Y. Zhang
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
18 records found
1
Deformable medical image registration is traditionally formulated as an optimization problem. While classical methods solve this problem iteratively, recent learning-based approaches use weight-tied neural networks to mimic this process by unrolling the prediction of deformation fields in a fixed number of steps. However, classical methods typically converge after sufficient iterations, but learning-based unrolling methods lack a theoretical convergence guarantee and show instability empirically. In addition, unrolling methods have a practical bottleneck at training time: GPU memory usage grows linearly with the unrolling steps due to backpropagation through time (BPTT). To address both theoretical and practical challenges, we propose DEQReg, a novel registration framework based on Deep Equilibrium Models (DEQ), which formulates registration as an equilibrium-seeking problem, establishing a natural connection between classical optimization and learning-based unrolling methods. DEQReg maintains constant memory usage, enabling theoretically unlimited iteration steps. Through extensive evaluation on the public brain MRI and lung CT datasets, we show that DEQReg can achieve competitive registration performance, while substantially reducing memory consumption compared to state-of-the-art unrolling methods. We also reveal an intriguing phenomenon: the performance of existing unrolling methods first increases slightly then degrades irreversibly when the inference steps go beyond the training configuration. In contrast, DEQReg achieves stable convergence with its inbuilt equilibrium-seeking mechanism, bridging the gap between classical optimization-based and modern learning-based registration methods.
...
Deformable medical image registration is traditionally formulated as an optimization problem. While classical methods solve this problem iteratively, recent learning-based approaches use weight-tied neural networks to mimic this process by unrolling the prediction of deformation fields in a fixed number of steps. However, classical methods typically converge after sufficient iterations, but learning-based unrolling methods lack a theoretical convergence guarantee and show instability empirically. In addition, unrolling methods have a practical bottleneck at training time: GPU memory usage grows linearly with the unrolling steps due to backpropagation through time (BPTT). To address both theoretical and practical challenges, we propose DEQReg, a novel registration framework based on Deep Equilibrium Models (DEQ), which formulates registration as an equilibrium-seeking problem, establishing a natural connection between classical optimization and learning-based unrolling methods. DEQReg maintains constant memory usage, enabling theoretically unlimited iteration steps. Through extensive evaluation on the public brain MRI and lung CT datasets, we show that DEQReg can achieve competitive registration performance, while substantially reducing memory consumption compared to state-of-the-art unrolling methods. We also reveal an intriguing phenomenon: the performance of existing unrolling methods first increases slightly then degrades irreversibly when the inference steps go beyond the training configuration. In contrast, DEQReg achieves stable convergence with its inbuilt equilibrium-seeking mechanism, bridging the gap between classical optimization-based and modern learning-based registration methods.
Comparison of Amplitude- and Amplitude-Frequency-Modulated Pulses for Outer Volume Suppression in Cardiac MRI
Application to Myocardial T 2 Mapping
Purpose
To evaluate single- and dual-band pulse modules for suppressing extra-cardiac signal in cardiac MRI, and test their applicability for time-efficient imaging with reduced FOV in myocardial 𝑇2 mapping at 3 T.
Methods
Outer Volume Suppression (OVS) was implemented using two regional saturation slabs, along the phase-encoding direction. Different Amplitude-Modulated (AM) and Amplitude-Frequency-Modulated (AM-FM) pulses were explored. AM sinc and AM-FM Hyperbolic Secant (HS) pulses, as single- (S:1b and HS:1b) and dual-band (S:2b and HS:2b) versions, were chosen for full evaluation. Pulses were individually optimized for minimal residual signal and optimal suppression-band profiles. Pulse performance was assessed through simulations and phantom experiments, and evaluated in seven healthy subjects using residual signal and quantitative myocardial 𝑇2 mapping with rFOV.
Results
In simulations, optimized HS pulses demonstrated superior suppression-band profile homogeneity and 𝐵0 robustness, while both pulse classes showed comparable 𝐵+1 sensitivity. Phantom data revealed better signal suppression with HS pulses (S:1b: 19.22% ± 3.23%, S:2b: 16.85% ± 6.80%, HS:1b: 4.66% ± 1.04%, HS:2b: 2.85% ± 1.01%). The residual difference across slabs was slightly reduced using simultaneous dual-band compared with sequential single-band approaches (S:1b: 3.81% vs. S:2b: 0.92%; HS:1b: 4.56% vs. HS:2b: 0.66%). In vivo, despite significantly lower residual with HS pulses (p = 0.0015) and visually observed artifacts with sinc pulses, no significant difference in 𝑇2 was observed across all configurations (p = 0.087) in myocardial 𝑇2 mapping.
Conclusion
AM-FM pulses achieved robust suppression of extra-cardiac signals, with dual-band implementations minimizing suppression-band differences. This can enable artifact-free rFOV imaging, for improved image quality in myocardial 𝑇2
mapping. ...
To evaluate single- and dual-band pulse modules for suppressing extra-cardiac signal in cardiac MRI, and test their applicability for time-efficient imaging with reduced FOV in myocardial 𝑇2 mapping at 3 T.
Methods
Outer Volume Suppression (OVS) was implemented using two regional saturation slabs, along the phase-encoding direction. Different Amplitude-Modulated (AM) and Amplitude-Frequency-Modulated (AM-FM) pulses were explored. AM sinc and AM-FM Hyperbolic Secant (HS) pulses, as single- (S:1b and HS:1b) and dual-band (S:2b and HS:2b) versions, were chosen for full evaluation. Pulses were individually optimized for minimal residual signal and optimal suppression-band profiles. Pulse performance was assessed through simulations and phantom experiments, and evaluated in seven healthy subjects using residual signal and quantitative myocardial 𝑇2 mapping with rFOV.
Results
In simulations, optimized HS pulses demonstrated superior suppression-band profile homogeneity and 𝐵0 robustness, while both pulse classes showed comparable 𝐵+1 sensitivity. Phantom data revealed better signal suppression with HS pulses (S:1b: 19.22% ± 3.23%, S:2b: 16.85% ± 6.80%, HS:1b: 4.66% ± 1.04%, HS:2b: 2.85% ± 1.01%). The residual difference across slabs was slightly reduced using simultaneous dual-band compared with sequential single-band approaches (S:1b: 3.81% vs. S:2b: 0.92%; HS:1b: 4.56% vs. HS:2b: 0.66%). In vivo, despite significantly lower residual with HS pulses (p = 0.0015) and visually observed artifacts with sinc pulses, no significant difference in 𝑇2 was observed across all configurations (p = 0.087) in myocardial 𝑇2 mapping.
Conclusion
AM-FM pulses achieved robust suppression of extra-cardiac signals, with dual-band implementations minimizing suppression-band differences. This can enable artifact-free rFOV imaging, for improved image quality in myocardial 𝑇2
mapping. ...
Purpose
To evaluate single- and dual-band pulse modules for suppressing extra-cardiac signal in cardiac MRI, and test their applicability for time-efficient imaging with reduced FOV in myocardial 𝑇2 mapping at 3 T.
Methods
Outer Volume Suppression (OVS) was implemented using two regional saturation slabs, along the phase-encoding direction. Different Amplitude-Modulated (AM) and Amplitude-Frequency-Modulated (AM-FM) pulses were explored. AM sinc and AM-FM Hyperbolic Secant (HS) pulses, as single- (S:1b and HS:1b) and dual-band (S:2b and HS:2b) versions, were chosen for full evaluation. Pulses were individually optimized for minimal residual signal and optimal suppression-band profiles. Pulse performance was assessed through simulations and phantom experiments, and evaluated in seven healthy subjects using residual signal and quantitative myocardial 𝑇2 mapping with rFOV.
