Ge Precision Healthcare LLC

General Electric Company

University of Southern California

Robert Marc Lebel

Daniel Vance Litwiller

Ersin Bayram

Xinzeng Wang

Graeme Colin McKinnon

Krishna S Nayak

Arnaud Guidon

Sagar Mandava

Yi Guo

Yinghua Zhu

Xucheng Zhu

Sajan Goud Lingala

Christopher Judson Hardy

Kevin Franklin King

Richard Scott Hinks

Florian Wiesinger

Luca Marinelli

Dawei Gui

Ek Tsoon Tan

Gaohong G Wu

Ali Ersoz

Zachary Slavens

Graeme C Mckinnon

Kang Wang

Moran Wei

Tim Sprenger

Suryanarayanan S Kaushik

Yannick Bliesener

Michael Carl

Embodiments of the present disclosure relate to systems and methods for reconstructing magnetic resonance imaging (MRI) images from under sampled k-space data. The method includes acquiring multi-coil under sampled k-space data with a fully-sampled calibration region, determining a k-space convolution kernel, and producing an initial multi-coil MRI image estimate via inverse Fourier transform. The estimate is regularized using a trained deep learning regularizer, transformed to synthetic fully-sampled k-space data, and convolved with the kernel to produce data-consistent synthetic k-space data. This is combined with the original under sampled data to produce an estimated fully-sampled k-space data, which is then inverse transformed to update the multi-coil MRI image estimate. The process iteratively improves image quality, reduces noise, and preserves features, enhancing MRI reconstruction from under sampled data.

Systems and methods for deep learning regularized reconstruction of under sampled magnetic resonance imaging data

A computer-implemented method for generating a chemical shift artifact corrected reconstructed image from magnetic resonance imaging (MRI) data includes inputting into a trained deep neural network an image generated from the MRI data acquired during a non-Cartesian MRI scan of a subject. The method also includes utilizing the trained deep neural network to generate the chemical shift artifact corrected reconstructed image from the image, wherein the trained deep neural network was trained utilizing a tissue mixing model that models interactions between different tissue types to mitigate chemical shift artifacts. The method further includes outputting from the trained deep neural network the chemical shift artifact corrected reconstructed image.

System and method for deep learning-based chemical shift artifact mitigation of non-Cartesian magnetic resonance imaging data

Methods and systems are provided for processing medical images using deep neural networks. In one embodiment, a medical image processing method comprises receiving a first medical image having a first characteristic and one or more acquisition parameters corresponding to acquisition of the first medical image, incorporating the one or more acquisition parameters into a trained deep neural network, and mapping, by the trained deep neural network, the first medical image to a second medical image having a second characteristic. The deep neural network may thereby receive at least partial information regarding the type, extent, and/or spatial distribution of the first characteristic in a first medical image, enabling the trained deep neural network to selectively convert the received first medical image.

Systems and methods for medical image processing using deep neural network

A computer-implemented method of reducing noise in magnetic resonance (MR) images is provided. The method includes executing a neural network model of analyzing MR images, wherein the neural network model is trained with a pair of pristine images and corrupted images. The pristine images are the corrupted images with noise reduced, and target output images of the neural network model are the pristine images. The method also includes receiving first MR signals and second MR signals, reconstructing first and second MR images based on the first MR signals and the second MR signals, and analyzing the first MR image and the second MR image using the neural network model. The method further includes deriving a denoised MR image based on the analysis, wherein the denoised MR image is a combined image based on the first MR image and the second MR image and outputting the denoised MR image.

Systems and methods of noise reduction in magnetic resonance images

A method for producing an image of a subject with a magnetic resonance imaging (MRI) comprises acquiring a first set of partial k-space data from the subject and generating a phase corrected image based on a phase correction factor and the first set of the partial k-space data. The method further includes transforming the phase corrected image into a second set of partial k-space data and reconstructing the image of the subject from the second set of the partial k-space data and a weighting function.

System and method for reconstruction using a high-resolution phase in magnetic resonance images

Methods and systems are provided for independently removing streak artifacts and noise from medical images, using trained deep neural networks. In one embodiment, streak artifacts and noise may be selectively and independently removed from a medical image by receiving the medical image comprising streak artifacts and noise, mapping the medical image to a streak residual and a noise residual using the trained deep neural network, subtracting the streak residual from the medical image to a first extent, and subtracting the noise residual from the medical image to a second extent, to produce a de-noised medical image, and displaying the de-noised medical image via a display device.

Methods and system for selective removal of streak artifacts and noise from images using deep neural networks

A magnetic resonance (MR) image processing system is provided. The system includes an MR image processing computing device that includes at least one processor. The processor is programmed to execute a neural network model configured to receive crude MR data as an input and output processed MR images associated with the crude MR data, the crude MR data and the processed MR images having the first number of dimensions. The processor is also programmed to receive a pair of pristine data and corrupted data both having a second number of dimensions lower than the first number of dimensions. The corrupted data are the pristine data added with primitive features. The processor is further programmed to train the neural network model using the pair of the pristine data and the corrupted data. The trained neural network model is configured to change primitive features associated with the crude MR data.

