Value Proposition
· Superior Image Quality – the proposed nonlinear DPS method outperforms traditional reconstruction methods and linear DPS models, offering better high-quality images across different acquisition protocols compared to conditionally trained deep learning approaches
· State-of-the-Art Reconstruction – cutting-edge implementations of combining generative AI with rigorous physical modeling for medical imaging systems, particularly CT
· Boosts Diagnostic Accuracy and Efficiency – improved the precision of medical image diagnosis and speeds up the workflow for radiologists, leading to better patient care and more efficient medical imaging processes
Technology Description
· The disclosed technology is a comprehensive package of software tools designed to enhance tomographic reconstruction. It leverages generative neural networks to improve the utility of CT information, enabling rapid processing of 3D data, clear differentiation between materials (such as implants), and initialization that reduces variability and enhances algorithm stability.
Unmet Need
· The disclosed technology addresses the need for improved tomographic reconstruction by leveraging advanced deep learning approaches. It enhances the accuracy of medical image diagnosis and increases throughput for radiologists, ultimately improving patient care and efficiency in medical imaging.
Stage of Development
· Current efforts are focused on creating a functional software development kit, which is expected to take a few months to complete.
Data Availability
· Data available upon request.
Publications
Li, et al. (2024). CT reconstruction using diffusion posterior sampling conditioned on a nonlinear measurement model. Journal of medical imaging (Bellingham, Wash.), 11(4), 043504. https://doi.org/10.1117/1.JMI.11.4.043504
Li, et al. (2024). CT Reconstruction using Nonlinear Diffusion Posterior Sampling with Detector Blur Modeling. Conference proceedings. International Conference on Image Formation in X-Ray Computed Tomography, 2024, 30–33. CT Reconstruction using Nonlinear Diffusion Posterior Sampling with Detector Blur Modeling - PubMed
Jiang, et al. (2024). Multi-Material Decomposition Using Spectral Diffusion Posterior Sampling. ArXiv [Preprint], arXiv:2408.01519v1. Multi-Material Decomposition Using Spectral Diffusion Posterior Sampling - PubMed
Jiang, et al. (2024). Strategies for CT Reconstruction using Diffusion Posterior Sampling with a Nonlinear Model. ArXiv [Preprint], arXiv:2407.12956v1. Strategies for CT Reconstruction using Diffusion Posterior Sampling with a Nonlinear Model - PMC
