Siemens Healthcare Gmbh

National Institutes of Health, a Component of the US Dept. of Health & Human Services

Waqas Majeed

Himanshu Bhat

David Grodzki

Rainer Schneider

Adrienne Campbell

Peter Schmitt

Gerhard Laub

Sunil Goraksha Patil

Florian Maier

Axel Joachim Krafft

Wuyi Zhao

In the field of MRI, a model-based MRI image reconstruction technique is provided. The model-based reconstruction technique increases the performance of Time-of-Flight MRA. In a learning phase, a model is calculated from a sufficiently large set of data acquired at both low and high magnetic fields, using deep learning strategies. In a clinical phase, the model is applied to measured data generating high MR image quality.

MRI image reconstruction using machine learning

Embodiments provide a computer-implemented method for selecting thermal images for generating a temperature difference map through proton resonance frequency (PRF) thermometry, including: acquiring a set of baseline images prior to a thermal treatment of an organ of interest; identifying a subset of baseline images in a most stable motion state from the set of baseline images; averaging the subset of baseline images to generate a template image; determining an acceptance threshold based on an image similarity measure (ISM) between each of the set of baseline images and the template image; acquiring a set of thermal images during the thermal treatment; and selecting a subset of thermal images from the set of thermal images, wherein each of the subset of thermal images has the image similarity measure above the acceptance threshold.

Image-based retrospective gating of MR images for PRF thermometry

The invention relates to a method for optimizing an MR imaging sequence with a plurality of sequence blocks where gradient requirements of logical gradients switched along axes including a read, slice, and phase axis are determined. Based on the requirements, a reference gradient is determined among the logical gradients having a longest minimum duration relative to remaining logical gradients within the sequence block during which the reference gradient is switched on. A scaling factor is determined for each of the logical gradients. The scaling factors for the logical gradient factors are updated taking into account the determined stretch ratios, and optimized amplitudes are determined for the logical gradients based on the updated scaling factors.

Optimization of MR imaging sequence

A method for phase correction in proton resonance frequency (PRF) thermometry application includes acquiring a series of magnetic resonance (MR) images comprising a first MR image and plurality of subsequent MR images depicting an anatomical area of interest. The MR images are acquired while tissue in the anatomical area of interest is undergoing a temperature change. Each subsequent MR image is registered to the first MR image to yield a plurality of registered images. A plurality of basis images are computed from the registered images using Principal Component Analysis (PCA). The basis images are used to remove motion-related phase changes from a second series of MR images, thereby yielding a motion corrected second series of MR images. One or more temperature difference maps are generated that depict a relative temperature change for the tissue in the anatomical area of interest based on the motion corrected second series.

Image registration and principal component analysis based multi-baseline phase correction method for proton resonance frequency thermometry

A method for decomposing noise into white and spatially correlated components during MR thermometry imaging includes acquiring a series of MR images of an anatomical object and generating a series of temperature difference maps of the anatomical object. The method further includes receiving a selection of a region of interest (ROI) within the temperature difference map and estimating total noise variance values depicting total noise variance in the temperature difference map. Each total noise variance value is determined using a random sampling of a pre-determined number of voxels from the ROI. A white noise component and a spatially correlated noise component of the total noise variance providing a best fit to the total noise variance values for all of the random samplings are identified. The white noise component and the spatially correlated noise component are displayed on a user interface.

Growing community of inventors

Ellicott City, MD, United States of America

Waqas Majeed