Title:
PET IMAGE RECONSTRUCTION METHOD BASED ON SWIN-TRANSFORMER REGULARIZATION
Document Type and Number:
WIPO Patent Application WO/2024/011797
Kind Code:
A1
Abstract:
Disclosed in the present invention is a PET image reconstruction method based on Swin-transformer regularization. A reconstruction model used in the method is composed of several iteration modules, each of which is composed of an EM iteration layer, a Swin-transformer-based regularization layer and a pixel-to-pixel image fusion layer, wherein the regularization layer is used for learning prior information which represents images, and the regularization layer comprises a convolutional layer for extracting shallow features of the images, a Swin-transformer layer for extracting deep features of the images, and a last convolutional layer which fuses the deep and shallow features by using residual connection; and the image fusion layer is used for fusing an EM iteration result and a regularization result. By means of the present invention, reconstruction can be performed on the basis of sinogram projection data, such that a high-quality PET image is obtained, structural information is reserved, and the noise level of the PET image is reduced to a great extent.
Inventors:
LIU HUAFENG (CN)
HU RUI (CN)
HU RUI (CN)
Application Number:
PCT/CN2022/130794
Publication Date:
January 18, 2024
Filing Date:
November 09, 2022
Export Citation:
Assignee:
UNIV ZHEJIANG (CN)
International Classes:
G06T11/00; G06N3/04; G06T5/50
Foreign References:
| CN115187689A | 2022-10-14 | |||
| CN114387236A | 2022-04-22 | |||
| CN113706388A | 2021-11-26 | |||
| US20170039706A1 | 2017-02-09 |
Other References:
vol. 18, 16 September 2022, SPRINGER INTERNATIONAL PUBLISHING, article HU RUI; LIU HUAFENG: "TransEM: Residual Swin-Transformer Based Regularized PET Image Reconstruction", pages: 184 - 193, XP047633535, DOI: 10.1007/978-3-031-16440-8_18
MEHRANIAN ABOLFAZL; READER ANDREW J.: "Model-Based Deep Learning PET Image Reconstruction Using Forward-Backward Splitting Expectation Maximisation", 2019 IEEE NUCLEAR SCIENCE SYMPOSIUM AND MEDICAL IMAGING CONFERENCE (NSS/MIC), IEEE, 26 October 2019 (2019-10-26), pages 1 - 4, XP033748113, DOI: 10.1109/NSS/MIC42101.2019.9059998
HUANG JIAHAO; FANG YINGYING; WU YINZHE; WU HUANJUN; GAO ZHIFAN; LI YANG; SER JAVIER DEL; XIA JUN; YANG GUANG: "Swin transformer for fast MRI", ARXIV:2201.03230V2, vol. 493, 12 April 2022 (2022-04-12), pages 281 - 304, XP087053571, DOI: 10.1016/j.neucom.2022.04.051
MEHRANIAN ABOLFAZL; READER ANDREW J.: "Model-Based Deep Learning PET Image Reconstruction Using Forward-Backward Splitting Expectation Maximisation", 2019 IEEE NUCLEAR SCIENCE SYMPOSIUM AND MEDICAL IMAGING CONFERENCE (NSS/MIC), IEEE, 26 October 2019 (2019-10-26), pages 1 - 4, XP033748113, DOI: 10.1109/NSS/MIC42101.2019.9059998
HUANG JIAHAO; FANG YINGYING; WU YINZHE; WU HUANJUN; GAO ZHIFAN; LI YANG; SER JAVIER DEL; XIA JUN; YANG GUANG: "Swin transformer for fast MRI", ARXIV:2201.03230V2, vol. 493, 12 April 2022 (2022-04-12), pages 281 - 304, XP087053571, DOI: 10.1016/j.neucom.2022.04.051
Attorney, Agent or Firm:
HANGZHOU TIANQIN INTELLECTUAL PROPERTY AGENCY CO., LTD. (CN)
Download PDF: