Koninklijke Philips Corporation N.V.

Koninklijke Philips Electronics N.v.

Philips Intellectual Property & Standards Gmbh

Daniel Bystrov

Stewart Young

Heinrich Schulz

Jens Von Berg

Torbjoern Vik

Thomas Netsch

Rafael Wiemker

Vladimir Pekar

Nicole Schadewaldt

Thomas Buelow

Harald Sepp Heese

Astrid Ruth Franz

Dominik Benjamin Kutra

Sven Kroenke

Axel Saalbach

Kongkuo Lu

Torbjorn Vik

Nataly Wieberneit

Cristian Lorenz

Jochen Peters

Julien Senegas

Yuechen Qian

Michael Kaus

Kirsten Regina Meetz

Sven Krönke

Juergen Weese

Tobias Klinder

Eric Cohen-Solal

Alexandra Groth

Martin Bergtholdt

Steffen Renisch

Prashant Kumar

Alfonso Agatino Isola

Thomas Blaffert

Peter Mazurkewitz

Pallavi Vajinepalli

Gabriel Ryan Mankovich

Christoph Neukirchen

Bernd Menser

Ingwer Curt Carlsen

Ranjith Naveen Tellis

Ulrich Neitzel

Johan Michiel Den Harder

Frederik Visser

Yogish Mallya

André Goossen, Sr

Rudolf Theodoor Springorum

Matthieu Frederic Bal

Biswaroop Chakrabarti

Ran Cohen

Marc Kouwenhoven

Bernd Lundt

Arjan Willem Simonetti

Arianne Van Muiswinkel

Soubhik Paul

Ruediger Grewer

Stefan Matthias Florian Schmidt

Torbjørn Vik

Torbjoern Vic

Bela Fadida

Lior Wolloch

Sebastian Dries

Cornelis Johannes Fransiscus Maria Bergmans

Cornelis Johannes Franciscus Maria Bergmans

Wendy De Kok

Heiner Matthias Brueck

An X-ray imaging system () includes an X-ray source (), an X-ray detector (), a depth camera (), and a processor (). The processor () receives depth camera image data from the depth camera, projects the depth camera image data onto a radiation-receiving surface of the X-ray detector (), from a perspective of the X-ray source (), and generates an image representation () of the projected depth camera image data on the radiation-receiving surface of the X-ray detector (), from a perspective of the depth camera ().

Positioning an object with respect to an X-ray detector

A computer-implemented method for positioning a subject in medical imaging, comprising: receiving a first image () of a region of interest () of the subject (S); determining first positioning data based on the first image (), wherein the first positioning data indicates an alignment of the region of interest () relative to a first image acquisition unit used to acquire the first image () (S); determining guidance data based on the first positioning data, wherein the guidance data comprises a guidance for an alignment of the region of interest () relative to a second image acquisition unit used to acquire a second image () from a current alignment to a target alignment, wherein the target alignment is to correspond to that derived from the first positioning data (S); providing the guidance data for acquiring the second image () (S).

Guidance for positioning a patient in medical imaging

An orientation of a patient's chest is determined by an apparatus and method. According to the invention, a camera image of a patient comprising image data of the patient's chest is received, wherein the camera image is acquired with an optical camera. Further, an X-ray radiograph of the patient's chest with an X-ray imaging axis extending from an X-ray source to an X-ray detector is received. Then, the camera image and the X-ray radiograph are provided to at least one processor. Further, information relating to a clavicle to spinous process distance of the patient is then determined using the camera image. Finally, an orientation of the patient's chest in the X-ray radiograph with respect to the X-ray imaging axis is determined using the camera image, the X-ray radiograph, and the information relating to the clavicle to spinal process distance of the patient.

Apparatus for determining an orientation of a patient's chest

The present invention relates to landmark and/or temporal event detection. It is proposed to utilize previously learned spatial statistical correlations between multiple landmarks in order to regularize convolutional neural networks (CNNs) either as a post-processing step or during training in order to utilize anatomical prior knowledge, reduce the false-positive prediction rate, and/or increase the accuracy and stability of the algorithm. The proposed apparatus and method may also be applied to improve the detection of correlated events in e.g., time-series by leveraging prior knowledge.

