arXiv Computer Vision
BrainTaskonomy: Learning How to Pretrain and What to Transfer in fMRI Foundation Models
Read the original on arXiv Computer Vision →The Flow has not summarised this story yet — read it at arXiv Computer Vision.
Related stories
arXiv Computer Vision
Sep 24
A generalizable structural brain MRI foundation model built through dual-priority federated pretraining
BrainFedFM is a structural brain MRI foundation model that was federatively pretrained on 164,707 3‑D scans from 42 sites using a dual‑priority approach that emphasizes informative anatomical regions locally and prioritizes site contributions globally. The model outperformed seven baseline models—including four centralized foundation models—across 20 downstream tasks (classification, regression, segmentation), achieving a mean rank of 1.68 and a 50% performance gain, especially in classification and regression and among underrepresented populations. These results demonstrate the model’s generalizability and show that federated pretraining can effectively develop neuroimaging foundation models without pooling raw images.
arXiv Machine Learning
Sep 21
BrainWideBench: Benchmarking large-scale pretraining and across-animal transfer in multi-region neural recordings
BrainWideBench is a benchmark that evaluates across‑animal transfer on multi‑region neural recordings from 139 mice, covering 276 brain regions. It comprises three task suites—behavior decoding, neural activity prediction, and anatomical organization recovery—to test whether learned representations support diverse downstream objectives. The benchmark shows that while pretraining improves performance over single‑session baselines, current methods vary in transfer ability and none perform uniformly well across all suites, highlighting the challenge of developing general‑purpose neural representations.
arXiv Computer Vision
Sep 7
A Generalizable Feature Extractor for Alzheimer's-Related Brain MRI Tasks
The study investigates whether a compact, supervised 3D CNN pretrained for brain‑age prediction can act as a reusable foundation model for various Alzheimer's‑related neuroimaging tasks. By freezing the 7.18 million weights and adding only ~1 % of trainable parameters via Low‑Rank Adaptation, the model achieved high performance across six experiments, including dementia classification, MCI progression prediction, amyloid positivity detection, and volume estimation of hippocampal and white matter hypointensities. The results demonstrate that the pretrained brain‑age model generalizes well to new datasets without retraining, offering a data‑efficient alternative to larger networks.
arXiv AI
Jun 4
The Variance Brain Foundation Models Forgot: Third-Order Statistics Predict Cognition Where Billion-Parameter Models Fail
arXiv:2606. 04010v1 Announce Type: cross Abstract: Brain foundation models (BFMs) are self-supervised Transformers pretrained on fMRI data.
arXiv Machine Learning
Sep 24
A Scaling Study for fMRI Foundation Models
The study investigates how data volume, model size, and training duration affect the performance of fMRI foundation models. Using over 200 datasets and 10,000 GPU‑hours, the authors find that larger models benefit more from additional data, and that at a fixed compute budget, increasing data yields greater gains than enlarging the model. By selecting optimal combinations of data, size, and duration, they produce models that outperform existing fMRI foundation models on out‑of‑distribution tasks while requiring less pretraining compute.
arXiv Machine Learning
Aug 4
EEG-FM-Compass: Progress, Benchmarking, and Future Directions for EEG Foundation Models
arXiv:2601. 17883v3 Announce Type: replace Abstract: Electroencephalography (EEG) foundation models (FMs) have recently emerged as a promising paradigm for brain-computer interfaces, aiming to learn transferable neural representations from large-scale heterogeneous recordings.