Accelerating the nonequispaced fast Fourier transform on commodity graphics hardware

Rapid CNN-based needle localization for automatic slice alignment in MR-guided interventions using 3D undersampled radial white-marker imaging

BACKGROUND

In MR-guided in-bore percutaneous needle interventions, typically 2D interactive real-time imaging is used for navigating the needle into the target. Misaligned 2D imaging planes can result in losing visibility of the needle in the 2D images, which impedes successful targeting. Necessary iterative manual slice adjustment can prolong interventional workflows. Therefore, rapid automatic alignment of the imaging planes with the needle would be preferable to improve such workflows.

PURPOSE

To investigate rapid 3D localization of needles in MR-guided interventions via a convolutional neural network (CNN)-based localization algorithm using an undersampled white-marker contrast acquisition for the purpose of automatic imaging slice alignment.

METHODS

A radial 3D rf-spoiled gradient echo MR pulse sequence with white-marker encoding was implemented and a CNN-based localization algorithm was employed to extract position and orientation of an aspiration needle from the undersampled white-marker images. The CNN was trained using porcine tissue phantoms (257 needle trajectories, four-fold data augmentation, 90%/10% split into training and validation dataset). Achievable localization times and accuracy were evaluated retrospectively in an ex vivo study (109 needle trajectories) for a range of needle orientations between 78° and 90° relative to the B0 field. A proof-of-concept in vivo experiment was performed in two porcine animal models and feasibility of automatic imaging slice alignment was evaluated retrospectively.

RESULTS

Ex vivo needle localization was achieved with a median localization accuracy of 1.9 mm (distance needle tip to detected needle axis) and a median angular deviation of 2.6° for needle orientations between 86° and 90° to the B0 field from fully sampled WM images (resolution of (4 mm)3, 6434 acquired radial k-space spokes, acquisition time of 80.4 s) in a field-of-view of (256 mm)3. Localization accuracy decreased with increasing undersampling and needle trajectory increasingly aligned with B0. For needle orientations between 86° and 90° to the B0 field, a highly accelerated acquisition of only 32 k-space spokes (acquisition time of 0.4 s) yielded a median localization accuracy of 3.1 mm and a median angular deviation of 4.7°. For needle orientations between 78° and 82° to the B0 field, a median accuracy and angular deviation of 3.5 mm and 6.8° could still be achieved with 64 sampled spokes (acquisition time of 0.8 s). In vivo, a localization accuracy of 1.4 mm and angular deviation of 3.4° was achieved sampling 32 k-space spokes (acquisition time of 0.48 s) with the needle oriented at 87.7° to the B0 field. For a needle oriented at 77.6° to the B0 field, localization accuracy of 5.3 mm and angular deviation of 6.8° were still achieved sampling 128 k-space spokes (acquisition time of 1.92 s), allowing for retrospective slice alignment.

CONCLUSION

The investigated approach enables passive biopsy needle localization in 3D. Acceleration of the localization to real-time applicability is feasible for needle orientations approximately perpendicular to B0. The method can potentially facilitate MR-guided needle interventions by enabling automatic imaging slice alignment with the needle.

Low-rank motion correction for accelerated free-breathing first-pass myocardial perfusion imaging

PURPOSE

Develop a novel approach for accelerated 2D free-breathing myocardial perfusion via low-rank motion-corrected (LRMC) reconstructions.

METHODS

Myocardial perfusion imaging requires high spatial and temporal resolution, despite scan time constraints. Here, we incorporate LRMC models into the reconstruction-encoding operator, together with high-dimensionality patch-based regularization, to produce high quality, motion-corrected myocardial perfusion series from free-breathing acquisitions. The proposed framework estimates beat-to-beat nonrigid respiratory (and any other incidental) motion and the dynamic contrast subspace from the actual acquired data, which are then incorporated into the proposed LRMC reconstruction. LRMC was compared with iterative SENSitivity Encoding (SENSE) (itSENSE) and low-rank plus sparse (LpS) reconstruction in 10 patients based on image-quality scoring and ranking by two clinical expert readers.

RESULTS

LRMC achieved significantly improved results relative to itSENSE and LpS in terms of image sharpness, temporal coefficient of variation, and expert reader evaluation. Left ventricle image sharpness was approximately 75%, 79%, and 86% for itSENSE, LpS and LRMC, respectively, indicating improved image sharpness for the proposed approach. Corresponding temporal coefficient of variation results were 23%, 11% and 7%, demonstrating improved temporal fidelity of the perfusion signal with the proposed LRMC. Corresponding clinical expert reader scores (1-5, from poor to excellent image quality) were 3.3, 3.9 and 4.9, demonstrating improved image quality with the proposed LRMC, in agreement with the automated metrics.

CONCLUSION

LRMC produces motion-corrected myocardial perfusion in free-breathing acquisitions with substantially improved image quality when compared with iterative SENSE and LpS reconstructions.

Multicenter Repeatability and Reproducibility of MR Fingerprinting in Phantoms and in Prostatic Tissue

PURPOSE

To evaluate multicenter repeatability and reproducibility of T1 and T2 maps generated using MR fingerprinting (MRF) in the International Society for Magnetic Resonance in Medicine/National Institute of Standards and Technology MRI system phantom and in prostatic tissues.

METHODS

MRF experiments were performed on 5 different 3 Tesla MRI scanners at 3 different institutions: University Hospitals Cleveland Medical Center (Cleveland, OH), Brigham and Women's Hospital (Boston, MA) in the United States, and Diagnosticos da America (Rio de Janeiro, RJ) in Brazil. Raw MRF data were reconstructed using a Gadgetron-based MRF online reconstruction pipeline to yield quantitative T1 and T2 maps. The repeatability of T1 and T2 values over 6 measurements in the International Society for Magnetic Resonance in Medicine/National Institute of Standards and Technology MRI system phantom was assessed to demonstrate intrascanner variation. The reproducibility between the 4 clinical scanners was assessed to demonstrate interscanner variation. The same-day test-retest normal prostate mean T1 and T2 values from peripheral zone and transitional zone were also compared using the intraclass correlation coefficient and Bland-Altman analysis.

RESULTS

The intrascanner variation of values measured using MRF was less than 2% for T1 and 4.7% for T2 for relaxation values, within the range of 307.7 to 2360 ms for T1 and 19.1 to 248.5 ms for T2 . Interscanner measurements showed that the T1 variation was less than 4.9%, and T2 variation was less than 8.1% between multicenter scanners. Both T1 and T2 values in in vivo prostatic tissue demonstrated high test-retest reliability (intraclass correlation coefficient > 0.92) and strong linear correlation (R2  > 0.840).

