RPI ID:
2023-078-301

Innovation Summary:
This technology introduces a tomographic image reconstruction system that leverages parallel denoising diffusion probabilistic models (DDPMs) to recover high-quality images from sparse projection data. The system uses a probabilistic framework to iteratively refine image estimates, enabling accurate reconstructions even with limited input data. Parallel processing accelerates computation, making the method suitable for real-time or large-scale imaging applications. The approach is particularly effective in reducing noise and artifacts common in low-dose or fast-scan tomography.

Challenges / Opportunities:
Traditional reconstruction methods struggle with sparse or noisy data, often producing low-resolution or artifact-laden images. This invention addresses the need for robust reconstruction techniques that maintain image fidelity under constrained acquisition conditions. It opens opportunities for safer medical imaging with reduced radiation exposure and faster scan times. The method also supports integration into existing imaging platforms with minimal hardware changes.

Key Benefits / Advantages:
✔ High-quality image reconstruction from sparse data
✔ Reduces noise and artifacts using DDPMs
✔ Enables low-dose and fast-scan tomography
✔ Parallel processing for computational efficiency
✔ Compatible with existing imaging systems

Applications:
• Medical imaging (CT, PET)
• Industrial and security tomography

Keywords:
Tomographic reconstruction, denoising diffusion models, sparse projections, medical imaging, parallel processing

Intellectual Property:
WO2024243250A2 (Application PCT/US2024/030455), Published May 22, 2024