Abstract: A system is provided for MRI coil sensitivity estimation and reconstruction At least two cascades of regularization networks are serially connected such that the output of a cascade is used as input of a following cascade, at least two deepsets coil sensitivity map networks are serially connected such that the output of a deepsets coil sensitivity map network is used as input of a following deepsets coil sensitivity map network (CR), and wherein the outputs of the deepsets coil sensitivity map networks are also used as inputs for the cascades.

Abstract: For machine training and application of a trained complex-valued machine learning model, an activation function of the machine learning model, such as a neural network, includes a learnable parameter that is complex or defined in a complex domain with two dimensions, such as real and imaginary or magnitude and phase dimensions. The complex learnable parameter is trained for any of various applications, such as MR fingerprinting, other medical imaging, or non-medical uses.

Abstract: Object specific in-homogeneities in an MRI system are corrected. Prescan information available at the MR imaging system is determined. The prescan information includes at least object specific information of an object located in the MR imaging system from which an MR image is to be generated. The prescan information does not include a B1 map of the MRI system with the object being present in the MR imaging system. The prescan information is applied to a trained machine learning module provided at the MRI system. The trained machine learning module determines and generates shimming information as output. The shimming information is applied to a shimming module of the MR imaging system, wherein the shimming module uses the shimming information to generate a corrected magnetic field B0.

Abstract: Systems and methods for generating a synthetic image are provided. An input medical image in a first modality is received. A synthetic image in a second modality is generated from the input medical image. The synthetic image is upsampled to increase a resolution of the synthetic image. An output image is generated to simulate image processing of the upsampled synthetic image. The output image is output.

Abstract: For radial sampling in magnetic resonance imaging (MRI), a rescaling factor is determined from k-space data for each coil. The rescale factor is inversely proportional to the streak energy in the k-space data. The k-space data from the coils is rescaled for reconstruction, such as weighting the k-space data by the rescale factor in a data consistency term of iterative reconstruction. The rescale factor is additionally or alternatively used to determine a correction field for correction of intensity bias applied to intensities in the image-object space after reconstruction. These approaches may result in a diagnostically useful bias-corrected image with reduced streak artifact while benefiting from the efficient computation (i.e., computer operates to reconstruct more quickly).

Abstract: For reconstruction in medical imaging, such as reconstruction in MR imaging, a high-resolution image is reconstructed using a generator of a progressive generative adversarial network (PGAN or progressive GAN). In machine training the network, both the generator and discriminator of the GAN are grown progressively: starting from a low resolution, new layers are added that model finer details as training progresses. The resulting generator may be better able to handle high-resolution information than a generator of a GAN.

Abstract: A system is provided for MRI coil sensitivity estimation and reconstruction At least two cascades of regularization networks are serially connected such that the output of a cascade is used as input of a following cascade, at least two deepsets coil sensitivity map networks are serially connected such that the output of a deepsets coil sensitivity map network is used as input of a following deepsets coil sensitivity map network (CR), and wherein the outputs of the deepsets coil sensitivity map networks are also used as inputs for the cascades.

Abstract: A computer-implemented method includes, based on an input dataset defining an input image, determining a reconstructed image using a reconstruction algorithm, and executing a data-consistency operation for enforcing consistency between the input image and the reconstructed image. The data-consistency operation determines, for multiple K-space positions at which the input dataset comprises respective source data, a contribution of respective K-space values associated with the input dataset to a K-space representation of the reconstructed image.

Abstract: A magnetic resonance (MR) image may be created from MR data by receiving the MR data, applying a transform to the MR data, where a result of the applying is an image space representation of the MR data, determining a wrapped phase map of the image space representation of the MR data, obtaining an unwrapped phase map based on the wrapped phase map, scaling the unwrapped phase map into a B0 field map, reconstructing the MR image based on the MR data, correcting the MR image based on the B0 field map, and outputting the MR image. The scaling may be free of accounting for effects on the MR data by artifact sources secondary to B0 field inhomogeneities.

