Single-shot fast spin-echo diffusion tensor imaging of the brain and spine with head and phased array coils at 1.5 T and 3.0 T

In this study, we investigated the use of a single-shot fast spin-echo-based sequence to perform diffusion tensor imaging (DTI) with improved anatomic fidelity through the entire brain and the cervical spine. Traditionally, diffusion tensor images have been acquired by single-shot echo-planar imaging (EPI) methods in which large distortions result from magnetic susceptibility effects, especially near air-tissue interfaces. These distortions can be problematic, especially in anterior and inferior portions of the brain, and they also can severely limit applications in the spine. At higher magnetic fields these magnetic susceptibility artifacts are increased. The single-shot fast spin-echo (SSFSE) method used in this study utilizes radiofrequency rephasing in the transverse plane and thus provides diffusion images with negligible distortion even at 3 Tesla. In addition, the SSFSE sequence does not require multiple fast-receivers, which are not available on many magnetic resonance (MR) systems. Phased array coils were used to increase the signal-to-noise ratio of the images, offering a major inherent advantage in diffusion tensor imaging of the spine and brain. The mean diffusion measurements obtained with the SSFSE acquisition were not statistically different (p > 0.05) from EPI-based acquisitions. Compared to routine T(2)-weighted MR images, the DTI-EPI sequence showed up to 20% in elongation of the brain in the anterior-posterior direction on a sagittal image due to magnetic susceptibility distortions, whereas in the DTI-SSFSE, the image distortions were negligible. The diffusion tensor SSFSE method was also able to assess diffusion abnormalities in a brain stem hemorrhage, unaffected by the spatial distortions that limited conventional EPI acquisition.     

Improved B0 field map estimation for high field EPI

Echo planar imaging (EPI) is an ultrafast magnetic resonance imaging (MRI) technique that allows one to acquire a 2D image in about 100 ms. Unfortunately, the standard EPI images suffer from substantial geometric distortions, mainly originating from susceptibility differences in adjacent tissues. To reduce EPI distortions, correction methods based on a field map, which is a map of the off-resonance frequencies, have been developed. In this work, a nonlinear least squares estimator is used to optimize the estimation of the field map of the B₀ field. The model of the EPI and reference data includes parameters for the phase evolution, the complex magnitude, the relaxation of the MRI signal and the EPI-specific phase difference between odd and even echoes, and from these parameters, additional corrections might be computed. The reference data required to estimate the field map can be acquired with a modified EPI-sequence. The proposed method is tested on simulated as well as experimental data and proves to be significantly more robust against noise, compared to the previously suggested method.     

Distortion correction for a double inversion-recovery sequence with an echo-planar imaging readout

A double inversion-recovery (DIR) sequence with an echo-planar imaging (EPI) readout can be used to image selectively the grey matter of the brain, and this has previously been applied to improve the sensitivity of the statistical analysis of functional magnetic resonance imaging (fMRI) data. If a procedure were to be implemented to remove the distortions that are inherent in the EPI-based fMRI data set, then a similar technique would have to be applied to the DIR-EPI image also to ensure that it matches the geometry of the functional data. A comparison of candidate methodologies for correcting distortions in DIR-EPI images, based on the reversed-gradient method, is presented. A corrected image could be calculated from two DIR-EPI images acquired with k-space traversal in opposite directions, but that method was not able to cope with the large regions of low signal intensity corresponding to the nulled white matter. It was found that the optimal procedure to apply the reversed-gradient method to DIR-EPI images was to acquire two additional EPI images (without the two inversion pulses) with opposite-direction k-space traversal; the distortion-correction information calculated from those EPI images was then applied to the DIR-EPI data.     

Combination of integrated dynamic shimming and readout-segmented echo planar imaging for diffusion weighted MRI of the head and neck region at 3 Tesla

