Koninklijke Philips Corporation N.V.

Steven Michael Cochoff

Yang-Ming Zhu

Ronald Sukalac

Charles Nortmann

Xiangyu Wu

L Alan Love

Richard Cheng-Hsiu Chen

Chris A Dauterman

Madhavi Ahuja

Dawn M Maniawski

Andriy Andreyev

Chuanyong Bai

Shekhar Dwivedi

Cheng-Hsiu Chen

Sydney Kaplan

Yogish Mallya

Narayan Ayyakad Krishnan

Manish Kumar Sharma

Emission imaging data are reconstructed to generate a low dose reconstructed image. Standardized uptake value (SUV) conversion () is applied to convert the low dose reconstructed image to a low dose SUV image. A neural network () is applied to the low dose SUV image to generate an estimated full dose SUV image. Prior to applying the neural network the low dose reconstructed image or the low dose SUV image is filtered using a low pass filter (). The neural network is trained on a set of training low dose SUV images and corresponding training full dose SUV images to transform the training low dose SUV images to match the corresponding training full dose SUV images, using a loss function having a mean square error loss component () and a loss component () that penalizes loss of image texture and/or a loss component () that promotes edge preservation.

Full dose PET image estimation from low-dose PET imaging using deep learning

An imaging device () includes a positron emission tomography (PET) scanner () including radiation detectors () and coincidence circuitry for detecting electron-positron annihilation events as 511 keV gamma ray pairs defining lines of response (LORs) with each event having a detection time difference At between the 511 keV gamma rays of the pair. At least one processor () is programmed to reconstruct a dataset comprising detected electron-positron annihilation events acquired for a region of interest by the PET scanner to form a reconstructed PET image wherein the reconstruction includes TOF localization of the events along respective LORs using a TOF kernel having a location parameter dependent on At and a TOF kernel width or shape that varies over the region of interest. A display device () is configured to display the reconstructed PET image.

Time-of-flight (TOF) PET image reconstruction using locally modified TOF kernels

A medical imaging system () comprises one or more displays (). A viewer device () generates an interactive user interface screen () on the display (), which viewer device () enables a user to simultaneously inspect selected image data of multiple patients or multiple images.

Multi-modality medical image viewing

A method and system for removing an object support from imaging data such as CT imaging data are provided. The automatic or semi-automatic removal process comprises identifying and locating the top edge of the object support in sagittal imaging plane data, and then removing the object support from transverse or volumetric imaging data.

Removing an object support from imaging data

In image registration, a similarity measure is computed of first and second images () offset at a plurality of relative axial offsets (). A starting relative axial offset () between the first and second images is identified based on the computed similarity measures. An iterative image registration process is performed to relatively register the first and second images () using the identified starting relative axial offset between the first and second images as an initial condition for the iterative image registration process. A starting relative in-slice offset () may also be identified as an in-slice offset effective to align corresponding slices of the first and second images () offset at the starting relative axial offset (), the identified starting relative in-slice offset also being used as an initial condition for the iterative image registration process.

Automatic pre-alignment for registration of medical images

An image processing system and method employs a registration processor ( ) to calculate a statistical measure of likelihood for two volumetric images ( ). The likelihood is calculated based on an assumption that the voxel values in two images in registration are probabilitically related. The likelihood is calculated for a plurality of relative transformations in iterative fashion until a transformation that maximizes the likelihood is found. The transformation that maximizes the likelihood provides an optimal registration and the parameters for the optimized transform are output to memory ( ) for use by a display system ( ) in aligning the images for display as a fused or composite image. If statistics about the relationship between the involved images are known, a mutation probability can be derived and used in the likelihood calculation. If there is no such prior knowledge, the mutation probability can be estimated purely from the image data. All voxel pairs in the overlapping volume or, alternatively, a portion of it can be used to compute the likelihood. The optimized likelihood has a simple and less abstract interpretation and can be further used to provide an indication of the quality of the registration.

Imaging registration system and method using likelihood maximization

Growing community of inventors

Hudson, OH, United States of America

Steven Michael Cochoff