Technical Description
The invention discloses an AI-based multi-contrast MRI reconstruction method that uses kernel-modulated CNNs and bi-level meta-learning to reconstruct multiple MRI contrasts from under-sampled k-space data. A single multi-task model adapts to diverse and even unseen contrast protocols, improving image quality and reducing scan time while remaining computationally and storage efficient.
Problems Addressed
- Long MRI Acquisitions
- Motion/Artifact Degradation
- Single-Contrast DL Models
- Poor Generalization to Protocols
- High Computation Retraining
- No Built-In Evaluation
Tech Features
- Multi-Task Single Model
- Kernel Modulation Network
- Bi-Level Meta-Learning
- Contrast-Aware Conditioning
- Under-Sampled K-Space Reconstruction
- Iterative Correction-Evaluation Loop
Target Audience
- MRI Scanner Manufacturers Industry
- Radiology Service Providers Sector
- Healthcare AI Software Vendors Industry
- PACS/RIS Imaging IT Companies Sector
- Research & Development
Tech ID: P21-1978 TRL 4 Patent Status: Granted Available For Exclusive and Non-exclusive License