PurposeTo evaluate the performance of combined integrated slice-by-slice shimming and readout-segmented EPI (irsEPI) for diffusion-weighted MR imaging (DWI) of the neck at 3 Tesla.MethodsThis study was approved by the local ethics committee. An anthropometric phantom of the head/neck region incorporating compartments with different diffusivities was constructed. In vivo measurements were performed in 10 healthy volunteers. DWI of the phantom and volunteers was performed on a 3 Tesla MR scanner using single shot EPI (sEPI), a prototype single shot EPI with integrated slice-by-slice shimming (iEPI), readout segmented EPI (rsEPI) and a prototype readout segmented EPI with integrated shimming irsEPI. Apparent diffusion coefficients (ADC) and spatial distortions of phantom compartments were quantified. For phantom and volunteer measurements, the presence of geometric distortions, signal losses, ghosting artifacts as well as overall image quality were visually assessed on a 4-point scale by two radiologists in consensus. In addition, failure of fat saturation was assessed in volunteer data.ResultsQuantification of ADC within the phantom compartments was comparable using the different EPI techniques without significant variations. Using irsEPI, spatial distortions in phase-encoding direction were markedly reduced compared to iEPI, rsEPI and especially sEPI. irsEPI yielded significantly better overall image quality compared to sEPI, iEPI and rsEPI in phantom data as well as volunteer measurements. Markedly reduced geometric distortions and signal loss as well as better fat saturation were observed using irsEPI.ConclusionThe use of irsEPI significantly improves image quality and reduces artifacts caused by magnetic field inhomogeneities in EPI based DWI of the head/neck at 3 Tesla.     

Single-shot fast spin-echo diffusion tensor imaging of the lumbar spine at 1.5 and 3 T

Diffusion tensor imaging (DTI) of the lumbar spine could improve diagnostic specificity. The purpose of this work was to determine the feasibility of and to validate DTI with single-shot fast spin-echo (SSFSE) for lumbar intervertebral discs at 1.5 and 3 T. Six normal volunteers were scanned with DTI-SSFSE using an eight- and a three-b-value protocol at 1.5 and 3 T, respectively. Apparent diffusion coefficient (ADC) values were computed and validated based on those obtained at 1.5 T from corresponding diffusion tensor scans using line scan diffusion imaging (LSDI), a technique that has been previously validated for use in the spine. Pearson correlation coefficients for LSDI and DTI-SSFSE ADC values were .88 and .89 for 1.5 and 3 T, respectively, with good quantitative agreement according to the Bland-Altman method. Results indicate that DTI-SSFSE is a candidate as a clinical sequence for obtaining diffusion tensor images of the lumbar intervertebral discs with scan times shorter than 4 min.     

Diffusion anisotropy indexes are sensitive to selecting the EPI readout-encoding bandwidth at high-field MRI

Diffusion tensor magnetic resonance imaging (DT-MRI) is generally performed using an echo planar imaging (EPI) acquisition to map directional water diffusion. However, the oscillating magnetic field gradients of the EPI acquisition can result in considerable mechanical vibrations, which lead, in turn, to magnetic field fluctuations causing Nyquist ghosting in the EPI data. The objective of this study was to investigate effects of EPI readout gradient modulation frequency, which is directly associated with the EPI readout bandwidth (BW), on the accuracy of DT-MRI measurements in a high magnetic field system. A spherical water phantom was used to study the relationship between the EPI BW and the Nyquist ghost for a spin-echo EPI acquisition with a matrix size of 128x128, complemented by diffusion sensitization gradients of up to b=800 s/mm(2) along six directions for DT-MRI. Nine volunteers (four males and five females) were studied using EPI at different BW acquisitions. Analysis of variance was used to investigate the EPI BW effects. The phantom studies demonstrated a systematic relationship between BWs and the intensities of Nyquist ghosts. In the human brain studies, EPI BW variations substantially corrupted diffusion anisotropy indexes (i.e., fractional anisotropy and relative anisotropy) (F=10.5, P=.0001) but were unrelated to diffusion-encoding directions (F=0.14, P=.98). It was possible to minimize BW dependence (F=1.48, P=.25) by tuning the modulation frequency of the EPI readout gradient. In conclusion, diffusion anisotropic indexes are sensitive to the readout BW of EPI due to associated Nyquist ghosting. However, the effect can be minimized by tuning the modulation frequency of the EPI readout gradient, that is, the EPI BW, to a range outside the harmonics of mechanical gradient vibrations.     

Correction of eddy current-induced artefacts in diffusion tensor imaging using iterative cross-correlation.

