General Electric Company

Ge Precision Healthcare LLC

University of Michigan

Purdue Research Foundation

Rensselaer Polytechnic Institute

University of Notre Dame Du Lac

Sathish Ramani

Bruno Kristiaan Bernard De Man

Lin Fu

Peter Michael Edic

Jed Douglas Pack

Jeffrey Allen Fessler

Jiang Hsieh

Jean-Baptiste Thibault

James Vradenburg Miller

Zhou Yu

Debashish Pal

Ge Wang

Charles Addison Bouman

Jeffrey Michael Ashe

Brian Edward Nett

Jiahua Fan

Graeme Colin McKinnon

Ken David Sauer

Ileana Hancu

Sangtae Ahn

Bruno De Man

Xue Rui

Mingye Wu

Jie Tang

Somesh Srivastava

Donghwan Kim

Eri Haneda

Rolf Feodor Schulte

Madison Gray McGaffin

Matthew Getzin

A computer-implemented method for image processing is provided. The method includes obtaining data acquired by a medical imaging system. The method also includes normalizing the data. The method further includes de-noising the normalized data utilizing a deep learning-based denoising network. The method even further includes de-normalizing the de-noised data. The method yet further includes generating blended data based on both the data and the de-normalized de-noised data.

System and method for normalizing dynamic range of data acquired utilizing medical imaging

Various methods and systems are provided for multi-material decomposition for computed tomography. In one embodiment, a method comprises acquiring, via an imaging system, projection data for a plurality of x-ray spectra, estimating path lengths for a plurality of materials based on the projection data and calibration data for the imaging system, iteratively refining the estimated path lengths based on a linearized model derived from the calibration data, and reconstructing material-density images for each material of the plurality of materials from the iteratively-refined estimated path lengths. By determining path-length estimates in this way without modeling the physics of the imaging system, accurate material decomposition may be performed more quickly and with less sensitivity to changes in physics of the system, and furthermore may be extended to more than two materials.

Methods and systems for multi-material decomposition

Various methods and systems are provided for spectral computed tomography (CT) imaging. In one embodiment, a method comprises performing a scan of a subject to acquire, with a detector array comprising a plurality of detector elements, projection data of the subject, generating corrected path-length estimates based on the projection data and one or more selected correction functions, and reconstructing at least one material density image based on the corrected path-length estimates. In this way, the fidelity of spectral information is improved, thereby increasing image quality for spectral computed tomography (CT) imaging systems, especially those configured with photon-counting detectors.

Methods and systems for improved spectral fidelity for material decomposition

The present disclosure relates to image reconstruction with favorable properties in terms of noise reduction, spatial resolution, detail preservation and computational complexity. The disclosed techniques may include some or all of: a first-pass reconstruction, a simplified datafit term, and/or a deep learning denoiser. In various implementations, the disclosed technique is portable to different CT platforms, such as by incorporating a first-pass reconstruction step.

Iterative image reconstruction framework

A computer-implemented method for image processing is provided. The method includes acquiring multiple multi-energy spectral scan datasets and computing basis material images representative of multiple basis materials from the multi-energy spectral scan datasets, wherein the multiple basis material images include correlated noise. The method also includes jointly denoising the multiple basis material images in at least a spectral domain utilizing a deep learning-based denoising network to generate multiple de-noised basis material images.

System and method for processing data acquired utilizing multi-energy computed tomography imaging

A system and method for localizing a deep brain stimulation electrode in vivo in a subject or object is provided. A magnetic resonance imaging system obtains MR image data from a volume-of-interest by way of a zero echo time (ZTE) or ultrashort echo time (UTE) pulse sequence, with one or more of a phase domain image and a magnitude domain image being analyzed from the MR image data acquired by the ZTE or UTE pulse sequence. One or more electrodes are localized within the volume-of-interest based on an analysis of the phase domain image and/or magnitude domain image. In localizing the electrodes, a multi-scale correlation-based analysis of the volume-of-interest is performed to estimate at least one of an electrode center and electrode contact locations of the electrode, with the localization being achieved with a fast scan-time and with a high level of accuracy and precision.

System and method for localization of deep brain stimulation electrode via magnetic resonance imaging

The present approach relates to the use of machine learning and deep learning systems suitable for solving large-scale, space-variant tomographic reconstruction and/or correction problems. In certain embodiments, a tomographic transform of measured data obtained from a tomography scanner is used as an input to a neural network. In accordance with certain aspects of the present approach, the tomographic transform operation(s) is performed separate from or outside the neural network such that the result of the tomographic transform operation is instead provided as an input to the neural network. In addition, in certain embodiments, one or more layers of the neural network may be provided as wavelet filter banks.

Tomographic reconstruction based on deep learning

A method for characterizing anatomical features includes receiving scanned data and image data corresponding to a subject. The scanned data comprises sinogram data. The method further includes identifying a first region in an image of the image data corresponding to a region of interest. The method also includes determining a second region in the scanned data. The second region corresponds to the first region. The method further includes identifying a sinogram trace corresponding to the region of interest. The sinogram trace comprises sinogram data present within the second region. The method includes determining a data feature of the subject based on the sinogram trace and a deep learning network. The method also includes determining a diagnostic condition corresponding to a medical condition of the subject based on the data feature.

System and method for characterizing anatomical features

According to some embodiments, system and methods are provided comprising determining one or more image locations at which motion occurred within an image volume of an object containing movable anatomical features prior to segmenting the movable anatomical features; estimating motion at each of the one or more image locations; correcting one or more motion artifacts in the image volume at each of the one or more image locations, where motion was estimated resulting in the generation of a final motion compensated image; and segmenting one or more features of interest in the final motion compensated image. Numerous other aspects are provided.

Method for segmenting small features in an image volume

Methods, systems, and non-transitory computer readable media for image reconstruction are presented. Measured data corresponding to a subject is received. A preliminary image update in a particular iteration is determined based on one or more image variables computed using at least a subset of the measured data in the particular iteration. Additionally, at least one momentum term is determined based on the one or more image variables computed in the particular iteration and/or one or more further image variables computed in one or more iterations preceding the particular iteration. Further, a subsequent image update is determined using the preliminary image update and the momentum term. The preliminary image update and/or the subsequent image update are iteratively computed for a plurality of iterations until one or more termination criteria are satisfied.

Ordered subsets with momentum for X-ray CT image reconstruction

The use of the channelized preconditioners in iterative reconstruction is disclosed. In certain embodiments, different channels correspond to different frequency sub-bands and the output of the different channels can be combined to update an image estimate used in the iterative reconstruction process. While individual channels may be relatively simple, the combined channels can represent complex spatial variant operations. The use of channelized preconditioners allows empirical adjustment of individual channels.

Iterative reconstruction in image formation

A method is provided for reconstructing an image of an object that includes image elements. The method includes accessing measurement data associated with the image elements, introducing an auxiliary variable to transform an original problem of reconstructing the image to a constrained optimization problem, and solving the constrained optimization problem using a method of multipliers to create a sequence of sub-problems and solve the sequence of sub-problems. Solving the sequence of sub-problems includes reconstructing the image by optimizing a first objective function. The first objective function is optimized by iteratively solving a nested sequence of approximate optimization problems. An inner loop iteratively optimizes a second objective function approximating the first objective function. An outer loop utilizes the solution of the second objective function to optimize the first objective function.

Method and apparatus for iterative reconstruction

Growing community of inventors

Niskayuna, NY, United States of America

Sathish Ramani