CONCLUSION

Prostate MRF measurements of T1 and T2 are repeatable and reproducible between MRI scanners at different centers on different continents for the above measurement ranges.

A System for Real-Time, Online Mixed-Reality Visualization of Cardiac Magnetic Resonance Images

Image-guided cardiovascular interventions are rapidly evolving procedures that necessitate imaging systems capable of rapid data acquisition and low-latency image reconstruction and visualization. Compared to alternative modalities, Magnetic Resonance Imaging (MRI) is attractive for guidance in complex interventional settings thanks to excellent soft tissue contrast and large fields-of-view without exposure to ionizing radiation. However, most clinically deployed MRI sequences and visualization pipelines exhibit poor latency characteristics, and spatial integration of complex anatomy and device orientation can be challenging on conventional 2D displays. This work demonstrates a proof-of-concept system linking real-time cardiac MR image acquisition, online low-latency reconstruction, and a stereoscopic display to support further development in real-time MR-guided intervention. Data are acquired using an undersampled, radial trajectory and reconstructed via parallelized through-time radial generalized autocalibrating partially parallel acquisition (GRAPPA) implemented on graphics processing units. Images are rendered for display in a stereoscopic mixed-reality head-mounted display. The system is successfully tested by imaging standard cardiac views in healthy volunteers. Datasets comprised of one slice (46 ms), two slices (92 ms), and three slices (138 ms) are collected, with the acquisition time of each listed in parentheses. Images are displayed with latencies of 42 ms/frame or less for all three conditions. Volumetric data are acquired at one volume per heartbeat with acquisition times of 467 ms and 588 ms when 8 and 12 partitions are acquired, respectively. Volumes are displayed with a latency of 286 ms or less. The faster-than-acquisition latencies for both planar and volumetric display enable real-time 3D visualization of the heart.

USING FAST FOURIER TRANSFORM AND POLYNOMIAL FITTING ON DORSAL FOOT KINEMATICS DATA TO IDENTIFY SIMULATED ANKLE SPRAIN MOTIONS FROM COMMON SPORTING MOTIONS

Ankle sprain is very common in sports, and a commonly suggested etiology is the delayed peroneal muscle reaction time. Recent studies showed the successful attempts to deliver electrical stimulation to the peroneal muscles externally to initiate contraction before it could react, however, the success relies on a workable method to detect ankle sprain injury in time. This study presented a fast Fourier transform and polynomial fitting method with dorsal foot kinematics data for quick ankle sprain detection. Five males performed 100 simulated ankle sprain and 250 common sporting motion trials. Eight gyrometers recorded the three-dimensional angular velocities at 500[Formula: see text]Hz. Data were trimmed with a 0.11[Formula: see text]s window size, the suggested duration of preinjury phase in ankle sprain, and were transformed from time to frequency domain by fast Fourier transform and fitted with a fifth-order polynomial. First-order coefficients from polynomial fitting on frequency space were obtained. The method achieved 97.0% sensitivity and 91.4% specificity in identifying simulated sprains, vertical jump–landing, cutting, stepping-down, running, and walking motions, with vertical jump–landing excluded due to its relatively low specificity (67.3%). The method can be used to detect ankle sprain in sports with mainly floor movements and minimal vertical jump–landing motion.

Accelerating the computation for real-time application of the sinc function using graphics processing units

In magnetic resonance imaging, the fidelity of image reconstruction is an important criterion. It has been suggested that the infinite-extent sinc kernel is the ideal interpolation kernel for ensuring the reconstruction quality of non-Cartesian trajectories. However, the application of the sinc function has been limited owing to its computational overheads. Recently, graphics processing units (GPUs) have been employed as fast computation tools because of their efficient and versatile parallel computation abilities. We implemented an accelerated convolution function with the sinc kernel using GPUs computing and evaluated the reconstruction performance. The computation time was significantly improved: Computation using the proposed method was approximately 270 times faster than that on a central processing unit (CPU) and approximately 4.6 times faster than that on a CPU optimized by level-3 Basic Linear Algebra Subprograms. The images reconstructed using the fast sinc function exhibited no adverse errors at all matrix sizes (resolutions). The total reconstruction time was approximately 0.3–3 s for all matrices, indicating that the sinc function could be a practical option for image reconstruction. Ultimately, its application would present a fundamental improvement to the performance of image reconstruction, and the GPU implementation of the convolution function with the sinc kernel could resolve various challenges in image data processing.

snapMRF: GPU-accelerated magnetic resonance fingerprinting dictionary generation and matching using extended phase graphs

PURPOSE

Magnetic resonance fingerprinting (MRF) is a state-of-the-art quantitative MRI technique with a computationally demanding reconstruction process, the accuracy of which depends on the accuracy of the signal model employed. Having a fast, validated, open-source MRF reconstruction would improve the dependability and accuracy of clinical applications of MRF.

METHODS

We parallelized both dictionary generation and signal matching on the GPU by splitting the simulation and matching of dictionary atoms across threads. Signal generation was modeled using both Bloch equation simulation and the extended phase graph (EPG) formalism. Unit tests were implemented to ensure correctness. The new package, snapMRF, was tested with a calibration phantom and an in vivo brain.

RESULTS

Compared with other online open-source packages, dictionary generation was accelerated by 10-1000× and signal matching by 10-100×. On a calibration phantom, T₁ and T₂ values were measured with relative errors that were nearly identical to those from existing packages when using the same sequence and dictionary configuration, but errors were much lower when using variable sequences that snapMRF supports but that competitors do not.

CONCLUSION

Our open-source package snapMRF was significantly faster and retrieved accurate parameters, possibly enabling real-time parameter map generation for small dictionaries. Further refinements to the acquisition scheme and dictionary setup could improve quantitative accuracy.

Trajectory optimized NUFFT: Faster non-Cartesian MRI reconstruction through prior knowledge and parallel architectures

Purpose

The non‐uniform fast Fourier transform (NUFFT) involves interpolation of non‐uniformly sampled Fourier data onto a Cartesian grid, an interpolation that is slowed by complex, non‐local data access patterns. A faster NUFFT would increase the clinical relevance of the plethora of advanced non‐Cartesian acquisition methods.

