YU SHI

Dr. YU SHI

Electrical Engineering and Automation, Changshu Institute of Technology, Changshu, Jiangsu, China
+1
orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
7 October 2022 in Physics in Medicine & Biology
Verified
Journal Article
Closed access
YU SHIYU SHI
YU SHI
YWYirong Wang
Yirong Wang
JZJianwei Zhou
Jianwei Zhou
YZYuzhen Zhu
Yuzhen Zhu
XLXudong Lyu
Xudong Lyu
SZSHOUPING ZHU
SHOUPING ZHU
Objective . To develop a simultaneous positron emission tomography-Optical (OPET) breast imaging dual-head PET subsystem, called DH-Mammo PET, for accurate, early diagnosis and efficacy assessment of breast cancer with high resolution and sensitivity. Approach . We developed a breast-dedicated PET based on LYSO crystal, silicon photomultiplier array and multi-voltage threshold sampling technique. It consists of two detector heads, each with a detection area of 216 mm × 145.5 mm. The distance between the detector heads is fixed at 120 mm. In order to extract coincidences and correct data, GPU-based software coincidence processing, random, scatter, normalization, gap-filling and attenuation corrections were applied in turn. The images were reconstructed using maximum likelihood expectation maximization with depth of interaction (DOI) modeling. The performance of DH-Mammo PET was evaluated referring to NEMA NU 4–2008, NU 2–2007 and Chinese industry recommended standard YY/T 1835–2022. Besides, several clinical patient images of DH-Mammo PET were compared with those of a whole-body PET/CT. Main results . The energy resolution was 14.5%, and time resolution was < 1.31 ns. Indicated by the 22 Na point source imaging, its spatial resolution was 2.60 mm (5.40 mm), 1.00 mm (1.04 mm), and 0.96 mm (0.93 mm) in the X , Y and Z directions, respectively, using the system response matrix with (without) DOI modeling. Indicated by the Derenzo phantom imaging, the spatial resolution was ∼3.0 mm, <1.2 mm, and <1.2 mm in the X , Y and Z directions. The system sensitivity was 6.87%, 4.89% and 3.37% with an energy window of 100–800, 250–750 and 350–650 keV, respectively. The scatter fraction was 26.43%, and the peak NECR was 162.6 kcps at 24.1 MBq for the modified rat-like phantom. As for the recovery coefficients, they ranged from 0.15 to 1.04 for rods between 1 and 5 mm obtained with a NEMA image quality phantom. The spill-over ratio for the air-filled and water-filled chamber was 0.05 and 0.11, respectively. DH-Mammo PET can provide more image details in clinical experiments and fulfil a fast scan with 60–120 s acquisition time. Significance . Good spatial resolution and high sensitivity of DH-Mammo PET would enable fast and accurate PET imaging of the breast. Besides, combining the DH-Mammo PET with the diffuse optical tomography would make full use of tumor metabolic imaging and tissue endogenous optical imaging, which would improve the accuracy of early clinical diagnosis of small lesions of breast cancers.

1 April 2026 in IEEE Transactions on Radiation and Plasma Medical Sciences
Verified
Journal Article
Closed access
CFChangsheng Fang
Changsheng Fang
YLYongtong Liu
Yongtong Liu
BMBahareh Morovati
Bahareh Morovati
SHShuo Han
Shuo Han
YU SHIYU SHI
YU SHI
HYHengyong Yu
Hengyong Yu

24 August 2019
Verified
Conference Paper
Open access
YU SHIYU SHI
YU SHI
FMFanzhen Meng
Fanzhen Meng
YMYe Mao
Ye Mao
CLChenfeng Li
Chenfeng Li
SZShouping Zhu
Shouping Zhu
Back projection filtering (BPF) and filtered back projection (FBP) are classic CT reconstruction analytic methods. However, when dealing with the reconstruction of incomplete datasets, analytic methods may suffer from artifacts due to the global nature of filtering. The iterative algorithm is another CT reconstruction method. Although the iterative algorithm can get excellent results, it requires high computation cost. In this work, inspired by the idea of BPF, we propose a CT image reconstruction strategy based on a new convolutional neural network that can capture features with different levels. The proposed method can learn the mapping between the direct back projection and the ideal CT images and then realize accurate image reconstruction. It is suitable for not only complete projection data but also transverse truncated projection data with different regions of interest (ROIs) sizes. The experiment results show that the proposed method outperforms the conventional FBP and other state-of-the-art deep learning methods in the truncated reconstruction. At the same time, the proposed method requires a lower computation cost which is helpful for fast reconstruction in the clinical research.

