Model-based Reconstruction Techniques Research Articles

Variational model-based reconstruction techniques for multi-patch data in Magnetic Particle Imaging

Magnetic Particle Imaging is an emerging imaging modality through which it is possible to detect tracers containing superparamagnetic nanoparticles. The exposure of the particles to dynamic magnetic fields generates a non-linear response that is used to locate the particles and produce an image of their distribution. The bounding box that can be covered by a single scan curve depends on the strength of the gradients of the magnetic fields applied, which is limited due to the risk of causing peripheral nerve stimulation (PNS) in the patients. To address this issue, multiple scans are performed. The scan data must be merged together to produce reconstructions of larger regions of interest. In this paper we propose a mathematical framework which can deal with rather general multi-patching scenarios including rigid transformations of the field of view (FoV), the specimen and of the scanner. We show the flexibility of this framework in a variety of different scanning scenarios. Moreover, we describe an iterative reconstruction algorithm that yields a reconstruction of the target distribution by minimizing a convex functional which includes positivity constraints and sparsity enforcing priors. We show its convergence to a minimizer and perform numerical experiments on simulated data.

Hybrid and Model-Based Iterative Reconstruction Influences the Volumetry of Visceral and Subcutaneous Adipose Tissue on Ultra-Low-Dose CT.

The aim of this study was to compare three different reconstruction algorithms for the volumetry of the visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) on ultra-low-dose computed tomography (CT) images. Thirty-seven male patients underwent ultra-low-dose CT at the level of the fourth lumbar vertebra (22.5 mm in z-axis). The acquisitions were reconstructed in 5-mm slices with 50% overlap using filtered back projection (FBP), hybrid iterative reconstruction (HIR), and iterative model-based reconstruction (IMR) techniques. The volume of VAT and SAT was measured using an interactive seed-growing segmentation and by thresholding (-30 to -190 HU). The volume of SAT measured by the interactive method was smaller in FBP compared with both HIR (P = 0.0011) and IMR (P = 0.0034), and the volume of VAT was greater in IMR compared with HIR (P = 0.0253) or FBP (P = 0.0065). Using the thresholding method, IMR volumes of VAT were greater compared with HIR (P < 0.0001), and volumes of SAT were greater compared with both HIR and FBP (both P ≤ 0.0001). The VAT to SAT ratio was greater in IMR compared with HIR or FBP (both P < 0.0001). There are significant differences among FBP, HIR, and IMR in the volumetry of SAT and VAT, their ratios, and attenuation measured on ultra-low-dose images.

SAR image despeckling using plug‐and‐play ADMM

The presence of speckle, a granular structure, seriously affects the visual interpretability of the synthetic aperture radar (SAR) images. Recently, model-based reconstruction techniques that use alternating direction method of multipliers (ADMM) have been widely utilised for denoising problems. Owing to the modular structure of ADMM, it is very simple to implement and also enables one to plug in any off-the-shelf denoising algorithm. However, the major limitations of these methods are mostly they have been established for Gaussian noise and the selection of a regulariser for a specific forward model is unknown. In this study, the despeckling of SAR images is addressed in the presence of multiplicative noise such as Rayleigh and Poisson. A mean filter is introduced for images corrupted by Rayleigh-based speckle, which transforms it to Nakagami distributed speckle, to improve the performance. Maximum a posteriori-based estimation involving Nakagami and Poisson distributions are applied to plug-and-play ADMM (PnP ADMM) framework, which provides enhanced despeckling ability along with the preservation of important details. The convergence analysis of the proposed PnP ADMM for multiplicative noise is also provided. The simulation studies are carried out taking several parameters for visual quality and statistical assessment to justify the effectiveness of the proposed method.

Multi-GPU Acceleration of Branchless Distance Driven Projection and Backprojection for Clinical Helical CT

Abstract Model-based image reconstruction (MBIR) techniques have the potential to generate high quality images from noisy measurements and a small number of projections which can reduce the x-ray dose in patients. These MBIR techniques rely on projection and backprojection to refine an image estimate. One of the widely used projectors for these modern MBIR based technique is called branchless distance driven (DD) projection and backprojection. While this method produces superior quality images, the computational cost of iterative updates keeps it from being ubiquitous in clinical applications. In this paper, we provide several new parallelization ideas for concurrent execution of the DD projectors in multi-GPU systems using CUDA programming tools. We have introduced some novel schemes for dividing the projection data and image voxels over multiple GPUs to avoid runtime overhead and inter-device synchronization issues. We have also reduced the complexity of overlap calculation of the algorithm by eliminating the common projection plane and directly projecting the detector boundaries onto image voxel boundaries. To reduce the time required for calculating the overlap between the detector edges and image voxel boundaries, we have proposed a pre-accumulation technique to accumulate image intensities in perpendicular 2D image slabs (from a 3D image) before projection and after backprojection to ensure our DD kernels run faster in parallel GPU threads. For the implementation of our iterative MBIR technique we use a parallel multi-GPU version of the alternating minimization (AM) algorithm with penalized likelihood update. The time performance using our proposed reconstruction method with Siemens Sensation 16 patient scan data shows an average of 24 times speedup using a single TITAN X GPU and 74 times speedup using 3 TITAN X GPUs in parallel for combined projection and backprojection.

