PPFM: Image denoising in photon-counting CT using single-step posterior sampling Poisson flow generative models

Dennis Hein,Staffan Holmin,Timothy Szczykutowicz,Jonathan S Maltz,Mats Danielsson,Ge Wang,Mats Persson

2023-12-19

Abstract:Diffusion and Poisson flow models have shown impressive performance in a wide range of generative tasks, including low-dose CT image denoising. However, one limitation in general, and for clinical applications in particular, is slow sampling. Due to their iterative nature, the number of function evaluations (NFE) required is usually on the order of $10-10^3$, both for conditional and unconditional generation. In this paper, we present posterior sampling Poisson flow generative models (PPFM), a novel image denoising technique for low-dose and photon-counting CT that produces excellent image quality whilst keeping NFE=1. Updating the training and sampling processes of Poisson flow generative models (PFGM)++, we learn a conditional generator which defines a trajectory between the prior noise distribution and the posterior distribution of interest. We additionally hijack and regularize the sampling process to achieve NFE=1. Our results shed light on the benefits of the PFGM++ framework compared to diffusion models. In addition, PPFM is shown to perform favorably compared to current state-of-the-art diffusion-style models with NFE=1, consistency models, as well as popular deep learning and non-deep learning-based image denoising techniques, on clinical low-dose CT images and clinical images from a prototype photon-counting CT system.

Image and Video Processing,Computer Vision and Pattern Recognition,Medical Physics

What problem does this paper attempt to address?

The paper attempts to address the problem of efficient denoising in low-dose and photon-counting CT (Computed Tomography) images. Specifically, the paper focuses on how to reduce the number of function evaluations (NFE) required to generate samples while maintaining high-quality images, thereby improving sampling speed to meet the real-time processing needs in clinical applications. Existing diffusion models and Poisson flow generative models, although performing well in generative tasks, have slower sampling speeds due to their iterative nature, typically requiring 10 to 1000 function evaluations. This is a significant limitation in clinical environments where rapid processing of large amounts of data is needed.

To this end, the paper proposes the Posterior Sampling Poisson Flow Model (PPFM), a novel image denoising technique. PPFM updates the training and sampling process of the Poisson Flow Generative Model (PFGM)++ to learn a conditional generator that defines the trajectory from the prior noise distribution to the target posterior distribution. Additionally, the paper proposes a method to "hijack" and regularize the sampling process to achieve single-step sampling (NFE=1). This method not only produces high-quality images but also greatly improves sampling efficiency, making it more practical for clinical applications.

The main contributions of the paper include:
1. Proposing the PPFM framework, which achieves single-step sampling while maintaining high-quality images.
2. Exploring the advantages of using the PFGM++ framework compared to diffusion models in image denoising tasks, especially in the performance when the variable D is used as a hyperparameter.
3. Experimental results demonstrate that the proposed method outperforms current state-of-the-art single-step sampling diffusion models and other popular deep learning and non-deep learning image denoising techniques on clinical low-dose CT images and clinical images acquired from photon-counting CT prototype systems.

These findings are significant for advancing low-dose CT and photon-counting CT technologies, helping to improve image quality while reducing radiation dose, thereby better serving clinical diagnosis and treatment planning.