Do Multiple Instance Learning Models Transfer?

Daniel Shao, Richard J. Chen, Andrew H. Song, Joel Runevic, Ming Y. Lu, Tong Ding, Faisal Mahmood
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:54219-54238, 2025.

Abstract

Multiple Instance Learning (MIL) is a cornerstone approach in computational pathology for distilling embeddings from gigapixel tissue images into patient-level representations to predict clinical outcomes. However, MIL is frequently challenged by the constraints of working with small, weakly-supervised clinical datasets. Unlike fields such as natural language processing and computer vision, which effectively use transfer learning to improve model quality in data-scarce environments, the transferability of MIL models remains largely unexplored. We conduct the first comprehensive investigation into transfer learning capabilities of pretrained MIL models, evaluating 11 MIL models across 19 pretraining tasks spanning tissue subtyping, cancer grading, and molecular subtype prediction. We observe a substantial performance boost with finetuning pretrained models over training from randomly initialized weights, even with domain differences between pretraining and target tasks. Pretraining on pan-cancer datasets enables consistent generalization across organs and task types compared to single-disease pretraining. Remarkably, this pan-cancer pretraining leads to better transfer than that of a state-of-the-art slide-level foundation model, while using only 6.5% of the training data. These findings indicate that MIL architectures exhibit robust adaptability, offering insights into the benefits of leveraging pretrained models to enhance performance in computational pathology.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-shao25a, title = {Do Multiple Instance Learning Models Transfer?}, author = {Shao, Daniel and Chen, Richard J. and Song, Andrew H. and Runevic, Joel and Lu, Ming Y. and Ding, Tong and Mahmood, Faisal}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {54219--54238}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/shao25a/shao25a.pdf}, url = {https://proceedings.mlr.press/v267/shao25a.html}, abstract = {Multiple Instance Learning (MIL) is a cornerstone approach in computational pathology for distilling embeddings from gigapixel tissue images into patient-level representations to predict clinical outcomes. However, MIL is frequently challenged by the constraints of working with small, weakly-supervised clinical datasets. Unlike fields such as natural language processing and computer vision, which effectively use transfer learning to improve model quality in data-scarce environments, the transferability of MIL models remains largely unexplored. We conduct the first comprehensive investigation into transfer learning capabilities of pretrained MIL models, evaluating 11 MIL models across 19 pretraining tasks spanning tissue subtyping, cancer grading, and molecular subtype prediction. We observe a substantial performance boost with finetuning pretrained models over training from randomly initialized weights, even with domain differences between pretraining and target tasks. Pretraining on pan-cancer datasets enables consistent generalization across organs and task types compared to single-disease pretraining. Remarkably, this pan-cancer pretraining leads to better transfer than that of a state-of-the-art slide-level foundation model, while using only 6.5% of the training data. These findings indicate that MIL architectures exhibit robust adaptability, offering insights into the benefits of leveraging pretrained models to enhance performance in computational pathology.} }
Endnote
%0 Conference Paper %T Do Multiple Instance Learning Models Transfer? %A Daniel Shao %A Richard J. Chen %A Andrew H. Song %A Joel Runevic %A Ming Y. Lu %A Tong Ding %A Faisal Mahmood %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-shao25a %I PMLR %P 54219--54238 %U https://proceedings.mlr.press/v267/shao25a.html %V 267 %X Multiple Instance Learning (MIL) is a cornerstone approach in computational pathology for distilling embeddings from gigapixel tissue images into patient-level representations to predict clinical outcomes. However, MIL is frequently challenged by the constraints of working with small, weakly-supervised clinical datasets. Unlike fields such as natural language processing and computer vision, which effectively use transfer learning to improve model quality in data-scarce environments, the transferability of MIL models remains largely unexplored. We conduct the first comprehensive investigation into transfer learning capabilities of pretrained MIL models, evaluating 11 MIL models across 19 pretraining tasks spanning tissue subtyping, cancer grading, and molecular subtype prediction. We observe a substantial performance boost with finetuning pretrained models over training from randomly initialized weights, even with domain differences between pretraining and target tasks. Pretraining on pan-cancer datasets enables consistent generalization across organs and task types compared to single-disease pretraining. Remarkably, this pan-cancer pretraining leads to better transfer than that of a state-of-the-art slide-level foundation model, while using only 6.5% of the training data. These findings indicate that MIL architectures exhibit robust adaptability, offering insights into the benefits of leveraging pretrained models to enhance performance in computational pathology.
APA
Shao, D., Chen, R.J., Song, A.H., Runevic, J., Lu, M.Y., Ding, T. & Mahmood, F.. (2025). Do Multiple Instance Learning Models Transfer?. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:54219-54238 Available from https://proceedings.mlr.press/v267/shao25a.html.

Related Material