Technical Description
This invention introduces a curriculum-based model-agnostic meta-learning (CMAML) framework for post-processing Magnetic Resonance images. It utilizes a single deep learning model to capture artifact-invariant latent representations, enabling adaptive restoration across multiple known and unseen artifact types and degradation levels.
Problems Addressed
- Highly Operator-Dependent
- Highly Patient-Dependent
- Economically Inefficient Hardware
- Computationally Inefficient Models
- Lacks Artifact-Invariant Data
- Lacks Artifact-Specific Restoration
- Tedious Training Processes
Tech Features
- Enhanced Scan Quality
- Efficient Single-Model Restoration
- Adaptive Meta-Learning Framework
- Enhanced Artifact Generalization
- Automated Latent Representation
- Optimized Training Efficiency
Target Audience
- Healthcare Clinical Sectors
- Medical Imaging Industries
- MRI Manufacturer Sectors
- Diagnostic Software Industries
- Research & Development
Tech ID: P24-2049 TRL 6 Patent Status: Published Available For Exclusive and Non-exclusive License
