RPI ID: 2022-041-301

Innovation Summary:
This invention introduces a novel motion correction framework for photon-counting computed tomography (PCCT) using locally linear embedding (LLE). The method addresses motion artifacts by modeling the high-dimensional projection data as a low-dimensional manifold and applying LLE to recover motion-free representations. Unlike traditional correction techniques that rely on rigid or affine transformations, this approach captures nonlinear motion patterns and preserves fine structural details. The algorithm is compatible with spectral CT data and enhances image quality across multiple energy channels, making it highly suitable for dynamic or motion-prone imaging scenarios.

Challenges / Opportunities:
Motion artifacts remain a major limitation in high-resolution CT imaging, especially in cardiac, pediatric, and trauma applications. Existing correction methods often fail to handle complex, non-rigid motion or require additional hardware and computational overhead. This invention offers a software-based, data-driven solution that improves diagnostic accuracy without altering the imaging hardware. It opens opportunities for real-time correction, spectral imaging enhancement, and AI-integrated reconstruction pipelines in next-generation CT systems.

Key Benefits / Advantages:
✔ Corrects nonlinear motion artifacts in photon-counting CT
✔ Preserves spectral and spatial resolution
✔ Hardware-agnostic and computationally efficient
✔ Enhances diagnostic accuracy in dynamic imaging
✔ Compatible with AI-based reconstruction workflows

Applications:
• Cardiac and thoracic CT imaging
• Pediatric and trauma diagnostics
• Spectral and multi-energy CT
• Motion-robust imaging in clinical and research settings

Keywords:
#motioncorrection #photoncountingCT #spectralimaging #LLE #medicalimagingAI #nonlinearmotion

Intellectual Property:
PCT Application Filed PCT/US2023/016359 (Publication No.: WO2023183640A1)