason grade from crowds and experts},
journal = {Computer Methods and Programs in Biomedicine},
volume = {257},
pages = {108472},
year = {2024},
issn = {0169-2607},
doi = {https://doi.org/10.1016/j.cmpb.2024.108472},
url = {https://www.sciencedirect.com/science/article/pii/S0169260724004656},
author = {Miguel López-Pérez and Alba Morquecho and Arne Schmidt and Fernando Pérez-Bueno and Aurelio Martín-Castro and Javier Mateos and Rafael Molina},
keywords = {Computational pathology, Crowdsourcing, Prostate cancer, Gleason grade, Gaussian processes, Medical image analysis},
}

Data

Dataset publicly available at: https://zenodo.org/records/14178894

Code

The code is included in the folder code/, we use GPflow 1 for the crowdsourcing methods and PyTorch 1 for feature extraction.

For feature extraction use pip install torch.

For classification and ablation study use requeriments.txt.

Feature Extraction

• code/configs/
  Contains the config files for the experiments.

• code/feature_extraction/
  Contains the code to train a CNN prostate classifier on SICAP and extract features from CrowdGleason.

  1. train_feat_extractor.py – Trains the classifier on SICAP.
  2. predict_features.py – Extracts features from SICAP and CrowdGleason using the trained model.

Classification

• code/classification/
  Contains scripts to run classification models and experiments.
  1. main.py – Runs the crowdsourcing methods on a single dataset.
  2. run_gp_both.py – Runs the Gaussian Process (GP) methods on the combined dataset.
  3. run_svgpmix.py – Runs the model combining expert and crowd annotations on the combined dataset.

Ablation Study

• code/ablation_study/
  Contains scripts to evaluate model robustness under different conditions.
  1. main.py – Runs ablation studies varying the number of annotators.
  2. run_svgpmix.py – Runs ablation studies varying the number of expert-labeled samples.
  3. senior_vs_junior.py – Compares crowdsourcing methods based on annotator expertise (senior vs. junior).