1 September 2026 in IEEE Transactions on Radiation and Plasma Medical Sciences
Verified
Journal Article
Closed access
SHShuo Han
Shuo Han
BMBahareh Morovati
Bahareh Morovati
YLYongtong Liu
Yongtong Liu
CFChangsheng Fang
Changsheng Fang
SFShuyi Fan
Shuyi Fan
HYHengyong Yu
Hengyong Yu
Cardiovascular diseases (CVDs) remain the leading cause of morbidity and mortality worldwide. Computed tomography (CT) plays an important role in the diagnosis and management of CVDs. However, the temporal resolution of current cardiac CT imaging techniques is still insufficient to freeze the beating heart. One solution to improve temporal resolution is fast data acquisition for a limited-angle range. Unfortunately, this approach introduces significant limited-angle artifacts that compromise image quality. Diffusion models have shown promising results to address this issue by effectively reconstructing missing data, but the process is typically time-consuming. In this paper, we propose a novel method for 3D limited-angle reconstruction using a diffusion model, optimized to significantly reduce computational time. Specifically, our approach integrates the regularized derivative least square (RDLS) algorithm with denoising diffusion implicit models. The numerical simulations and clinical data experiments demonstrate that the proposed model outperforms the existing methods both quantitatively and qualitatively. Compared with DDS, our method delivers a 1.8× speedup on the simulated dataset and a 3.4× speedup on the real clinical dataset.

1 December 2020 in Journal of Instrumentation
Verified
Journal Article
Closed access
PDF file
FMF. Meng
F. Meng
JCJ. Cheng
J. Cheng
CLC. Li
C. Li
YU SHIYU SHI
YU SHI
GZG. Zhang
G. Zhang
SZSHOUPING ZHU
SHOUPING ZHU
Compact quad-head PET systems have been developed by many researchers. This compact configuration leads to depth of interaction (DOI) blurring, which greatly reduces the quality of the reconstructed image. How to accurately model the photon propagation is the key to relieving DOI blurring. In this study, photon propagation is modelled by using the multi-ray method. As we all know, the precision of the multi-ray model is proportional to its computational complexity. Focusing on this problem, we have developed the symmetries of the quad-head PET system according to the different structural features of the opposite and adjacent detector heads. Given all the symmetrical properties, the construction of the photon propagation model becomes a trade-off between the calculation precision and computational complexity of the multi-ray method. In addition, when photon propagation is stored in the system response matrix (SRM), the storage cost is also reduced largely. In the proposed quad-head PET system, the SRM is compressed about 559 times, and the matrix scale is approximately 1.7 GB. The calculation of the SRM can be finished in 60 s when the graphics processing unit (GPU) acceleration technology is adopted. In order to verify the imaging performances, several simulated experiments are carried out. The metrics FWHM, location accuracy, Contrast Recovery (CR), and Coefficient of Variation (COV) are adopted to quantify the imaging results. We find that the mean location difference can reach 1.29% based on a simulated multi-spheres experiment. Through this experiment, the FWHMs can reach to about 1.0 mm in the center or close to the edge of the field of view (FOV). In the image quality experiment, the CR can up to 0.95 in the hot region, and the COV reaches 5.82% in the background region.

1 May 2025 in Journal of Information and Intelligence
Verified
Journal Article
Open access
YLYanyun Liu
Yanyun Liu
XHXiumeng Hua
Xiumeng Hua
XLXiangyu Liu
Xiangyu Liu
CWCongrui Wang
Congrui Wang
XCXiao Chen
Xiao Chen
SZSHOUPING ZHU
SHOUPING ZHU