Results
In simulations, optimized HS pulses demonstrated superior suppression-band profile homogeneity and 𝐵0 robustness, while both pulse classes showed comparable 𝐵+1 sensitivity. Phantom data revealed better signal suppression with HS pulses (S:1b: 19.22% ± 3.23%, S:2b: 16.85% ± 6.80%, HS:1b: 4.66% ± 1.04%, HS:2b: 2.85% ± 1.01%). The residual difference across slabs was slightly reduced using simultaneous dual-band compared with sequential single-band approaches (S:1b: 3.81% vs. S:2b: 0.92%; HS:1b: 4.56% vs. HS:2b: 0.66%). In vivo, despite significantly lower residual with HS pulses (p = 0.0015) and visually observed artifacts with sinc pulses, no significant difference in 𝑇2 was observed across all configurations (p = 0.087) in myocardial 𝑇2 mapping.
Conclusion
AM-FM pulses achieved robust suppression of extra-cardiac signals, with dual-band implementations minimizing suppression-band differences. This can enable artifact-free rFOV imaging, for improved image quality in myocardial 𝑇2
mapping.
To evaluate single- and dual-band pulse modules for suppressing extra-cardiac signal in cardiac MRI, and test their applicability for time-efficient imaging with reduced FOV in myocardial 𝑇2 mapping at 3 T.
Methods
Outer Volume Suppression (OVS) was implemented using two regional saturation slabs, along the phase-encoding direction. Different Amplitude-Modulated (AM) and Amplitude-Frequency-Modulated (AM-FM) pulses were explored. AM sinc and AM-FM Hyperbolic Secant (HS) pulses, as single- (S:1b and HS:1b) and dual-band (S:2b and HS:2b) versions, were chosen for full evaluation. Pulses were individually optimized for minimal residual signal and optimal suppression-band profiles. Pulse performance was assessed through simulations and phantom experiments, and evaluated in seven healthy subjects using residual signal and quantitative myocardial 𝑇2 mapping with rFOV.
Results
In simulations, optimized HS pulses demonstrated superior suppression-band profile homogeneity and 𝐵0 robustness, while both pulse classes showed comparable 𝐵+1 sensitivity. Phantom data revealed better signal suppression with HS pulses (S:1b: 19.22% ± 3.23%, S:2b: 16.85% ± 6.80%, HS:1b: 4.66% ± 1.04%, HS:2b: 2.85% ± 1.01%). The residual difference across slabs was slightly reduced using simultaneous dual-band compared with sequential single-band approaches (S:1b: 3.81% vs. S:2b: 0.92%; HS:1b: 4.56% vs. HS:2b: 0.66%). In vivo, despite significantly lower residual with HS pulses (p = 0.0015) and visually observed artifacts with sinc pulses, no significant difference in 𝑇2 was observed across all configurations (p = 0.087) in myocardial 𝑇2 mapping.
Conclusion
AM-FM pulses achieved robust suppression of extra-cardiac signals, with dual-band implementations minimizing suppression-band differences. This can enable artifact-free rFOV imaging, for improved image quality in myocardial 𝑇2
mapping.
Journal article
(2026)
-
Y. Zhao, Y. Zhang, Joao Tourais, S.D. Weingärtner, Avan Suinesiaputra, Alistair Young, Yuchi Han, Orlando Simonetti, Q. Tao
Accurate segmentation of cardiac MRI is essential for assessment of cardiac function through biomarkers such as the left and right ventricular ejection fraction (LVEF, RVEF). Although AI methods have achieved high average segmentation accuracy, the precision of biomarkers for individual patients–quantified by estimation variance, remains critical for reliable diagnosis. Calibrated biomarkers, whose uncertainty accurately reflects the true variability, are highly desirable. However, existing evaluations predominantly focus on population-level segmentation accuracy, leaving biomarker-level uncertainty and calibration largely underexplored. Intrinsic anatomical ambiguity and annotation variability are major sources of biomarker variability and cannot be fully eliminated, even when training on a single annotation set. To address this, we propose a probabilistic segmentation framework that explicitly models aleatoric uncertainty with the goal of improving calibration in the biomarker space. The framework disentangles two key sources of uncertainty: 1) detection uncertainty, arising from ambiguous inclusion of basal or apical slices in 2D cardiac MRI, modeled via objectness probabilities; and 2) contour uncertainty, reflecting variability in ventricular boundary delineation, modeled through mean–variance regression of elliptic Fourier descriptors, a compact representation of closed contours. By propagating these uncertainties to derived biomarkers, the proposed method produces more informative and better-calibrated confidence estimates for ejection fraction. Compared to conventional pixel-wise approaches, our framework improves biomarker reliability, particularly in realistic settings dominated by annotation ambiguity and limited domain shift.
...
Accurate segmentation of cardiac MRI is essential for assessment of cardiac function through biomarkers such as the left and right ventricular ejection fraction (LVEF, RVEF). Although AI methods have achieved high average segmentation accuracy, the precision of biomarkers for individual patients–quantified by estimation variance, remains critical for reliable diagnosis. Calibrated biomarkers, whose uncertainty accurately reflects the true variability, are highly desirable. However, existing evaluations predominantly focus on population-level segmentation accuracy, leaving biomarker-level uncertainty and calibration largely underexplored. Intrinsic anatomical ambiguity and annotation variability are major sources of biomarker variability and cannot be fully eliminated, even when training on a single annotation set. To address this, we propose a probabilistic segmentation framework that explicitly models aleatoric uncertainty with the goal of improving calibration in the biomarker space. The framework disentangles two key sources of uncertainty: 1) detection uncertainty, arising from ambiguous inclusion of basal or apical slices in 2D cardiac MRI, modeled via objectness probabilities; and 2) contour uncertainty, reflecting variability in ventricular boundary delineation, modeled through mean–variance regression of elliptic Fourier descriptors, a compact representation of closed contours. By propagating these uncertainties to derived biomarkers, the proposed method produces more informative and better-calibrated confidence estimates for ejection fraction. Compared to conventional pixel-wise approaches, our framework improves biomarker reliability, particularly in realistic settings dominated by annotation ambiguity and limited domain shift.
Conference paper
(2026)
-
Yidong Zhao, Peter Kellman, Hui Xue, Tongyun Yang, Yi Zhang, Yuchi Han, Orlando Simonetti, Qian Tao
Pretrained segmentation models for cardiac magnetic resonance imaging (MRI) struggle to generalize across different imaging sequences due to significant variations in image contrast. These variations arise from changes in imaging protocols, yet the same fundamental spin properties, including proton density, T1, and T2 values, govern all acquired images. With this core principle, we introduce Reverse Imaging, a novel physics-driven method for cardiac MRI data augmentation and domain adaptation to fundamentally solve the generalization problem. Our method reversely infers the underlying spin properties from observed cardiac MRI images, by solving ill-posed nonlinear inverse problems regularized by the prior distribution of spin properties. We acquire this “spin prior” by learning a generative diffusion model from the multiparametric SAturation-recovery single-SHot acquisition sequence (mSASHA) dataset, which offers joint cardiac T1 and T2 maps. Our method enables approximate but meaningful spin-property estimates from MR images, which provide an interpretable “latent variable” that lead to highly flexible image synthesis of arbitrary novel sequences. We show that Reverse Imaging enables highly accurate segmentation across vastly different image contrasts and imaging protocols, realizing wide-spectrum generalization of cardiac MRI segmentation.
