Towards Brain MRI Foundation Models for the Clinic: Findings from the FOMO25 Challenge

TL;DR

This paper presents the FOMO25 challenge, demonstrating that self-supervised foundation models trained on large, heterogeneous clinical brain MRI data can outperform supervised models, with task-specific pretraining objectives and small models showing strong results.

Contribution

The paper introduces the FOMO25 challenge, providing a large pretraining dataset and benchmarking foundation models on clinical brain MRI tasks with diverse data and evaluation settings.

Findings

Self-supervised pretraining improves generalization on clinical data under domain shift.

Out-of-domain trained models surpass supervised in-domain baselines.

Different pretraining objectives benefit specific tasks, e.g., MAE for segmentation.

Abstract

Clinical deployment of automated brain MRI analysis faces a fundamental challenge: clinical data is heterogeneous and noisy, and high-quality labels are prohibitively costly to obtain. Self-supervised learning (SSL) can address this by leveraging the vast amounts of unlabeled data produced in clinical workflows to train robust \textit{foundation models} that adapt out-of-domain with minimal supervision. However, the development of foundation models for brain MRI has been limited by small pretraining datasets and in-domain benchmarking focused on high-quality, research-grade data. To address this gap, we organized the FOMO25 challenge as a satellite event at MICCAI 2025. FOMO25 provided participants with a large pretraining dataset, FOMO60K, and evaluated models on data sourced directly from clinical workflows in few-shot and out-of-domain settings. Tasks covered infarct classification, meningioma segmentation, and brain age regression, and considered both models trained on FOMO60K (method track) and any data (open track). Nineteen foundation models from sixteen teams were evaluated using a standardized containerized pipeline. Results show that (a) self-supervised pretraining improves generalization on clinical data under domain shift, with the strongest models trained \textit{out-of-domain} surpassing supervised baselines trained \textit{in-domain}. (b) No single pretraining objective benefits all tasks: MAE favors segmentation, hybrid reconstruction-contrastive objectives favor classification, and (c) strong performance was achieved by small pretrained models, and improvements from scaling model size and training duration did not yield reliable benefits.