Leland Stanford Junior University

General Electric Company

University of California

Ge Precision Healthcare LLC

The United States Government As Represented by the Department of Veterans Affairs

Shreyas S Vasanawala

John Mark Pauly

Joseph Yitan Cheng

Tao Zhang

Michael Lustig

Christopher Michael Sandino

Marcus Tedrow Alley

Brian A Hargreaves

Dwight G Nishimura

Albert Hsiao

Feiyu Chen

Frank Ong

Fraser Robb

Greig Scott

Thomas Grafendorfer

Valentina Taviani

David Y Zeng

Morteza Mardani Korani

Yuxin Chen

Julio A Oscanoa Aida

David Leigh Donoho

James S Tropp

Bruce L Daniel

Suchandrima Banerjee

Paul David Calderon

Peng Lai

Yan Wu

Daniel B Ennis

Cagan Alkan

Kanghyun Ryu

Jiacheng Jason He

Zheng Zhong

Mert Pilanci

Elizabeth K Cole

Matthew J Middione

Brain A Hargreaves

A method for magnetic resonance imaging (MRI) includes acquiring under-sampled k-space measurements from an MRI apparatus using multiple receiver coils; reconstructing an MRI image from the under-sampled k-space measurements and coil sensitivity maps using an unrolled neural network; generating reconstructed multi-coil k-space data from the MRI image by multiplying the MRI image by the coil sensitivity maps followed by performing a Fourier transform; estimating a k-space null-space convolutional kernel from fully-sampled k-space measurements in autocalibration signal lines of the under-sampled k-space measurements; solving a convex optimization problem to produce refined k-space data from the k-space null-space kernel, the under-sampled k-space measurements, and the reconstructed multi-coil k-space data; and producing a refined MRI image from the refined k-space data by performing an inverse Fourier transform followed by a coil combination using the coil sensitivity maps.

Method for improving high frequency image features and details in deep learning MRI reconstructions

A method for magnetic resonance imaging (MRI) performs a spoiled gradient-recalled (SPGR) MRI scan with an MRI scanner to produce MRI data; and reconstructs an MRI image from the MRI data; wherein performing the SPGR MRI scan comprises playing an interleaved-randomized spoiler (IRS) gradient after every M-th acquisition block, where M≥2, and where an absolute area of the IRS gradient of each IRS is randomized between zero and a maximum gradient area achievable on the MRI scanner.

Magnetic resonance imaging using 3D spoiled gradient-recalled sequence

In a method for magnetic resonance imaging pseudorandomly undersampled k-space imaging data is acquired with multiple receiver coils of an MRI imaging apparatus. MR image reconstruction is performed to produce a reconstructed MR image from the k-space imaging data by iteratively solving sketched approximations of an original reconstruction problem. The sketched approximations use a sketched model matrix Athat is a lower-dimensional version of an original model matrix A of the original reconstruction problem. The sketched model matrix Apreserves the Fourier structure of the MR reconstruction problem and reduces the number of coils actively used during reconstruction.

Randomized dimension reduction for magnetic resonance image iterative reconstruction

A method for magnetic resonance imaging performs chemical shift encoded imaging to produce complex dual-echo images which are then applied (with imaging parameters) as input to a deep neural network to produce as output water-only and fat-only images. The deep neural network can be trained with ground truth water/fat images derived from chemical shift encoded images using a conventional water-fat separation algorithm such as projected power approach, IDEAL, or VARPRO. The chemical shift encoded imaging comprises performing an image acquisition with the MRI scanner via a spoiled-gradient echo sequence or a spin-echo sequence.

Deep learning-based water-fat separation from dual-echo chemical shift encoded imaging

A method for MR imaging includes acquiring with an MR imaging apparatus undersampled k-space imaging data having one or more temporal dimensions and two or more spatial dimensions; transforming the undersampled k-space imaging data to image space data using zero-filled or sliding window reconstruction and sensitivity maps; decomposing the image space data into a compressed representation comprising a product of spatial and temporal parts, where the spatial part comprises spatial basis functions and the temporal part comprises temporal basis functions; processing the spatial basis functions and temporal basis functions to produce reconstructed spatial basis functions and reconstructed temporal basis functions, wherein the processing iteratively applies conjugate gradient and convolutional neural network updates using 2D or 3D spatial and 1D temporal networks; and decompressing the reconstructed spatial basis functions and reconstructed temporal basis functions to produce a reconstructed MRI image having one or more temporal dimensions and two or more spatial dimensions.

Method for high-dimensional image reconstruction using low-dimensional representations and deep learning

A method is disclosed for generating pixel risk maps for diagnostic image reconstruction. The method produces uncertainty/variance maps by feeding into a trained encoder a short-scan image acquired from a medical imaging scan to generate latent code statistics including the mean &#x3bc;and variance &#x3c3;; selecting random samples based on the latent code statistics, z&#x2dc;N(&#x3bc;,&#x3c3;); feeding the random samples into a trained decoder to generate a pool of reconstructed images; and calculating, for each pixel of the pool of reconstructed images, the pixel mean and variance statistics across the pool of reconstructed images. The risk of each pixel may be calculated using the Stein's Unbiased Risk Estimator on the input density compensated data, that involves calculating the end-to-end divergence of the deep neural network.