Methods

Here we customize the NUFFT procedure for a radial trajectory and GPU architecture to eliminate the bottlenecks encountered when allowing for arbitrary trajectories and hardware. We call the result TRON, for TRajectory Optimized NUFFT. We benchmark the speed and accuracy TRON on a Shepp‐Logan phantom and on whole‐body continuous golden‐angle radial MRI.

Results

TRON was 6–30× faster than the closest competitor, depending on test data set, and was the most accurate code tested.

Conclusions

Specialization of the NUFFT algorithm for a particular trajectory yielded significant speed gains. TRON can be easily extended to other trajectories, such as spiral and PROPELLER. TRON can be downloaded at http://github.com/davidssmith/TRON.

Accelerated magnetic resonance fingerprinting using soft-weighted key-hole (MRF-SOHO)

Object

To develop a novel approach for highly accelerated Magnetic Resonance Fingerprinting (MRF) acquisition.

Materials and methods

The proposed method combines parallel imaging, soft-gating and key-hole approaches to highly accelerate MRF acquisition. Slowly varying flip angles (FA), commonly used during MRF acquisition, lead to a smooth change in the signal contrast of consecutive time-point images. This assumption enables sharing of high frequency data between different time-points, similar to what is done in some dynamic MR imaging methods such as key-hole. The proposed approach exploits this information using a SOft-weighted key-HOle (MRF-SOHO) reconstruction to achieve high acceleration factors and/or increased resolution without compromising image quality or increasing scan time. MRF-SOHO was validated on a standard T₁/T₂ phantom and in in-vivo brain acquisitions reconstructing T₁, T₂ and proton density parametric maps.

Results

Accelerated MRF-SOHO using less data per time-point and less time-point images enabled a considerable reduction in scan time (up to 4.6x), while obtaining similar T₁ and T₂ accuracy and precision when compared to zero-filled MRF reconstruction. For the same number of spokes and time-points, the proposed method yielded an enhanced performance in quantifying parameters than the zero-filled MRF reconstruction, which was verified with 2, 1 and 0.7 (sub-millimetre) resolutions.

Conclusion

The proposed MRF-SOHO enabled a 4.6x scan time reduction for an in-plane spatial resolution of 2x2 mm² when compared to zero-filled MRF and enabled sub-millimetric (0.7x0.7 mm²) resolution MRF.

A survey of GPU-based acceleration techniques in MRI reconstructions

Image reconstruction in magnetic resonance imaging (MRI) clinical applications has become increasingly more complicated. However, diagnostic and treatment require very fast computational procedure. Modern competitive platforms of graphics processing unit (GPU) have been used to make high-performance parallel computations available, and attractive to common consumers for computing massively parallel reconstruction problems at commodity price. GPUs have also become more and more important for reconstruction computations, especially when deep learning starts to be applied into MRI reconstruction. The motivation of this survey is to review the image reconstruction schemes of GPU computing for MRI applications and provide a summary reference for researchers in MRI community.

A review of GPU-based medical image reconstruction

Tomographic image reconstruction is a computationally demanding task, even more so when advanced models are used to describe a more complete and accurate picture of the image formation process. Such advanced modeling and reconstruction algorithms can lead to better images, often with less dose, but at the price of long calculation times that are hardly compatible with clinical workflows. Fortunately, reconstruction tasks can often be executed advantageously on Graphics Processing Units (GPUs), which are exploited as massively parallel computational engines. This review paper focuses on recent developments made in GPU-based medical image reconstruction, from a CT, PET, SPECT, MRI and US perspective. Strategies and approaches to get the most out of GPUs in image reconstruction are presented as well as innovative applications arising from an increased computing capacity. The future of GPU-based image reconstruction is also envisioned, based on current trends in high-performance computing.

Technical Note: Evaluation of pre-reconstruction interpolation methods for iterative reconstruction of radial k-space data

PURPOSE

To evaluate the use of three different pre-reconstruction interpolation methods to convert non-Cartesian k-space data to Cartesian samples such that iterative reconstructions can be performed more simply and more rapidly.

METHODS

Phantom as well as cardiac perfusion radial datasets were reconstructed by four different methods. Three of the methods used pre-reconstruction interpolation once followed by a fast Fourier transform (FFT) at each iteration. The methods were: bilinear interpolation of nearest-neighbor points (BINN), 3-point interpolation, and a multi-coil interpolator called GRAPPA Operator Gridding (GROG). The fourth method performed a full non-Uniform FFT (NUFFT) at each iteration. An iterative reconstruction with spatiotemporal total variation constraints was used with each method. Differences in the images were quantified and compared.

RESULTS

The GROG multicoil interpolation, the 3-point interpolation, and the NUFFT-at-each-iteration approaches produced high quality images compared to BINN, with the GROG-derived images having the fewest streaks among the three preinterpolation approaches. However, all reconstruction methods produced approximately equal results when applied to perfusion quantitation tasks. Pre-reconstruction interpolation gave approximately an 83% reduction in reconstruction time.

CONCLUSION

Image quality suffers little from using a pre-reconstruction interpolation approach compared to the more accurate NUFFT-based approach. GROG-based pre-reconstruction interpolation appears to offer the best compromise by using multicoil information to perform the interpolation to Cartesian sample points prior to image reconstruction. Speed gains depend on the implementation and relatively standard optimizations on a MATLAB platform result in preinterpolation speedups of ~ 6 compared to using NUFFT at every iteration, reducing the reconstruction time from around 42 min to 7 min.

Accelerated motion corrected three-dimensional abdominal MRI using total variation regularized SENSE reconstruction

PURPOSE

Develop a nonrigid motion corrected reconstruction for highly accelerated free-breathing three-dimensional (3D) abdominal images without external sensors or additional scans.

METHODS

The proposed method accelerates the acquisition by undersampling and performs motion correction directly in the reconstruction using a general matrix description of the acquisition. Data are acquired using a self-gated 3D golden radial phase encoding trajectory, enabling a two stage reconstruction to estimate and then correct motion of the same data. In the first stage total variation regularized iterative SENSE is used to reconstruct highly undersampled respiratory resolved images. A nonrigid registration of these images is performed to estimate the complex motion in the abdomen. In the second stage, the estimated motion fields are incorporated in a general matrix reconstruction, which uses total variation regularization and incorporates k-space data from multiple respiratory positions. The proposed approach was tested on nine healthy volunteers and compared against a standard gated reconstruction using measures of liver sharpness, gradient entropy, visual assessment of image sharpness and overall image quality by two experts.

