Siemens Aktiengesellschaft

Siemens Medical Solutions Usa, Inc.

Siemens Healthcare Gmbh

Siemens Corporation

National Institutes of Health

Siemens Energy, Inc.

Ti-chiun Chang

Tong Fang

Mariappan S Nadar

Yunqiang Chen

Michael Zenge

Joachim Bamberger

Peter Durlak

Jan Ernst

Peter Speier

Stefan Böhm

Jason Jenn-Kwei Tyan

Jun Liu

Edgar Müller

Michelle Xiaohong Yan

Alban Lefebvre

Patrick Reeb

Edgar Mueller

Gregory G Slabaugh

Jens Gühring

Kai Tobias Block

Jeremy Rapin

Ziyan Wu

Marcel Dominik Nickel

Justinian Rosca

Heiko Claussen

Peter Schmitt

Gozde Unal

Xiaoguang Lu

Arun Innanje

Qiu Wang

Hui Xue

Andrei Szabo

Stefan Boehm

Kuan-Chuan Peng

Markus Lendl

Nirmal Janardhanan

Sandra Martin

Thomas Schiwietz

Thomas Pohl

Jeremy Ralph Wiles

Zhili Yang

Michelle Xiao-Hong Yan

Rameswar Panda

Markus Roβmeier

Jeremy-Ralph Wiles

Michael S Hansen

Sandra Knoth

Jiayu Song

Lin He

Aditi Roy

Ramesh Nair

Rameswar Rameswar

Deep learning is used to train a neural network for end-to-end prediction of short term (e.g., 20 minutes or less) solar irradiation based on camera images and metadata. The architecture of the neural network includes a recurrent network for temporal considerations. The images and metadata are input at different locations in the neural network. The resulting machine-learned neural network predicts solar irradiation based on camera images and metadata so that a solar plant and back-up power source may be controlled to minimize output power variation.

Solar irradiation prediction using deep learning with end-to-end training

A system and method for correcting short term irradiance prediction for control of a photovoltaic power station includes an irradiance data manager that generates an observed irradiance curve based on irradiance measurements received from an irradiation sensor, and a reference curve generation engine that generates a reference curve representative of irradiance values for a clear sky day. An irradiance prediction engine generates an irradiance prediction curve based on image segmentation and motion filtering of cloud pixels in sky images. A prediction correction engine corrects the irradiance prediction curve within a short term future time interval based on probabilistic analysis of time segmented intervals of the observed irradiance curve, the reference curve, and the irradiance prediction curve.

Automatic generation of reference curves for improved short term irradiation prediction in PV power generation

Systems and methods are provided for adapting images from different cameras so that a single trained classifier or an analyzer may be used. The classifier or analyzer operates on images that include a particular color distribution or characteristic. A generative network is used to adapt images from other cameras to have a similar color distribution or characteristic for use by the classifier or analyzer. A generative adversarial process is used to train the generative network.

Color adaptation using adversarial training networks

A method for estimating a coil sensitivity map for a magnetic resonance (MR) image includes providing a matrix A of sliding blocks of a 3D image of coil calibration data, calculating a left singular matrix Vfrom a singular value decomposition of A corresponding to &#x3c4; leading singular values, calculating P=VV, calculating a matrix that is an inverse Fourier transform of a zero-padded matrix P, and solving Mc=(S)cfor c, where cis a vector of coil sensitivity maps for all coils at spatial location r, and 

Eigen-vector approach for coil sensitivity maps estimation

Systems and methods are disclosed for processing an image to detect anomalous pixels. An image classification is received from a trained convolutional neural network (CNN) for an input image with a positive classification being defined to represent detection of an anomaly in the image and a negative classification being defined to represent absence of an anomaly. A backward propagation analysis of the input image for each layer of the CNN generates an attention mapping that includes a positive attention map and a negative attention map. A positive mask is generated based on intensity thresholds of the positive attention map and a negative mask is generated based on intensity thresholds of the negative attention map. An image of segmented anomalous pixels is generated based on an aggregation of the positive mask and the negative mask.

Weakly supervised anomaly detection and segmentation in images

A short-term cloud forecasting system includes a cloud segmentation processor that receives image data from images captured by an all sky camera. The cloud segmentation processor calculates a probability for each pixel in an image that the pixel is representative of a cloud. A cloud motion estimation processor calculates a motion vector representing estimated cloud motion calculates weights representing the likelihood that the cloud motion will cause a cloud to occlude the sun at a time in the near future. An uncertainty processor calculates one or more uncertainty indexes that quantify the confidence that a cloud forecast is accurate. Combining the cloud probabilities, the global motion vector and the at least one uncertainty index, in a sun occlusion prediction processor produces a short-term cloud forecast based on the image data that may be used as input to control systems for controlling a power grid.

Short term cloud forecast, improved cloud recognition and prediction and uncertainty index estimation

A computer-implemented method for performing isotropic reconstruction of Magnetic Resonance Imaging (MRI) data includes receiving a stack of slices acquired by an MRI device in two or more directions and reslicing the stack of slices into (i) an acquired view stack comprising high-resolution slices acquired in-plane, and (ii) a reslice stack comprising degraded slices acquired out of plane. An estimated slice profile is generated based on the stack of slices and the acquired view stack is convolved with the estimated slice profile to yield a simulated distorted slice stack. The simulated distorted slice stack is subtracted from the acquired view stack to yield a high-frequency band estimate and the high-frequency band estimate is combined with the reslice stack to yield isotropic reconstruction results.