...
Pretrained segmentation models for cardiac magnetic resonance imaging (MRI) struggle to generalize across different imaging sequences due to significant variations in image contrast. These variations arise from changes in imaging protocols, yet the same fundamental spin properties, including proton density, T1, and T2 values, govern all acquired images. With this core principle, we introduce Reverse Imaging, a novel physics-driven method for cardiac MRI data augmentation and domain adaptation to fundamentally solve the generalization problem. Our method reversely infers the underlying spin properties from observed cardiac MRI images, by solving ill-posed nonlinear inverse problems regularized by the prior distribution of spin properties. We acquire this “spin prior” by learning a generative diffusion model from the multiparametric SAturation-recovery single-SHot acquisition sequence (mSASHA) dataset, which offers joint cardiac T1 and T2 maps. Our method enables approximate but meaningful spin-property estimates from MR images, which provide an interpretable “latent variable” that lead to highly flexible image synthesis of arbitrary novel sequences. We show that Reverse Imaging enables highly accurate segmentation across vastly different image contrasts and imaging protocols, realizing wide-spectrum generalization of cardiac MRI segmentation.
Beyond the LUMIR challenge
The pathway to foundational registration models
Medical image challenges have played a transformative role in advancing the field, catalyzing innovation and establishing new performance benchmarks. Image registration, a foundational task in neuroimaging, has similarly advanced through the Learn2Reg initiative. Building on this, we introduce the L arge-scale U nsupervised Brain M RI I mage R egistration (LUMIR) challenge, a next-generation benchmark for unsupervised brain MRI registration. Previous challenges relied upon anatomical label maps, however LUMIR provides 4,014 unlabeled T1-weighted MRIs for training, encouraging biologically plausible deformation modeling through self-supervision. Evaluation includes 590 in-domain test subjects and extensive zero-shot tasks across disease populations, imaging protocols, and species. Deep learning methods consistently achieved state-of-the-art performance and produced anatomically plausible, diffeomorphic deformation fields. They outperformed several leading optimization-based methods and remained robust to most domain shifts. These findings highlight the growing maturity of deep learning in neuroimaging registration and its potential to serve as a foundation model for general-purpose medical image registration.
...
Medical image challenges have played a transformative role in advancing the field, catalyzing innovation and establishing new performance benchmarks. Image registration, a foundational task in neuroimaging, has similarly advanced through the Learn2Reg initiative. Building on this, we introduce the L arge-scale U nsupervised Brain M RI I mage R egistration (LUMIR) challenge, a next-generation benchmark for unsupervised brain MRI registration. Previous challenges relied upon anatomical label maps, however LUMIR provides 4,014 unlabeled T1-weighted MRIs for training, encouraging biologically plausible deformation modeling through self-supervision. Evaluation includes 590 in-domain test subjects and extensive zero-shot tasks across disease populations, imaging protocols, and species. Deep learning methods consistently achieved state-of-the-art performance and produced anatomically plausible, diffeomorphic deformation fields. They outperformed several leading optimization-based methods and remained robust to most domain shifts. These findings highlight the growing maturity of deep learning in neuroimaging registration and its potential to serve as a foundation model for general-purpose medical image registration.
Deformable image registration remains a central challenge in medical image analysis, particularly under multi-modal scenarios where intensity distributions vary significantly across scans. While deep learning methods provide efficient feed-forward predictions, they often fail to generalize robustly under distribution shifts at test time. A straightforward remedy is full network fine-tuning, yet for modern architectures such as Transformers or deep U-Nets, this adaptation is prohibitively expensive in both memory and runtime when operating in 3D. Meanwhile, the naive fine-tuning struggles more with potential degradation in performance in the existence of drastic domain shifts. In this work, we propose a registration framework that integrates a frozen pretrained mono-modal registration model with a lightweight adaptation pipeline for multi-modal image registration. Specifically, we employ style transfer based on contrast-agnostic representation generation and refinement modules to bridge modality and domain gaps with instance optimization at test time. This design is orthogonal to the choice of backbone mono-modal model, thus avoids the computational burden of full fine-tuning while retaining the flexibility to adapt to unseen domains. We evaluate our approach on the Learn2Reg 2025 LUMIR validation set and observe consistent improvements over the pretrained state-of-the-art mono-modal backbone. In particular, the method ranks second on the multi-modal subset, third on the out-of-domain subset, and achieves fourth place overall in Dice score. These results demonstrate that combining frozen mono-modal models with modality adaptation and lightweight instance optimization offers an effective and practical pathway toward robust multi-modal registration.
...
Deformable image registration remains a central challenge in medical image analysis, particularly under multi-modal scenarios where intensity distributions vary significantly across scans. While deep learning methods provide efficient feed-forward predictions, they often fail to generalize robustly under distribution shifts at test time. A straightforward remedy is full network fine-tuning, yet for modern architectures such as Transformers or deep U-Nets, this adaptation is prohibitively expensive in both memory and runtime when operating in 3D. Meanwhile, the naive fine-tuning struggles more with potential degradation in performance in the existence of drastic domain shifts. In this work, we propose a registration framework that integrates a frozen pretrained mono-modal registration model with a lightweight adaptation pipeline for multi-modal image registration. Specifically, we employ style transfer based on contrast-agnostic representation generation and refinement modules to bridge modality and domain gaps with instance optimization at test time. This design is orthogonal to the choice of backbone mono-modal model, thus avoids the computational burden of full fine-tuning while retaining the flexibility to adapt to unseen domains. We evaluate our approach on the Learn2Reg 2025 LUMIR validation set and observe consistent improvements over the pretrained state-of-the-art mono-modal backbone. In particular, the method ranks second on the multi-modal subset, third on the out-of-domain subset, and achieves fourth place overall in Dice score. These results demonstrate that combining frozen mono-modal models with modality adaptation and lightweight instance optimization offers an effective and practical pathway toward robust multi-modal registration.
Spin-lattice relaxation time (T1) is an important biomarker in cardiac parametric mapping for characterizing myocardial tissue and diagnosing cardiomyopathies. Conventional Modified Look-Locker Inversion Recovery (MOLLI) acquires 11 breath-hold baseline images with interleaved rest periods to ensure mapping accuracy. However, prolonged scanning can be challenging for patients with poor breathholds, often leading to motion artifacts that degrade image quality. In addition, T1 mapping requires a voxel-wise nonlinear fitting to a signal recovery model involving an iterative estimation process. Recent studies have proposed deep-learning approaches for rapid T1 mapping using shortened sequences to reduce acquisition time for patient comfort. Nevertheless, existing methods overlook important physics constraints, limiting interpretability and generalization. In this work, we present an accelerated, end-to-end T1 mapping framework leveraging Physics-Informed Neural Ordinary Differential Equations (ODEs) to model temporal dynamics and address these challenges. Our method achieves high-accuracy T1 estimation from a sparse subset of baseline images and ensures efficient null index estimation at the test time. Specifically, we develop a continuous-time LSTM-ODE model to enable selective Look-Locker (LL) data acquisition with arbitrary time lags. Experimental results show superior performance in T1 estimation for both native and post-contrast sequences and demonstrate the strong benefit of our physics-based formulation over direct data-driven T1 priors.
...