RESULTS

The proposed method achieves similar quality to the gated reconstruction with nonsignificant differences for liver sharpness (1.18 and 1.00, respectively), gradient entropy (1.00 and 1.00), visual score of image sharpness (2.22 and 2.44), and visual rank of image quality (3.33 and 3.39). An average reduction of the acquisition time from 102 s to 39 s could be achieved with the proposed method.

CONCLUSION

In vivo results demonstrate the feasibility of the proposed method showing similar image quality to the standard gated reconstruction while using data corresponding to a significantly reduced acquisition time. Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance.

Magnetic Resonance Sequences and Rapid Acquisition for MR-Guided Interventions

Interventional MR uses rapid imaging to guide diagnostic and therapeutic procedures. One of the attractions of MR-guidance is the abundance of inherent contrast mechanisms available. Dynamic procedural guidance with real-time imaging has pushed the limits of MR technology, demanding rapid acquisition and reconstruction paired with interactive control and device visualization. This article reviews the technical aspects of real-time MR sequences that enable MR-guided interventions.

A review of 3D first-pass, whole-heart, myocardial perfusion cardiovascular magnetic resonance

A comprehensive review is undertaken of the methods available for 3D whole-heart first-pass perfusion (FPP) and their application to date, with particular focus on possible acceleration techniques. Following a summary of the parameters typically desired of 3D FPP methods, the review explains the mechanisms of key acceleration techniques and their potential use in FPP for attaining 3D acquisitions. The mechanisms include rapid sequences, non-Cartesian k-space trajectories, reduced k-space acquisitions, parallel imaging reconstructions and compressed sensing. An attempt is made to explain, rather than simply state, the varying methods with the hope that it will give an appreciation of the different components making up a 3D FPP protocol. Basic estimates demonstrating the required total acceleration factors in typical 3D FPP cases are included, providing context for the extent that each acceleration method can contribute to the required imaging speed, as well as potential limitations in present 3D FPP literature. Although many 3D FPP methods are too early in development for the type of clinical trials required to show any clear benefit over current 2D FPP methods, the review includes the small but growing quantity of clinical research work already using 3D FPP, alongside the more technical work. Broader challenges concerning FPP such as quantitative analysis are not covered, but challenges with particular impact on 3D FPP methods, particularly with regards to motion effects, are discussed along with anticipated future work in the field.

An augmented Lagrangian based compressed sensing reconstruction for non-Cartesian magnetic resonance imaging without gridding and regridding at every iteration

Background

Non-Cartesian trajectories are used in a variety of fast imaging applications, due to the incoherent image domain artifacts they create when undersampled. While the gridding technique is commonly utilized for reconstruction, the incoherent artifacts may be further removed using compressed sensing (CS). CS reconstruction is typically done using conjugate-gradient (CG) type algorithms, which require gridding and regridding to be performed at every iteration. This leads to a large computational overhead that hinders its applicability.

Methods

We sought to develop an alternative method for CS reconstruction that only requires two gridding and one regridding operation in total, irrespective of the number of iterations. This proposed technique is evaluated on phantom images and whole-heart coronary MRI acquired using 3D radial trajectories, and compared to conventional CS reconstruction using CG algorithms in terms of quantitative vessel sharpness, vessel length, computation time, and convergence rate.

Results

Both CS reconstructions result in similar vessel length (P = 0.30) and vessel sharpness (P = 0.62). The per-iteration complexity of the proposed technique is approximately 3-fold lower than the conventional CS reconstruction (17.55 vs. 52.48 seconds in C++). Furthermore, for in-vivo datasets, the convergence rate of the proposed technique is faster (60±13 vs. 455±320 iterations) leading to a ∼23-fold reduction in reconstruction time.

Conclusions

The proposed reconstruction provides images of similar quality to the conventional CS technique in terms of removing artifacts, but at a much lower computational complexity.

Image reconstruction: an overview for clinicians

Image reconstruction plays a critical role in the clinical use of magnetic resonance imaging (MRI). The MRI raw data is not acquired in image space and the role of the image reconstruction process is to transform the acquired raw data into images that can be interpreted clinically. This process involves multiple signal processing steps that each have an impact on the image quality. This review explains the basic terminology used for describing and quantifying image quality in terms of signal-to-noise ratio and point spread function. In this context, several commonly used image reconstruction components are discussed. The image reconstruction components covered include noise prewhitening for phased array data acquisition, interpolation needed to reconstruct square pixels, raw data filtering for reducing Gibbs ringing artifacts, Fourier transforms connecting the raw data with image space, and phased array coil combination. The treatment of phased array coils includes a general explanation of parallel imaging as a coil combination technique. The review is aimed at readers with no signal processing experience and should enable them to understand what role basic image reconstruction steps play in the formation of clinical images and how the resulting image quality is described.

Algebraic reconstruction technique for parallel imaging reconstruction of undersampled radial data: application to cardiac cine

PURPOSE

To investigate algebraic reconstruction technique (ART) for parallel imaging reconstruction of radial data, applied to accelerated cardiac cine.

METHODS

A graphics processing unit (GPU)-accelerated ART reconstruction was implemented and applied to simulations, point spread functions and in 12 subjects imaged with radial cardiac cine acquisitions. Cine images were reconstructed with radial ART at multiple undersampling levels (192 Nr × Np  = 96 to 16). Images were qualitatively and quantitatively analyzed for sharpness and artifacts, and compared to filtered back-projection, and conjugate gradient SENSE.

RESULTS

Radial ART provided reduced artifacts and mainly preserved spatial resolution, for both simulations and in vivo data. Artifacts were qualitatively and quantitatively less with ART than filtered back-projection using 48, 32, and 24 Np , although filtered back-projection provided quantitatively sharper images at undersampling levels of 48-24 Np (all P < 0.05). Use of undersampled radial data for generating auto-calibrated coil-sensitivity profiles resulted in slightly reduced quality. ART was comparable to conjugate gradient SENSE. GPU-acceleration increased ART reconstruction speed 15-fold, with little impact on the images.

CONCLUSION

GPU-accelerated ART is an alternative approach to image reconstruction for parallel radial MR imaging, providing reduced artifacts while mainly maintaining sharpness compared to filtered back-projection, as shown by its first application in cardiac studies.

