Zijian Zhou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Zijian Zhou has a total of 55 co-authors. Here are the top co-authors by number of shared publications:
Peng Hu
ShanghaiTech University
3 shared publications- Thomas Theis
Department of Chemistry, North Carolina State University, Raleigh, North Carolina, 27965-8204, USA
3 shared publications Jacob R. Lindale
Department of Chemistry, Duke University, Durham, USA
2 shared publications- Peng Hu
School of Biomedical Engineering, ShanghaiTech University, Shanghai, China
2 shared publications - Warren Warren
Department of Chemistry, Duke University, Durham, NC 27708, USA
2 shared publications Wenjian Liu
ShanghaiTech University
2 shared publications
Latest Publications
- 26 September 2026
- Published by Springer Nature
- p. 615-624
Unsupervised and Segmentation-Free Phase Unwrapping for 4D Flow MRI Using Deep Image Prior
- 26 September 2026
- Published by Springer Nature
- p. 615-624
Unsupervised and Segmentation-Free Phase Unwrapping for 4D Flow MRI Using Deep Image Prior
- 26 September 2026
- Published by Springer Nature
- p. 146-155
ECHO: Estimated Composite Hybrid Observation for Zero-Shot Self-Supervised MRI Reconstruction
- 26 September 2026
- Published by Springer Nature
- p. 146-155
ECHO: Estimated Composite Hybrid Observation for Zero-Shot Self-Supervised MRI Reconstruction
- 12 June 2026
- Published by Wiley in Magnetic Resonance in Medicine
Multi‐Echo SSFP: A Method for Synthetic bSSFP MRI Contrast Without Banding Artifacts at High Field
Purpose: Balanced steady‐state free precession (bSSFP) is an important MRI sequence paradigm with a high signal‐to‐noise ratio (SNR) and inherent contrast, but suffers from banding artifacts due to inhomogeneity, especially in high‐field MRI. This study introduces a multi‐echo SSFP (MESS) sequence that synthesizes tunable bSSFP contrast without banding artifacts. Methods: In bSSFP imaging, the observed signal can be formulated as a superposition of multiple echo orders, where off‐resonance phase divergences produce destructive interference, leading to dark bands. In contrast, the MESS sequence acquires these echoes separately. Extended phase graph (EPG) modeling demonstrates that by summing the echo magnitudes, MESS can reconstruct a signal that mirrors bSSFP contrast without banding artifacts. Image contrast can generally be tuned retrospectively by varying the virtual phase‐cycle increment . Phantom and in vivo experiments on a 5T scanner were performed to validate this approach. Results: Phantom studies confirmed that MESS reconstructs the bSSFP contrast free of banding artifacts. In vivo scans further demonstrated robust artifact suppression while maintaining bSSFP‐like image contrast, and permitted flexible contrast variation by tuning . Conclusion: MESS enables robust generation of banding artifact‐free bSSFP contrast in high‐field MRI by leveraging echo magnitude and virtual phase‐cycling, indicating a strong potential for robust image contrast synthesis.
- 12 June 2026
- Published by Wiley in Magnetic Resonance in Medicine
Multi‐Echo SSFP: A Method for Synthetic bSSFP MRI Contrast Without Banding Artifacts at High Field
Purpose: Balanced steady‐state free precession (bSSFP) is an important MRI sequence paradigm with a high signal‐to‐noise ratio (SNR) and inherent contrast, but suffers from banding artifacts due to inhomogeneity, especially in high‐field MRI. This study introduces a multi‐echo SSFP (MESS) sequence that synthesizes tunable bSSFP contrast without banding artifacts. Methods: In bSSFP imaging, the observed signal can be formulated as a superposition of multiple echo orders, where off‐resonance phase divergences produce destructive interference, leading to dark bands. In contrast, the MESS sequence acquires these echoes separately. Extended phase graph (EPG) modeling demonstrates that by summing the echo magnitudes, MESS can reconstruct a signal that mirrors bSSFP contrast without banding artifacts. Image contrast can generally be tuned retrospectively by varying the virtual phase‐cycle increment . Phantom and in vivo experiments on a 5T scanner were performed to validate this approach. Results: Phantom studies confirmed that MESS reconstructs the bSSFP contrast free of banding artifacts. In vivo scans further demonstrated robust artifact suppression while maintaining bSSFP‐like image contrast, and permitted flexible contrast variation by tuning . Conclusion: MESS enables robust generation of banding artifact‐free bSSFP contrast in high‐field MRI by leveraging echo magnitude and virtual phase‐cycling, indicating a strong potential for robust image contrast synthesis.