Spin-lattice relaxation time (T1) is an important biomarker in cardiac parametric mapping for characterizing myocardial tissue and diagnosing cardiomyopathies. Conventional Modified Look-Locker Inversion Recovery (MOLLI) acquires 11 breath-hold baseline images with interleaved rest periods to ensure mapping accuracy. However, prolonged scanning can be challenging for patients with poor breathholds, often leading to motion artifacts that degrade image quality. In addition, T1 mapping requires a voxel-wise nonlinear fitting to a signal recovery model involving an iterative estimation process. Recent studies have proposed deep-learning approaches for rapid T1 mapping using shortened sequences to reduce acquisition time for patient comfort. Nevertheless, existing methods overlook important physics constraints, limiting interpretability and generalization. In this work, we present an accelerated, end-to-end T1 mapping framework leveraging Physics-Informed Neural Ordinary Differential Equations (ODEs) to model temporal dynamics and address these challenges. Our method achieves high-accuracy T1 estimation from a sparse subset of baseline images and ensures efficient null index estimation at the test time. Specifically, we develop a continuous-time LSTM-ODE model to enable selective Look-Locker (LL) data acquisition with arbitrary time lags. Experimental results show superior performance in T1 estimation for both native and post-contrast sequences and demonstrate the strong benefit of our physics-based formulation over direct data-driven T1 priors.
The low degree of labeling and limited photon count of fluorescent emitters in single molecule localization microscopy results in poor quality images of macro-molecular complexes. Particle fusion provides a single reconstruction with high signal-to-noise ratio by combining many single molecule localization microscopy images of the same structure. The underlying assumption of homogeneity is not always valid, heterogeneity can arise due to geometrical shape variations or distinct conformational states. We introduce a Point Cloud Variational Auto-Encoder that works directly on 2D and 3D localization data, to detect multiple modes of variation in such datasets. The computing time is on the order of a few minutes, enabled by the linear scaling with dataset size, and fast network training in just four epochs. The use of lists of localization data instead of pixelated images leads to just minor differences in computational burden between 2D and 3D cases. With the proposed method, we detected radius variation in 2D Nuclear Pore Complex data, height variations in 3D DNA origami tetrahedron data, and both radius and height variations in 3D Nuclear Pore Complex data. In all cases, the detected variations were on the few nanometer scale.
...
The low degree of labeling and limited photon count of fluorescent emitters in single molecule localization microscopy results in poor quality images of macro-molecular complexes. Particle fusion provides a single reconstruction with high signal-to-noise ratio by combining many single molecule localization microscopy images of the same structure. The underlying assumption of homogeneity is not always valid, heterogeneity can arise due to geometrical shape variations or distinct conformational states. We introduce a Point Cloud Variational Auto-Encoder that works directly on 2D and 3D localization data, to detect multiple modes of variation in such datasets. The computing time is on the order of a few minutes, enabled by the linear scaling with dataset size, and fast network training in just four epochs. The use of lists of localization data instead of pixelated images leads to just minor differences in computational burden between 2D and 3D cases. With the proposed method, we detected radius variation in 2D Nuclear Pore Complex data, height variations in 3D DNA origami tetrahedron data, and both radius and height variations in 3D Nuclear Pore Complex data. In all cases, the detected variations were on the few nanometer scale.
Journal article
(2026)
-
A. Arami, Omer Burak Demirel, Toygan Kilic, Steen Moeller, Y. Zhao, Y. Zhang, Q. Tao, Hildo J. Lamb, Mehmet Akcakaya, S.D. Weingärtner
In this work, we aimed to develop and evaluate multi-band outer volume suppression pulses for increased acceleration rates in simultaneous multi-slice accelerated cardiac MRI. MB-OVS pulses were constructed from a multi-band combination of two slab-selective saturation pulses and tested for various pulse shapes using Bloch simulation and phantom experiment. The MB-OVS pulses were interleaved between imaging pulses to ensure homogeneous suppression throughout the cardiac cycle/imaging window in vivo. Simultaneous multi-slice (SMS) CINE and first-pass myocardial perfusion scans with and without the proposed MB-OVS pulses were compared in terms of residual artifacts at high acceleration rates. Among the tested pulses, both Bloch simulation and phantom experiments showed that amplitude-optimized sinc pulses provided the best trade-off in suppression efficiency, the required B+ 1 , SAR, and slab profile. CINE imaging with 5-fold SMS-OVS acceleration significantly outperformed imaging without MB-OVS, maintaining leakage-free image quality, even when adding 2-fold in-plane acceleration. SMS-OVS also enabled perfusion imaging in 9 slices with 1.7 × 1.7 mm2 resolution, achieving a 16-fold spatial-only acceleration while ensuring accurate contrast dynamics without leakage artifacts. Interleaved MB-OVS modules enabled thorough leakage artifact suppression in cardiac SMS-accelerated CINE and perfusion imaging, particularly at high acceleration rates. The proposed approach may be promising for unlocking further acceleration potential of SMS in cardiac imaging.
...
In this work, we aimed to develop and evaluate multi-band outer volume suppression pulses for increased acceleration rates in simultaneous multi-slice accelerated cardiac MRI. MB-OVS pulses were constructed from a multi-band combination of two slab-selective saturation pulses and tested for various pulse shapes using Bloch simulation and phantom experiment. The MB-OVS pulses were interleaved between imaging pulses to ensure homogeneous suppression throughout the cardiac cycle/imaging window in vivo. Simultaneous multi-slice (SMS) CINE and first-pass myocardial perfusion scans with and without the proposed MB-OVS pulses were compared in terms of residual artifacts at high acceleration rates. Among the tested pulses, both Bloch simulation and phantom experiments showed that amplitude-optimized sinc pulses provided the best trade-off in suppression efficiency, the required B+ 1 , SAR, and slab profile. CINE imaging with 5-fold SMS-OVS acceleration significantly outperformed imaging without MB-OVS, maintaining leakage-free image quality, even when adding 2-fold in-plane acceleration. SMS-OVS also enabled perfusion imaging in 9 slices with 1.7 × 1.7 mm2 resolution, achieving a 16-fold spatial-only acceleration while ensuring accurate contrast dynamics without leakage artifacts. Interleaved MB-OVS modules enabled thorough leakage artifact suppression in cardiac SMS-accelerated CINE and perfusion imaging, particularly at high acceleration rates. The proposed approach may be promising for unlocking further acceleration potential of SMS in cardiac imaging.
Journal article
(2025)
-
Maša Božić-Iven, Stanislas Rapacchi, Yi Zhang, Qian Tao, Lothar Rudi Schad, Sebastian Weingärtner
Purpose: To introduce Double Inversion Recovery (DIR) preparations for myocardial Arterial Spin Labeling (myoASL) for mitigation of heart rate (HR) variability induced physiological noise (PN).
Methods: DIR-labeling was implemented for double ECG-gated myoASLsequences and compared with conventional Flow-sensitive Alternating Inversion Recovery (FAIR) labeling using single inversions. In DIR-preparations, the FAIR-inversion pulses were immediately followed by an identical reinversion pulse, applied either slice-selectively or nonselectively. Bloch-equation-based simulation and phantom experiments were performed to evaluate the PN and SNR across a range of HR variabilities. Data from six healthy subjects were acquired to evaluate myocardial blood flow (MBF), PN, and SNR in vivo.
Results: Simulation experiments showed that the averageMBFvalues remained nearly constant across the range of HR variabilities and were comparable across all three sequences. However, DIR-labeling allowed for greater recovery of the myocardial background signal, which mitigates the sensitivity to HR-dependent changes in the inversion time. Consequently, PN in the presence of HR variability was substantially reduced with DIR-labeling. For HR variabilities corresponding to the mean value observed in vivo, this resulted in a simulated SNR gain of 1.79 ± 0.90 for selective and 1.55 ± 0.77 for nonselective DIR-labeling. In vivo, DIR-labeling showed reduced PN, with 53% (p < 0.05)/44% (p = 0.16) less PN compared with conventional FAIR-myoASL, leading to an average SNR gain of 1.47 ± 0.63 (p = 0.09)/1.32 ± 0.57 (p = 0.84) with selective/nonselective reinversions.
Conclusion: The proposed DIR-preparations reduce sensitivity to HR variations and alleviate PN in double ECG-gated myoASL, improving the precision of myoASL-based perfusion quantification.
...
Methods: DIR-labeling was implemented for double ECG-gated myoASLsequences and compared with conventional Flow-sensitive Alternating Inversion Recovery (FAIR) labeling using single inversions. In DIR-preparations, the FAIR-inversion pulses were immediately followed by an identical reinversion pulse, applied either slice-selectively or nonselectively. Bloch-equation-based simulation and phantom experiments were performed to evaluate the PN and SNR across a range of HR variabilities. Data from six healthy subjects were acquired to evaluate myocardial blood flow (MBF), PN, and SNR in vivo.
Results: Simulation experiments showed that the averageMBFvalues remained nearly constant across the range of HR variabilities and were comparable across all three sequences. However, DIR-labeling allowed for greater recovery of the myocardial background signal, which mitigates the sensitivity to HR-dependent changes in the inversion time. Consequently, PN in the presence of HR variability was substantially reduced with DIR-labeling. For HR variabilities corresponding to the mean value observed in vivo, this resulted in a simulated SNR gain of 1.79 ± 0.90 for selective and 1.55 ± 0.77 for nonselective DIR-labeling. In vivo, DIR-labeling showed reduced PN, with 53% (p < 0.05)/44% (p = 0.16) less PN compared with conventional FAIR-myoASL, leading to an average SNR gain of 1.47 ± 0.63 (p = 0.09)/1.32 ± 0.57 (p = 0.84) with selective/nonselective reinversions.
Conclusion: The proposed DIR-preparations reduce sensitivity to HR variations and alleviate PN in double ECG-gated myoASL, improving the precision of myoASL-based perfusion quantification.
...
Purpose: To introduce Double Inversion Recovery (DIR) preparations for myocardial Arterial Spin Labeling (myoASL) for mitigation of heart rate (HR) variability induced physiological noise (PN).
Methods: DIR-labeling was implemented for double ECG-gated myoASLsequences and compared with conventional Flow-sensitive Alternating Inversion Recovery (FAIR) labeling using single inversions. In DIR-preparations, the FAIR-inversion pulses were immediately followed by an identical reinversion pulse, applied either slice-selectively or nonselectively. Bloch-equation-based simulation and phantom experiments were performed to evaluate the PN and SNR across a range of HR variabilities. Data from six healthy subjects were acquired to evaluate myocardial blood flow (MBF), PN, and SNR in vivo.
Results: Simulation experiments showed that the averageMBFvalues remained nearly constant across the range of HR variabilities and were comparable across all three sequences. However, DIR-labeling allowed for greater recovery of the myocardial background signal, which mitigates the sensitivity to HR-dependent changes in the inversion time. Consequently, PN in the presence of HR variability was substantially reduced with DIR-labeling. For HR variabilities corresponding to the mean value observed in vivo, this resulted in a simulated SNR gain of 1.79 ± 0.90 for selective and 1.55 ± 0.77 for nonselective DIR-labeling. In vivo, DIR-labeling showed reduced PN, with 53% (p < 0.05)/44% (p = 0.16) less PN compared with conventional FAIR-myoASL, leading to an average SNR gain of 1.47 ± 0.63 (p = 0.09)/1.32 ± 0.57 (p = 0.84) with selective/nonselective reinversions.
Conclusion: The proposed DIR-preparations reduce sensitivity to HR variations and alleviate PN in double ECG-gated myoASL, improving the precision of myoASL-based perfusion quantification.
Methods: DIR-labeling was implemented for double ECG-gated myoASLsequences and compared with conventional Flow-sensitive Alternating Inversion Recovery (FAIR) labeling using single inversions. In DIR-preparations, the FAIR-inversion pulses were immediately followed by an identical reinversion pulse, applied either slice-selectively or nonselectively. Bloch-equation-based simulation and phantom experiments were performed to evaluate the PN and SNR across a range of HR variabilities. Data from six healthy subjects were acquired to evaluate myocardial blood flow (MBF), PN, and SNR in vivo.
Results: Simulation experiments showed that the averageMBFvalues remained nearly constant across the range of HR variabilities and were comparable across all three sequences. However, DIR-labeling allowed for greater recovery of the myocardial background signal, which mitigates the sensitivity to HR-dependent changes in the inversion time. Consequently, PN in the presence of HR variability was substantially reduced with DIR-labeling. For HR variabilities corresponding to the mean value observed in vivo, this resulted in a simulated SNR gain of 1.79 ± 0.90 for selective and 1.55 ± 0.77 for nonselective DIR-labeling. In vivo, DIR-labeling showed reduced PN, with 53% (p < 0.05)/44% (p = 0.16) less PN compared with conventional FAIR-myoASL, leading to an average SNR gain of 1.47 ± 0.63 (p = 0.09)/1.32 ± 0.57 (p = 0.84) with selective/nonselective reinversions.
Conclusion: The proposed DIR-preparations reduce sensitivity to HR variations and alleviate PN in double ECG-gated myoASL, improving the precision of myoASL-based perfusion quantification.
Image registration is essential for medical image applications where alignment of voxels across multiple images is needed for qualitative or quantitative analysis. With recent advances in deep neural networks and parallel computing, deep learning-based medical image registration methods become competitive with their flexible modeling and fast inference capabilities. However, compared to traditional optimization-based registration methods, the speed advantage may come at the cost of registration performance at inference time. Besides, deep neural networks ideally demand large training datasets while optimization-based methods are training-free. To improve registration accuracy and data efficiency, we propose a novel image registration method, termed Recurrent Inference Image Registration (RIIR) network. RIIR is formulated as a meta-learning solver for the registration problem in an iterative manner. RIIR addresses the accuracy and data efficiency issues, by learning the update rule of optimization, with implicit regularization combined with explicit gradient input.
We extensively evaluated RIIR on brain MRI, lung CT, and quantitative cardiac MRI datasets, in terms of both registration accuracy and training data efficiency. Our experiments showed that RIIR outperformed a range of deep learning-based methods, even with only 5% of the training data, demonstrating high data efficiency. Key findings from our ablation studies highlighted the important added value of the hidden states introduced in the recurrent inference framework for meta-learning. Our proposed RIIR offers a highly data-efficient framework for deep learning-based medical image registration. ...
We extensively evaluated RIIR on brain MRI, lung CT, and quantitative cardiac MRI datasets, in terms of both registration accuracy and training data efficiency. Our experiments showed that RIIR outperformed a range of deep learning-based methods, even with only 5% of the training data, demonstrating high data efficiency. Key findings from our ablation studies highlighted the important added value of the hidden states introduced in the recurrent inference framework for meta-learning. Our proposed RIIR offers a highly data-efficient framework for deep learning-based medical image registration. ...
