Abstract
Low count PET (positron emission tomography) imaging is often desirable in clinical diagnosis and biomedical research, but its images are generally very noisy, due to the very weak signals in the sinograms used in image reconstruction. To address this issue, this paper presents a novel kernel graph filtering method for dynamic PET sinogram denoising. This method is derived from treating the dynamic sinograms as the signals on a graph, and learning the graph adaptively from the kernel principal components of the sinograms to construct a lowpass kernel graph spectrum filter. The kernel graph filter thus obtained is then used to filter the original sinogram time frames to obtain the denoised sinograms for PET image reconstruction. Extensive tests and comparisons on the simulated and real life in-vivo dynamic PET datasets show that the proposed method outperforms the existing methods in sinogram denoising and image enhancement of dynamic PET at all count levels, especially at low count, with a great potential in real life applications of dynamic PET imaging.
Highlights
Positron emission tomography (PET) is a functional imaging modality
Inspired by the success of Graph signal processing (GSP), we present in this paper a novel graph filtering method for dynamic sinogram denoising to significantly improve PET image quality
We propose a novel kernel graph filtering method to denoise the dynamic sinograms for dynamic PET image reconstruction
Summary
Positron emission tomography (PET) is a functional imaging modality. By monitoring the distribution of radioactive tracers, it can detect the early onset of diseases [1], such as Alzheimer’s disease. To show the rapid change of in-vivo radioactive tracer accurately, time frame with short scan durations should be applied, especially in early frame. This leads to lower counts in each frame, which results in lower signal-to-noise ratio of reconstructed PET image. To improve the quality of PET image with low counts, three types of methods have been proposed in the literature. Compared to the reconstruction methods without using priors, these methods can significantly reduce the noise level of PET image. These methods generally increase computational cost and may introduce artifacts. When magnetic resonance (MR) images are used as prior information in PET image reconstruction, some MR-only information, such as the bone or lesion existing only in MR image [4], may be introduced in the reconstructed PET image
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