Geometric distortions of echo-planar images produced by the strong eddy currents present in the diffusion tensor imaging experiment are a major confound to the accurate quantification of diffusion coefficients, and measures of diffusion anisotropy based upon them. Here we investigate how the method of iterative cross-correlation (ICC) of baseline and diffusion-weighted images (DWIs) originally proposed by Haselgrove and Moore (Magn. Reson. Med. 36:960-964; 1996) can be extended to correct high b-value DWIs, without the need for extrapolation of distortion parameters determined from low b-value images. Monte Carlo simulations of synthetic brain images show that the maximum value of the trace of the b-matrix, Tr(b), at which distorted DWIs can be accurately corrected by direct comparison with the undistorted baseline image is approximately 300 s mm(-2). Removal of the cerebrospinal fluid signal greatly extends this value of Tr(b) (up to approximately 2000 s mm(-2)), thereby allowing direct comparison of baseline and distorted images. The use of ICC distortion parameters determined from separate calibrations of water phantom images is also investigated, and found to be effective in correcting geometric distortions observed in the DWIs collected as part of a human volunteer diffusion tensor imaging study. This work suggests that distorted DWIs acquired at high values of b may be corrected using the ICC algorithm without collecting additional low b-value images, thus allowing simplified methods of measuring the apparent diffusion tensor D, based on collecting a small number of DWIs, to be implemented in quantitative patient examinations.     

Double-spin-echo diffusion weighting with a modified eddy current adjustment

Magnetic field inhomogeneities like eddy current-related gradient fields cause geometric distortions in echo-planar imaging (EPI). This in particular affects diffusion-weighted imaging where these distortions vary with the direction of the diffusion weighting and hamper the accurate determination of diffusion parameters. The double-spin-echo preparation often used aims to reduce the cumulative eddy current effect by adjusting the diffusion-weighting gradient pulse durations to the time constant of the dominant eddy current contribution. However, eddy currents with a variety of time constants may be present and cause residual distortions. Here, a modification is proposed where the two bipolar gradient pairs of the preparation are adjusted independently to different time constants. At the expense of a slightly prolonged echo time, residual geometric distortions and correspondingly increased values of the diffusion anisotropy can be reduced as is demonstrated in phantoms and the human brain. Thus, it may help to improve the reliability of diffusion-weighted EPI.     

Surface-based functional magnetic resonance imaging analysis of partial brain echo planar imaging data at 1.5 T

Surface-based functional magnetic resonance imaging (fMRI) analysis is more sensitive and accurate than volume-based analysis for detecting neural activation. However, these advantages are less important in practical fMRI experiments with commonly used 1.5-T magnetic resonance devices because of the resolution gap between the echo planar imaging data and the cortical surface models. We expected high-resolution segmented partial brain echo planar imaging (EPI) data to overcome this problem, and the activation patterns of the high-resolution data could be different from the low-resolution data. For the practical applications of surface-based fMRI analysis using segmented EPI techniques, the effects of some important factors (e.g., activation patterns, registration and local distortions) should be intensively evaluated because the results of surface-based fMRI analyses could be influenced by them. In this study, we demonstrated the difference between activations detected from low-resolution EPI data, which were covering whole brain, and high-resolution segmented EPI data covering partial brain by volume- and surface-based analysis methods. First, we compared the activation maps of low- and high-resolution EPI datasets detected by volume- and surface-based analyses, with the spatial patterns of activation clusters, and analyzed the distributions of activations in occipital lobes. We also analyzed the high-resolution EPI data covering motor areas and fusiform gyri of human brain, and presented the differences of activations detected by volume- and surface-based methods.     

Quantitative characterization of tissue microstructure with temporal diffusion spectroscopy

The signals recorded by diffusion-weighted magnetic resonance imaging (DWI) are dependent on the micro-structural properties of biological tissues, so it is possible to obtain quantitative structural information non-invasively from such measurements. Oscillating gradient spin echo (OGSE) methods have the ability to probe the behavior of water diffusion over different time scales and the potential to detect variations in intracellular structure. To assist in the interpretation of OGSE data, analytical expressions have been derived for diffusion-weighted signals with OGSE methods for restricted diffusion in some typical structures, including parallel planes, cylinders and spheres, using the theory of temporal diffusion spectroscopy. These analytical predictions have been confirmed with computer simulations. These expressions suggest how OGSE signals from biological tissues should be analyzed to characterize tissue microstructure, including how to estimate cell nuclear sizes. This approach provides a model to interpret diffusion data obtained from OGSE measurements that can be used for applications such as monitoring tumor response to treatment in vivo.     

