# Revisiting Automatic Data Curation for Vision Foundation Models in Digital Pathology

arXiv: https://arxiv.org/abs/2503.18709 · HTML: https://mlsanity.com/paper/2503.18709/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 19 of its 26 references that are in the map.
- First posted on arXiv 24 Mar 2025.

## Builds on (most cited first, 19 of 19)

- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [Emerging Properties in Self-Supervised Vision Transformers](https://mlsanity.com/paper/2104.14294/) (2021)
- [Attention-based Deep Multiple Instance Learning](https://mlsanity.com/paper/1802.04712/) (2018)
- [An Empirical Study of Example Forgetting during Deep Neural Network Learning](https://mlsanity.com/paper/1812.05159/) (2018)
- [Beyond neural scaling laws: beating power law scaling via data pruning](https://mlsanity.com/paper/2206.14486/) (2022)
- [BRACS: A Dataset for BReAst Carcinoma Subtyping in H&E Histology Images](https://mlsanity.com/paper/2111.04740/) (2021)
- [BACH: Grand Challenge on Breast Cancer Histology Images](https://mlsanity.com/paper/1808.04277/) (2018)
- [CellViT: Vision Transformers for Precise Cell Segmentation and Classification](https://mlsanity.com/paper/2306.15350/) (2023)
- [Virchow: A Million-Slide Digital Pathology Foundation Model](https://mlsanity.com/paper/2309.07778/) (2023)
- [Deep Active Learning over the Long Tail](https://mlsanity.com/paper/1711.00941/) (2017)
- [Automatic Data Curation for Self-Supervised Learning: A Clustering-Based Approach](https://mlsanity.com/paper/2405.15613/) (2024)
- [Hard Sample Aware Noise Robust Learning for Histopathology Image Classification](https://mlsanity.com/paper/2112.03694/) (2021)
- [RudolfV: A Foundation Model by Pathologists for Pathologists](https://mlsanity.com/paper/2401.04079/) (2024)
- [WSSS4LUAD: Grand Challenge on Weakly-supervised Tissue Semantic Segmentation for Lung Adenocarcinoma](https://mlsanity.com/paper/2204.06455/) (2022)
- [UniToPatho, a labeled histopathological dataset for colorectal polyps classification and adenoma dysplasia grading](https://mlsanity.com/paper/2101.09991/) (2021)
- [Active Learning Strategies for Weakly-supervised Object Detection](https://mlsanity.com/paper/2207.12112/) (2022)
- [Benchmarking Embedding Aggregation Methods in Computational Pathology: A Clinical Data Perspective](https://mlsanity.com/paper/2407.07841/) (2024)

## Built on by (most cited first, 4 of 4)

- [SPADE: Spatial Transcriptomics and Pathology Alignment Using a Mixture of Data Experts for an Expressive Latent Space](https://mlsanity.com/paper/2506.21857/) (2025)
- [CleanPatrick: A Benchmark for Image Data Cleaning](https://mlsanity.com/paper/2505.11034/) (2025)
- [LEMON: a foundation model for nuclear morphology in Computational Pathology](https://mlsanity.com/paper/2603.25802/) (2026)
- [HERO: Histology Encoder for Robust Representation in Oncology](https://mlsanity.com/paper/2609.35943/) (2026)

## Abstract (arXiv)

Vision foundation models (FMs) are accelerating the development of digital pathology algorithms and transforming biomedical research. These models learn, in a self-supervised manner, to represent histological features in highly heterogeneous tiles extracted from whole-slide images (WSIs) of real-world patient samples. The performance of these FMs is significantly influenced by the size, diversity, and balance of the pre-training data. However, data selection has been primarily guided by expert knowledge at the WSI level, focusing on factors such as disease classification and tissue types, while largely overlooking the granular details available at the tile level. In this paper, we investigate the potential of unsupervised automatic data curation at the tile-level, taking into account 350 million tiles. Specifically, we apply hierarchical clustering trees to pre-extracted tile embeddings, allowing us to sample balanced datasets uniformly across the embedding space of the pretrained FM. We further identify these datasets are subject to a trade-off between size and balance, potentially compromising the quality of representations learned by FMs, and propose tailored batch sampling strategies to mitigate this effect. We demonstrate the effectiveness of our method through improved performance on a diverse range of clinically relevant downstream tasks.