niclasclassen/robustness-of-transferability-estimation-metrics-for-medical-imaging

This repository contains the data splits of target datasets used in the following work:

@misc{claßen2026robustnesstransferabilityestimationmetrics,
      title={Robustness of transferability estimation metrics for medical imaging}, 
      author={Niclas Claßen and Théo Sourget and Dovile Juodelyte and Rob van der Goot and Veronika Cheplygina},
      year={2026},
      eprint={2608.09999},
      archivePrefix={arXiv},
      primaryClass={eess.IV},
      url={https://arxiv.org/abs/2608.09999}, 
}

One .npz file consists of:

# The original images and labels
'train_images', 'train_labels', 'val_images', 'val_labels', 'test_images', 'test_labels',

# The indices of the used miniature populations
'train_idx_run1_split-75pct', 'val_idx_run1_split-75pct', 'train_idx_run1_split-50pct', 'val_idx_run1_split-50pct', 'train_idx_run1_split-25pct', 'val_idx_run1_split-25pct', 'train_idx_run1_split-10pct', 'val_idx_run1_split-10pct', 'train_idx_run1_split-5pct', 'val_idx_run1_split-5pct', 'train_idx_run2_split-75pct', 'val_idx_run2_split-75pct', 'train_idx_run2_split-50pct', 'val_idx_run2_split-50pct', 'train_idx_run2_split-25pct', 'val_idx_run2_split-25pct', 'train_idx_run2_split-10pct', 'val_idx_run2_split-10pct', 'train_idx_run2_split-5pct', 'val_idx_run2_split-5pct', 'train_idx_run3_split-75pct', 'val_idx_run3_split-75pct', 'train_idx_run3_split-50pct', 'val_idx_run3_split-50pct', 'train_idx_run3_split-25pct', 'val_idx_run3_split-25pct', 'train_idx_run3_split-10pct', 'val_idx_run3_split-10pct', 'train_idx_run3_split-5pct', 'val_idx_run3_split-5pct', 'train_idx_run4_split-75pct', 'val_idx_run4_split-75pct', 'train_idx_run4_split-50pct', 'val_idx_run4_split-50pct', 'train_idx_run4_split-25pct', 'val_idx_run4_split-25pct', 'train_idx_run4_split-10pct', 'val_idx_run4_split-10pct', 'train_idx_run4_split-5pct', 'val_idx_run4_split-5pct', 'train_idx_run5_split-75pct', 'val_idx_run5_split-75pct', 'train_idx_run5_split-50pct', 'val_idx_run5_split-50pct', 'train_idx_run5_split-25pct', 'val_idx_run5_split-25pct', 'train_idx_run5_split-10pct', 'val_idx_run5_split-10pct', 'train_idx_run5_split-5pct', 'val_idx_run5_split-5pct', 'train_idx_run6_split-75pct', 'val_idx_run6_split-75pct', 'train_idx_run6_split-50pct', 'val_idx_run6_split-50pct', 'train_idx_run6_split-25pct', 'val_idx_run6_split-25pct', 'train_idx_run6_split-10pct', 'val_idx_run6_split-10pct', 'train_idx_run6_split-5pct', 'val_idx_run6_split-5pct', 'train_idx_run7_split-75pct', 'val_idx_run7_split-75pct', 'train_idx_run7_split-50pct', 'val_idx_run7_split-50pct', 'train_idx_run7_split-25pct', 'val_idx_run7_split-25pct', 'train_idx_run7_split-10pct', 'val_idx_run7_split-10pct', 'train_idx_run7_split-5pct', 'val_idx_run7_split-5pct', 'train_idx_run8_split-75pct', 'val_idx_run8_split-75pct', 'train_idx_run8_split-50pct', 'val_idx_run8_split-50pct', 'train_idx_run8_split-25pct', 'val_idx_run8_split-25pct', 'train_idx_run8_split-10pct', 'val_idx_run8_split-10pct', 'train_idx_run8_split-5pct', 'val_idx_run8_split-5pct', 'train_idx_run9_split-75pct', 'val_idx_run9_split-75pct', 'train_idx_run9_split-50pct', 'val_idx_run9_split-50pct', 'train_idx_run9_split-25pct', 'val_idx_run9_split-25pct', 'train_idx_run9_split-10pct', 'val_idx_run9_split-10pct', 'train_idx_run9_split-5pct', 'val_idx_run9_split-5pct', 'train_idx_run10_split-75pct', 'val_idx_run10_split-75pct', 'train_idx_run10_split-50pct', 'val_idx_run10_split-50pct', 'train_idx_run10_split-25pct', 'val_idx_run10_split-25pct', 'train_idx_run10_split-10pct', 'val_idx_run10_split-10pct', 'train_idx_run10_split-5pct', 'val_idx_run10_split-5pct'

The data originates from MedMNIST V2:

@article{medmnistv2,
    title={MedMNIST v2-A large-scale lightweight benchmark for 2D and 3D biomedical image classification},
    author={Yang, Jiancheng and Shi, Rui and Wei, Donglai and Liu, Zequan and Zhao, Lin and Ke, Bilian and Pfister, Hanspeter and Ni, Bingbing},
    journal={Scientific Data},
    volume={10},
    number={1},
    pages={41},
    year={2023},
    publisher={Nature Publishing Group UK London}
}