Functional MRI of the motor cortex using a conventional gradient system: comparison of FLASH and EPI techniques

Gradient echo (GE) and echo planar imaging (EPI) techniques are two different approaches to functional MRI (fMRI). In contrast to GE sequences, the ultra short EPI technique facilitates fMRI experiments with high spatial and temporal resolution or mapping of the whole brain. Although it has become the method of choice for fMRI, EPI is generally restricted to modern scanners with a strong gradient system. The aim of our study was to evaluate the applicability of EPI for fMRI of the motor cortex using a 1.5 T scanner with a conventional gradient system of 10 mT/m (rise time: 1 ms). Therefore, EPI was compared with a well-established high resolution fast low angle shot (FLASH) technique (matrix size 128²). The FLASH technique was applied additionally with a 64² matrix size to exclude influences caused by different spatial resolution, because the EPI sequence was restricted to a 64² matrix size. A total of 35 healthy volunteers were included in this study. The task consisted of clenching and spreading of the right hand. FLASH and EPI techniques were compared regarding geometric distortions as well as qualitative and quantitative fMRI criteria: Mean signal increase between activation and rest and the area of activation were measured within the contralateral, ipsilateral, and supplementary motor cortex. The quality of subtraction images between activation and rest, as well as the quality of z-maps and time course within activated regions of interest, was evaluated visually. EPI revealed significant distortions of the anterior and postior brain margins; lateral distortions (relevant for the motor cortex) could be neglected in most cases. The mean signal increase was significantly higher using FLASH 128² compared to FLASH 64² and EPI 64², whereas the activated areas proved to be smaller in FLASH 128² functional images. Both results can be explained by well-documented partial volume effects, caused by different voxel size. Similar quality of the subtraction images and of the time courses in different regions of interest were found for all techniques under investigation, but slightly reduced quality of z-map in FLASH 128². Within the limits of reproducibility and measurement accuracy, the location of contralateral activation was similar using FLASH and EPI sequences. In conclusion, EPI proved to be a reliable technique for fMRI of the motor cortex, even on an MR scanner with a conventional gradient system.     

Computer simulation studies of the effects of dynamic shimming on susceptibility artifacts in EPI at high field

Dynamic shimming in multi-slice imaging aims to achieve optimal magnetic field homogeneity by updating the shim coil currents for each slice in real time. Dynamic shimming may reduce the signal loss and geometric distortion caused by magnetic susceptibility variations between tissues and is likely to be valuable for fast T2*-sensitive imaging techniques like EPI. A computer simulation of dynamic shimming using real image data has been developed to demonstrate the effectiveness of higher order dynamic shimming for echo planar imaging at high magnetic field, and to investigate the potential benefits of different orders of shim coil. Geometric distortions and signal intensities for different degrees of dynamic shimming were simulated and the results are compared with the images obtained with a conventional shimming technique. These results demonstrate the effectiveness, necessity and difficulty of high order dynamic shimming.     

Artifacts caused by transcranial magnetic stimulation coils and EEG electrodes in T(2)*-weighted echo-planar imaging

We investigated the effects of transcranial magnetic stimulation (TMS) coils and electroencephalographic (EEG) electrodes on T(2)*-weighted echo-planar images (EPI) at 2.0 T (gradient-echo EPI, mean TE = 53 ms, 2x2x4 mm(3)). In comparison with anatomic gradient-echo images (3D FLASH, TE = 4 ms, 1x1x1 mm(3)), T(2)*-weighted EPI acquisitions of a water-filled spherical phantom revealed severe signal losses and geometric distortions in the vicinity of TMS coils. Even remote effects were observed for image orientations perpendicular to the coil plane. EEG electrodes and the fixation gel caused milder localized distortions. In humans, complications were avoided by the large distance between the TMS coil and the cortical surface and when using an EPI orientation parallel to the plane of the coil. It is concluded that T(2)*-weighted EPI studies of human brain function may be performed without distortions caused by TMS coils and EEG electrodes.     

