Siemens Medical Solutions Usa, Inc.

Xinhong Ding

Alexander Hans Vija

Ronald E Malmin

Michal Cachovan

Manojeet Bhattacharya

Francesc dAssis Massanes Basi

Jun Ma

Shuchen Zhang

A method for automatically classifying emission tomographic images includes receiving original images and a plurality of class labels designating each original image as belonging to one of a plurality of possible classifications and utilizing a data generator to create generated images based on the original images. The data generator shuffles the original images. The number of generated images is greater than the number of original images. One or more geometric transformations are performed on the generated images. A binomial sub-sampling operation is applied to the transformed images to yield a plurality of sub-sampled images for each original image. A multi-layer convolutional neural network (CNN) is trained using the sub-sampled images and the class labels to classify input images as corresponding to one of the possible classifications. A plurality of weights corresponding to the trained CNN are identified and those weights are used to create a deployable version of the CNN.

Binomial subsample data augmented CNN for image classification

A system and method include acquisition of a plurality of projection images of a subject, each of the projection images associated with a respective projection angle, determination, for each of the projection images, of a center-of-light location in a first image region, determination of a local fluctuation measure based on the determined center-of-light locations, and determination of a quality measure associated with the plurality of projection images based on the local fluctuation measure.

Systems to assess projection data inconsistency

A system and method include training of an artificial neural network to generate a simulated attenuation-corrected reconstructed volume from an input non-attenuation-corrected reconstructed volume, the training based on a plurality of non-attenuation-corrected volumes generated from respective ones of a plurality of sets of two-dimensional emission data and on a plurality of attenuation-corrected reconstructed volumes generated from respective ones of the plurality of sets of two-dimensional emission data.

Multi-focal non-parallel collimator-based imaging

A system and method includes input of a plurality of sets of training data to a neural network to generate a plurality of sets of output data, determination of a first loss based on the plurality of sets of output data and on the plurality of sets of ground truth data, determination if a second loss based on the plurality of sets of output data and one or more physics-based constraints, and modification of the neural network based on the first loss and the second loss.

Physics-constrained network and training thereof

During calibration of a SPECT system, system-specific sensitivities and cross-calibration factors for multiple isotopes for correcting for dose are determined for various combinations of options, including the option of which specific well counter with which to measure the dose. The options may include selected energy windows for isotopes with multiple energy windows. This arrangement allows for custom-specified isotopes not included in standard listings. For use with a particular patient, the cross-calibration factor for the well counter used to measure the dosage for the patient is accessed and used for dose correction. More accurate quantitative functional information may result from the corrected dose. The cross-calibration may be more easily implemented despite the options using the sensitivities and cross-calibrations provided for various combinations.

Flexible application of cross-calibration for quantitative functional imaging

A system and method includes acquisition of a plurality of images depicting a respective scintillator crystal, determination of a plurality of categories based on the plurality of images, determination of a crystal quality value associated with each of the plurality of categories, training of a network to receive an input image and output an indication of one of the plurality of categories based on the input image, the training based on the plurality of images and the at least one category associated with each pf the plurality of images, operation of the trained network to receive a first image of a first scintillator crystal and output a first one of the plurality of categories based on the first image, and determination of a quality of the first scintillator crystal based on the first one of the plurality of categories and a first crystal quality value associated with the first one of the plurality of categories.

Scintillator crystal evaluation system

A system and method includes acquisition of a plurality of non-attenuation-corrected volumes, each of the non-attenuation-corrected volumes based on a respective one of a plurality of sets of two-dimensional emission data, acquisition of a plurality of attenuation coefficient maps, each of the plurality of attenuation coefficient maps corresponding to a respective one of the plurality of sets of two-dimensional emission data, training of a convolutional network to generate a generated attenuation coefficient map from an input image volume, the training based on the plurality of non-attenuation-corrected volumes and respective ones of the plurality of attenuation coefficient maps, and output of trained kernels of the trained convolutional network to an emission imaging system.

Network training and architecture for medical imaging

Systems and methods for generating corrected emission tomography images are provided. A method includes obtaining a reconstructed image based on emission tomography data of a head of a patient and defining a boundary region in the reconstructed image estimating a position of a skull of the patient in the reconstructed image. The method also includes generating a map of attenuation coefficient values for the reconstructed image based on the boundary region. The reconstructed image can then be adjusted based on the map. In the method, the attenuation coefficient values within the boundary region are selected to correspond to an attenuation coefficient value for bone and the attenuation coefficient values for the portion of the image surrounded by the boundary region are selected to correspond to an attenuation value for tissue.

Doseless emission tomography attenuation correction

A method of clinical collaboration between a clinical site and an analysis site includes receiving scan data from a scanner via a first reconstruction computer system at the clinical site, implementing a reconstruction procedure on the received scan data using a second reconstruction computer system at the analysis site and configured in accordance with a reconstruction configuration parameter, the analysis site being remote from the clinical site, and transmitting data indicative of the reconstruction configuration parameter to the first reconstruction computer system to configure the first reconstruction computer system in accordance with the reconstruction configuration parameter.

Clinical collaboration and medical computing framework

A method for measuring a SPECT collimator's hole orientation angles includes obtaining a set of stepped radiation line images of a line radiation source by scanning/stepping the line radiation source across a first collimator in a first direction; obtaining a second set of stepped radiation line images of the line radiation source across the first collimator in a second direction that is perpendicular to the first direction; and obtaining two sets of stepped radiation line images for a second collimator, wherein one of the two collimators is a reference collimator and the other is a collimator being measured. Calculating the collimator hole orientation angles requires determining offset distances along the two directions for each pair of lines between the reference collimator's line images and the measured collimator's line images by identifying and pairing the lines from the reference collimator line images and the measured collimator line images. The method provides an automated way of identifying and pairing the lines.

Automatic line identification and pairing for nuclear imaging collimator vector map characterization

A method and a system for implementing the method for simultaneously registering and zipping a multiple scan whole body SPECT/CT image. The method includes the steps of simultaneously registering and zipping multiple input images and re-sampling the registered images. The step of simultaneously registering and zipping multiple input images is accomplished by initially aligning the images to be registered with each other, aligning the images with a reference image, and adjusting the alignment of the images with each other.

Growing community of inventors

Buffalo Grove, IL, United States of America

Xinhong Ding