General Methods Make Great Domain-specific Foundation Models: A Case-study on Fetal Ultrasound
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912446983176192 |
|---|---|
| author | Ambsdorf, Jakob Munk, Asbjørn Llambias, Sebastian Christensen, Anders Nymark Mikolaj, Kamil Balestriero, Randall Tolsgaard, Martin Feragen, Aasa Nielsen, Mads |
| author_facet | Ambsdorf, Jakob Munk, Asbjørn Llambias, Sebastian Christensen, Anders Nymark Mikolaj, Kamil Balestriero, Randall Tolsgaard, Martin Feragen, Aasa Nielsen, Mads |
| contents | With access to large-scale, unlabeled medical datasets, researchers are confronted with two questions: Should they attempt to pretrain a custom foundation model on this medical data, or use transfer-learning from an existing generalist model? And, if a custom model is pretrained, are novel methods required? In this paper we explore these questions by conducting a case-study, in which we train a foundation model on a large regional fetal ultrasound dataset of 2M images. By selecting the well-established DINOv2 method for pretraining, we achieve state-of-the-art results on three fetal ultrasound datasets, covering data from different countries, classification, segmentation, and few-shot tasks. We compare against a series of models pretrained on natural images, ultrasound images, and supervised baselines. Our results demonstrate two key insights: (i) Pretraining on custom data is worth it, even if smaller models are trained on less data, as scaling in natural image pretraining does not translate to ultrasound performance. (ii) Well-tuned methods from computer vision are making it feasible to train custom foundation models for a given medical domain, requiring no hyperparameter tuning and little methodological adaptation. Given these findings, we argue that a bias towards methodological innovation should be avoided when developing domain specific foundation models under common computational resource constraints. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_19552 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | General Methods Make Great Domain-specific Foundation Models: A Case-study on Fetal Ultrasound Ambsdorf, Jakob Munk, Asbjørn Llambias, Sebastian Christensen, Anders Nymark Mikolaj, Kamil Balestriero, Randall Tolsgaard, Martin Feragen, Aasa Nielsen, Mads Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning I.4 With access to large-scale, unlabeled medical datasets, researchers are confronted with two questions: Should they attempt to pretrain a custom foundation model on this medical data, or use transfer-learning from an existing generalist model? And, if a custom model is pretrained, are novel methods required? In this paper we explore these questions by conducting a case-study, in which we train a foundation model on a large regional fetal ultrasound dataset of 2M images. By selecting the well-established DINOv2 method for pretraining, we achieve state-of-the-art results on three fetal ultrasound datasets, covering data from different countries, classification, segmentation, and few-shot tasks. We compare against a series of models pretrained on natural images, ultrasound images, and supervised baselines. Our results demonstrate two key insights: (i) Pretraining on custom data is worth it, even if smaller models are trained on less data, as scaling in natural image pretraining does not translate to ultrasound performance. (ii) Well-tuned methods from computer vision are making it feasible to train custom foundation models for a given medical domain, requiring no hyperparameter tuning and little methodological adaptation. Given these findings, we argue that a bias towards methodological innovation should be avoided when developing domain specific foundation models under common computational resource constraints. |
| title | General Methods Make Great Domain-specific Foundation Models: A Case-study on Fetal Ultrasound |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning I.4 |
| url | https://arxiv.org/abs/2506.19552 |