Koninklijke Philips Corporation N.V.

Karsten Sommer

Thomas Erik Amthor

Peter Koken

Axel Saalbach

Peter Boernert

Mariya Ivanova Doneva

Michael Gunter Helle

Jan Jakob Meineke

Steffen Weiss

Rafael Wiemker

Sascha Krueger

Jochen Keupp

Tim Philipp Harder

Tom Brosch

Christophe Michael Jean Schulke

Tobias Klinder

Thomas Buelow

Johan Samuel Van Den Brink

Tim Nielsen

Jochen Peters

Julien Senegas

Irina Waechter-Stehle

Ulrich Wolfgang Katscher

Holger Eggers

Frank Michael Weber

Ingmar Graesslin

Rolf Jürgen Weese

Arne Ewald

Tanja Nordhoff

George Randall Duensing

Christophe Michael Jean Schuelke

Julian Senegas

Christophe Schuelke

Christophe Michael Jean Schülke

The invention provides for a method of training a neural network () configured for providing a further processing location (). The method comprises providing () a labeled medical image (), wherein the labeled medical image comprises multiple labels each indicating a truth processing location (). The method further comprises inputting () the labeled medical image into the neural network to obtain one trial processing location. The one trial processing location comprises a most likely trial processing location (). The method further comprises determine () the closest truth processing location () for the most likely trial processing location. The method further comprises calculating () an error vector () using the closest truth processing location and the most likely trial processing location. The method further comprises training () the neural network using the error vector.

Determination of a further processing location in magnetic resonance imaging

Disclosed herein is a medical system () comprising a memory () storing machine executable instructions () and an image generating neural network (). The image generating neural network is configured for outputting synthetic magnetic resonance image data () in response to receiving reference magnetic resonance image data () as input. The synthetic magnetic resonance image data is a simulation of magnetic resonance image data acquired according to a first configuration of a magnetic resonance imaging system when the reference magnetic resonance image data is acquired according to a second configuration of the magnetic resonance imaging system. Execution of the machine executable instructions causes a computational system () to: receive () measured k-space data () acquired according to the first configuration of the magnetic resonance imaging system; receive () the reference magnetic resonance image data; receive () the synthetic magnetic resonance image data by inputting the reference magnetic resonance image data into the image generating neural network; and reconstruct () corrected magnetic resonance image data () from the measured k-space data and the synthetic magnetic resonance image data.

Correction of magnetic resonance images using multiple magnetic resonance imaging system configurations

A computer-implemented method for preparing a subject in medical imaging, comprising: obtaining a series of images of a region of interest comprising at least a part of the subject, wherein the series of images comprises at least a first image and at least a subsequent, second image (S); determining a position of at least one landmark from the series of images, wherein the at least one landmark is anatomically related to a target anatomy (S); determining a confidence level assigned to the position of the at least one landmark (S); determining the position of the target anatomy based on the position of the at least one landmark, and the confidence level (S); providing the position of the target anatomy for preparing the subject in medical imaging (S).

Patient preparation for medical imaging

Disclosed herein is a medical system () comprising a memory () storing machine executable instructions () and a convolutional neural network (). The convolutional neural network is configured for receiving an initial Dixon magnetic resonance image () as input. The convolutional neural network is configured for identifying one or more water-fat swap regions () in the initial Dixon magnetic resonance image. The medical system further comprises a processor () for controlling the medical system. Execution of the machine executable instructions causes the processor to: receive () the initial Dixon magnetic resonance image; and receive () the one or more water-fat swap regions from the convolutional neural network in response to inputting the initial Dixon magnetic resonance image into the convolutional neural network.

Automated detection of water-fat swaps in Dixon magnetic resonance imaging

The invention relates to a method of MR imaging of an object () positioned in an examination volume of a MR device (). It is an object of the invention to enable efficient and high-quality non-Cartesian MR imaging, even in situations of strong Binhomogeneity. In accordance with the invention, the method comprises: —subjecting the object to an imaging sequence comprising at least one RF excitation pulse and modulated magnetic field gradients, —acquiring MR signals along at least one non-Cartesian k-space trajectory, —reconstructing an MR image from the acquired MR signals, and —detecting one or more mal-sampling artefacts caused by Binhomogeneity induced insufficient k-space sampling in the MR image using a deep learning network. Moreover, the invention relates to a MR device () and to a computer program.