- Open Access
- 3 June 2026
- Published by Elsevier in Journal of Cardiovascular Magnetic Resonance
- Vol. 28 (2) , 102750
Fully self-gated dual-venc five-dimensional flow magnetic resonance imaging for simultaneous imaging of the cardiovascular and portal venous systems: A feasibility study
- 3 June 2026
- Published by Elsevier in Journal of Cardiovascular Magnetic Resonance
- Vol. 28 (2) , 102750
Open AccessFully self-gated dual-venc five-dimensional flow magnetic resonance imaging for simultaneous imaging of the cardiovascular and portal venous systems: A feasibility study
- 4 September 2025
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 95 (2) , 756-769
Rapid free‐breathing myocardial T 1 mapping with free‐running interleaved multi‐slice acquisition and respiratory motion correction using self‐navigated auto calibration
Purpose: To develop a rapid 2D free‐running myocardial mapping technique that is robust to through‐plane respiratory motion. Methods: A free‐running golden angle radial sequence consisting of encoding and self‐navigated auto motion calibration (SNAC) was developed. The encoding adopted inversion recovery (IR) prepared interleaved multi‐slice acquisition with optimized inter‐slice gap to ensure a uniform excitation of the middle slice regardless of through‐plane respiratory motion. The flip angles were alternated between the odd and even IR readouts to correct flip‐angle errors. SNAC was designed to calibrate the through‐plane motion with the respiratory self‐navigation signal extracted from the free‐running sequence, and integrate the multi‐slice data into a through‐plane motion‐corrected 2D slice for mapping reconstruction. Numerical simulations were performed to optimize the key sequence parameters, followed by phantom and in‐vivo imaging to validate the accuracy and repeatability. Results: Numerical simulations yielded the dual flip angles minimizing estimation errors and the adjacent slice offset achieving uniformity in the superimposed slice profile. Phantom experiments demonstrated a strong correlation between the proposed and reference method, and no dependence of the free‐running estimation on heart rates. Adding through‐plane respiratory motion correction significantly improved the in‐vivo mapping sharpness and visual quality. Validated against the conventional breath‐hold mapping technique, the motion corrected free‐running method achieved comparable mapping quality and repeatability. Conclusion: Respiratory motion is effectively suppressed in the proposed 2D free‐running method, which achieves superior mapping with acceptable repeatability in a short scan time of 46 s.
- 4 September 2025
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 95 (2) , 756-769
Rapid free‐breathing myocardial T 1 mapping with free‐running interleaved multi‐slice acquisition and respiratory motion correction using self‐navigated auto calibration
Purpose: To develop a rapid 2D free‐running myocardial mapping technique that is robust to through‐plane respiratory motion. Methods: A free‐running golden angle radial sequence consisting of encoding and self‐navigated auto motion calibration (SNAC) was developed. The encoding adopted inversion recovery (IR) prepared interleaved multi‐slice acquisition with optimized inter‐slice gap to ensure a uniform excitation of the middle slice regardless of through‐plane respiratory motion. The flip angles were alternated between the odd and even IR readouts to correct flip‐angle errors. SNAC was designed to calibrate the through‐plane motion with the respiratory self‐navigation signal extracted from the free‐running sequence, and integrate the multi‐slice data into a through‐plane motion‐corrected 2D slice for mapping reconstruction. Numerical simulations were performed to optimize the key sequence parameters, followed by phantom and in‐vivo imaging to validate the accuracy and repeatability. Results: Numerical simulations yielded the dual flip angles minimizing estimation errors and the adjacent slice offset achieving uniformity in the superimposed slice profile. Phantom experiments demonstrated a strong correlation between the proposed and reference method, and no dependence of the free‐running estimation on heart rates. Adding through‐plane respiratory motion correction significantly improved the in‐vivo mapping sharpness and visual quality. Validated against the conventional breath‐hold mapping technique, the motion corrected free‐running method achieved comparable mapping quality and repeatability. Conclusion: Respiratory motion is effectively suppressed in the proposed 2D free‐running method, which achieves superior mapping with acceptable repeatability in a short scan time of 46 s.
