Peng Hu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Peng Hu has a total of 36 co-authors. Here are the top co-authors by number of shared publications:
- Haikun Qi
ShanghaiTech University
3 shared publications Zijian Zhou
ShanghaiTech University 393 Huaxia Middle Road 201210 Shanghai China
2 shared publications- Zijian Zhou
ShanghaiTech University
2 shared publications - Bingbing Cheng
TRUTH Lab, School of Biomedical Engineering, ShanghaiTech University, Shanghai, China
1 shared publication Dongmei Wu
Shanghai Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai, China
1 shared publication- Fei Shan
Department of Radiology, Shanghai Public Health Clinical Center, Fudan University, Shanghai, China
1 shared publication
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
- Open Access
- 11 September 2026
- Published by Springer Nature in Communications Engineering
Image-guided cardiac focused ultrasound neuromodulation regulates heart rate
- 11 September 2026
- Published by Springer Nature in Communications Engineering
Open AccessImage-guided cardiac focused ultrasound neuromodulation regulates heart rate
- 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
- Open Access
Branch Learning in MRI: More Data, More Models, More Training
We investigated two complementary strategies for multicontrast cardiac MR reconstruction: physics-consistent data-space augmentation (DualSpaceCMR) and parameter-efficient capacity scaling via VQPrompt and Moero. DualSpaceCMR couples image-level transforms with kspace noise and motion simulations while preserving forwardmodel consistency. VQPrompt adds a lightweight bottleneck prompt; Moero embeds a sparse mixture of experts within a deep unrolled network with histogram-based routing. In the multivendor, multisite CMRxRecon25 benchmark, we evaluate fewshot and out-of-distribution generalization. On small datasets, k-space motion-plus-noise improves reconstruction; on the large benchmark it degrades performance, revealing sensitivity to augmentation ratio and schedule. VQPrompt produces modest and consistent gains with negligible memory overhead. Moero continues to improve after early plateaus and maintains baseline-like fewshot and out-of-distribution behavior despite mild overfitting, but sparse routing lowers PyTorch throughput and makes wall clock time the main bottleneck. These results motivate scale-aware augmentation and suggest prompt-based capacity scaling as a practical path, while efficiency improvements are crucial for sparse expert models.
Open AccessBranch Learning in MRI: More Data, More Models, More Training
We investigated two complementary strategies for multicontrast cardiac MR reconstruction: physics-consistent data-space augmentation (DualSpaceCMR) and parameter-efficient capacity scaling via VQPrompt and Moero. DualSpaceCMR couples image-level transforms with kspace noise and motion simulations while preserving forwardmodel consistency. VQPrompt adds a lightweight bottleneck prompt; Moero embeds a sparse mixture of experts within a deep unrolled network with histogram-based routing. In the multivendor, multisite CMRxRecon25 benchmark, we evaluate fewshot and out-of-distribution generalization. On small datasets, k-space motion-plus-noise improves reconstruction; on the large benchmark it degrades performance, revealing sensitivity to augmentation ratio and schedule. VQPrompt produces modest and consistent gains with negligible memory overhead. Moero continues to improve after early plateaus and maintains baseline-like fewshot and out-of-distribution behavior despite mild overfitting, but sparse routing lowers PyTorch throughput and makes wall clock time the main bottleneck. These results motivate scale-aware augmentation and suggest prompt-based capacity scaling as a practical path, while efficiency improvements are crucial for sparse expert models.
- Open Access
Sampling Density Compensation using Fast Fourier Deconvolution
Density Compensation Function (DCF) is widely used in non-Cartesian MRI reconstruction, either for direct Non-Uniform Fast Fourier Transform (NUFFT) reconstruction or for iterative undersampled reconstruction. Current state-of-the-art methods involve time-consuming tens of iterations, which is one of the main hurdles for widespread application of the highly efficient non-Cartesian MRI. In this paper, we propose an efficient, non-iterative method to calculate DCF for arbitrary non-Cartesian $k$-space trajectories using Fast Fourier Deconvolution. Simulation experiments demonstrate that the proposed method is able to yield DCF for 3D non-Cartesian reconstruction in around 20 seconds, achieving orders of magnitude speed improvement compared to the state-of-the-art method while achieving similar reconstruction quality.
Open AccessSampling Density Compensation using Fast Fourier Deconvolution
Density Compensation Function (DCF) is widely used in non-Cartesian MRI reconstruction, either for direct Non-Uniform Fast Fourier Transform (NUFFT) reconstruction or for iterative undersampled reconstruction. Current state-of-the-art methods involve time-consuming tens of iterations, which is one of the main hurdles for widespread application of the highly efficient non-Cartesian MRI. In this paper, we propose an efficient, non-iterative method to calculate DCF for arbitrary non-Cartesian $k$-space trajectories using Fast Fourier Deconvolution. Simulation experiments demonstrate that the proposed method is able to yield DCF for 3D non-Cartesian reconstruction in around 20 seconds, achieving orders of magnitude speed improvement compared to the state-of-the-art method while achieving similar reconstruction quality.