Correction of excitation profile in Zero Echo Time (ZTE) imaging using quadratic phase-modulated RF pulse excitation and iterative reconstruction

Zero-echo Time (ZTE) imaging is a promising technique for magnetic resonance imaging (MRI) of short-T2 tissue nuclei in tissues. A problem inherent to the method currently hindering its translation to the clinic is the presence of a spatial encoding gradient during excitation, which causes the hard pulse to become spatially selective, resulting in blurring and shadow artifacts in the image. While shortening radio-frequency (RF) pulse duration alleviates this problem the resulting elevated RF peak power and specific absorption rate (SAR) in practice impede such a solution. In this work, an approach is described to correct the artifacts by applying quadratic phase-modulated RF excitation and iteratively solving an inverse problem formulated from the signal model of ZTE imaging. A simple pulse sequence is also developed to measure the excitation profile of the RF pulse. Results from simulations, phantom and in vivo studies, demonstrate the effectiveness of the method in correcting image artifacts caused by inhomogeneous excitation. The proposed method may contribute toward establishing ZTE MRI as a routine 3D pulse sequence for imaging protons and other nuclei with quasi solid-state behavior on clinical scanners.

Distributed MRI reconstruction using Gadgetron-based cloud computing

PURPOSE

To expand the open source Gadgetron reconstruction framework to support distributed computing and to demonstrate that a multinode version of the Gadgetron can be used to provide nonlinear reconstruction with clinically acceptable latency.

METHODS

The Gadgetron framework was extended with new software components that enable an arbitrary number of Gadgetron instances to collaborate on a reconstruction task. This cloud-enabled version of the Gadgetron was deployed on three different distributed computing platforms ranging from a heterogeneous collection of commodity computers to the commercial Amazon Elastic Compute Cloud. The Gadgetron cloud was used to provide nonlinear, compressed sensing reconstruction on a clinical scanner with low reconstruction latency (eg, cardiac and neuroimaging applications).

RESULTS

The proposed setup was able to handle acquisition and 11 -SPIRiT reconstruction of nine high temporal resolution real-time, cardiac short axis cine acquisitions, covering the ventricles for functional evaluation, in under 1 min. A three-dimensional high-resolution brain acquisition with 1 mm(3) isotropic pixel size was acquired and reconstructed with nonlinear reconstruction in less than 5 min.

CONCLUSION

A distributed computing enabled Gadgetron provides a scalable way to improve reconstruction performance using commodity cluster computing. Nonlinear, compressed sensing reconstruction can be deployed clinically with low image reconstruction latency.

Real-time imaging with radial GRAPPA: Implementation on a heterogeneous architecture for low-latency reconstructions

Combination of non-Cartesian trajectories with parallel MRI permits to attain unmatched acceleration rates when compared to traditional Cartesian MRI during real-time imaging. However, computationally demanding reconstructions of such imaging techniques, such as k-space domain radial generalized auto-calibrating partially parallel acquisitions (radial GRAPPA) and image domain conjugate gradient sensitivity encoding (CG-SENSE), lead to longer reconstruction times and unacceptable latency for online real-time MRI on conventional computational hardware. Though CG-SENSE has been shown to work with low-latency using a general purpose graphics processing unit (GPU), to the best of our knowledge, no such effort has been made for radial GRAPPA. Radial GRAPPA reconstruction, which is robust even with highly undersampled acquisitions, is not iterative, requiring only significant computation during initial calibration while achieving good image quality for low-latency imaging applications. In this work, we present a very fast, low-latency, reconstruction framework based on a heterogeneous system using multi-core CPUs and GPUs. We demonstrate an implementation of radial GRAPPA that permits reconstruction times on par with or faster than acquisition of highly accelerated datasets in both cardiac and dynamic musculoskeletal imaging scenarios. Acquisition and reconstruction times are reported.

MRISIMUL: a GPU-based parallel approach to MRI simulations

A new step-by-step comprehensive MR physics simulator (MRISIMUL) of the Bloch equations is presented. The aim was to develop a magnetic resonance imaging (MRI) simulator that makes no assumptions with respect to the underlying pulse sequence and also allows for complex large-scale analysis on a single computer without requiring simplifications of the MRI model. We hypothesized that such a simulation platform could be developed with parallel acceleration of the executable core within the graphic processing unit (GPU) environment. MRISIMUL integrates realistic aspects of the MRI experiment from signal generation to image formation and solves the entire complex problem for densely spaced isochromats and for a densely spaced time axis. The simulation platform was developed in MATLAB whereas the computationally demanding core services were developed in CUDA-C. The MRISIMUL simulator imaged three different computer models: a user-defined phantom, a human brain model and a human heart model. The high computational power of GPU-based simulations was compared against other computer configurations. A speedup of about 228 times was achieved when compared to serially executed C-code on the CPU whereas a speedup between 31 to 115 times was achieved when compared to the OpenMP parallel executed C-code on the CPU, depending on the number of threads used in multithreading (2-8 threads). The high performance of MRISIMUL allows its application in large-scale analysis and can bring the computational power of a supercomputer or a large computer cluster to a single GPU personal computer.

Medical image processing on the GPU - past, present and future

Graphics processing units (GPUs) are used today in a wide range of applications, mainly because they can dramatically accelerate parallel computing, are affordable and energy efficient. In the field of medical imaging, GPUs are in some cases crucial for enabling practical use of computationally demanding algorithms. This review presents the past and present work on GPU accelerated medical image processing, and is meant to serve as an overview and introduction to existing GPU implementations. The review covers GPU acceleration of basic image processing operations (filtering, interpolation, histogram estimation and distance transforms), the most commonly used algorithms in medical imaging (image registration, image segmentation and image denoising) and algorithms that are specific to individual modalities (CT, PET, SPECT, MRI, fMRI, DTI, ultrasound, optical imaging and microscopy). The review ends by highlighting some future possibilities and challenges.