- 23 August 2025
- Published by Springer Nature in Magnetic Resonance Materials in Physics, Biology and Medicine
- Vol. 39 (1) , 97-108
Epicardial and paracardial adipose tissue quantification in short-axis cardiac cine MRI using deep learning
Objective Epicardial and paracardial adipose tissues (EAT and PAT) are two types of fat depots around the heart and they have important roles in cardiac physiology. Manual quantification of EAT and PAT from cardiac MR (CMR) is time-consuming and prone to human bias. Leveraging the cardiac motion, we aimed to develop deep learning neural networks for automated segmentation and quantification of EAT and PAT in short-axis cine CMR. Materials and methods A modified U-Net equipped with modules of multi-resolution convolution, motion information extraction, feature fusion, and dual attention mechanisms, was developed. Multiple steps of ablation studies were performed to verify the efficacy of each module. The performance of different networks was also compared. Results The final network incorporating all modules achieved segmentation Dice indices of 77.72% ± 2.53% and 77.18% ± 3.54% for EAT and PAT, respectively, which were significantly higher than the baseline U-Net. It also achieved the highest performance compared to other networks. With our model, the determination coefficients of EAT and PAT volumes to the reference were 0.8550 and 0.8025, respectively. Conclusion Our proposed network can provide accurate and quick quantification of EAT and PAT on routine short-axis cine CMR, which can potentially aid cardiologists in clinical settings.
- 23 August 2025
- Published by Springer Nature in Magnetic Resonance Materials in Physics, Biology and Medicine
- Vol. 39 (1) , 97-108
Epicardial and paracardial adipose tissue quantification in short-axis cardiac cine MRI using deep learning
Objective Epicardial and paracardial adipose tissues (EAT and PAT) are two types of fat depots around the heart and they have important roles in cardiac physiology. Manual quantification of EAT and PAT from cardiac MR (CMR) is time-consuming and prone to human bias. Leveraging the cardiac motion, we aimed to develop deep learning neural networks for automated segmentation and quantification of EAT and PAT in short-axis cine CMR. Materials and methods A modified U-Net equipped with modules of multi-resolution convolution, motion information extraction, feature fusion, and dual attention mechanisms, was developed. Multiple steps of ablation studies were performed to verify the efficacy of each module. The performance of different networks was also compared. Results The final network incorporating all modules achieved segmentation Dice indices of 77.72% ± 2.53% and 77.18% ± 3.54% for EAT and PAT, respectively, which were significantly higher than the baseline U-Net. It also achieved the highest performance compared to other networks. With our model, the determination coefficients of EAT and PAT volumes to the reference were 0.8550 and 0.8025, respectively. Conclusion Our proposed network can provide accurate and quick quantification of EAT and PAT on routine short-axis cine CMR, which can potentially aid cardiologists in clinical settings.
- 13 April 2020
- Published by Royal Society of Chemistry (RSC) in Chemical Communications
- Vol. 56 (65) , 9336-9339
Rational ligand choice extends the SABRE substrate scope
Here we report on chelating ligands for Signal Amplification By Reversible Exchange (SABRE) catalysts that permit hyperpolarisation on otherwise sterically hindered substrates. We demonstrate 1H enhancements of ∼100-fold over 8.5 T thermal for 2-substituted pyridines, and smaller, yet significant enhancements for provitamin B6 and caffeine. We also show 15N-enhancements of ∼1000-fold and 19F-enhancements of 30-fold.