- 25 April 2024
- Published by Springer Nature in La radiologia medica
- Vol. 129 (6) , 834-844
To characterize small renal cell carcinoma using diffusion relaxation correlation spectroscopic imaging and apparent diffusion coefficient based histogram analysis: a preliminary study
Purpose To study the capability of diffusion-relaxation correlation spectroscopic imaging (DR-CSI) on subtype classification and grade differentiation for small renal cell carcinoma (RCC). Histogram analysis for apparent diffusion coefficient (ADC) was studied for comparison. Materials and methods A total of 61 patients with small RCC (< 4 cm) were included in the retrospective study. MRI data were reviewed, including a multi-b (0–1500 s/mm2) multi-TE (51–200 ms) diffusion weighted imaging (DWI) sequence. Region of interest (ROI) was delineated manually on DWI to include solid tumor. For each patient, a D-T2 spectrum was fitted and segmented into 5 compartments, and the volume fractions VA, VB, VC, VD, VE were obtained. ADC mapping was calculated, and histogram parameters ADC 90th, 10th, median, standard deviation, skewness and kurtosis were obtained. All MRI metrices were compared between clear cell RCC (ccRCC) and non-ccRCC group, and between high-grade and low-grade group. Receiver operator curve analysis was used to assess the corresponding diagnostic performance. Results Significantly higher ADC 90th, ADC 10th and ADC median, and significantly lower DR-CSI VB was found for ccRCC compared to non-ccRCC. Significantly lower ADC 90th, ADC median and significantly higher VB was found for high-grade RCC compared to low-grade. For identifying ccRCC from non-ccRCC, VB showed the highest area under curve (AUC, 0.861) and specificity (0.882). For differentiating high- from low-grade, ADC 90th showed the highest AUC (0.726) and specificity (0.786), while VB also displayed a moderate AUC (0.715). Conclusion DR-CSI may offer improved accuracy in subtype identification for small RCC, while do not show better performance for small RCC grading compared to ADC histogram.
- 25 April 2024
- Published by Springer Nature in La radiologia medica
- Vol. 129 (6) , 834-844
To characterize small renal cell carcinoma using diffusion relaxation correlation spectroscopic imaging and apparent diffusion coefficient based histogram analysis: a preliminary study
Purpose To study the capability of diffusion-relaxation correlation spectroscopic imaging (DR-CSI) on subtype classification and grade differentiation for small renal cell carcinoma (RCC). Histogram analysis for apparent diffusion coefficient (ADC) was studied for comparison. Materials and methods A total of 61 patients with small RCC (< 4 cm) were included in the retrospective study. MRI data were reviewed, including a multi-b (0–1500 s/mm2) multi-TE (51–200 ms) diffusion weighted imaging (DWI) sequence. Region of interest (ROI) was delineated manually on DWI to include solid tumor. For each patient, a D-T2 spectrum was fitted and segmented into 5 compartments, and the volume fractions VA, VB, VC, VD, VE were obtained. ADC mapping was calculated, and histogram parameters ADC 90th, 10th, median, standard deviation, skewness and kurtosis were obtained. All MRI metrices were compared between clear cell RCC (ccRCC) and non-ccRCC group, and between high-grade and low-grade group. Receiver operator curve analysis was used to assess the corresponding diagnostic performance. Results Significantly higher ADC 90th, ADC 10th and ADC median, and significantly lower DR-CSI VB was found for ccRCC compared to non-ccRCC. Significantly lower ADC 90th, ADC median and significantly higher VB was found for high-grade RCC compared to low-grade. For identifying ccRCC from non-ccRCC, VB showed the highest area under curve (AUC, 0.861) and specificity (0.882). For differentiating high- from low-grade, ADC 90th showed the highest AUC (0.726) and specificity (0.786), while VB also displayed a moderate AUC (0.715). Conclusion DR-CSI may offer improved accuracy in subtype identification for small RCC, while do not show better performance for small RCC grading compared to ADC histogram.
- 22 February 2024
- Published by Springer Nature in Magnetic Resonance Materials in Physics, Biology and Medicine
- Vol. 37 (3) , 397-409
Stop moving: MR motion correction as an opportunity for artificial intelligence
Subject motion is a long-standing problem of magnetic resonance imaging (MRI), which can seriously deteriorate the image quality. Various prospective and retrospective methods have been proposed for MRI motion correction, among which deep learning approaches have achieved state-of-the-art motion correction performance. This survey paper aims to provide a comprehensive review of deep learning-based MRI motion correction methods. Neural networks used for motion artifacts reduction and motion estimation in the image domain or frequency domain are detailed. Furthermore, besides motion-corrected MRI reconstruction, how estimated motion is applied in other downstream tasks is briefly introduced, aiming to strengthen the interaction between different research areas. Finally, we identify current limitations and point out future directions of deep learning-based MRI motion correction.
- 22 February 2024
- Published by Springer Nature in Magnetic Resonance Materials in Physics, Biology and Medicine
- Vol. 37 (3) , 397-409
Stop moving: MR motion correction as an opportunity for artificial intelligence
Subject motion is a long-standing problem of magnetic resonance imaging (MRI), which can seriously deteriorate the image quality. Various prospective and retrospective methods have been proposed for MRI motion correction, among which deep learning approaches have achieved state-of-the-art motion correction performance. This survey paper aims to provide a comprehensive review of deep learning-based MRI motion correction methods. Neural networks used for motion artifacts reduction and motion estimation in the image domain or frequency domain are detailed. Furthermore, besides motion-corrected MRI reconstruction, how estimated motion is applied in other downstream tasks is briefly introduced, aiming to strengthen the interaction between different research areas. Finally, we identify current limitations and point out future directions of deep learning-based MRI motion correction.
- Open Access
- 9 November 2023
- Published by Elsevier in Journal of Cardiovascular Magnetic Resonance
- Vol. 25 (1) , 63-17
Free-breathing simultaneous native myocardial T1, T2 and T1ρ mapping with Cartesian acquisition and dictionary matching
- 9 November 2023
- Published by Elsevier in Journal of Cardiovascular Magnetic Resonance
- Vol. 25 (1) , 63-17
Open AccessFree-breathing simultaneous native myocardial T1, T2 and T1ρ mapping with Cartesian acquisition and dictionary matching