Probabilistic regularization of convolutional neural networks for multiple-feature detection based on correlations

In order to enhance enhanced X-ray image inhalation quality monitoring, a metric is proposed hat reproducibly provides an index of ribs visible to be used in the assessment of the inhalation state. In an example, a detected diaphragm in a chest X-ray image may be projected into an atlas that contains labels for all intercostal spaces, namely spaces between rib centerlines. A spatial representation of both the clavicle and the ribs is provided in the atlas, a cumulative histogram is built for all points, i.e. pixels, of the diaphragm, for every point a rib label counter of the rib in the rib label map at that point is incremented as well as all ribs above it, the rib label counter is normalized by a division by the number of points, the median (or a different quantile) may be taken of this distribution serving as an inhalation index. An objective metric of inhalation state is thus achieved.

Inhalation metric for chest x-ray images

The invention relates to a method () for supervised training of an artificial neural network for medical image analysis. The method comprises acquiring (SI) first and second sets of training samples, wherein the training samples comprise feature vectors and associated predetermined labels, the feature vectors being indicative of medical images and the labels pertaining to anatomy detection, to semantic segmentation of medical images, to classification of medical images, to computer-aided diagnosis, to detection and/or localization of biomarkers or to quality assessment of medical images. The accuracy of predetermined labels may be better for the second set of training samples than for the first set of training samples. The neural network is trained (S) by reducing a cost function, which comprises a first and a second part. The first part of the cost function depends on the first set of training samples, and the second part of the cost function depends on a first subset of training samples, the first subset being a subset of the second set of training samples. In addition, the second part of the cost function depends on an upper bound for the average prediction performance of the neural network for the first subset of training samples and the second part of the cost function is configured for preventing that the average prediction performance for the first subset of training samples exceeds the upper bound.

Constrained training of artificial neural networks using labelled medical data of mixed quality

An apparatus () for generating a training set of anonymized images () for training an artificial intelligence (AI) component () from images () of a plurality of persons. The apparatus includes at least one electronic processor () programmed to: spatially map the images of the plurality of persons to a reference image () to generate images () in a common reference frame; partition the images in the common reference frame into P spatial regions () to generate P sets of image patches () corresponding to the P spatial regions; assemble a set of training images () in the common reference frame by, for each training image in the common reference frame, selecting an image patch from each of the P sets of image patches and assembling the selected image patches into the training image in the common reference frame; and process the training images in the common reference frame to generate the training set of anonymized images including applying statistical inverse spatial mappings to the training images in the common reference frame, wherein the statistical inverse spatial mappings are derived from spatial mappings () of the images of the plurality of persons to the reference image.

Anatomical encryption of patient images for artificial intelligence

A resource management system for better operation of a plurality of devices. The system comprises an input interface (IN) for receiving input data including one or more characteristics of at least one work object (P-P) and/or including context data. The at least one work object (P-P) can be processed by one or more processing devices (M). The said processing is specified in a respective work specification (S-S). A predictor component (PC) of the system is configured to predict, based on the input data, a change to the work specification. The system comprises an output interface (OUT) for providing output data that represents said predicted change.

Resource scheduling in adaptive radiation therapy planning

The field of view of an X-ray imaging system should be set appropriately to ensure that anatomical information of interest is not omitted. In particular, it is necessary to ensure that the operator of an X-ray system does not allow a patient to leave the X-ray imaging system until it is certain that the correct anatomy has been imaged. This application discusses a technique enable the visualization of a field of view boundary error caused by the incorrect configuration of an X-ray imaging system. Optionally, the boundary error is displayed either on a user display of a system console, or by projecting the field of view error onto the patient in the X-ray system. Thus, an operator of the system may be alerted to the presence of a boundary error, enabling a new X-ray exposure to be taken, if necessary.