Methods for risk map prediction in AI-based MRI reconstruction

A method for magnetic resonance imaging corrects non-stationary off-resonance image artifacts. A magnetic resonance imaging (MRI) apparatus performs an imaging acquisition using non-Cartesian trajectories and processes the imaging acquisitions to produce a final image. The processing includes reconstructing a complex-valued image and using a convolutional neural network (CNN) to correct for non-stationary off-resonance artifacts in the image. The CNN is preferably a residual network with multiple residual layers.

Deep learning method for nonstationary image artifact correction

A method for phase-contrast magnetic resonance imaging (PC-MRI) acquires undersampled PC-MRI data using a magnetic resonance imaging scanner and reconstructs MRI images from the undersampled PC-MRI data by reconstructing a first flow-encoded image using a first convolutional neural network, reconstructing a complex difference image using a second convolutional neural network, combining the complex difference image and the first flow-encoded image to obtain a second flow-encoded image, and generating a velocity encoded image from the first flow-encoded image and second flow-encoded image using phase difference processing.

Artificial intelligence based reconstruction for phase contrast magnetic resonance imaging

A method of magnetic resonance imaging acquires undersampled MRI data and generates by an adversarially trained generative neural network MRI data having higher quality without using any fully-sampled data as a ground truth. The generative neural network is adversarially trained using a discriminative neural network that distinguishes between undersampled MRI training data and candidate undersampled MRI training data produced by applying an MRI measurement function containing an undersampling mask to generated MRI training data produced by the generative neural network from the undersampled MRI training data.

MRI image reconstruction from undersampled data using adversarially trained generative neural network

A method is disclosed for phase contrast magnetic resonance imaging (MRI) comprising: acquiring phase contrast 3D spatiotemporal MRI image data; inputing the 3D spatiotemporal MRI image data to a three-dimensional spatiotemporal convolutional neural network to produce a phase unwrapping estimate; generating from the phase unwrapping estimate an integer number of wraps per pixel; and combining the integer number of wraps per pixel with the phase contrast 3D spatiotemporal MRI image data to produce final output.

Method for correction of phase-contrast magnetic resonance imaging data using a neural network

A method for magnetic resonance imaging (MRI) includes steps of acquiring by an MRI scanner undersampled magnetic-field-gradient-encoded k-space data; performing a self-calibration of a magnetic-field-gradient-encoding point-spread function using a first neural network to estimate systematic waveform errors from the k-space data, and computing the magnetic-field-gradient-encoding point-spread function from the systematic waveform errors; reconstructing an image using a second neural network from the magnetic-field-gradient-encoding point-spread function and the k-space data.

Method for estimating systematic imperfections in medical imaging systems with deep learning

A magnetic resonance imaging scan performs an MRI acquisition using an undersampling pattern to produce undersampled k-space data; adds the undersampled k-space data to aggregate undersampled k-space data for the scan; reconstructs an image from the aggregate undersampled k-space data; updates the undersampling pattern from the reconstructed image and aggregate undersampled k-space data using a deep reinforcement learning technique defined by an environment, reward, and agent, where the environment comprises an MRI reconstruction technique, where the reward comprises an image quality metric, and where the agent comprises a deep convolutional neural network and fully connected layers; and repeats these steps to produce a final reconstructed MRI image for the scan.

Reinforcement learning for online sampling trajectory optimization for magnetic resonance imaging

A method for providing magnetic resonance imaging with dynamic contrast and 4D flow of a volume of an object in a magnetic resonance imaging (MRI) system is provided. Contrast agent is provided to the volume of the object. Magnetic resonance excitation from the MRI system is applied to the volume of the object. The MRI system reads out a subsample of less than 10% of spatially resolved data and velocity encoded data with respect to time. The readout subsample is used to determine both dynamic contrast and 4D flow.

Dynamic contrast enhanced magnetic resonance imaging with flow encoding

A method for magnetic resonance imaging performs unsupervised training of a deep neural network of an MRI apparatus using a training set of under-sampled MRI scans, where each scan comprises slices of under-sampled, unclassified k-space MRI measurements. The MRI apparatus performs an under-sampled scan to produce under-sampled k-space data, updates the deep neural network with the under-sampled scan, and processes the under-sampled k-space data by the updated deep neural network of the MRI apparatus to reconstruct a final MRI image.

Method for performing magnetic resonance imaging reconstruction with unsupervised deep learning

Various methods and systems are provided for reconstructing magnetic resonance images from accelerated magnetic resonance imaging (MRI) data. In one embodiment, a method for reconstructing a magnetic resonance (MR) image includes: estimating multiple sets of coil sensitivity maps from undersampled k-space data, the undersampled k-space data acquired by a multi-coil radio frequency (RF) receiver array; reconstructing multiple initial images using the undersampled k-space data and the estimated multiple sets of coil sensitivity maps; iteratively reconstructing, with a trained deep neural network, multiple images by using the initial images and the multiple sets of coil sensitivity maps to generate multiple final images, each of the multiple images corresponding to a different set of the multiple sets of sensitivity maps; and combining the multiple final images output from the trained deep neural network to generate the MR image.

Growing community of inventors

Stanford, CA, United States of America

Shreyas S Vasanawala