Isotropic volume reconstruction from multiple 2D scans using super-resolution techniques

A method for parallel magnetic resonance imaging (MRI) reconstruction of digital images includes providing a set of acquired k-space MR image data v, a redundant Haar wavelet matrix W satisfying WW=I, wherein I is an identity matrix, a regularization parameter &#x3bb;&#x2267;0, and a counter limit k, initializing a variable z=Wv, and intermediate quantities p=q=0, calculating y=arg min&#xbd;&#x2225;z&#x2212;(p+z)&#x2225;+&#x3bb;&#x2225;z&#x2225;for 0&#x2266;i&#x2266;k, wherein z denotes values of an MR image sought to be reconstructed, updating p=(p+z)&#x2212;y, updating z=arg min&#xbd;&#x2225;z&#x2212;(q+z)&#x2225;+g(z), wherein and updating q=(q+y)&#x2212;z, wherein x=Wz is a solution of that specifies a reconstruction of the MR image.

Efficient redundant haar minimization for parallel MRI reconstruction

A system and methods to deblend seismic data from a plurality of sources and received by a plurality of sensors as shot gathers are disclosed. The deblending is performed by a Mutual Interdependence Analysis Method to separate contributions of different shots. Deblending is also performed by applying a measure of coherence in parallel data domains such as Common Shot Gather and Common Midpoint. Deblending is also achieved by using the hyperbolic nature of seismic data in the common midpoint domain. Deblended signals are estimated and are applied to create a seismic image. Also, Bergman iteration based migration is applied directly on the blended seismic shot gathers without first deblending as an alternative method. The methods are applied in seismic imaging for exploration of natural resources.

Methods for deblending of seismic shot gathers

A method for image reconstruction includes receiving under-sampled k-space data, determining a data fidelity term of a first image of the under-sampled k-space data in view of a second image of the under-sampled k-space data, wherein a time component separated the first image and the second image, determining a spatial penalization on redundant Haar wavelet coefficients of the first image in view of the second image, and optimizing the first image according the data fidelity term and the spatial penalization, wherein the spatial penalization selectively penalizes temporal coefficients and an optimized image of the first image is output.

Image reconstruction using redundant Haar wavelets

A reconstructed image is rendered of a patient by a processor from a set of undersampled MRI data by first subtracting two repetitions of the acquired data in k-space to create a third dataset. The processor reconstructs the image by minimizing an objective function under a constraint related to the third dataset, wherein the objective function includes applying a Karhunen-Loeve Transform (KLT) to a temporal dimension of data. The objective function under the constraint is expressed as arg min{&#x2225;&#x3c6;(f)&#x2225;subject to &#x2225;Af&#x2212;y&#x2225;&#x2266;&#x3b5;}. The reconstructed image is an angiogram which may be a 4D angiogram. The angiogram is used to diagnose a vascular disease.

Non-contrast-enhanced 4D MRA using compressed sensing reconstruction

A reconstructed image is rendered from a set of MRI data by first estimating an image with an area which does not contain artifacts or has an artifact with a relative small magnitude. Corresponding data elements in the estimated image and a trial image are processed, for instance by multiplication, to generate an intermediate data set. The intermediate data set is transformed and minimized iteratively to generate a reconstructed image that is free or substantially free of artifacts. In one embodiment a Karhunen-Loeve Transform (KLT) is used. A sparsifying transformation may be applied to generate the reconstructed image. The sparsifying transformation may be also not be applied.

Prior enhanced compressed sensing (PRINCE-CS) reconstruction for dynamic 2D-radial cardiac MRI

For resonance image data of an imaged subject, a method that first detects and estimates the dominant motions of k-space data (i.e., the motion vectors) and then constructs a graphical model for each estimated motion vector. The segments of the k-space that are determined to be corrupted by motion are restored by minimizing the energy associated with the corresponding graphical model. Consequently, the MR image of the imaged subject becomes free of motion artifacts.

Method for correcting motion artifacts in magnetic resonance images

A method for enhancing stent visibility in digital medical images includes providing a time series of 2-dimensional (2D) images of a stent in a vessel, estimating motion of the stent in a subset of images of the time series of images, estimating motion of clutter in the subset of images, where clutter comprises anatomical structures other than the stent, estimating a clutter layer in the subset of images from the estimated clutter motion, estimating a stent layer in the subset of images from the clutter layer and the estimated clutter motion, and minimizing a functional of the estimated stent motion, the estimated stent layer, the estimated clutter motion, and the estimated clutter layer to in calculate a refined stent layer image, where the refined stent layer image has enhanced visibility of the stent.

System and method for multi-image based stent visibility enhancement

A method for detecting markers within X-ray images includes applying directional filters to a sequence of X-ray image frames. Marker candidate pixels are determined based on the output of the directional filters. Candidate pixels are grouped into clusters and distances between each possible pair of clusters is determined and the most frequently occurring distance is considered an estimated distance between markers. A first marker is detected at the cluster that most closely resembles a marker based on certain criteria and a second marker is then detected at a cluster that is the estimated distance from the first marker. The pair of first and second marker detections is scored to determine detection quality. If the detected marker pair has an acceptable score then the detected marker pair is used.

Marker detection in X-ray images

Growing community of inventors

Princeton Junction, NJ, United States of America

Ti-chiun Chang