Yanjie Zhu

Adaptive Dictionary Learning in Sparse Gradient Domain for Image Recovery

Image recovery from undersampled data has always been challenging due to its implicit ill-posed nature but becomes fascinating with the emerging compressed sensing (CS) theory. This paper proposes a novel gradient based dictionary learning method for image recovery, which effectively integrates the popular total variation (TV) and dictionary learning technique into the same framework. Specifically, we first train dictionaries from the horizontal and vertical gradients of the image and then reconstruct the desired image using the sparse representations of both derivatives. The proposed method enables local features in the gradient images to be captured effectively, and can be viewed as an adaptive extension of the TV regularization. The results of various experiments on MR images consistently demonstrate that the proposed algorithm efficiently recovers images and presents advantages over the current leading CS reconstruction approaches.

PANDA- T1ρ: Integrating principal component analysis and dictionary learning for fast T1ρ mapping

Long scanning time greatly hinders the widespread application of spin‐lattice relaxation in rotating frame ( T1ρ ) in clinics. In this study, a novel method is proposed to reconstruct the T1ρ ‐weighted images from undersampled k‐space data and hence accelerate the acquisition of T1ρ imaging.

Incorporating reference in parallel imaging and compressed sensing

To develop a new compressed sensing parallel imaging technique called READ‐PICS that can effectively incorporate prior information from a reference scan for MR image reconstruction from highly undersampled multichannel measurements.

A model-based method with joint sparsity constraint for direct diffusion tensor estimation

Diffusion tensor imaging (DTI) has been widely used for nondestructive characterization of microstructures of myocardium or brain connectivity. It requires repeated acquisition with different diffusion gradients. The long acquisition time greatly limits the clinical application of DTI. In this paper, a novel method, named model-based method with joint sparsity constraint (MB-JSC), effectively incorporates the prior information on the joint sparsity of different diffusion-weighted images in direct estimation of the diffusion tensor from highly undersampled k-space data. Experimental results demonstrate that the proposed method is able to estimate the diffusion tensors more accurately than the existing method when a high net reduction factor is used.

Direct diffusion tensor estimation using a model‐based method with spatial and parametric constraints

Purpose: To develop a new model‐based method with spatial and parametric constraints (MB‐SPC) aimed at accelerating diffusion tensor imaging (DTI) by directly estimating the diffusion tensor from highly undersampled k‐space data. Methods: The MB‐SPC method effectively incorporates the prior information on the joint sparsity of different diffusion‐weighted images using an L1–L2 norm and the smoothness of the diffusion tensor using a total variation seminorm. The undersampled k‐space datasets were obtained from fully sampled DTI datasets of a simulated phantom and an ex‐vivo experimental rat heart with acceleration factors ranging from 2 to 4. The diffusion tensor was directly reconstructed by solving a minimization problem with a nonlinear conjugate gradient descent algorithm. The reconstruction performance was quantitatively assessed using the normalized root mean square error (nRMSE) of the DTI indices. Results: The MB‐SPC method achieves acceptable DTI measures at an acceleration factor up to 4. Experimental results demonstrate that the proposed method can estimate the diffusion tensor more accurately than most existing methods operating at higher net acceleration factors. Conclusion: The proposed method can significantly reduce artifact, particularly at higher acceleration factors or lower SNRs. This method can easily be adapted to MR relaxometry parameter mapping and is thus useful in the characterization of biological tissue such as nerves, muscle, and heart tissue.

SCOPE: signal compensation for low-rank plus sparse matrix decomposition for fast parameter mapping

Magnetic resonance (MR) parameter mapping is useful for many clinical applications. However, its practical utility is limited by the long scan time. To address this problem, this paper developed a novel image reconstruction method for fast MR parameter mapping. The proposed method (SCOPE) used a low-rank plus sparse model to reconstruct the parameter-weighted images from highly undersampled acquisitions. A signal compensation strategy was introduced to promote low rankness along the parametric direction and thus improve the reconstruction accuracy. Specifically, compensation was performed by multiplying the original signal by the inversion of the mono-exponential decay at each voxel. The performance of SCOPE was evaluated via quantitative T 1ρ mapping. The results of the simulation and in vivo experiments with acceleration factors from 3 to 5 are shown. The performance of SCOPE was verified via comparisons with several low-rank and sparsity-based methods. The experimental results showed that the T 1ρ maps obtained using SCOPE were more accurate than those obtained using competing methods and were comparable to the reference, even when the acceleration factor reached 5. SCOPE can greatly reduce the scan time of parameter mapping while still achieving high accuracy. This technique might therefore help facilitate fast MR parameter mapping in clinical use.

Improved parallel image reconstruction using feature refinement: Improved Parallel Imaging by Feature Refinement

The aim of this study was to develop a novel feature refinement MR reconstruction method from highly undersampled multichannel acquisitions for improving the image quality and preserve more detail information.

Fast T1 mapping of the brain at high field using Look-Locker and fast imaging

This study aims to develop and evaluate a new method for fast high resolution T1 mapping of the brain based on the Look-Locker technique. Single-shot turboflash sequence with high temporal acceleration is used to sample the recovery of inverted magnetization. Multi-slice interleaved acquisition within one inversion slab is used to reduce the number of inversion pulses and hence SAR. Accuracy of the proposed method was studied using simulation and validated in phantoms. It was then evaluated in healthy volunteers and stroke patients. In-vivo results were compared to values obtained by inversion recovery fast spin echo (IR-FSE) and literatures. With the new method, T1 values in phantom experiments agreed with reference values with median error <3%. For in-vivo experiments, a T1 map was acquired in 3.35s and the T1 maps of the whole brain were acquired in 2min with two-slice interleaving, with a spatial resolution of 1.1×1.1×4mm3. The T1 values obtained were comparable to those measured with IR-FSE and those reported in literatures. These results demonstrated the feasibility of the proposed method for fast T1 mapping of the brain in both healthy volunteers and stroke patients at 3T.