High-throughput optogenetic functional magnetic resonance imaging with parallel computations

Optogenetic functional magnetic resonance imaging (of MRI) technology enables cell-type-specific, temporally precise neuronal control and the accurate, in vivo readout of the resulting activity across the entire brain. With the ability to precisely control excitation and inhibition parameters and accurately record the resulting activity, there is an increased need for a high-throughput method to bring the of MRI studies to their full potential. In this paper, an advanced system facilitating real-time fMRI with interactive control and analysis in a fraction of the MRI acquisition repetition time (TR) is proposed. With high-processing speed, sufficient time will be available for the integration of future developments that further enhance of MRI data or streamline the study. We designed and implemented a highly optimised, massively parallel system using graphics processing units (GPUs), which achieves the reconstruction, motion correction, and analysis of 3D volume data in approximately 12.80 ms. As a result, with a 750 ms TR and 4 interleaf fMRI acquisition, we can now conduct sliding window reconstruction, motion correction, analysis and display in approximately 1.7% of the TR. Therefore, a significant amount of time can now be allocated to integrating advanced but computationally intensive methods that improve image quality and enhance the analysis results within a TR. Utilising the proposed high-throughput imaging platform with sliding window reconstruction, we were also able to observe the much-debated initial dips in our of MRI data. Combined with methods to further improve SNR, the proposed system will enable efficient real-time, interactive, high-throughput of MRI studies.

More IMPATIENT: A Gridding-Accelerated Toeplitz-based Strategy for Non-Cartesian High-Resolution 3D MRI on GPUs

Several recent methods have been proposed to obtain significant speed-ups in MRI image reconstruction by leveraging the computational power of GPUs. Previously, we implemented a GPU-based image reconstruction technique called the Illinois Massively Parallel Acquisition Toolkit for Image reconstruction with ENhanced Throughput in MRI (IMPATIENT MRI) for reconstructing data collected along arbitrary 3D trajectories. In this paper, we improve IMPATIENT by removing computational bottlenecks by using a gridding approach to accelerate the computation of various data structures needed by the previous routine. Further, we enhance the routine with capabilities for off-resonance correction and multi-sensor parallel imaging reconstruction. Through implementation of optimized gridding into our iterative reconstruction scheme, speed-ups of more than a factor of 200 are provided in the improved GPU implementation compared to the previous accelerated GPU code.

Accelerating fibre orientation estimation from diffusion weighted magnetic resonance imaging using GPUs

With the performance of central processing units (CPUs) having effectively reached a limit, parallel processing offers an alternative for applications with high computational demands. Modern graphics processing units (GPUs) are massively parallel processors that can execute simultaneously thousands of light-weight processes. In this study, we propose and implement a parallel GPU-based design of a popular method that is used for the analysis of brain magnetic resonance imaging (MRI). More specifically, we are concerned with a model-based approach for extracting tissue structural information from diffusion-weighted (DW) MRI data. DW-MRI offers, through tractography approaches, the only way to study brain structural connectivity, non-invasively and in-vivo. We parallelise the Bayesian inference framework for the ball & stick model, as it is implemented in the tractography toolbox of the popular FSL software package (University of Oxford). For our implementation, we utilise the Compute Unified Device Architecture (CUDA) programming model. We show that the parameter estimation, performed through Markov Chain Monte Carlo (MCMC), is accelerated by at least two orders of magnitude, when comparing a single GPU with the respective sequential single-core CPU version. We also illustrate similar speed-up factors (up to 120x) when comparing a multi-GPU with a multi-CPU implementation.

MRI temporal acceleration techniques

In recent years, there has been an explosive growth of magnetic resonance imaging (MRI) techniques that allow faster scan speed by exploiting temporal or spatiotemporal redundancy of the images. These techniques improve the performance of dynamic imaging significantly across multiple clinical applications, including cardiac functional examinations, perfusion imaging, blood flow assessment, contrast-enhanced angiography, functional MRI, and interventional imaging, among others. The scan acceleration permits higher spatial resolution, increased temporal resolution, shorter scan duration, or a combination of these benefits. Along with the exciting developments is a dizzying proliferation of acronyms and variations of the techniques. The present review attempts to summarize this rapidly growing topic and presents conceptual frameworks to understand these techniques in terms of their underlying mechanics and connections. Techniques from view sharing, keyhole, k-t, to compressed sensing are covered.

Compressed sensing reconstruction for whole-heart imaging with 3D radial trajectories: a graphics processing unit implementation

A disadvantage of three-dimensional (3D) isotropic acquisition in whole-heart coronary MRI is the prolonged data acquisition time. Isotropic 3D radial trajectories allow undersampling of k-space data in all three spatial dimensions, enabling accelerated acquisition of the volumetric data. Compressed sensing (CS) reconstruction can provide further acceleration in the acquisition by removing the incoherent artifacts due to undersampling and improving the image quality. However, the heavy computational overhead of the CS reconstruction has been a limiting factor for its application. In this article, a parallelized implementation of an iterative CS reconstruction method for 3D radial acquisitions using a commercial graphics processing unit is presented. The execution time of the graphics processing unit-implemented CS reconstruction was compared with that of the C++ implementation, and the efficacy of the undersampled 3D radial acquisition with CS reconstruction was investigated in both phantom and whole-heart coronary data sets. Subsequently, the efficacy of CS in suppressing streaking artifacts in 3D whole-heart coronary MRI with 3D radial imaging and its convergence properties were studied. The CS reconstruction provides improved image quality (in terms of vessel sharpness and suppression of noise-like artifacts) compared with the conventional 3D gridding algorithm, and the graphics processing unit implementation greatly reduces the execution time of CS reconstruction yielding 34-54 times speed-up compared with C++ implementation.

Plane-wave ultrasound beamforming using a nonuniform fast Fourier transform

Beamforming of plane-wave ultrasound echo signals in the Fourier domain provides fast and accurate image reconstruction. Conventional implementations perform a k-space interpolation from the uniform sampled grid to a nonuniform acoustic dispersion grid. In this paper, we demonstrate that this step can be replaced by a nonuniform Fourier transform. We study the performance of the nonuniform fast Fourier transform (NUFFT) in terms of signal-to-noise ratio and computational cost, and show that the NUFFT offers an advantage in the trade-off between speed and accuracy, compared with other frequency-domain beamforming strategies.

Fast direct fourier reconstruction of radial and PROPELLER MRI data using the chirp transform algorithm on graphics hardware

PURPOSE

To develop and test a new algorithm for fast direct Fourier transform (DrFT) reconstruction of MR data on non-Cartesian trajectories composed of lines with equally spaced points.