Spiral MR imaging with off-resonance artefact correction

A system () and computer-implemented method are provided for data collection for distributed machine learning of a machine learnable model. A privacy policy data () is provided defining computer-readable criteria for limiting a selection of medical image data () to a subset of the medical image data to obfuscate an identity of the at least one patient. The medical image data is selected based on the computer-readable criteria to obtain privacy policy-compliant training data () for transmission to another entity. The system and method enable medical data collection at clinical sites without requiring manual oversight, and enables such selections to be made automatically, e.g., based on a request for medical image data which may be received from outside of the clinical site.

Medical data collection for machine learning

The invention relates to a system for assessing a pulmonary image which allows for an improved assessment with respect to lung nodules detectability. The pulmonary image is smoothed for providing different pulmonary images (20, 21, 22) with different degrees of smoothing, wherein signal values and noise values, which are indicative of the lung vessel detectability and the noise in these images, are determined and used for determining an image quality being indicative of the usability of the pulmonary image to be assessed for detecting lung nodules. Since a pulmonary image shows lung vessels with many different vessel sizes and with many different image values, which cover the respective ranges of potential lung nodules generally very well, the image quality determination based on the different pulmonary images with different degrees of smoothing allows for a reliable assessment of the pulmonary image's usability for detecting lung nodules. The image quality is used to determine a radiation dose level to be applied for generating a next pulmonary image.

System and method for assessing a pulmonary image

A magnetic resonance imaging system including a memory configured to store machine executable instructions, pulse sequence commands, and a first machine learning model including a first deep learning network. The pulse sequence commands are configured for controlling the magnetic resonance imaging system to acquire a set of magnetic resonance imaging data. The first machine learning model includes a first input and a first output, a processor, wherein execution of the machine executable instructions causes the processor to control the magnetic resonance imaging system to repeatedly perform an acquisition and analysis process including: acquiring a dataset including a subset of the set of magnetic resonance imaging data from an imaging zone of the magnetic resonance imaging system according to the pulse sequence commands, providing the dataset to the first input of the first machine learning model, in response to the providing, receiving a prediction of a motion artifact level of the acquired magnetic resonance imaging data from the first output of the first machine learning model, the motion artifact level characterizing a number and/or extent of motion artifacts present in the acquired magnetic resonance imaging data.

Motion artifact prediction during data acquisition

The invention provides for a magnetic resonance imaging system () for acquiring magnetic resonance data () from a subject () within a measurement zone (). The magnetic resonance imaging system () comprises: a processor () for controlling the magnetic resonance imaging system () and a memory () storing machine executable instructions (), pulse sequence commands () and a dictionary (). The pulse sequence commands () are configured for controlling the magnetic resonance imaging system () to acquire the magnetic resonance data () of multiple steady state free precession (SSFP) states per repetition time. The pulse sequence commands () are further configured for controlling the magnetic resonance imaging system () to acquire the magnetic resonance data () of the multiple steady state free precession (SSFP) states according to a magnetic resonance fingerprinting protocol. The dictionary () comprises a plurality of tissue parameter sets. Each tissue parameter set is assigned with signal evolution data pre-calculated for multiple SSFP states.

Multi-state magnetic resonance fingerprinting

The present disclosure relates to a computer implemented medical analysis method for predicting metastases () in a test tissue sample, the method comprising: providing a first machine learning model () having an input and an output, receiving a description () of a tumor () and first image data () of a test tissue sample of an anatomy region (), the test tissue sample being free of metastases (), providing the first image data () and the tumor description () to the input of the first machine learning model (), in response to the providing, receiving from the output of the first machine learning model () a prediction of occurrence of metastases () originating from the tumor () in the test tissue sample, and providing the prediction.

Medical analysis method for predicting metastases in a test tissue sample

The invention provides for a magnetic resonance imaging system (). The execution of machine executable instructions causes a processor () controlling the magnetic resonance imaging system to control () the magnetic resonance imaging system to acquire the magnetic resonance imaging data () using pulse sequence commands () and reconstruct () a magnetic resonance image (). Execution of the machine executable instructions causes the processor to receive () a list of suggested pulse sequence command changes () by inputting the magnetic resonance image and image metadata () into an MRI artifact detection module (&#x2033;). The MRI artifact detection module comprises at least one neural network, which has been trained using images from failed magnetic resonance imaging protocols and/or magnetic resonance data extracted from the magnetic resonance imaging protocols labeled as failed accessed from a log file () which logs the execution of previous magnetic resonance imaging protocols. Execution of the machine executable instructions further causes the processor to receive () a selection of a chosen pulse sequence command change () from the list of suggested pulse sequence command changes. Execution of the machine executable instructions further causes the processor to modify () the pulse sequence commands using the chosen pulse sequence command change.