Mapping eddy current induced fields for the correction of diffusion-weighted echo planar images

Diffusion-weighted echo-planar magnetic resonance imaging is potentially of great importance as a diagnostic imaging tool; however, the technique currently suffers a number of limitations, including the image distortion caused by the eddy current induced fields when the diffusion-weighting magnetic field gradient pulses are applied. The distortions cause mis-registration between images with different diffusion-weighting, that then results in artifacts in quantitative diffusion images. A method is presented to measure the magnetic fields generated from the eddy currents for each of three orthogonal gradient pulse vectors, and then to use these to ascertain the image distortion that occurs in subsequent diffusion-weighted images with arbitrary gradient pulse vector amplitude and direction, and image plane orientation. The image distortion can then be reversed. Both temporal and spatial dependence of the residual eddy current induced fields are included in the analysis. Image distortion was substantially reduced by the correction scheme, for arbitrary slice position and angulation. This method of correction is unaffected by the changes in image contrast that occur due to diffusion weighting, and does not need any additional scanning time during the patient scan. It is particularly suitable for use with single-shot echo planar imaging.     

Simulation and experimental verification of the diffusion in an anisotropic fiber phantom

Diffusion weighted magnetic resonance imaging enables the visualization of fibrous tissues such as brain white matter. The validation of this non-invasive technique requires phantoms with a well-known structure and diffusion behavior. This paper presents anisotropic diffusion phantoms consisting of parallel fibers. The diffusion properties of the fiber phantoms are measured using diffusion weighted magnetic resonance imaging and bulk NMR measurements. To enable quantitative evaluation of the measurements, the diffusion in the interstitial space between fibers is modeled using Monte Carlo simulations of random walkers. The time-dependent apparent diffusion coefficient and kurtosis, quantifying the deviation from a Gaussian diffusion profile, are simulated in 3D geometries of parallel fibers with varying packing geometries and packing densities. The simulated diffusion coefficients are compared to the theory of diffusion in porous media, showing a good agreement. Based on the correspondence between simulations and experimental measurements, the fiber phantoms are shown to be useful for the quantitative validation of diffusion imaging on clinical MRI-scanners.     

Functional MRI using super-resolved spatiotemporal encoding

Recently, new ultrafast imaging sequences such as rapid acquisition by sequential excitation and refocusing (RASER) and hybrid spatiotemporal encoding (SPEN) magnetic resonance imaging (MRI) have been proposed, in which the phase encoding of conventional echo planar imaging (EPI) is replaced with a SPEN. In contrast to EPI, SPEN provides significantly higher immunity to frequency heterogeneities including those caused by B₀ inhomogeneities and chemical shift offsets. Utilizing the inherent robustness of SPEN, it was previously shown that RASER can be used to successfully perform functional MRI (fMRI) experiments in the orbitofrontal cortex — a task which is challenging using EPI due to strong magnetic susceptibility variation near the air-filled sinuses. Despite this superior performance, systematic analyses have shown that, in its initial implementation, the use of SPEN was penalized by lower signal-to-noise ratio (SNR) and higher radiofrequency power deposition as compared to EPI-based methods. A recently developed reconstruction algorithm based on super-resolution principles is able to alleviate both of these shortcomings; the use of this algorithm is hereby explored within an fMRI context. Specifically, a series of fMRI measurements on the human visual cortex confirmed that the super-resolution algorithm retains the statistical significance of the blood oxygenation level dependent (BOLD) response, while significantly reducing the power deposition associated with SPEN and restoring the SNR to levels that are comparable with those of EPI.     

On the accuracy of EPI-based phase contrast velocimetry

For over a decade, echo-planar imaging (EPI) has been used in both the medical and applied sciences to capture velocity fields of fluid flows. However, previous studies have not rigorously confirmed the accuracy of the measurements or sought to understand the limitations of the technique. In this study, a bipolar gradient was added to a flow-compensated EPI pulse sequence to obtain rapid phase contrast images of steady and unsteady flows through two step stenoses. For steady Re = 100 and 258 flows, accuracy was measured through systematic comparisons with CFD simulations, mass flow rate measurements, and spin echo phase contrast images. On average, the EPI image data exhibited velocity errors of 5 to 10 percent, while mass was conserved to within 5.6 percent at each axial position. Compared to spin-echo phase contrast images, the EPI images have 50 percent lower signal-to-noise ratio, larger local velocity errors, and similar mass conservation characteristics. An unsteady flow was then examined by starting a pump and allowing it to reach a steady Re = 100 flow. Accuracy in this case was measured by the consistency between mass flow rate measurements at different axial positions. Images taken at 0.3 s intervals captured the velocity field evolution and showed that 50 to 100 percent errors occur when the flow changes on a time scale faster than the image acquisition time.     