- 13 April 2020
- Published by Royal Society of Chemistry (RSC) in Chemical Communications
- Vol. 56 (65) , 9336-9339
Rational ligand choice extends the SABRE substrate scope
Here we report on chelating ligands for Signal Amplification By Reversible Exchange (SABRE) catalysts that permit hyperpolarisation on otherwise sterically hindered substrates. We demonstrate 1H enhancements of ∼100-fold over 8.5 T thermal for 2-substituted pyridines, and smaller, yet significant enhancements for provitamin B6 and caffeine. We also show 15N-enhancements of ∼1000-fold and 19F-enhancements of 30-fold.
- 24 July 2019
- Published by AIP Publishing in The Journal of Chemical Physics
- Vol. 151 (4) , 044201
Selective hyperpolarization of heteronuclear singlet states via pulsed microtesla SABRE
Signal Amplification By Reversible Exchange (SABRE) and its heteronuclear variant SABRE in SHield Enables Alignment Transfer to Heteronuclei create large nuclear magnetization in target ligands, exploiting level crossings in an iridium catalyst that transiently binds both the ligands and parahydrogen. This requires a specific, small magnetic field to match Zeeman splittings to scalar couplings. Here, we explore a different strategy, direct creation of heteronuclear singlet states in the target ligands, which produces enhanced signals at other field strengths, including zero field. We also show that pulsed methods (including pulsed field nulling) coherently and selectively pump such singlets, affording a significant enhancement on the resulting hyperpolarization.
- 24 July 2019
- Published by AIP Publishing in The Journal of Chemical Physics
- Vol. 151 (4) , 044201
Selective hyperpolarization of heteronuclear singlet states via pulsed microtesla SABRE
Signal Amplification By Reversible Exchange (SABRE) and its heteronuclear variant SABRE in SHield Enables Alignment Transfer to Heteronuclei create large nuclear magnetization in target ligands, exploiting level crossings in an iridium catalyst that transiently binds both the ligands and parahydrogen. This requires a specific, small magnetic field to match Zeeman splittings to scalar couplings. Here, we explore a different strategy, direct creation of heteronuclear singlet states in the target ligands, which produces enhanced signals at other field strengths, including zero field. We also show that pulsed methods (including pulsed field nulling) coherently and selectively pump such singlets, affording a significant enhancement on the resulting hyperpolarization.
- 1 May 2019
- Published by Radiological Society of North America (RSNA) in Radiology: Artificial Intelligence
- Vol. 1 (3) , e180084
Evaluating a Fully Automated Pulmonary Nodule Detection Approach and Its Impact on Radiologist Performance
A deep learning model showed improved overall sensitivity compared with manual identification of pulmonary nodules and was insensitive to radiation dose, patient age, or CT manufacturer; the model ... Purpose To compare sensitivity in the detection of lung nodules between the deep learning (DL) model and radiologists using various patient population and scanning parameters and to assess whether ...
- 1 May 2019
- Published by Radiological Society of North America (RSNA) in Radiology: Artificial Intelligence
- Vol. 1 (3) , e180084
Evaluating a Fully Automated Pulmonary Nodule Detection Approach and Its Impact on Radiologist Performance
A deep learning model showed improved overall sensitivity compared with manual identification of pulmonary nodules and was insensitive to radiation dose, patient age, or CT manufacturer; the model ... Purpose To compare sensitivity in the detection of lung nodules between the deep learning (DL) model and radiologists using various patient population and scanning parameters and to assess whether ...
- Open Access
- 2 March 2018
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 4 (3) , eaar2978
15 N 4 -1,2,4,5-tetrazines as potential molecular tags: Integrating bioorthogonal chemistry with hyperpolarization and unearthing para -N 2
A bioorthogonal reaction–based HP-MR strategy affords novel hyperpolarized products from tetrazine, including para - 15 N 2 gas.
- 2 March 2018
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 4 (3) , eaar2978
Open Access15 N 4 -1,2,4,5-tetrazines as potential molecular tags: Integrating bioorthogonal chemistry with hyperpolarization and unearthing para -N 2
A bioorthogonal reaction–based HP-MR strategy affords novel hyperpolarized products from tetrazine, including para - 15 N 2 gas.