Pith. sign in

REVIEW 6 cited by

MRI-CORE: A Foundation Model for Magnetic Resonance Imaging

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2506.12186 v2 pith:TCV6A32V submitted 2025-06-13 eess.IV cs.AIcs.CVcs.LG

MRI-CORE: A Foundation Model for Magnetic Resonance Imaging

classification eess.IV cs.AIcs.CVcs.LG
keywords datafoundationmodelmodelsmri-coreimaginglearningmagnetic
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

The widespread use of Magnetic Resonance Imaging (MRI) in combination with deep learning shows promise for many high-impact automated diagnostic and prognostic tools. However, training new models requires large amounts of labeled data, a challenge due to high cost of precise annotations and data privacy. To address this issue, we introduce the MRI-CORE, a vision foundation model trained using more than 6 million slices from over 110 thousand MRI volumes across 18 body locations. Our experiments show notable improvements in performance over state-of-the-art methods in 13 data-restricted segmentation tasks, as well as in image classification, and zero-shot segmentation, showing the strong potential of MRI-CORE to enable data-efficient development of artificial intelligence models. We also present data on which strategies yield most useful foundation models and a novel analysis relating similarity between pre-training and downstream task data with transfer learning performance. Our model is publicly available with a permissive license.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. How Much MRI Preprocessing Is Enough? A Cost-Utility Study for Brain MRI Foundation Models

    cs.CV 2026-06 conditional novelty 6.0

    For MAE and JEPA brain MRI pretraining, moving from the cheapest stable preprocessing level (P2) to the best heavier pipeline buys only 3.4 and 1.8 percentage points of aggregate downstream utility, and most gains are...

  2. LegSegNet: A Public Deep Learning System for Lower Extremity CT Tissue Segmentation and Quantification

    cs.CV 2026-05 unverdicted novelty 6.0

    LegSegNet is the first public end-to-end deep learning system for lower extremity CT tissue segmentation and body composition quantification, reporting an average Dice score of 89.31 on held-out test slices.

  3. Revisiting 2D Foundation Models for Scalable 3D Medical Image Classification

    cs.CV 2025-12 conditional novelty 5.0

    A frozen 2D vision foundation model with lightweight LoRA adapters and attention-based slice fusion achieves state-of-the-art 3D medical image classification across 12 tasks with about 1M trainable parameters per task.

  4. A Benchmark of (MRI-) Foundation Models to Predict IDH Mutational Status in Glioma

    eess.IV 2026-06 accept novelty 4.0

    Radiomics TabPFN matches or outperforms image foundation models for IDH prediction in glioma MRI, with results sensitive to cohort shifts and representation type.

  5. A Benchmark of (MRI-) Foundation Models to Predict IDH Mutational Status in Glioma

    eess.IV 2026-06 unverdicted novelty 4.0

    TabPFN on radiomic features matched or outperformed image foundation models for IDH mutational status prediction in glioma MRI, with BiomedCLIP strongest among visual encoders and performance sensitive to cohort shift...

  6. How Much MRI Preprocessing Is Enough? A Cost-Utility Study for Brain MRI Foundation Models

    cs.CV 2026-06 accept novelty 4.0

    Empirical comparison of graded MRI preprocessing levels for MAE and JEPA pretraining on brain scans shows moderate levels (P2) are often sufficient, with limited additional utility from stronger preprocessing on downs...