Fast Automatic Parameter Selection for MRI Reconstruction

Toma, Tanjin Taher; Weller, Daniel S.

doi:10.1109/isbi45749.2020.9098569

Cited by 4 publications

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Automatic parameter tuning techniques [58][59][60] could be adopted to reach a more optimum combination of these weights.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Joint MAPLE: Accelerated joint T₁ and T2*$$ {{\mathrm{T}}_2}^{\ast } $$ mapping with scan‐specific self‐supervised networks

Heydari,

Ahmadi,

Kim

et al. 2024

Magnetic Resonance in Med

View full text Add to dashboard Cite

PurposeQuantitative MRI finds important applications in clinical and research studies. However, it is encoding intensive and may suffer from prohibitively long scan times. Accelerated MR parameter mapping techniques have been developed to help address these challenges. Here, an accelerated joint T1, , frequency and proton density mapping technique with scan‐specific self‐supervised network reconstruction is proposed to synergistically combine parallel imaging, model‐based, and deep learning approaches to speed up parameter mapping.MethodsProposed framework, Joint MAPLE, includes parallel imaging, signal modeling, and data consistency blocks which are optimized jointly in a combined loss function. A scan‐specific self‐supervised reconstruction is embedded into the framework, which takes advantage of multi‐contrast data from a multi‐echo, multi‐flip angle, gradient echo acquisition.ResultsIn comparison with parallel reconstruction techniques powered by low‐rank methods, emerging scan specific networks, and model‐based estimation approaches, the proposed framework reduces the reconstruction error in parameter maps by approximately two‐fold on average at acceleration rates as high as R = 16 with uniform sampling. It can outperform evaluated parallel reconstruction techniques up to four‐fold on average in the presence of challenging sub‐sampling masks. It is observed that Joint MAPLE performs well at extreme acceleration rates of R = 25 and R = 36 with error values less than 20%.ConclusionJoint MAPLE enables higher fidelity parameter estimation at high acceleration rates by synergistically combining parallel imaging and model‐based parameter mapping and exploiting multi‐echo, multi‐flip angle datasets. Utilizing a scan‐specific self‐supervised reconstruction obviates the need for large data sets for training while improving the parameter estimation ability.

show abstract

“…Automatic parameter tuning techniques [58][59][60] could be adopted to reach a more optimum combination of these weights.…”

Section: Discussionmentioning

confidence: 99%

“…Based on a heuristic approach, for each experiment a limited set of values for each regularization weight were evaluated to minimize NRMSE and obtain better reconstruction and mapping performance. Automatic parameter tuning techniques 58–60 could be adopted to reach a more optimum combination of these weights.…”

Section: Discussionmentioning

confidence: 99%

Joint MAPLE: Accelerated joint T₁ and T2*$$ {{\mathrm{T}}_2}^{\ast } $$ mapping with scan‐specific self‐supervised networks

Heydari,

Ahmadi,

Kim

et al. 2024

Magnetic Resonance in Med

View full text Add to dashboard Cite

show abstract

“…Algorithms on the choice of regularization strength for general imaging inverse problems and especially magnetic resonance imaging are broadly studied in academic research. There are different strategies to emphasize: the classical methods based on the L-curve and the condition of the system (Hansen and O'Leary 1993, Lin et al 2004, Ying et al 2004, SURE-based methods (Ramani et al 2012), Bayesian approaches (Saquib et al 1998, Chaabene et al 2020 and learning-based methods (Toma and Weller 2020). However, all of these methods tackle only the regularization term and not the choice of the frequency selection.…”

Section: Introductionmentioning

confidence: 99%

Solving the MPI reconstruction problem with automatically tuned regularization parameters

Scheffler,

Boberg,

Knopp

2024

Phys. Med. Biol.

View full text Add to dashboard Cite

In the field of medical imaging, Magnetic Particle Imaging (MPI) poses a promising non-ionizing tomographic technique with high spatial and temporal resolution. In MPI, iterative solvers are used to reconstruct the particle distribution out of the measured voltage signal based on a system matrix. The amount of regularization needed to reconstruct an image of good quality differs from measurement to measurement, depending on the MPI system and the measurement settings. Finding the right choice for the three major parameters controlling the regularization is commonly done by hand and requires time and experience. In this work, we study the reduction to a single regularization parameter and propose a method that enables automatic reconstruction. The method is qualitatively and quantitatively validated on several MPI data sets showing promising results.

show abstract

“…Thus, an adaptive multi-parameter selection method is needed, regarding the gradually widely use of combined penalty terms. Though deep learning approaches [24]- [27] can deal with multiple regularization parameters selection, the access to the ground truth image cannot be granted in real-life applications, such as in SAR imaging. In addition, deep learning approaches also can be very computationally demanding, but the training results sometimes only apply to limited situations, such as a fixed signal-to-noise ratio.…”

Section: Introductionmentioning

confidence: 99%

L-Hypersurface Based Parameters Selection in Composite Regularization Models With Application to SAR and TomoSAR Imaging

Fan,

Wang,

et al. 2023

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Composite regularization models are widely used in sparse signal processing, making multiple regularization parameters selection a significant problem to be solved. Variety kinds of composite regularization models are used in sparse microwave imaging, including ℓ1 and ℓ2 penalty, nonconvex and TV penalty, combined dictionary, etc. In this article, a new adaptive multiple regularization parameters selection method named L-hypersurface is proposed. The effectiveness of the proposed method is verified by experiments. Simulation experiments indicate that the selected optimal regularization parameters have satisfied reconstruction results, both visually and numerically. Furthermore, experiments on Gaofen-3 SAR satellite data are also exploited to show the performance of the proposed method.

show abstract

Fast Automatic Parameter Selection for MRI Reconstruction

Cited by 4 publications

References 12 publications

Joint MAPLE: Accelerated joint T₁ and T2*$$ {{\mathrm{T}}_2}^{\ast } $$ mapping with scan‐specific self‐supervised networks

Joint MAPLE: Accelerated joint T₁ and T2*$$ {{\mathrm{T}}_2}^{\ast } $$ mapping with scan‐specific self‐supervised networks

Solving the MPI reconstruction problem with automatically tuned regularization parameters

L-Hypersurface Based Parameters Selection in Composite Regularization Models With Application to SAR and TomoSAR Imaging

Contact Info

Product

Resources

About

Fast Automatic Parameter Selection for MRI Reconstruction

Cited by 4 publications

References 12 publications

Joint MAPLE: Accelerated joint T1 and T2*$$ {{\mathrm{T}}_2}^{\ast } $$ mapping with scan‐specific self‐supervised networks

Joint MAPLE: Accelerated joint T1 and T2*$$ {{\mathrm{T}}_2}^{\ast } $$ mapping with scan‐specific self‐supervised networks

Solving the MPI reconstruction problem with automatically tuned regularization parameters

L-Hypersurface Based Parameters Selection in Composite Regularization Models With Application to SAR and TomoSAR Imaging

Contact Info

Product

Resources

About

Joint MAPLE: Accelerated joint T₁ and T2*$$ {{\mathrm{T}}_2}^{\ast } $$ mapping with scan‐specific self‐supervised networks

Joint MAPLE: Accelerated joint T₁ and T2*$$ {{\mathrm{T}}_2}^{\ast } $$ mapping with scan‐specific self‐supervised networks