Multi-GPU Acceleration of Branchless Distance Driven Projection and Backprojection for Clinical Helical CT.

Model-based image reconstruction (MBIR) techniques have the potential to generate high quality images from noisy measurements and a small number of projections which can reduce the x-ray dose in patients. These MBIR techniques rely on projection and backprojection to refine an image estimate. One of the widely used projectors for these modern MBIR based technique is called branchless distance driven (DD) projection and backprojection. While this method produces superior quality images, the computational cost of iterative updates keeps it from being ubiquitous in clinical applications. In this paper, we provide several new parallelization ideas for concurrent execution of the DD projectors in multi-GPU systems using CUDA programming tools. We have introduced some novel schemes for dividing the projection data and image voxels over multiple GPUs to avoid runtime overhead and inter-device synchronization issues. We have also reduced the complexity of overlap calculation of the algorithm by eliminating the common projection plane and directly projecting the detector boundaries onto image voxel boundaries. To reduce the time required for calculating the overlap between the detector edges and image voxel boundaries, we have proposed a pre-accumulation technique to accumulate image intensities in perpendicular 2D image slabs (from a 3D image) before projection and after backprojection to ensure our DD kernels run faster in parallel GPU threads. For the implementation of our iterative MBIR technique we use a parallel multi-GPU version of the alternating minimization (AM) algorithm with penalized likelihood update. The time performance using our proposed reconstruction method with Siemens Sensation 16 patient scan data shows an average of 24 times speedup using a single TITAN X GPU and 74 times speedup using 3 TITAN X GPUs in parallel for combined projection and backprojection.

Computationally efficient error estimate for evaluation of regularization in photoacoustic tomography.

The model-based image reconstruction techniques for photoacoustic (PA) tomography require an explicit regularization. An error estimate (?2) minimization-based approach was proposed and developed for the determination of a regularization parameter for PA imaging. The regularization was used within Lanczos bidiagonalization framework, which provides the advantage of dimensionality reduction for a large system of equations. It was shown that the proposed method is computationally faster than the state-of-the-art techniques and provides similar performance in terms of quantitative accuracy in reconstructed images. It was also shown that the error estimate (?2) can also be utilized in determining a suitable regularization parameter for other popular techniques such as Tikhonov, exponential, and nonsmooth (?1 and total variation norm based) regularization methods.

Exponential filtering of singular values improves photoacoustic image reconstruction

Model-based image reconstruction techniques yield better quantitative accuracy in photoacoustic image reconstruction. In this work, an exponential filtering of singular values was proposed for carrying out the image reconstruction in photoacoustic tomography. The results were compared with widely popular Tikhonov regularization, time reversal, and the state of the art least-squares QR-based reconstruction algorithms for three digital phantom cases with varying signal-to-noise ratios of data. It was shown that exponential filtering provides superior photoacoustic images of better quantitative accuracy. Moreover, the proposed filtering approach was observed to be less biased toward the regularization parameter and did not come with any additional computational burden as it was implemented within the Tikhonov filtering framework. It was also shown that the standard Tikhonov filtering becomes an approximation to the proposed exponential filtering.

Patient-adapted reconstruction and acquisition dynamic imaging method (PARADIGM) for MRI

Dynamic magnetic resonance imaging (MRI) is a challenging problem because the MR data acquisition is often not fast enough to meet the combined spatial and temporal Nyquist sampling rate requirements. Current approaches to this problem include hardware-based acceleration of the acquisition, and model-based image reconstruction techniques. In this paper we propose an alternative approach, called PARADIGM, which adapts both the acquisition and reconstruction to the spatio-temporal characteristics of the imaged object. The approach is based on time-sequential sampling theory, addressing the problem of acquiring a spatio-temporal signal under the constraint that only a limited amount of data can be acquired at a time instant. PARADIGM identifies a model class for the particular imaged object using a scout MR scan or auxiliary data. This object-adapted model is then used to optimize MR data acquisition, such that the imaging constraints are met, acquisition speed requirements are minimized, essentially perfect reconstruction of any object in the model class is guaranteed, and the inverse problem of reconstructing the dynamic object has a condition number of one. We describe spatio-temporal object models for various dynamic imaging applications including cardiac imaging. We present the theory underlying PARADIGM and analyze its performance theoretically and numerically. We also propose a practical MR imaging scheme for 2D dynamic cardiac imaging based on the theory. For this application, PARADIGM is predicted to provide a 10–25 × acceleration compared to the optimal non-adaptive scheme. Finally we present generalized optimality criteria and extend the scheme to dynamic imaging with three spatial dimensions.