Visualizing collimation errors

Medical radiography requires specialist control of radiography equipment to achieve good imaging results. Typical errors that can occur consist of an inappropriate field of view being accidentally applied. This results in a 'cropping effect' in which portions of the region of interest of a patient which would be of clinical use are omitted from the image. Conventionally, the only solution is to re-take the entire image with a more appropriate (and inevitably larger) field of view selected. This is undesirable, because it might require recall of the patient from another location, and the patient will be subject to two exposures, thus undesirably increasing their X-ray dosage. The present application proposes to use an anatomical atlas to analyse an X-ray image output from an initial exposure, in particular to assess whether significant anatomical elements are missing from the image. If elements are missing, the field of view of the X-ray imager is recalibrated, using collimation or pan/tilt adjustments, for example. A subsequent X-ray image is obtained, and combined with the initial image, to provide an output image. Because only a small area of the region of interest may subsequently need to be exposed, a smaller additional dose results.

X-ray apparatus having a composite field of view

A system () for detecting and labeling structures of interest includes a current patient study database () containing a current patient study () with clinical contextual information (), a statistical model patient report database () containing at least one or more prior patient documents containing clinical contextual information (), an image metadata processing engine () configured to extract metadata for preparing an input for an anatomical structure classifier (), a natural language processing engine () configured to extract clinical context information () from the prior patient documents, an anatomical structure detection and labeling engine () or processor (), and a display device () configured to display findings from the current patient study. The anatomical structure detection and labeling engine () or processor () is configured to identify and label one or more structures of interest () from the extracted metadata and clinical context information () and aggregate series level data.

Optimized anatomical structure of interest labelling

The present invention relates to an apparatus for the detection of opacities in X-ray images. It is described to provide () an analysis X-ray image of a region of interest of an analyzed body part. A model of a normal region of interest is provided (), wherein the model is based on a plurality of X-ray images of the region of interest. At least one abnormality is detected () in the region of interest of the analyzed body part, the detection comprising comparing the analysis X-ray image of the region of interest and the model of the normal region of interest. Information is output () on the at least one abnormality.

Apparatus for the detection of opacities in X-ray images

When delineating anatomical structures in a medical image of a patient for radiotherapy planning, a processor () detects landmarks () in a low-resolution image (e.g., MRI or low-dose CT) and maps the detected landmarks to reference landmarks () in a reference contour of the anatomical structure. The mapped landmarks facilitate adjusting the reference contour to fit the anatomical structure. The adjusted reference contour data is transformed and applied to a second image using a thin-plate spline, and the adjusted high-resolution image is used for radiotherapy planning.

Automated anatomy delineation for image guided therapy planning

The quality of thoracic X-rays is depends on the inspiration state of an imaged patient. If the lungs of a patient are not significantly inflated during a thoracic X-ray, tissue from surrounding organs can cause the image to become cloudier, and the lung parenchyma is not effectively shown. Thus, there is a reliance on a medical professional to ensure that the patient is in an appropriate posture and inspiration state at the point of image exposure. This can be difficult with patients suffering from chronic conditions, though. Therefore, this application discusses a means to assess the inspiration state using the position of rib bones, and the diaphragm line in the X-ray image. Thus, an accurate estimate of inspiration state can automatically be reported to a medical professional, before a patient leaves the examination room, allowing an X-ray to be retaken, if necessary.

X-ray image inhalation quality monitoring

A radiation planning system includes a predictor-corrector optimizer unit which computes a predicted dose based on a collection of control points with a current approximate dose, each control point with a corresponding set of leaf positions, and determines an additional control point with a corresponding set of leaf positions based on a difference of the predicted fluence and the current approximate fluence through a least cost or shortest path in a layered graph structure of realizable leaf positions. Tools are described to help a planner to evaluate the effect of parameter changes to the current plan based on an identified zone of influence. The planner interactively views the current plan based on a visualization of the plan objectives and correlations between the objectives.

Radiation therapy planning optimization and visualization

Growing community of inventors

Hamburg, Germany

Daniel Bystrov