October 12, 2025Open Access

MM-DINOv2: Adapting Foundation Models for Multi-Modal Medical Image Analysis

View Full Paper

Ask AI

Bookmark

MM-DINOv2 achieves a Matthews Correlation Coefficient (MCC) of 0.6 in glioma subtype classification, improving accuracy by 11.1%.
By using multi-modal patch embeddings, the framework effectively processes various imaging data for clinical applications.
Full-modality masking helps the model learn robust relationships across missing modalities, addressing a common clinical challenge.
This framework leverages large unlabeled datasets, enhancing prediction reliability in real-world medical settings.

Discussion

No takes yet. Share an insight, caveat, or question.

Novel approach enhances medical imaging by employing multi-modal patch embeddings and semi-supervised learning techniques.

Scholz et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark