GPU based parallel framework for receiver coil sensitivity estimation in SENSE reconstruction

FPGA implementation of real-time SENSE reconstruction using pre-scan and Emaps sensitivities

In parallel magnetic resonance imaging (MRI), the quality of image reconstruction is very important. The signal to noise ratio (SNR) of reconstruction image would be obviously reduced with the increase of the acceleration factors because of the ill-posed problem in the sensitivity encoding (SENSE) reconstruction. Through in-depth analysis of the total variation (TV) regularized SENSE reconstruction model, an efficient Bregman iteration algorithm was introduced to obtain the optimal solution and improve the image reconstruction quality. In this paper, the simulation experiments were performed on the real cardiac data and brain data of MRI, respectively. The experimental results demonstrated that the proposed algorithm can alleviate the aliasing artifacts and has a better image reconstruction effect. Compared with the conventional TV regularized SENSE reconstruction algorithm, under the condition of the higher acceleration factors, the SNR has also been improved and the normalized mean squared error (NMSE) of reconstruction image is also decreased. This shows that the proposed iteration TV SENSE image reconstruction algorithm is effective and feasible.

The application of parallel magnetic resonance (MR) imaging is increasing as clinicians continue to strive for improved spatial and temporal resolution, benefits that arise from the use of fewer phase encodings during imaging. To reconstruct images, extra information is needed to map the spatial sensitivity of each coil element, which may be accomplished by acquiring a calibration image in one common implementation of parallel MR imaging. Although obtaining a quick calibration image is an efficient method for gathering this information, corruption of the image or disharmony with subsequent images may lead to errors in reconstruction. Although conventional MR imaging sequences may be employed with parallel MR imaging, the altered image reconstruction introduces several new artifacts and changes the appearance of conventional artifacts. The altered appearance of traditional artifacts may obscure the source of the problem, and, in some cases, the severity of artifacts associated with parallel MR imaging may be exacerbated, hindering image interpretation. Several artifacts arise in the context of parallel MR imaging, including both traditional artifacts and those associated with parallel MR imaging.

Parallel magnetic resonance imaging (pMRI) reduces MRI acquisition time, with Sensitivity Encoding (SENSE) being a widely used method that exploits coil sensitivity maps for efficient image reconstruction. However, SENSE can introduce aliasing and noise artifacts, especially at high acceleration factors. To address this limitation, we propose a deep learning‐based postprocessing framework that enhances SENSE‐reconstructed images using a Convolutional Autoencoder (CAE). The CAE is applied after the SENSE reconstruction to reduce artifacts and improve image quality, without modifying the original reconstruction pipeline. A dataset of 842 fully sampled anatomical images is used to simulate 8‐channel coil data, with both uniform and variable‐density (VD) undersampling applied at different acceleration factors. The CAE is trained on paired inputs (SENSE‐reconstructed images) and targets (fully sampled references) to learn the mapping from degraded to high‐quality images. Quantitative evaluation using Peak Signal‐to‐Noise Ratio (PSNR) and Normalized Mean Squared Error (NMSE), along with qualitative visual assessment, shows that the proposed CAE‐SENSE framework significantly improves image fidelity, particularly with variable‐density undersampling. These results demonstrate the potential of deep learning as a complementary tool to enhance conventional parallel imaging methods in accelerated MRI.

Both parallel Magnetic Resonance Imaging (pMRI) and Compressed Sensing (CS) are emerging techniques to accelerate conventional MRI by reducing the number of acquired data in the k-space. So far, first attempts to combine sensitivity encoding (SENSE) imaging in pMRI with CS have been proposed in the context of Cartesian trajectories. Here, we extend these approaches to non-Cartesian trajectories by jointly formulating the CS and SENSE recovery in a hybrid Fourier/wavelet framework and optimizing a convex but nons-mooth criterion. On anatomical MRI data, we show that HYR <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> PICS outperforms wavelet-based regularized SENSE reconstruction. Our results are also in agreement with the Transform Point Spread Function (TPSF) criterion that measures the degree of incoherence of k-space undersampling schemes.

Fast acquisition in magnetic resonance imaging (MRI) is considered in this paper. Often, fast acquisition is achieved using parallel imaging (pMRI) techniques. It has been shown recently that the combination of pMRI and compressed sensing (CS), which enables exact reconstruction of sparse or compressible signals from a small number of random measurements, can accelerate the speed of MRI acquisition because the number of measurements are much smaller than that by pMRI per se. Also recently in CS, chaos filters were designed to obtain chaotic measurements. This chaotic CS approach potentially offers simpler hardware implementation. In this paper, we combine chaotic CS and pMRI. However, instead of using chaos filters, the measurements are obtained by chaotically undersampling the k-space. MRI image reconstruction is then performed by using nonlinear conjugate gradient optimization. For pMRI, we use the well-known approach SENSE - sensitivity encoding -, which requires an estimation of the sensitivity maps. The performance of the proposed method is analyzed using the point spread function, the transform point spread function, and the reconstruction error measure.

Parallel magnetic resonance imaging (MRI) techniques use spatial information from arrays of radiofrequency (RF) detector coils to accelerate imaging. A number of parallel MRI techniques have been described in recent years, and numerous clinical applications are currently being explored. The advent of practical parallel imaging presents various challenges for image reconstruction and RF system design. Recent advances in tailored SiMultaneous Acquisition of Spatial Harmonics (SMASH) image reconstructions are summarized. These advances enable robust SMASH imaging in arbitrary image planes with a wide range of coil array geometries. A generalized formalism is described which may be used to understand the relations between SMASH and SENSE, to derive typical implementations of each as special cases, and to form hybrid techniques combining some of the advantages of both. Accurate knowledge of coil sensitivities is crucial for parallel MRI, and errors in calibration represent one of the most common and the most pernicious sources of error in parallel image reconstructions. As one example, motion of the patient and/or the coil array between the sensitivity reference scan and the accelerated acquisition can lead to calibration errors and reconstruction artifacts. Self-calibrating parallel MRI approaches that address this problem by eliminating the need for external sensitivity references are reviewed. The ultimate achievable signal-to-noise ratio (SNR) for parallel MRI studies is closely tied to the geometry and sensitivity patterns of the coil arrays used for spatial encoding. Several parallel imaging array designs that depart from the traditional model of overlapped adjacent loop elements are described.

Magnetic Resonance Imaging (MRI) is one of well-known imaging modalities because it is noninvasive, does not produce ionization and offers various contrast mechanisms for all kinds of tissues, especially soft tissues. Faster image acquisition is indispensable in current imaging technology, especially for dynamic object. Parallel MRI (pMRI) offers such possibility to faster acquire the reconstructed image by using several receiver coils with varying spatial sensitivity instead of a single coil with homogeneous sensitivity. This paper proposes a rotary phased array head coil owing to the phased array coil advantages for rapid imaging with parallel processing and reducing image artifact due to its larger field of view. The proposed array coils consist of eight independently single-coil, where each single-coil is made from some pieces of conductors onto which directly connected each other with twelve capacitors so thus the resonant frequency can be tuned precisely at 128 MHz for 3T MRI system. Each coil is circularly-arrayed and sequentially-activated to form a volume RF coil in order to achieve relatively homogeneous H-fields. The numerical results show good reflection coefficient at 128 MHz and relatively uniform H-field strength distribution. Moreover, the specific absorption rate (SAR) is relatively low inside the head phantom model up to 50W of input power.

Simultaneous Multi-Slice (SMS) magnetic resonance imaging (MRI) is a rapidly evolving technique for increasing imaging speed. Controlled aliasing techniques utilize periodic undersampling patterns to help mitigate the loss in signal-to-noise ratio (SNR) in SMS MRI. To evaluate the performance of different undersampling patterns, a quantitative description of the image SNR loss is needed. Additionally, eddy current effects in echo planar imaging (EPI) lead to slice-specific Nyquist ghosting artifacts. These artifacts cannot be accurately corrected for each individual slice before or after slice-unaliasing. In this work, we propose a hybrid-space sensitivity encoding (SENSE) reconstruction framework for SMS MRI by adopting a three-dimensional representation of the SMS acquisition. Analytical SNR loss maps are derived for SMS acquisitions with arbitrary phase encoding undersampling patterns. Moreover, we propose a matrix-decoding correction method that corrects the slice-specific Nyquist ghosting artifacts in SMS EPI acquisitions. Brain images demonstrate that the proposed hybrid-space SENSE reconstruction generates images with comparable quality to commonly used split-slice-generalized autocalibrating partially parallel acquisition reconstruction. The analytical SNR loss maps agree with those calculated by a Monte Carlo based method, but require less computation time for high quality maps. The analytical maps enable a fair comparison between the performances of coherent and incoherent SMS undersampling patterns. Phantom and brain SMS EPI images show that the matrix-decoding method performs better than the single-slice and slice-averaged Nyquist ghosting correction methods under the hybrid-space SENSE reconstruction framework.

The application of partially parallel imaging techniques to regular clinical MRI studies has brought about the benefit of significantly faster acquisitions but at the cost of amplified and non-uniformly distributed noise, especially, for high acceleration factors. In this work, denoising of the images reconstructed by the sensitivity- encoding (SENSE) algorithm is presented. To efficiently remove noise and simultaneously preserve the image details, weighted total variation smoothing with weighted edge restoration has been developed. Denoising is guided by the g-factor of the SENSE reconstruction and image gradient that is representative of edge information. An automatic iteration termination scheme is proposed to balance denoising and edge preservation, and reduce the difficulty of parameter decision in conventional approaches. The proposed g-factor and gradient weighted denoising with edge restoration (g-DENOISER) method retains the image details better than the denoising techniques guided by only the g-factor and simultaneously suppresses the noise stronger than the techniques where g-factor was not used to adaptively adjust the denoiser parameters according to local noise characteristics. It can be effectively used to reduce noise in images acquired using high acceleration factors and reconstructed using SENSE.

