A5022205605- Advanced MRI Techniques and Applications
- Cardiac Imaging and Diagnostics
- Cardiovascular Function and Risk Factors
- Atomic and Subatomic Physics Research
- Medical Imaging Techniques and Applications
- Advanced X-ray and CT Imaging
- Cardiac Valve Diseases and Treatments
- Radiomics and Machine Learning in Medical Imaging
- Glioma Diagnosis and Treatment
- Medical Image Segmentation Techniques
- Coronary Interventions and Diagnostics
- Long-Term Effects of COVID-19
- Cardiovascular Effects of Exercise
- Characterization and Applications of Magnetic Nanoparticles
- Ultrasound and Hyperthermia Applications
- Cardiovascular Health and Disease Prevention
- MRI in cancer diagnosis
- Cardiac Arrhythmias and Treatments
- AI in cancer detection
- Cerebrovascular and Carotid Artery Diseases
- Cardiovascular Disease and Adiposity
Tsinghua University
2020-2024
King's College London
2024
Beijing Institute of Technology
2019
Recent innovations in Magnetic Resonance Imaging (MRI) hardware and software have reignited interest low-field ($<1\,\mathrm{T}$) ultra-low-field MRI ($<0.1\,\mathrm{T}$). These technologies offer advantages such as lower power consumption, reduced specific absorption rate, field-inhomogeneities, cost-effectiveness, presenting a promising alternative for resource-limited point-of-care settings. However, faces inherent challenges like signal-to-noise ratio therefore, potentially spatial...
In myocardial T1 mapping, undesirable motion poses significant challenges because uncorrected can affect estimation accuracy and cause incorrect diagnosis. this study, we propose evaluate a correction method for mapping using self-supervised deep learning based registration with contrast separation (SDRAP). A sparse coding was first proposed to separate the component from -weighted (T1w) images. Then, neural network cross-correlation (SDRAP-CC) or mutual information as similarity measurement...
To develop and validate a highly efficient motion compensated free-breathing isotropic resolution 3D whole-heart joint T
Quantitative myocardial tissue characterization with T1 and T2 parametric mapping can provide an accurate complete assessment of abnormalities across a broad range cardiomyopathies. However, current clinical tools rely predominantly on two-dimensional (2D) breath-hold sequences. Clinical adoption three-dimensional (3D) techniques is limited by long scan duration. The aim this study to develop validate time-efficient 3D free-breathing simultaneous sequence using multi-parametric...
Background Quantitative T1, T2, and T2* measurements of carotid atherosclerotic plaque are important in evaluating vulnerability monitoring its progression. Purpose To develop a sequence to simultaneously quantify plaque. Materials Methods The simultaneous mapping (SIMPLE*) is composed three modules with different T2 preparation pulses, inversion-recovery acquisition schemas. Single-echo data were used for T1 quantification, while the multiecho (ME) quantification. quantitative accuracy...
Background Radiofrequency ablation (RFA) is a widely used treatment for atrial fibrillation, reducing the risk of cardiac arrhythmia. Detailed visualization and quantification scarring has potential to improve preprocedural decision-making postprocedural prognosis. Conventional bright-blood late gadolinium enhancement (LGE) MRI can help detect scars; however, its suboptimal myocardium blood contrast inhibits accurate scar estimation. Purpose To develop test free-breathing LGE approach that...
To develop and evaluate a deep neural network (DeepFittingNet) for T1 /T2 estimation of the most commonly used cardiovascular MR mapping sequences to simplify data processing improve robustness.DeepFittingNet is 1D composed recurrent (RNN) fully connected (FCNN) network, in which RNN adapts different number input signals from various FCNN subsequently predicts A, B, Tx three-parameter model. DeepFittingNet was trained using Bloch-equation simulations MOLLI saturation-recovery single-shot...
The purpose of the current study was to develop and validate a three‐dimensional (3D) free‐breathing cardiac T 1 ‐mapping sequence using SAturation‐recovery Variable‐flip‐Angle (SAVA). SAVA sequentially acquires multiple electrocardiogram‐triggered volumes multishot spoiled gradient‐echo sequence. first volume samples equilibrium signal longitudinal magnetization, where flip angle 2° is used reduce time for magnetization return equilibrium. succeeding three are saturation prepared with...