Comparison of multiple sclerosis clinical subgroups using navigated spin echo diffusion-weighted imaging.

The apparent diffusion coefficient (ADC) of tissue provides an indication of the size, shape, and orientation of the water spaces in tissue. Thus, pathologic differences between lesions in multiple sclerosis (MS) patients with different clinical courses may be reflected by changes in ADC measurements in lesions and white matter. Twelve healthy subjects and 35 MS patients with a relapsing-remitting (n = 10), benign (n = 8), secondary progressive (n = 8) and primary progressive (n = 9) clinical course were studied. T2-weighted and post-gadolinium T1-weighted images were obtained using a 1.5 T Signa Echospeed magnetic resonance imaging (MRI) system. Diffusion-weighted imaging was implemented using a pulsed gradient spin echo (PGSE) sequence with diffusion gradients applied in turn along three orthogonal directions in order to obtain the average apparent diffusion coefficient (ADCav). Navigator echo correction and cardiac gating were used to reduce motion artifact. ADC maps were derived using a two point calculation based on the Stejskal-Tanner formula. Diffusion anisotropy was estimated using the van Gelderen formula to calculate an anisotropy index. MS lesions had a higher ADC and reduced anisotropy compared with normal appearing white matter. Highest ADC values were found in gadolinium enhancing lesions and non-enhancing hypointense lesions on T1-weighted imaging. MS white matter had a slightly higher ADC and lower anisotropy than white matter of healthy subjects. Lesion and white matter ADC values did not differ between patients with different clinical courses of MS. There was no correlation between lesion ADC and disability. Diffusion-weighted imaging with measurement of ADC using the PGSE method provides quantitative information on acute edematous MS lesions and chronic lesions associated with demyelination and axonal loss but does not distinguish between clinical subtypes of MS.     

High-resolution diffusion imaging using phase-corrected segmented echo-planar imaging

Diffusion magnetic resonance imaging (MRI) was performed with a high-resolution segmented echo-planar imaging technique, which provided images with substantially less susceptibility artifacts than images obtained with single-shot echo-planar imaging (EPI). Diffusion imaging performed with any multishot pulse sequence is inherently sensitive to motion artifacts and in order to reduce motion artifacts, the presented method utilizes navigator echo phase corrections, performed after a one-dimensional Fourier transform along the frequency-encoding direction. Navigator echo phases were fitted to a straight line prior to phase correction to avoid errors from internal motion. In vivo imaging was performed using electro cardiographic (ECG) triggering. Apparent diffusion coefficient (ADC) maps were calculated on a pixel-by-pixel basis using up to seven diffusion sensitivities, ranging from b = 0 to 1129 x 10(6) s/m(2).     

An unsupervised deep learning technique for susceptibility artifact correction in reversed phase-encoding EPI images

Echo planar imaging (EPI) is a fast and non-invasive magnetic resonance imaging technique that supports data acquisition at high spatial and temporal resolutions. However, susceptibility artifacts, which cause the misalignment to the underlying structural image, are unavoidable distortions in EPI. Traditional susceptibility artifact correction (SAC) methods estimate the displacement field by optimizing an objective function that involves one or more pairs of reversed phase-encoding (PE) images. The estimated displacement field is then used to unwarp the distorted images and produce the corrected images. Since this conventional approach is time-consuming, we propose an end-to-end deep learning technique, named S-Net, to correct the susceptibility artifacts the reversed-PE image pair. The proposed S-Net consists of two components: (i) a convolutional neural network to map a reversed-PE image pair to the displacement field; and (ii) a spatial transform unit to unwarp the input images and produce the corrected images. The S-Net is trained using a set of reversed-PE image pairs and an unsupervised loss function, without ground-truth data. For a new image pair of reversed-PE images, the displacement field and corrected images are obtained simultaneously by evaluating the trained S-Net directly. Evaluations on three different datasets demonstrate that S-Net can correct the susceptibility artifacts in the reversed-PE images. Compared with two state-of-the-art SAC methods (TOPUP and TISAC), the proposed S-Net runs significantly faster: 20 times faster than TISAC and 369 times faster than TOPUP, while achieving a similar correction accuracy. Consequently, S-Net accelerates the medical image processing pipelines and makes the real-time correction for MRI scanners feasible. Our proposed technique also opens up a new direction in learning-based SAC.