High-resolution MRI using compressed sensing-sensitivity encoding (CS-SENSE) for patients with suspected neurovascular compression syndrome: comparison with the conventional SENSE parallel acquisition technique

In order to improve the image quality of BLADE magnetic resonance imaging (MRI) using the index tensor solvers and to evaluate MRI image quality in a clinical setting, we implemented BLADE MRI reconstructions using two tensor solvers (the least-squares solver and the L1 total-variation regularized least absolute deviation (L1TV-LAD) solver) on a graphics processing unit (GPU). The BLADE raw data were prospectively acquired and presented in random order before being assessed by two independent radiologists. Evaluation scores were examined for consistency and then by repeated measures analysis of variance (ANOVA) to identify the superior algorithm. The simulation showed the structural similarity index (SSIM) of various tensor solvers ranged between 0.995 and 0.999. Inter-reader reliability was high (Intraclass correlation coefficient (ICC) = 0.845, 95% confidence interval: 0.817, 0.87). The image score of L1TV-LAD was significantly higher than that of vendor-provided image and the least-squares method. The image score of the least-squares method was significantly lower than that of the vendor-provided image. No significance was identified in L1TV-LAD with a regularization strength of 0.4–1.0. The L1TV-LAD with a regularization strength of 0.4–0.7 was found consistently better than least-squares and vendor-provided reconstruction in BLADE MRI with a SENSitivity Encoding (SENSE) factor of 2. This warrants further development of the integrated computing system with the scanner.

With the successful application of deep learning to magnetic resonance (MR) imaging, parallel imaging techniques based on neural networks have attracted wide attention. However, in the absence of high-quality, fully sampled datasets for training, the performance of these methods is limited. And the interpretability of models is not strong enough. To tackle this issue, this paper proposes a Physics-bAsed unsupeRvised Contrastive rEpresentation Learning (PARCEL) method to speed up parallel MR imaging. Specifically, PARCEL has a parallel framework to contrastively learn two branches of model-based unrolling networks from augmented undersampled multi-coil k-space data. A sophisticated co-training loss with three essential components has been designed to guide the two networks in capturing the inherent features and representations for MR images. And the final MR image is reconstructed with the trained contrastive networks. PARCEL was evaluated on two vivo datasets and compared to five state-of-the-art methods. The results show that PARCEL is able to learn essential representations for accurate MR reconstruction without relying on fully sampled datasets. The code will be made available at https://github.com/ternencewu123/PARCEL.

To evaluate partially parallel three-dimensional (3D) magnetic resonance (MR) imaging for assessment of regional lung perfusion in healthy volunteers and patients suspected of having lung cancer or metastasis. Seven healthy volunteers and 20 patients suspected of having lung cancer or metastasis were examined with 3D gradient-echo MR imaging with partially parallel image acquisitions (fast low-angle shot 3D imaging; repetition time msec/echo time msec, 1.9/0.8; flip angle, 40 degrees; acceleration factor, two; number of reference k-space lines for calibration, 24; field of view, 500 x 440 mm; matrix, 256 x 123; slab thickness, 160 mm; number of partitions, 32; voxel size, 3.6 x 2.0 x 5.0 mm(3); acquisition time, 1.5 seconds) after administration of 0.1 mmol/kg of gadobenate dimeglumine. In volunteers, 3D MR perfusion data sets were assessed for topographic and temporal distribution of regional lung perfusion. Sensitivity, specificity, accuracy, and positive and negative predictive values for perfusion MR imaging for detecting perfusion abnormalities in patients were calculated, with conventional radionuclide perfusion scintigraphy as the standard of reference. Interobserver and intermodality agreement was determined by using kappa statistics. Topographic analysis of lung perfusion in volunteers revealed a significantly higher signal-to-noise ratio (SNR) of up to 327% in gravity-dependent lung areas. Temporal analysis similarly revealed much shorter lag time to peak enhancement in gravity-dependent lung areas. In patients, perfusion MR imaging achieved high sensitivity (88%-94%), specificity (100%), and accuracy (90%-95%) for detection of perfusion abnormalities. Interobserver agreement (kappa = 0.86) was very good and intermodality agreement (kappa = 0.69-0.83) was good to very good for detection of perfusion defects. A significant difference (P <.0001) in SNR was observed between normally perfused lung (14 +/- 7 [SD]) and perfusion defects (7 +/- 4) in patients. Partially parallel MR imaging with high spatial and temporal resolution allows assessment of regional lung perfusion and has high diagnostic accuracy for detecting perfusion abnormalities.

Dual double arterial phase dynamic MR imaging with sensitivity encoding (SENSE): which is better for diagnosing hypervascular hepatocellular carcinomas, in-phase or opposed-phase imaging?