Abstract The purpose of the current study was to develop and evaluate a three‐dimensional single Breath‐hOLd cardiac T 2 mapping sequence (3D BOLT) with low‐rank plus sparse (L + S) reconstruction for rapid whole‐heart measurement. 3D BOLT collects three highly accelerated electrocardiogram‐triggered volumes coverage, all within 12‐heartbeat breath‐hold. Saturation pulses are performed every heartbeat prepare longitudinal magnetization before preparation (T ‐prep) or readout, echo time ‐prep...
MR parametric mapping including T1 and T2 enabled quantitative evaluation of changes myocardium. We previously proposed a time-efficient technique for 3D free-breathing simultaneous based on multi-parametric SAturation recovery Variable flip Angle (mSAVA). This study evaluated the accuracy, precision, reproducibility mSAVA in comparison with conventional 2D sequences. achieved good between that MOLLI SASHA, better precision than SASHA measurements. measured by had both GraSE bSSFP T2. offers...
SASHA with a 3-parameter fitting model has high T1 accuracy but low precision due to SNR in saturation-recovery T1-weighted images. Alternatively, two-parameter could improve the penalty of losing accuracy. In this study, we developed 1D neural network (DeepFittingNet) predict and alleviate impaction from noise. We trained DeepFittingNet using simulation different Rician noise levels tested it in-vivo MR data. Results showed that had comparable fitting.
Multi-parametric SAturation recovery and Variable flip Angle (mSAVA) acquires four 3D volumes with different T1 T2 weightings during free-breathing to simultaneously generate whole heart parametric maps. We proposed a fast simple CMR protocol using pre- post-contrast mSAVA acquisitions additionally ECV maps, bright-blood dark-blood LGE images. could comprehensively assess the myocardium over within ~6+6 min in cohort of twenty patients.
The most used curve-fitting method for map reconstruction of the cardiovascular magnetic resonance mapping is sensitive to initial conditions, time-consuming, and prone fitting error. In this study, we sought develop a deep-learning approach (DeepFittingNet) perform T1 T2 calculations clinically cardiac parametric mappings, simplify clinical workflow T1/T2 measurements improve robustness. testing, DeepFittingNet could estimation tasks MOLLI, SASHA, T2-prep bSSFP. Compared algorithm,...
Abstract Purpose This study aims to develop and evaluate a novel cardiovascular MR sequence, MyoFold, designed for the simultaneous quantifications of myocardial tissue composition wall motion. Methods MyoFold is as 2D single breathing‐holding integrating joint T 1 /T 2 mapping cine imaging. The sequence uses 2‐fold accelerated balanced SSFP (bSSFP) data readout incorporates electrocardiogram synchronization align with cardiac cycle. initially acquires six single‐shot inversion‐recovery...
Motivation: Myocardial T1 and T2 mapping has emerged as a useful clinical tool for the diagnosis of different heart disease. However, current sequences were mostly developed with 2D breathhold acquisitions validated at 1.5T or 3T. The investigation myocardial techniques on more affordable low-field MRI systems is scarce. Goal(s): To develop highly-efficient free-breathing 3D whole-heart joint T1/T2 sequence isotropic-resolution 0.55T. Approach: proposed acquires 3 interleaved volumes...
Motivation: 3D whole-heart late gadolinium enhancement imaging has previously been demonstrated at 1.5T, but not low-field (0.55T). Goal(s): To develop a novel free breathing, whole-heart, framework low field. Approach: Patients with known ischaemic heart disease were scanned and results of the proposed sequence compared 2D LGE images. Results: There is excellent agreement between datasets in detection myocardial scar. Impact: Preliminary demonstrate feasibility for comprehensive scar 0.55T.
This paper presents a multi-view learning based method for left atrial cavity segmentation in 3D Late Gadolinium Enhanced Magnetic Resonance Imaging (LGE-MRI). Segmenting atrium is challenging due to the low intensity contrast, motion artifacts, and extremely thin walls. Since spatial consistency of could help alleviate ambiguity caused by those problems, proposed consists three deep convolutional streams which construct likelihood maps from different views, i.e., axial view, coronal...