Image registration is essential for medical image applications where alignment of voxels across multiple images is needed for qualitative or quantitative analysis. With recent advances in deep neural networks and parallel computing, deep learning-based medical image registration methods become competitive with their flexible modeling and fast inference capabilities. However, compared to traditional optimization-based registration methods, the speed advantage may come at the cost of registration performance at inference time. Besides, deep neural networks ideally demand large training datasets while optimization-based methods are training-free. To improve registration accuracy and data efficiency, we propose a novel image registration method, termed Recurrent Inference Image Registration (RIIR) network. RIIR is formulated as a meta-learning solver for the registration problem in an iterative manner. RIIR addresses the accuracy and data efficiency issues, by learning the update rule of optimization, with implicit regularization combined with explicit gradient input.
We extensively evaluated RIIR on brain MRI, lung CT, and quantitative cardiac MRI datasets, in terms of both registration accuracy and training data efficiency. Our experiments showed that RIIR outperformed a range of deep learning-based methods, even with only 5% of the training data, demonstrating high data efficiency. Key findings from our ablation studies highlighted the important added value of the hidden states introduced in the recurrent inference framework for meta-learning. Our proposed RIIR offers a highly data-efficient framework for deep learning-based medical image registration.
We extensively evaluated RIIR on brain MRI, lung CT, and quantitative cardiac MRI datasets, in terms of both registration accuracy and training data efficiency. Our experiments showed that RIIR outperformed a range of deep learning-based methods, even with only 5% of the training data, demonstrating high data efficiency. Key findings from our ablation studies highlighted the important added value of the hidden states introduced in the recurrent inference framework for meta-learning. Our proposed RIIR offers a highly data-efficient framework for deep learning-based medical image registration.
Quantitative cardiac magnetic resonance imaging (MRI) is an increasingly important diagnostic tool for cardiovascular diseases. Yet, co-registration of all baseline images within the quantitative MRI sequence is essential for the accuracy and precision of quantitative maps. However, co-registering all baseline images from a quantitative cardiac MRI sequence remains a nontrivial task because of the simultaneous changes in intensity and contrast, in combination with cardiac and respiratory motion. To address the challenge, we propose a novel motion correction framework based on robust principle component analysis (rPCA) that decomposes quantitative cardiac MRI into low-rank and sparse components, and we integrate the groupwise CNN-based registration backbone within the rPCA framework. The low-rank component of rPCA corresponds to the quantitative mapping (i.e. limited degree of freedom in variation), while the sparse component corresponds to the residual motion, making it easier to formulate and solve the groupwise registration problem. We evaluated our proposed method on cardiac T1 mapping by the modified Look-Locker inversion recovery (MOLLI) sequence, both before and after the Gadolinium contrast agent administration. Our experiments showed that our method effectively improved registration performance over baseline methods without introducing rPCA, and reduced quantitative mapping error in both in-domain (pre-contrast MOLLI) and out-of-domain (post-contrast MOLLI) inference. The proposed rPCA framework is generic and can be integrated with other registration backbones.
...
Quantitative cardiac magnetic resonance imaging (MRI) is an increasingly important diagnostic tool for cardiovascular diseases. Yet, co-registration of all baseline images within the quantitative MRI sequence is essential for the accuracy and precision of quantitative maps. However, co-registering all baseline images from a quantitative cardiac MRI sequence remains a nontrivial task because of the simultaneous changes in intensity and contrast, in combination with cardiac and respiratory motion. To address the challenge, we propose a novel motion correction framework based on robust principle component analysis (rPCA) that decomposes quantitative cardiac MRI into low-rank and sparse components, and we integrate the groupwise CNN-based registration backbone within the rPCA framework. The low-rank component of rPCA corresponds to the quantitative mapping (i.e. limited degree of freedom in variation), while the sparse component corresponds to the residual motion, making it easier to formulate and solve the groupwise registration problem. We evaluated our proposed method on cardiac T1 mapping by the modified Look-Locker inversion recovery (MOLLI) sequence, both before and after the Gadolinium contrast agent administration. Our experiments showed that our method effectively improved registration performance over baseline methods without introducing rPCA, and reduced quantitative mapping error in both in-domain (pre-contrast MOLLI) and out-of-domain (post-contrast MOLLI) inference. The proposed rPCA framework is generic and can be integrated with other registration backbones.
Quantitative T1 mapping by MRI is an increasingly important tool for clinical assessment of cardiovascular diseases. The cardiac T1 map is derived by fitting a known signal model to a series of baseline images, while the quality of this map can be deteriorated by involuntary respiratory and cardiac motion. To correct motion, a template image is often needed to register all baseline images, but the choice of template is nontrivial, leading to inconsistent performance sensitive to image contrast. In this work, we propose a novel deep-learning-based groupwise registration framework, which omits the need for a template, and registers all baseline images simultaneously. We design two groupwise losses for this registration framework: the first is a linear principal component analysis (PCA) loss that enforces alignment of baseline images irrespective of the intensity variation, and the second is an auxiliary relaxometry loss that enforces adherence of intensity profile to the signal model. We extensively evaluated our method, termed “PCA-Relax”, and other baseline methods on an in-house cardiac MRI dataset including both pre- and post-contrast T1 sequences. All methods were evaluated under three distinct training-and-evaluation strategies, namely, standard, one-shot, and test-time-adaptation. The proposed PCA-Relax showed further improved performance of registration and mapping over well-established baselines. The proposed groupwise framework is generic and can be adapted to applications involving multiple images.
...
Quantitative T1 mapping by MRI is an increasingly important tool for clinical assessment of cardiovascular diseases. The cardiac T1 map is derived by fitting a known signal model to a series of baseline images, while the quality of this map can be deteriorated by involuntary respiratory and cardiac motion. To correct motion, a template image is often needed to register all baseline images, but the choice of template is nontrivial, leading to inconsistent performance sensitive to image contrast. In this work, we propose a novel deep-learning-based groupwise registration framework, which omits the need for a template, and registers all baseline images simultaneously. We design two groupwise losses for this registration framework: the first is a linear principal component analysis (PCA) loss that enforces alignment of baseline images irrespective of the intensity variation, and the second is an auxiliary relaxometry loss that enforces adherence of intensity profile to the signal model. We extensively evaluated our method, termed “PCA-Relax”, and other baseline methods on an in-house cardiac MRI dataset including both pre- and post-contrast T1 sequences. All methods were evaluated under three distinct training-and-evaluation strategies, namely, standard, one-shot, and test-time-adaptation. The proposed PCA-Relax showed further improved performance of registration and mapping over well-established baselines. The proposed groupwise framework is generic and can be adapted to applications involving multiple images.
Journal article
(2024)
-
Paulina Šiurytė, Joao Tourais, Yi Zhang, Chiara Coletti, Christal van de Steeg-Henzen, Stefano Mandija, Qian Tao, Markus Henningsson, Sebastian Weingärtner
Purpose: To develop and evaluate a robust cardiac B+1 mapping sequence at 3 T, using Bloch–Siegert shift (BSS)-based preparations.
Methods: A longitudinal magnetization preparation module was designed to encode |B+1 |. After magnetization tip-down, off-resonant Fermi pulses, placed symmetrically around two refocusing pulses, induced BSS, followed by tipping back of the magnetization. Bloch simulations were used to optimize refocusing pulse parameters and to assess the mapping sensitivity. Relaxation-induced B+1 error was simulated for various T 1 /T 2 times. The effective mapping range was determined in phantom experiments, and |B+1 | maps were compared to the conventional BSS method and subadiabatic hyperbolic-secant 8 (HS8) pulse-sensitized method. Cardiac B+1 maps were acquired in healthy subjects, and evaluated for repeatability and imaging plane intersection consistency. The technique was modified for three-dimensional (3D) acquisition of the whole heart in a single breath-hold, and compared to two-dimensional (2D) acquisition.
