Methods for reconstructing images from undersampled k-space data using a machine learning approach to learn non-linear mapping functions from acquired k-space lines to generate unacquired target points across multiple coils are described.

A system includes a machine readable storage medium storing instructions and a processor to execute the instructions. The processor executes the instructions to receive radial k-space magnetic resonance imaging (MRI) data of a patient and determine a series of dipole sources via direct dipole decomposition of the radial k-space MRI data. The processor executes the instructions to identify an activation within the patient based on the series of dipole sources.

In one embodiment, a magnetic resonance imaging apparatus includes: a scanner that includes a static magnetic field magnet configured to generate a static magnetic field, a gradient coil configured to generate a gradient magnetic field, and a WB (Whole Body) coil configured to apply an RF pulse to an object; and processing circuitry. The processing circuitry is configured to: set (i) a pulse sequence in which a sequence element is repeated, the sequence element including at least an inversion pulse and (ii) a data acquisition sequence executed after a delay time from the inversion pulse; and cause the scanner to execute the pulse sequence by using virtual gating.

The present invention is to acquire a multiphase image while avoiding extension of imaging time and excluding an influence of displacement of an image of each multiphase due to a motion. A method for collecting measurement data is to repeat sampling such that low-frequency data and high-frequency data have different densities. At this time, a sampling interval is set shorter than a motion cycle. Motion information is acquired in parallel with imaging, and measurement data obtained in time series is divided into a plurality of time phases based on the motion information so as to obtain a multiphase image. Displacement correction between multiphase images is performed, and then the multiphase images are integrated. Alternatively, measurement data after the displacement correction is used to generate a time-series image.

Systems and methods for generative adversarial networks (GANs) to remove artifacts from undersampled magnetic resonance (MR) images are described. The process of training the GAN can include providing undersampled 3D MR images to the generator model, providing the generated example and a real example to the discriminator model, applying adversarial loss, L2 loss, and structural similarity index measure loss to the generator model based on a classification output by the discriminator model, and repeating until the generator model has been trained to remove the artifacts from the undersampled 3D MR images. At runtime, the trained generator model of the GAN can be generate artifact-free images or parameter maps from undersampled MRI data of a patient.

The present disclosure provides a system for MRI. The system may obtain a plurality of auxiliary signals and a plurality of imaging signals collected by applying an MRI pulse sequence simultaneously to a plurality of slice locations of a subject. For each of at least one target slice location of the plurality of slice locations, the system may generate at least one target image of the target slice location based on the plurality of auxiliary signals and the plurality of imaging signals. During the application of the MRI pulse sequence, phase modulation may be applied to at least one of the plurality of slice locations so that the plurality of slice locations have different phases during the readout of at least one of the plurality of imaging signals.

Methods, computing devices, and magnetic resonance imaging systems that improve image quality in turbo spiral echo (TSE) imaging are disclosed. With this technology, a TSE pulse sequence is generated that includes a series of radio frequency (RF) refocusing pulses to produce a corresponding series of nuclear magnetic resonance (NMR) spin echo signals. A gradient waveform including a plurality of segments is generated. The plurality of segments collectively comprise a spiral ring retraced in-out trajectory. During an interval adjacent to each of the series of RF refocusing pulses, a first gradient pulse is generated according to the gradient waveform. The first gradient pulses encode the NMR spin echo signals. An image is then constructed from digitized samples of the NMR spin echo signals obtained based at least in part on the encoding.

Training a neural network to correct motion-induced artifacts in magnetic resonance images includes acquiring motion-free magnetic resonance image (MRI) data of a target object and applying a spatial transformation matrix to the motion-free MRI data. Multiple frames of MRI data are produced having respective motion states. A Non-uniform Fast Fourier Transform (NUFFT) can be applied to generate respective k-space data sets corresponding to each of the multiple frames of MRI; the respective k-space data sets can be combined to produce a motion-corrupted k-space data set and an adjoint NUFFT can be applied to the motion-corrupted k-space data set. Updated frames of motion-corrupted MRI data can be formed. Using the updated frames of motion corrupted MRI data, a neural network can be trained that generates output frames of motion free MRI data; and the neural network can be saved.

The present disclosure provides a system and method for magnetic resonance imaging. The method may include obtaining first k-space data collected from a subject in a non-Cartesian sampling manner. The method may also include generating second k-space data by regridding the first k-space data. The method may further include generating third k-space data by calibrating the second k-space data, wherein a calibrated field of view (FOV) corresponding to the third k-space data is constituted by a central portion of an intermediate FOV corresponding to the second k-space data. The method may still further include reconstructing, using at least one of a compressed sensing algorithm or a parallel imaging algorithm, a magnetic resonance (MR) image of the subject based at least in part on the third k-space data.

Systems and methods for determining a distribution map of susceptibility property of an object are provided. The method may include one or more of the following operations. A phase diagram corresponding to a magnetic resonance (MR) signal of the object may be obtained. A preliminary field map may be determined based on the phase diagram. Preliminary error limiting information associated with the preliminary field map may be obtained. A preliminary distribution map of susceptibility property of the object may be determined based on the preliminary field map and the preliminary error limiting information. An iteration process including at least one iteration may be performed to determine a target distribution map of susceptibility property of the object.