Whole-body PET self-attenuation correction based on dual-branch residual diffusion model(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2026年第4期
- Page:
- 445-454
- Research Field:
- 医学影像物理
- Publishing date:
Info
- Title:
- Whole-body PET self-attenuation correction based on dual-branch residual diffusion model
- Author(s):
- LI Shengjun1; FENG Qianjin1; 2
- 1. School of Biomedical Engineering, Southern Medical University, Guangzhou 510515, China 2. Guangdong Provincial Key Laboratory of Medical Image Processing, Guangzhou 510515, China
- Keywords:
- Keywords: positron emission computed tomography attenuation correction residual diffusion model mutual guidance attenuation prior
- PACS:
- R318
- DOI:
- DOI:10.3969/j.issn.1005-202X.2026.04.004
- Abstract:
- Abstract: Objective A dual-branch residual diffusion model (DRDM) is proposed to perform whole-body PET self-attenuation correction, enabling the simultaneous generation of attenuation-corrected PET (AC-PET) and computed tomography (CT) images using only non-attenuation-corrected PET (NAC-PET) images. Methods The proposed DRDM adopts a coarse-to-fine dual-branch architecture to jointly perform PET self-attenuation correction and CT generation. The coarse mapping relationships from NAC-PET to AC-PET and from NAC-PET to CT are independently established based on the residual diffusion framework. Then, the generated coarse AC-PET and CT images undergo a joint refinement process. Specifically, a dual-branch mutual guidance module is utilized to facilitate the exchange and fusion of AC-PET and CT features during the reverse diffusion process. Additionally, attenuation prior knowledge learned from real CT images is incorporated to further enhance the fidelity of the generated images. Results Experiments conducted on public PET/CT dataset demonstrate that the AC-PET generated by DRDM achieves a mean absolute error of 0.049 2±0.030 6, a peak signal-to-noise ratio of (39.22±2.24) dB, a structural similarity index measure of 0.981 4±0.015 6, and a learned perceptual image patch similarity of 0.039 6±0.031 6, relative to the real AC-PET. Conclusion The proposed DRDM effectively achieves PET self-attenuation correction, outperforming existing methods. It provides a novel solution for PET attenuation correction without additional CT scans.
Last Update: 2026-04-28