Foundation Multi-modal Model for Modality Synthesis & Segmentation of Magnetic Resonance Images

Project Details

Description

Our project focuses on creating a cutting-edge tool for analyzing and synthesizing MRI images, which are used in medical imaging to diagnose and monitor various conditions. Specifically, we're developing a system that can accurately identify different parts of cancer tumors in brain MRI scans and even generate new images across different MRI protocols.
To achieve this, we're building on existing technology called the Mixture of Experts architecture where we combine multiple machine learning models that are trained to specialize in different part of the brain images. We're training our system using a method called Multimodal self-supervised learning, which is a way of teaching it to recognize patterns in the images without needing lots of labeled examples. Instead, it learns by comparing similar and different images to understand what's important.
Overall, our goal is to make MRI analysis and synthesis more accurate and efficient, which could ultimately help doctors make better diagnoses and treatment plans for patients.
AcronymFM2MRI
StatusActive
Effective start/end date5/01/255/10/27

Funding

  • FNR - Fonds National de la Recherche: €685,000.00

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.