General Electric Company

Ge Precision Healthcare LLC

Leland Stanford Junior University

Zhye Yin

Jed Douglas Pack

Bruno Kristiaan Bernard De Man

Kai Zeng

Bruno De Man

Jiang Hsieh

Gopal Biligeri Avinash

Samit Kumar Basu

Saad Ahmed Sirohey

Xin Wang

Peter Michael Edic

Brian Edward Nett

Jiahua Fan

Bruno Kristiaan Bernard DeMan

Maria Iatrou

Bruno K B De Man

Jie Tang

Xiaoyu Tian

Kyle Morgan Champley

Norbert Joseph Pelc

Mark Ernest Vermilyea

Xiaoye Wu

Daniel David Harrison

Yanfeng Du

Uwe Wiedmann

James Walter LeBlanc

Colin Richard Wilson

Francis Howard Little

Souma Sengupta

Forrest Frank Hopkins

Yannan Jin

Xue Rui

Paulo Ricardo Mendonca

Thomas Matthew Benson

Peter Claudius Sanza

Adam S Wang

Roy Arnulf Helge Nilsen

Dirk Bequé

Brian Patrick Smyth

Jong Chul Ye

Sen Wang

Paul Francis Fitzgerald

Bernhard Claus

Yirong Yang

Yangyang Yao

John Charles Hill

Michael Chunhe Gong

Ming Yan

Systems/techniques that facilitate generation of image denoising training data via photon-count splitting are provided. In various embodiments, a system can access a set of sinograms generated by a photon-counting computed tomography scanner. In various aspects, the system can split the set of sinograms into a first reduced-photon-count set of sinograms and a second reduced-photon-count set of sinograms. In various instances, the system can convert, via image reconstruction, the first reduced-photon-count set of sinograms into at least one training input image and the second reduced-photon-count set of sinograms into at least one training output image. In various cases, the system can train a deep learning neural network based on the at least one training input image and the at least one training output image.

Generation of image denoising training data via photon-count splitting

Systems and methods are provided for computed tomography (CT) imaging. In one embodiment, a method comprises adaptively blending at least two input image volumes with different spatially-variant noise characteristics to generate an output image volume with uniform noise throughout the output image volume. In this way, images may be reconstructed from projection data with data redundancy without introducing image artifacts from stitching images or variance in image noise due to the data redundancy.

Systems and methods for adaptive blending in computed tomography imaging

Computer processing techniques are described for reducing streaks in computed tomography (CT) images. According to an embodiment, computer-implemented method comprises obtaining, by a system comprising a processor, a pair of CT images reconstructed from a same set of projection data, the pair comprising a first image reconstructed from the projection data using a standard reconstruction process and a second image reconstructed from the projection data using a filtering reconstruction process that results in the second image comprising a first reduced level of streaks relative to the first image. The method further comprises generating, by the system, a third image by fusing a first subset of pixels extracted from one or more non-uniform areas in the first image and a second subset of pixels extracted from one or more uniform areas in the second image, wherein the third image comprises a second reduced level of streaks relative to the first image.

Computer processing techniques for streak reduction in computed tomography images

Methods and apparatus to automatically generate an image quality metric for an image are provided. An example method includes automatically processing a first medical image using a deployed learning network model to generate an image quality metric for the first medical image, the deployed learning network model generated from a digital learning and improvement factory including a training network, wherein the training network is tuned using a set of labeled reference medical images of a plurality of image types, and wherein a label associated with each of the labeled reference medical images indicates a central tendency metric associated with image quality of the image. The example method includes computing the image quality metric associated with the first medical image using the deployed learning network model by leveraging labels and associated central tendency metrics to determine the associated image quality metric for the first medical image.

Deep learning medical systems and methods for image reconstruction and quality evaluation

An imaging method includes executing a low-dose preparatory scan to an object by applying tube voltages and tube currents in an x-ray source, and generating a first image of the object corresponding to the low-dose preparatory scan. The method further includes generating image quality estimates and dose estimates view by view at least based on the first image. The method includes optimizing the tube voltages and the tube currents to generate optimal profiles for the tube voltage and the tube current. At least one of the optimal profiles for the tube voltage and the tube current is generated based on the image quality estimates and the dose estimates. The method includes executing an acquisition scan by applying the tube voltages and the tube currents based on the optimal profiles and generating a second image of the object corresponding to the acquisition scan. An imaging system is also provided.

Imaging method and system of tube voltage and current optimization

Approaches for acquiring CT image data corresponding to a full scan, but at a reduced dose are disclosed. In one implementation, X-ray tube current modulation is employed to reduce the effective dose. In other implementations, acquisition of sparse views, z-collimation, and two-rotation acquisition protocols may be employed to achieve a reduced dose relative to a full-scan acquisition protocol.

Method and system for reduced dose X-ray imaging

A system and method for CT projection extrapolation are provided. The method comprises receiving a CT projection for extrapolation. The method also comprises selecting a target patch comprising at least one pixel of a row to be extrapolated. The method further comprises generating a correlation profile between the target patch and one or more source patches, wherein the source patches comprise measured pixels in the CT projection in one or more rows adjacent to the target patch. The projection data is generated for at least one pixel of the target patch based on the correlation profile and the measured pixels of at least one of the source patches.

Methods and systems for CT projection domain extrapolation

Embodiments of the disclosure relate to projection-based volumetric dose estimation for X-ray systems, such as X-ray imaging systems. For example, in one embodiment, an X-ray system is capable of estimating an X-ray dose based on an energy interaction of the X-rays with respective portions of an object. In another embodiment, the X-ray dose estimate may be provided on a per voxel, per region, and/or per organ basis.

X-ray dose estimation technique

Embodiments of the disclosure relate to projection-based dose estimation for X-ray systems, such as X-ray imaging systems. For example, in one embodiment, an X-ray system is capable of estimating an X-ray dose based on an intensity profile of the detected X-rays that have passed through a scanned object and an estimated mass of the object. In one embodiment, the intensity profile may be compared to a baseline scan to acquire an estimate of energy interaction with the object.

An approach is disclosed for acquiring multi-sector computed tomography scan data. The approach includes activating an X-ray source during heartbeats of a patient to acquire projection data over a limited angular range for each heartbeat. The projection data acquired over the different is combined. An image having good temporal resolution is reconstructed using the combined projection data.

Multi-sector computed tomography image acquisition

Algorithms are disclosed that recombine acquired data so as to generate a substantially uniform and complete set of frequency data where frequency data might otherwise be incomplete. This process, or its equivalent, may be accomplished in a computationally efficient manner using filtering steps in one or both of the reconstruction space and/or the post-processing space.

Reconstruction of projection data to generate tomographic images having improved frequency characteristics

Growing community of inventors

Schenectady, NY, United States of America

Zhye Yin