Results: Simulations indicate that the proposed preparation can be tailored to achieve high mapping sensitivity across various B+1 ranges, with maximum sensitivity at the upper B+1 range. T 1 /T 2 -induced bias did not exceed 5.2%. Experimentally reproduced B+1 sensitization closely matched simulations for B+1 ≥ 0.3B+1, max (mean difference 0.031±0.022, compared to 0.018±0.025 in the HS8-sensitized method), and showed 20-fold reduction in the standard deviation of repeated scans, compared with conventional BSS B+1 mapping, and an equivalent 2-fold reduction compared with HS8-sensitization. Robust cardiac B+1 map quality was obtained, with an average test-retest variability of 0.027±0.043 relative to normalized B+1 magnitude, and plane intersection bias of 0.052±0.031. 3D acquisitions showed good agreement with2D scans (mean absolute deviation 0.055±0.061).
Conclusion: BSS-based preparations enable robust and tailorable 2D/3D cardiac B+1 mapping at 3 T in a single breath-hold. ...
Methods: A longitudinal magnetization preparation module was designed to encode |B+1 |. After magnetization tip-down, off-resonant Fermi pulses, placed symmetrically around two refocusing pulses, induced BSS, followed by tipping back of the magnetization. Bloch simulations were used to optimize refocusing pulse parameters and to assess the mapping sensitivity. Relaxation-induced B+1 error was simulated for various T 1 /T 2 times. The effective mapping range was determined in phantom experiments, and |B+1 | maps were compared to the conventional BSS method and subadiabatic hyperbolic-secant 8 (HS8) pulse-sensitized method. Cardiac B+1 maps were acquired in healthy subjects, and evaluated for repeatability and imaging plane intersection consistency. The technique was modified for three-dimensional (3D) acquisition of the whole heart in a single breath-hold, and compared to two-dimensional (2D) acquisition.
Results: Simulations indicate that the proposed preparation can be tailored to achieve high mapping sensitivity across various B+1 ranges, with maximum sensitivity at the upper B+1 range. T 1 /T 2 -induced bias did not exceed 5.2%. Experimentally reproduced B+1 sensitization closely matched simulations for B+1 ≥ 0.3B+1, max (mean difference 0.031±0.022, compared to 0.018±0.025 in the HS8-sensitized method), and showed 20-fold reduction in the standard deviation of repeated scans, compared with conventional BSS B+1 mapping, and an equivalent 2-fold reduction compared with HS8-sensitization. Robust cardiac B+1 map quality was obtained, with an average test-retest variability of 0.027±0.043 relative to normalized B+1 magnitude, and plane intersection bias of 0.052±0.031. 3D acquisitions showed good agreement with2D scans (mean absolute deviation 0.055±0.061).
Conclusion: BSS-based preparations enable robust and tailorable 2D/3D cardiac B+1 mapping at 3 T in a single breath-hold. ...
Purpose: To develop and evaluate a robust cardiac B+1 mapping sequence at 3 T, using Bloch–Siegert shift (BSS)-based preparations.
Methods: A longitudinal magnetization preparation module was designed to encode |B+1 |. After magnetization tip-down, off-resonant Fermi pulses, placed symmetrically around two refocusing pulses, induced BSS, followed by tipping back of the magnetization. Bloch simulations were used to optimize refocusing pulse parameters and to assess the mapping sensitivity. Relaxation-induced B+1 error was simulated for various T 1 /T 2 times. The effective mapping range was determined in phantom experiments, and |B+1 | maps were compared to the conventional BSS method and subadiabatic hyperbolic-secant 8 (HS8) pulse-sensitized method. Cardiac B+1 maps were acquired in healthy subjects, and evaluated for repeatability and imaging plane intersection consistency. The technique was modified for three-dimensional (3D) acquisition of the whole heart in a single breath-hold, and compared to two-dimensional (2D) acquisition.
Results: Simulations indicate that the proposed preparation can be tailored to achieve high mapping sensitivity across various B+1 ranges, with maximum sensitivity at the upper B+1 range. T 1 /T 2 -induced bias did not exceed 5.2%. Experimentally reproduced B+1 sensitization closely matched simulations for B+1 ≥ 0.3B+1, max (mean difference 0.031±0.022, compared to 0.018±0.025 in the HS8-sensitized method), and showed 20-fold reduction in the standard deviation of repeated scans, compared with conventional BSS B+1 mapping, and an equivalent 2-fold reduction compared with HS8-sensitization. Robust cardiac B+1 map quality was obtained, with an average test-retest variability of 0.027±0.043 relative to normalized B+1 magnitude, and plane intersection bias of 0.052±0.031. 3D acquisitions showed good agreement with2D scans (mean absolute deviation 0.055±0.061).
Conclusion: BSS-based preparations enable robust and tailorable 2D/3D cardiac B+1 mapping at 3 T in a single breath-hold.
Methods: A longitudinal magnetization preparation module was designed to encode |B+1 |. After magnetization tip-down, off-resonant Fermi pulses, placed symmetrically around two refocusing pulses, induced BSS, followed by tipping back of the magnetization. Bloch simulations were used to optimize refocusing pulse parameters and to assess the mapping sensitivity. Relaxation-induced B+1 error was simulated for various T 1 /T 2 times. The effective mapping range was determined in phantom experiments, and |B+1 | maps were compared to the conventional BSS method and subadiabatic hyperbolic-secant 8 (HS8) pulse-sensitized method. Cardiac B+1 maps were acquired in healthy subjects, and evaluated for repeatability and imaging plane intersection consistency. The technique was modified for three-dimensional (3D) acquisition of the whole heart in a single breath-hold, and compared to two-dimensional (2D) acquisition.
Results: Simulations indicate that the proposed preparation can be tailored to achieve high mapping sensitivity across various B+1 ranges, with maximum sensitivity at the upper B+1 range. T 1 /T 2 -induced bias did not exceed 5.2%. Experimentally reproduced B+1 sensitization closely matched simulations for B+1 ≥ 0.3B+1, max (mean difference 0.031±0.022, compared to 0.018±0.025 in the HS8-sensitized method), and showed 20-fold reduction in the standard deviation of repeated scans, compared with conventional BSS B+1 mapping, and an equivalent 2-fold reduction compared with HS8-sensitization. Robust cardiac B+1 map quality was obtained, with an average test-retest variability of 0.027±0.043 relative to normalized B+1 magnitude, and plane intersection bias of 0.052±0.031. 3D acquisitions showed good agreement with2D scans (mean absolute deviation 0.055±0.061).
Conclusion: BSS-based preparations enable robust and tailorable 2D/3D cardiac B+1 mapping at 3 T in a single breath-hold.
Lost in Tracking
Uncertainty-Guided Cardiac Cine MRI Segmentation at Right Ventricle Base
Accurate biventricular segmentation of cardiac magnetic resonance (CMR) cine images is essential for the clinical evaluation of heart function. However, compared to left ventricle (LV), right ventricle (RV) segmentation is still more challenging and less reproducible. Degenerate performance frequently occurs at the RV base, where the in-plane anatomical structures are complex (with atria, valve, and aorta) and vary due to the strong interplanar motion. In this work, we propose to address the currently unsolved issues in CMR segmentation, specifically at the RV base, with two strategies: first, we complemented the public resource by reannotating the RV base in the ACDC dataset, with refined delineation of the right ventricle outflow tract (RVOT), under the guidance of an expert cardiologist. Second, we proposed a novel dual encoder U-Net architecture that leverages temporal incoherence to inform the segmentation when interplanar motions occur. The inter-planar motion is characterized by loss-of-tracking, via Bayesian uncertainty of a motion-tracking model. Our experiments showed that our method significantly improved RV base segmentation taking into account temporal incoherence. Furthermore, we investigated the reproducibility of deep learning-based segmentation and showed that the combination of consistent annotation and loss of tracking could enhance the reproducibility of RV segmentation, potentially facilitating a large number of clinical studies focusing on RV.