Systems and methods of magnetic resonance image processing using neural networks having reduced dimensionality

A computer-implemented method of removing truncation artifacts in magnetic resonance (MR) images is provided. The method includes receiving a crude image that is based on partial k-space data from a partial k-space that is asymmetrically truncated in at least one k-space dimension. The method also includes analyzing the crude image using a neural network model trained with a pair of pristine images and corrupted images. The corrupted images are based on partial k-space data from partial k-spaces truncated in one or more partial sampling patterns. The pristine images are based on full k-space data corresponding to the partial k-space data of the corrupted images, and target output images of the neural network model are the pristine images. The method further includes deriving an improved image of the crude image based on the analysis, wherein the derived improved image includes reduced truncation artifacts and increased high spatial frequency data.

Deep learning systems and methods of removal of truncation artifacts in magnetic resonance images

Tracer kinetic models are utilized as temporal constraints for highly under-sampled reconstruction of DCE-MRI data. In one embodiment, a method for improving dynamic contrast enhanced imaging. The method includes steps of administering a magnetic resonance contrast agent to a subject and then collecting magnetic resonance contrast agent from the subject. A tracer kinetic model (i.e. eTofts or Patlak) is selected to be applied to the magnetic resonance imaging data. The tracer kinetic model is applied to the magnetic resonance imaging data. Tracer kinetic maps and dynamic images are simultaneously reconstructed and a consistency constraint is applied. The proposed method allows for easy use of different tracer kinetic models in the formulation and estimation of patient-specific arterial input functions jointly with tracer kinetic maps.

Method for joint arterial input function and tracer kinetic parameter estimation in accelerated DCE-MRI using a model consistency constraint

Methods and systems are provided for de-noising medical images using deep neural networks. In one embodiment, a method comprises receiving a medical image acquired by an imaging system, wherein the medical image comprises colored noise; mapping the medical image to a de-noised medical image using a trained convolutional neural network (CNN); and displaying the de-noised medical image via a display device. The deep neural network may thereby reduce colored noise in the acquired noisy medical image, increasing a clarity and diagnostic quality of the image.

Systems and methods for reducing colored noise in medical images using deep neural network

Tracer kinetic models are utilized as temporal constraints for highly under-sampled reconstruction of DCE-MRI data. The method is flexible in handling any TK model, does not rely on tuning of regularization parameters, and in comparison to existing compressed sensing approaches, provides robust mapping of TK parameters at high under-sampling rates. In summary, the method greatly improves the robustness and ease-of-use while providing better quality of TK parameter maps than existing methods. In another embodiment, TK parameter maps are directly reconstructed from highly under-sampled DCE-MRI data. This method provides more accurate TK parameter values and higher under-sampling rates. It does not require tuning parameters and there are not additional intermediate steps. The proposed method greatly improves the robustness and ease-of-use while providing better quality of TK parameter maps than conventional indirect methods.

Method for improved dynamic contrast enhanced imaging using tracer-kinetic models as constraints

A computer-implemented method of correcting phase and reducing noise in magnetic resonance (MR) phase images is provided. The method includes executing a neural network model for analyzing MR images, wherein the neural network model is trained with a pair of pristine images and corrupted images, wherein the corrupted images include corrupted phase information, the pristine images are the corrupted images with the corrupted phase information reduced, and target output images of the neural network model are the pristine images. The method further includes receiving MR images including corrupted phase information, and analyzing the received MR images using the neural network model. The method also includes deriving pristine phase images of the received MR images based on the analysis, wherein the derived pristine phase images include reduced corrupted phase information, compared to the received MR images, and outputting MR images based on the derived pristine phase images.

Systems and methods of generating robust phase images in magnetic resonance images

Methods and systems are provided for deblurring medical images using deep neural networks. In one embodiment, a method for deblurring a medical image comprises receiving a blurred medical image and one or more acquisition parameters corresponding to acquisition of the blurred medical image, incorporating the one or more acquisition parameters into a trained deep neural network, and mapping, by the trained deep neural network, the blurred medical image to a deblurred medical image. The deep neural network may thereby receive at least partial information regarding the type, extent, and/or spatial distribution of blurring in a blurred medical image, enabling the trained deep neural network to selectively deblur the received blurred medical image.

Systems and methods for deblurring medical images using deep neural network

Methods and systems are provided for selectively denoising medical images. In an exemplary method, one or more deep learning networks are trained to map corrupted images onto a first type and a second type of artifacts present in corresponding corrupted images. Then the one or more trained learning networks are used to single out the first and second types of artifacts from a particular medical image. The first type of artifacts is removed to a first extent and the second type of artifacts is removed to a second extent. The first and second extents may be different. For example, one type of artifacts can be fully suppressed while the other can be partially removed form the medical image.

Growing community of inventors

Calgary, Canada

Robert Marc Lebel