Automatic artifact detection and pulse sequence modification in magnetic resonance imaging

The invention relates to a magnetic resonance imaging data processing system () for processing motion artifacts in magnetic resonance imaging data sets using a deep learning network () trained for the processing of motion artifacts in magnetic resonance imaging data sets. The magnetic resonance imaging data processing system () comprises a memory () storing machine executable instructions () and the trained deep learning network (). Furthermore, the magnetic resonance imaging data processing system () comprises a processor () for controlling the magnetic resonance imaging data processing system. Execution of the machine executable instructions () causes the processor () to control the magnetic resonance imaging data processing system () to: receive a magnetic resonance imaging data set (), apply the received magnetic resonance imaging data set () as an input to the trained deep learning network (), process one or more motion artifacts present in the received magnetic resonance imaging data set () using the trained deep learning network ().

Deep learning based processing of motion artifacts in magnetic resonance imaging data

The invention provides for a medical instrument () comprising a medical imaging system () configured for acquiring medical imaging data () from a subject (); a subject support () configured for supporting the subject during acquisition of the medical imaging data; and an optical imaging system (') configured for acquiring optical imaging data () of the subject on the subject support. The execution of the machine executable instructions causes a processor () controlling the medical instrument to: control () the optical imaging system to acquire the optical imaging data; generate () the initial vector () using the optical imaging data; generate () the synthetic image by inputting the initial vector into a generator neural network; calculate () a difference () between the synthetic image and the optical imaging data; and provide () a warning signal () if the difference differs by a predetermined threshold. The generator neural network is trained to generate a synthetic image () of the subject on the subject support in response to inputting an initial vector.

Automated detection of abnormal subject configuration for medical imaging

The invention provides for a method of operating a magnetic resonance imaging system for imaging a subject. The method comprises acquiring () tagged magnetic resonance data () and a first portion () of fingerprinting magnetic resonance data by controlling the magnetic resonance imaging system with tagging pulse sequence commands (). The tagging pulse sequence commands comprise a tagging inversion pulse portion () for spin labeling a tagging location within the subject. The tagging pulse sequence commands comprise a background suppression portion (). The background suppression portion comprises MRF pulse sequence commands for acquiring fingerprinting magnetic resonance data according to a magnetic resonance fingerprinting protocol. The tagging pulse sequence commands comprise an image acquisition portion (). The method comprises acquiring () control magnetic resonance data () and a second portion () of the fingerprinting magnetic resonance data by controlling the magnetic resonance imaging system with control pulse sequence commands. The control pulse sequence commands comprise a control inversion pulse portion (). The control pulse sequence commands comprise the background suppression portion ('). The control pulse sequence commands comprise the image acquisition portion (). The method comprises reconstructing () tagged magnitude images () using the tagged magnetic resonance data. The method comprises reconstructing () a control magnitude images () using the control magnetic resonance data. The method comprises constructing () an ASL image by subtracting the control magnitude images and the tagged magnitude images from each other. The method comprises reconstructing () a series of magnetic resonance fingerprinting images () using the first portion of the fingerprinting magnetic resonance data and/or the second portion of the fingerprinting magnetic resonance data. The method comprises generating () at least one magnetic resonance parametric map () by comparing the series of magnetic resonance fingerprinting images with a magnetic resonance fingerprinting dictionary.

Combined arterial spin labeling and magnetic resonance fingerprinting

A magnetic resonance imaging (MRI) system () includes a memory () for storing machine executable instructions () and magnetic resonance fingerprinting (MRF) pulse sequence commands () which cause the MRI system to acquire MRF magnetic resonance data () according to an MRF protocol. The pulse sequence commands are configured for acquiring the MRF magnetic resonance data in two-dimensional slices (), having a slice selection direction. A train of pulse sequence repetitions includes a sampling event where the MRF data is repeatedly sampled. Execution of the machine executable instructions causes a processor to control the MRI system to: acquire () the MRF magnetic resonance data; construct () a series () of at least one magnetic resonance parameter value for each voxel of the two dimensional slices; and calculate () a composition () of each of a set of predetermined substances within two or more sub-voxels () for each voxel of the two dimensional slices using a sub-voxel magnetic resonance fingerprinting dictionary () for each of the two or more sub-voxels and the series of the at least one magnetic resonance parameter value. Each voxel in the slice selection direction is divided into two or more sub-voxels.

Growing community of inventors

Hamburg, Germany

Karsten Sommer