THEORY AND METHODS

The DrFT, which is normally used as a reference in evaluating the accuracy of other reconstruction methods, can reconstruct images directly from non-Cartesian MR data without interpolation. However, DrFT reconstruction involves substantially intensive computation, which makes the DrFT impractical for clinical routine applications. In this article, the Chirp transform algorithm was introduced to accelerate the DrFT reconstruction of radial and Periodically Rotated Overlapping ParallEL Lines with Enhanced Reconstruction (PROPELLER) MRI data located on the trajectories that are composed of lines with equally spaced points. The performance of the proposed Chirp transform algorithm-DrFT algorithm was evaluated by using simulation and in vivo MRI data.

RESULTS

After implementing the algorithm on a graphics processing unit, the proposed Chirp transform algorithm-DrFT algorithm achieved an acceleration of approximately one order of magnitude, and the speed-up factor was further increased to approximately three orders of magnitude compared with the traditional single-thread DrFT reconstruction.

CONCLUSION

Implementation the Chirp transform algorithm-DrFT algorithm on the graphics processing unit can efficiently calculate the DrFT reconstruction of the radial and PROPELLER MRI data.

Fast l₁-SPIRiT compressed sensing parallel imaging MRI: scalable parallel implementation and clinically feasible runtime

We present l₁-SPIRiT, a simple algorithm for auto calibrating parallel imaging (acPI) and compressed sensing (CS) that permits an efficient implementation with clinically-feasible runtimes. We propose a CS objective function that minimizes cross-channel joint sparsity in the wavelet domain. Our reconstruction minimizes this objective via iterative soft-thresholding, and integrates naturally with iterative self-consistent parallel imaging (SPIRiT). Like many iterative magnetic resonance imaging reconstructions, l₁-SPIRiT's image quality comes at a high computational cost. Excessively long runtimes are a barrier to the clinical use of any reconstruction approach, and thus we discuss our approach to efficiently parallelizing l₁-SPIRiT and to achieving clinically-feasible runtimes. We present parallelizations of l₁-SPIRiT for both multi-GPU systems and multi-core CPUs, and discuss the software optimization and parallelization decisions made in our implementation. The performance of these alternatives depends on the processor architecture, the size of the image matrix, and the number of parallel imaging channels. Fundamentally, achieving fast runtime requires the correct trade-off between cache usage and parallelization overheads. We demonstrate image quality via a case from our clinical experimentation, using a custom 3DFT spoiled gradient echo (SPGR) sequence with up to 8× acceleration via Poisson-disc undersampling in the two phase-encoded directions.

Real-Time Compressive Sensing MRI Reconstruction Using GPU Computing and Split Bregman Methods

Compressive sensing (CS) has been shown to enable dramatic acceleration of MRI acquisition in some applications. Being an iterative reconstruction technique, CS MRI reconstructions can be more time-consuming than traditional inverse Fourier reconstruction. We have accelerated our CS MRI reconstruction by factors of up to 27 by using a split Bregman solver combined with a graphics processing unit (GPU) computing platform. The increases in speed we find are similar to those we measure for matrix multiplication on this platform, suggesting that the split Bregman methods parallelize efficiently. We demonstrate that the combination of the rapid convergence of the split Bregman algorithm and the massively parallel strategy of GPU computing can enable real-time CS reconstruction of even acquisition data matrices of dimension 4096² or more, depending on available GPU VRAM. Reconstruction of two-dimensional data matrices of dimension 1024² and smaller took ~0.3 s or less, showing that this platform also provides very fast iterative reconstruction for small-to-moderate size images.

Retrospective reconstruction of high temporal resolution cine images from real-time MRI using iterative motion correction

Cardiac function has traditionally been evaluated using breath-hold cine acquisitions. However, there is a great need for free breathing techniques in patients who have difficulty in holding their breath. Real-time cardiac MRI is a valuable alternative to the traditional breath-hold imaging approach, but the real-time images are often inferior in spatial and temporal resolution. This article presents a general method for reconstruction of high spatial and temporal resolution cine images from a real-time acquisition acquired over multiple cardiac cycles. The method combines parallel imaging and motion correction based on nonrigid registration and can be applied to arbitrary k-space trajectories. The method is demonstrated with real-time Cartesian imaging and Golden Angle radial acquisitions, and the motion-corrected acquisitions are compared with raw real-time images and breath-hold cine acquisitions in 10 (N = 10) subjects. Acceptable image quality was obtained in all motion-corrected reconstructions, and the resulting mean image quality score was (a) Cartesian real-time: 2.48, (b) Golden Angle real-time: 1.90 (1.00-2.50), (c) Cartesian motion correction: 3.92, (d) Radial motion correction: 4.58, and (e) Breath-hold cine: 5.00. The proposed method provides a flexible way to obtain high-quality, high-resolution cine images in patients with difficulty holding their breath.

Selection of convolution kernel in non-uniform fast Fourier transform for Fourier domain optical coherence tomography

Gridding based non-uniform fast Fourier transform (NUFFT) has recently been shown as an efficient method of processing non-linearly sampled data from Fourier-domain optical coherence tomography (FD-OCT). This method requires selecting design parameters, such as kernel function type, oversampling ratio and kernel width, to balance between computational complexity and accuracy. The Kaiser-Bessel (KB) and Gaussian kernels have been used independently on the NUFFT algorithm for FD-OCT. This paper compares the reconstruction error and speed for the optimization of these design parameters and justifies particular kernel choice for FD-OCT applications. It is found that for on-the-fly computation of the kernel function, the simpler Gaussian function offers a better accuracy-speed tradeoff. The KB kernel, however, is a better choice in the pre-computed kernel mode of NUFFT, in which the processing speed is no longer dependent on the kernel function type. Finally, the algorithm is used to reconstruct in-vivo images of a human finger at a camera limited 50k A-line/s.

Accelerating EPI distortion correction by utilizing a modern GPU-based parallel computation

BACKGROUND AND PURPOSE

The combination of phase demodulation and field mapping is a practical method to correct echo planar imaging (EPI) geometric distortion. However, since phase dispersion accumulates in each phase-encoding step, the calculation complexity of phase modulation is Ny-fold higher than conventional image reconstructions. Thus, correcting EPI images via phase demodulation is generally a time-consuming task.

METHODS

Parallel computing by employing general-purpose calculations on graphics processing units (GPU) can accelerate scientific computing if the algorithm is parallelized. This study proposes a method that incorporates the GPU-based technique into phase demodulation calculations to reduce computation time. The proposed parallel algorithm was applied to a PROPELLER-EPI diffusion tensor data set.

