RPI ID: 2018-008-401, 2018-008-601
Innovation Summary:
This invention from Rensselaer Polytechnic Institute introduces a deep learning-based framework for reducing metal artifacts in computed tomography (CT) images. The system leverages a neural network trained on paired CT datasets—with and without metal artifacts—to reconstruct high-fidelity images that preserve anatomical detail while eliminating distortions caused by metallic implants. The approach integrates both projection-domain and image-domain correction strategies, enabling robust performance across a wide range of clinical scenarios. It is designed to be compatible with existing CT systems and workflows, offering a practical upgrade path for improving diagnostic accuracy in patients with metal implants.
Challenges / Opportunities:
Metal artifacts in CT imaging—caused by dental fillings, orthopedic implants, or surgical hardware—can obscure critical anatomical structures and compromise diagnostic accuracy. Traditional correction methods often fall short in balancing artifact suppression with detail preservation. This invention addresses these limitations by applying deep neural networks that intelligently distinguish and correct artifact regions, opening opportunities for safer, more accurate imaging in orthopedics, oncology, and trauma care.
Key Benefits / Advantages:
✔ Significant reduction of metal-induced artifacts in CT images
✔ Preservation of anatomical detail and diagnostic integrity
✔ Dual-domain correction: projection and image space
✔ Compatible with existing CT hardware and software
✔ Scalable for real-time clinical deployment
Applications:
• Orthopedic and dental imaging
• Oncology and radiation therapy planning
• Emergency and trauma diagnostics
• Post-operative imaging
• AI-powered radiology tools
Keywords:
#CT metal artifact reduction #deep learning #medical imaging #neural networks #artifact correction #AI in radiology
Intellectual Property:
US Issued patent(s) US 11,589,834 B2, 11,872,070 B2