Abstract: In one approach, VDAMP is improved to allow multiple coils. The aliasing is modeled in the wavelet domain with spatial modulation for each of the frequency subbands. The spatial modulation uses the coil sensitivities. As a result of the spatial modulation, the aliasing modeling more closely models the variance allowing the regularization to use denoising operations. In another approach, the regularization computation may be simplified by using a machine-learned network in VDAMP. To account for the aliasing modeling of VDAMP, a convolutional neural network is trained with input of both the noisy image and the covariances of the aliasing model.

Abstract: For k-space trajectory infidelity correction, a model is machine trained to correct k-space measurements in k-space. K-space trajectory infidelity correction uses deep learning. Trajectory infidelity is corrected from a k-space point of view. Since the image artifacts arise from k-space acquisition distortion, a machine learning model is trained to correct in k-space, either changing values of k-space measurements or estimating the trajectory shifts in k-space.

Abstract: Machine training a network for and use of the machine-trained network are provided for tissue parameter estimation for a magnetic scanner using magnetic resonance fingerprinting. The machine-trained network is trained to both reconstruct a fingerprint image or fingerprint and to estimate values for multiple tissue parameters in magnetic resonance fingerprinting. The reconstruction of the fingerprint image or fingerprint may reduce noise, such as aliasing, allowing for more accurate estimation of the values of the multiple tissue parameters from the under sampled magnetic resonance fingerprinting information.

Abstract: For reconstruction in medical imaging, such as reconstruction in MR imaging, an iterative, hierarchal network for regularization may decrease computational complexity. The machine-learned network of the regularizer is unrolled or made iterative. For each iteration, nested U-blocks form a hierarchy so that some of the down-sampling and up-sampling of some U-blocks begin and end with lower resolution data or features, reducing computational complexity.

Abstract: For medical imaging such as MRI, machine training is used to train a network for segmentation using both the imaging data and protocol data (e.g., meta-data). The network is trained to segment based, in part, on the configuration and/or scanner, not just the imaging data, allowing the trained network to adapt to the way each image is acquired. In one embodiment, the network architecture includes one or more blocks that receive both types of data as input and output both types of data, preserving relevant features for adaptation through at least part of the trained network.

Abstract: For medical imaging such as MRI, machine training is used to train a network for segmentation using both the imaging data and protocol data (e.g., meta-data). The network is trained to segment based, in part, on the configuration and/or scanner, not just the imaging data, allowing the trained network to adapt to the way each image is acquired. In one embodiment, the network architecture includes one or more blocks that receive both types of data as input and output both types of data, preserving relevant features for adaptation through at least part of the trained network.

Abstract: For reconstruction in medical imaging, such as reconstruction in MR imaging, the number of iterations in deep learning-based reconstruction may be reduced by including a learnable extrapolation in one or more iterations. Regularization may be provided in fewer than all of the iterations of the reconstruction. The result of either approach alone or both together is better quality reconstruction and/or less computationally expensive reconstruction.

Abstract: For medical imaging such as MRI, machine training is used to train a network for segmentation using both the imaging data and protocol data (e.g., meta-data). The network is trained to segment based, in part, on the configuration and/or scanner, not just the imaging data, allowing the trained network to adapt to the way each image is acquired. In one embodiment, the network architecture includes one or more blocks that receive both types of data as input and output both types of data, preserving relevant features for adaptation through at least part of the trained network.

Abstract: Magnetic resonance imaging (MRI) image reconstruction using machine learning is described. A variational or unrolled deep neural network can be used in the context of an iterative optimization. In particular, a regularization operation can be based on a deep neural network. The deep neural network can take, as an input, an aliasing data structure being indicative of aliasing artifacts in one or prior images of the iterative optimization. The deep neural networks can be trained to suppress aliasing artifacts.

Abstract: For k-space trajectory infidelity correction, a model is machine trained to correct k-space measurements in k-space. K-space trajectory infidelity correction uses deep learning. Trajectory infidelity is corrected from a k-space point of view. Since the image artifacts arise from k-space acquisition distortion, a machine learning model is trained to correct in k-space, either changing values of k-space measurements or estimating the trajectory shifts in k-space.

Abstract: Systems and methods are provided for iterative reconstruction of a magnetic resonance image using magnetic resonance fingerprinting. An image series is estimated according to the following four steps: a gradient step to improve data consistency, fingerprint matching, spatial regularization, and a merging step. The fingerprint matching and spatial regularization steps are performed in parallel.