RESULTS

The GPU-based phase demodulation method reduced the EPI distortion correctly, and accelerated the computation. The total reconstruction time of the 16-slice PROPELLER-EPI diffusion tensor images with matrix size of 128 × 128 was reduced from 1,754 seconds to 101 seconds by utilizing the parallelized 4-GPU program.

CONCLUSIONS

GPU computing is a promising method to accelerate EPI geometric correction. The resulting reduction in computation time of phase demodulation should accelerate postprocessing for studies performed with EPI, and should effectuate the PROPELLER-EPI technique for clinical practice.

GPU computing in medical physics: a review

The graphics processing unit (GPU) has emerged as a competitive platform for computing massively parallel problems. Many computing applications in medical physics can be formulated as data-parallel tasks that exploit the capabilities of the GPU for reducing processing times. The authors review the basic principles of GPU computing as well as the main performance optimization techniques, and survey existing applications in three areas of medical physics, namely image reconstruction, dose calculation and treatment plan optimization, and image processing.

Split-acquisition real-time CINE phase-contrast MR flow measurements

The temporal and spatial resolution of real-time phase-contrast magnetic resonance (PCMR) is restricted by the need to acquire two interleaved phase images. In this article, we propose a split-acquisition real-time CINE PCMR technique, where the acquisition of flow-encoded and flow-compensated data is divided into separate blocks. By comparing magnitude images, automatic matching of data in cardio-respiratory space allows subtraction of background phase offsets. Thus, the data is acquired in real-time but with phase correction originating from a different heart beat. This effectively doubles the frame rate, allowing either higher temporal or spatial resolution. Two split-acquisition sequences were tested: one with high-temporal resolution and one with high-spatial resolution. Both sequences showed excellent agreement in stroke volumes in 20 adults when validated against cardiac-gated PCMR and interleaved real-time PCMR (cardiac gated: 95.2 ± 20.0 mL, interleaved real-time: 96.2 ± 20.7 mL, high-temporal resolution: 95.6 ± 20.1 mL, high-spatial resolution: 95.5 ± 20.4 mL). In six children, the high-spatial resolution sequence provided more accurate flow measurements than interleaved real-time PCMR, when compared with cardiac-gated PCMR (cardiac gated: 20.6 ± 7.6 mL, interleaved real-time: 24.3 ± 9.2 mL, high-spatial resolution: 20.8 ± 7.8 mL), due to the increased spatial resolution. The matching technique is shown to be accurate (truth: 94.6 ± 21.8, split-acquisition: 95.0 ± 21.9 mL) and quantitative image quality (signal-to-noise ratio, velocity-to-noise ratio and edge sharpness) is acceptable.

GPU-accelerated FDTD modeling of radio-frequency field-tissue interactions in high-field MRI

The analysis of high-field RF field-tissue interactions requires high-performance finite-difference time-domain (FDTD) computing. Conventional CPU-based FDTD calculations offer limited computing performance in a PC environment. This study presents a graphics processing unit (GPU)-based parallel-computing framework, producing substantially boosted computing efficiency (with a two-order speedup factor) at a PC-level cost. Specific details of implementing the FDTD method on a GPU architecture have been presented and the new computational strategy has been successfully applied to the design of a novel 8-element transceive RF coil system at 9.4 T. Facilitated by the powerful GPU-FDTD computing, the new RF coil array offers optimized fields (averaging 25% improvement in sensitivity, and 20% reduction in loop coupling compared with conventional array structures of the same size) for small animal imaging with a robust RF configuration. The GPU-enabled acceleration paves the way for FDTD to be applied for both detailed forward modeling and inverse design of MRI coils, which were previously impractical.

Assessing vascular response to exercise using a combination of real-time spiral phase contrast MR and noninvasive blood pressure measurements

PURPOSE

To measure the hemodynamic response to exercise using real-time velocity mapping magnetic resonance imaging (MRI), incorporating a high temporal resolution spiral phase contrast (PC) sequence accelerated with sensitivity encoding (SENSE).

MATERIALS AND METHODS

Twenty healthy adults underwent MRI at rest and during supine exercise at two different exercise levels. Flow volumes were assessed in the ascending aorta using a spiral SENSE real-time PC sequence. The sequence was validated at rest against a vendor supplied gated PC sequence, and also at rest and during exercise against left ventricular volumes assessed using a radial k-t SENSE real-time sequence. Combining the measured flow volumes with simultaneous oscillometric blood pressure measurements, enabled the noninvasive calculations of systemic vascular resistance (SVR) and arterial compliance (C).

RESULTS

Measured flow volumes correlated very well between the sequences at rest and during exercise. Cardiac output (CO) and heart rate were found to significantly increase during exercise, while SVR and C were found to decrease significantly.

CONCLUSION

Hemodynamic response to exercise can be accurately quantified using a high temporal resolution spiral SENSE real-time flow imaging. This may allow early detection of hypertension and a greater understanding of the early changes in this condition.

Reconstruction of dynamic contrast enhanced magnetic resonance imaging of the breast with temporal constraints

A number of methods using temporal and spatial constraints have been proposed for reconstruction of undersampled dynamic magnetic resonance imaging (MRI) data. The complex data can be constrained or regularized in a number of different ways, for example, the time derivative of the magnitude and phase image voxels can be constrained separately or jointly. Intuitively, the performance of different regularizations will depend on both the data and the chosen temporal constraints. Here, a complex temporal total variation (TV) constraint was compared to the use of separate real and imaginary constraints, and to a magnitude constraint alone. Projection onto Convex Sets (POCS) with a gradient descent method was used to implement the diverse temporal constraints in reconstructions of DCE MRI data. For breast DCE data, serial POCS with separate real and imaginary TV constraints was found to give relatively poor results while serial/parallel POCS with a complex temporal TV constraint and serial POCS with a magnitude-only temporal TV constraint performed well with an acceleration factor as large as R=6. In the tumor area, the best method was found to be parallel POCS with complex temporal TV constraint. This method resulted in estimates for the pharmacokinetic parameters that were linearly correlated to those estimated from the fully-sampled data, with K(trans,R=6)=0.97 K(trans,R=1)+0.00 with correlation coefficient r=0.98, k(ep,R=6)=0.95 k(ep,R=1)+0.00 (r=0.85). These results suggest that it is possible to acquire highly undersampled breast DCE-MRI data with improved spatial and/or temporal resolution with minimal loss of image quality.