WSI Classification Dataset for AEM

Dataset Summary

This dataset is derived from the publicly available CAMELYON16 and CAMELYON17 datasets. It consists of feature embeddings extracted from tissue patches of whole slide images (WSIs) using various pre-trained models. The dataset is designed for use in multiple instance learning (MIL) based WSI classification tasks, particularly for the Attention Entropy Maximization (AEM) method.

Usage

For detailed instructions on using this dataset with the Attention Entropy Maximization (AEM) method, please refer to the official AEM GitHub repository and arXiv paper:

• AEM GitHub Repository
• AEM: Attention Entropy Maximization for Multiple Instance Learning based Whole Slide Image Classification

These resources provide implementation details, examples, and documentation on applying AEM to WSI classification tasks using this dataset.

Dataset Creation

Source Data

• CAMELYON16: 400 WSIs of sentinel lymph node sections. More info
• CAMELYON17: 500 WSIs with slide-level annotations, selected from the CAMELYON17 training set. More information

Data Processing

1. Tissue patches were extracted from the WSIs using the CLAM toolkit.

2. Feature embeddings were generated for each patch using these pre-trained models:

   • ResNet18: PyTorch pre-trained weights
   • Lunit pre-trained DINO: Checkpoint
   • PathGen-CLIP: Pre-trained checkpoints

Considerations for Using the Data

Intended Uses

This dataset is primarily intended for research in computational pathology, specifically for developing and evaluating MIL-based WSI classification methods.

Social Impact and Biases

While this dataset aims to advance research in computational pathology and potentially improve diagnostic tools, users should be aware of potential biases inherent in the original CAMELYON datasets. These biases may affect the generalizability of models trained on this data.

Additional Information

Licensing Information

This dataset is released under the Apache 2.0 license.

Citation Information

If you use this dataset, please cite:

@article{zhang2023attention,
  title={Attention-challenging multiple instance learning for whole slide image classification},
  author={Zhang, Yunlong and Li, Honglin and Sun, Yuxuan and Zheng, Sunyi and Zhu, Chenglu and Yang, Lin},
  journal={arXiv preprint arXiv:2311.07125},
  year={2023}
}

@misc{zhang2024aemattentionentropymaximization,
      title={AEM: Attention Entropy Maximization for Multiple Instance Learning based Whole Slide Image Classification}, 
      author={Yunlong Zhang and Zhongyi Shui and Yunxuan Sun and Honglin Li and Jingxiong Li and Chenglu Zhu and Lin Yang},
      year={2024},
      eprint={2406.15303},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2406.15303}
}