...
Accurate biventricular segmentation of cardiac magnetic resonance (CMR) cine images is essential for the clinical evaluation of heart function. However, compared to left ventricle (LV), right ventricle (RV) segmentation is still more challenging and less reproducible. Degenerate performance frequently occurs at the RV base, where the in-plane anatomical structures are complex (with atria, valve, and aorta) and vary due to the strong interplanar motion. In this work, we propose to address the currently unsolved issues in CMR segmentation, specifically at the RV base, with two strategies: first, we complemented the public resource by reannotating the RV base in the ACDC dataset, with refined delineation of the right ventricle outflow tract (RVOT), under the guidance of an expert cardiologist. Second, we proposed a novel dual encoder U-Net architecture that leverages temporal incoherence to inform the segmentation when interplanar motions occur. The inter-planar motion is characterized by loss-of-tracking, via Bayesian uncertainty of a motion-tracking model. Our experiments showed that our method significantly improved RV base segmentation taking into account temporal incoherence. Furthermore, we investigated the reproducibility of deep learning-based segmentation and showed that the combination of consistent annotation and loss of tracking could enhance the reproducibility of RV segmentation, potentially facilitating a large number of clinical studies focusing on RV.
Journal article
(2024)
-
Niall Taylor, Yi Zhang, Dan W. Joyce, Ziming Gao, Andrey Kormilitzin, Alejo Nevado-Holgado
When the first transformer-based language models were published in the late 2010s, pretraining with general text and then fine-tuning the model on a task-specific dataset often achieved the state-of-the-art performance. However, more recent work suggests that for some tasks, directly prompting the pretrained model matches or surpasses fine-tuning in performance with few or no model parameter updates required. The use of prompts with language models for natural language processing (NLP) tasks is known as prompt learning. We investigated the viability of prompt learning on clinically meaningful decision tasks and directly compared this with more traditional fine-tuning methods. Results show that prompt learning methods were able to match or surpass the performance of traditional fine-tuning with up to 1000 times fewer trainable parameters, less training time, less training data, and lower computation resource requirements. We argue that these characteristics make prompt learning a very desirable alternative to traditional fine-tuning for clinical tasks, where the computational resources of public health providers are limited, and where data can often not be made available or not be used for fine-tuning due to patient privacy concerns. The complementary code to reproduce the experiments presented in this work can be found at https://github.com/NtaylorOX/Public_Clinical_Prompt.
...
When the first transformer-based language models were published in the late 2010s, pretraining with general text and then fine-tuning the model on a task-specific dataset often achieved the state-of-the-art performance. However, more recent work suggests that for some tasks, directly prompting the pretrained model matches or surpasses fine-tuning in performance with few or no model parameter updates required. The use of prompts with language models for natural language processing (NLP) tasks is known as prompt learning. We investigated the viability of prompt learning on clinically meaningful decision tasks and directly compared this with more traditional fine-tuning methods. Results show that prompt learning methods were able to match or surpass the performance of traditional fine-tuning with up to 1000 times fewer trainable parameters, less training time, less training data, and lower computation resource requirements. We argue that these characteristics make prompt learning a very desirable alternative to traditional fine-tuning for clinical tasks, where the computational resources of public health providers are limited, and where data can often not be made available or not be used for fine-tuning due to patient privacy concerns. The complementary code to reproduce the experiments presented in this work can be found at https://github.com/NtaylorOX/Public_Clinical_Prompt.
Deep learning-based methods have achieved prestigious performance for magnetic resonance imaging (MRI) reconstruction, enabling fast imaging for many clinical applications. Previous methods employ convolutional networks to learn the image prior as the regularization term. In quantitative MRI, the physical model of nuclear magnetic resonance relaxometry is known, providing additional prior knowledge for image reconstruction. However, traditional reconstruction networks are limited to learning the spatial domain prior knowledge, ignoring the relaxometry prior. Therefore, we propose a relaxometry-guided quantitative MRI reconstruction framework to learn the spatial prior from data and the relaxometry prior from MRI physics. Additionally, we also evaluated the performance of two popular reconstruction backbones, namely, recurrent variational networks (RVN) and variational networks (VN) with U-Net. Experiments demonstrate that the proposed method achieves highly promising results in quantitative MRI reconstruction.
...
Deep learning-based methods have achieved prestigious performance for magnetic resonance imaging (MRI) reconstruction, enabling fast imaging for many clinical applications. Previous methods employ convolutional networks to learn the image prior as the regularization term. In quantitative MRI, the physical model of nuclear magnetic resonance relaxometry is known, providing additional prior knowledge for image reconstruction. However, traditional reconstruction networks are limited to learning the spatial domain prior knowledge, ignoring the relaxometry prior. Therefore, we propose a relaxometry-guided quantitative MRI reconstruction framework to learn the spatial prior from data and the relaxometry prior from MRI physics. Additionally, we also evaluated the performance of two popular reconstruction backbones, namely, recurrent variational networks (RVN) and variational networks (VN) with U-Net. Experiments demonstrate that the proposed method achieves highly promising results in quantitative MRI reconstruction.
The euclidean space is evil
Hyperbolic attribute editing for few-shot image generation
Conference paper
(2023)
-
Lingxiao Li, Yi Zhang, Shuhui Wang
Few-shot image generation is a challenging task since it aims to generate diverse new images for an unseen category with only a few images. Existing methods suffer from the trade-off between the quality and diversity of generated images. To tackle this problem, we propose Hyperbolic Attribute Editing (HAE), a simple yet effective method. Unlike other methods that work in Euclidean space, HAE captures the hierarchy among images using data from seen categories in hyperbolic space. Given a well-trained HAE, images of unseen categories can be generated by moving the latent code of a given image toward any meaningful directions in the Poincaré disk with a fixing radius. Most importantly, the hyperbolic space allows us to control the semantic diversity of the generated images by setting different radii in the disk. Extensive experiments and visualizations demonstrate that HAE is capable of not only generating images with promising quality and diversity using limited data but achieving a highly controllable and interpretable editing process. Code is available at https://github.com/lingxiao-li/HAE.
...
Few-shot image generation is a challenging task since it aims to generate diverse new images for an unseen category with only a few images. Existing methods suffer from the trade-off between the quality and diversity of generated images. To tackle this problem, we propose Hyperbolic Attribute Editing (HAE), a simple yet effective method. Unlike other methods that work in Euclidean space, HAE captures the hierarchy among images using data from seen categories in hyperbolic space. Given a well-trained HAE, images of unseen categories can be generated by moving the latent code of a given image toward any meaningful directions in the Poincaré disk with a fixing radius. Most importantly, the hyperbolic space allows us to control the semantic diversity of the generated images by setting different radii in the disk. Extensive experiments and visualizations demonstrate that HAE is capable of not only generating images with promising quality and diversity using limited data but achieving a highly controllable and interpretable editing process. Code is available at https://github.com/lingxiao-li/HAE.