Newest papers

Brain2Vec: A Self-Supervised 3D Foundation Model for Structural Brain MRI

Spiros Maggioros, Güray Erus, Gareth A. Harman, George Aidinis, Alexander Getka, Aristeidis Sotiras, Χρήστος Δαβατζίκος · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28411 · cited by 0 works (OpenAlex) · open-access PDF

French

Les modèles d'apprentissage profond à travers les régions sont devenus des outils essentiels dans l'imagerie des données médicales, mais leur formation de bout en bout pour diverses tâches reste difficile en raison des données étiquetées limitées et des contraintes de calcul. Pour y remédier, nous présentons Brain2Vec, un modèle de base préliminaire formé sur 74,425 T1 pesées Imagerie par résonance magnétique prédictive (IRM) scans regroupés à partir de plusieurs études. Brain2Vec peut servir à la fois comme un backbone d'initialisation de poids et de fine-tuning sur les données d'installation pour finaliser les tâches de prédiction de l'âge du cerveau et comme un extracteur de fonctionnalités

Machine translation (Meta M2M100); the English original follows.

Abstract

Deep learning models have become essential tools in medical imaging, yet training them end-to-end for diverse tasks remains challenging due to limited labeled data and computational constraints. To address this, we introduce Brain2Vec, a foundation model pre-trained on 74,425 T1-weighted Magnetic Resonance Imaging (MRI) scans pooled from multiple studies. Brain2Vec can serve both as a weight initialization backbone for fine-tuning downstream tasks and as a standalone feature extractor. We evaluated Brain2Vec in three tasks: downstream predictive performance, convergence speed, and embedding quality. Using Brain2Vec's pretrained weights as initialization and fine-tuning on labeled data for a brain age prediction task, results demonstrate improved final accuracy and faster convergence compared to random initialization. Additionally, its frozen embeddings outperform anatomically defined volumetric features on four of five downstream classification tasks and perform competitively with a state-of-the-art foundation model pre-trained on substantially more data. Lastly, we show that Brain2Vec encodes information primarily from a spatially localized and neuroanatomically meaningful set of brain regions, and that manipulating a subset of latent features produces monotonic, region-specific effects on the reconstructed anatomy. Brain2Vec code and trained models are available as an open-source package on https://github.com/CBICA/Brain2Vec. Users can apply Brain2Vec locally by installing the package, or directly through the NiChart cloud platform https://cloud.neuroimagingchart.com, where Brain2Vec features can be derived without requiring local installation or specialized infrastructure.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Initialization, Computer science, Downstream (manufacturing), Artificial intelligence, Set (abstract data type), Feature (linguistics)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange