Siemens Healthcare Gmbh

Siemens Healthineers Ag

The University of Queensland

Kieran O'Brien

Shivraman Giri

Feng Liu

Markus Barth

Steffen Bollmann

Viktor Vegh

Benedicte Marechal

Francesco Cognolato

Jin Jin

A training method for training neural networks to determine a magnetic susceptibility distribution of a sample may include: storing a simulated magnetic susceptibility map of the sample, generating a modified magnetic susceptibility map by combining an influence of one or more external magnetic susceptibility sources with the simulated magnetic susceptibility map and storing the modified magnetic susceptibility maps. The method may include generating a first training image by applying a quantitative susceptibility mapping model the modified magnetic susceptibility map and storing the first training image, applying the first neural network to the first image and a second neural network to an output of the first neural network and changing network parameters of the first and the second neural network depending on a deviation of an output of the second artificial neural network from the simulated magnetic susceptibility map.

Method and system for determining a magnetic susceptibility distribution

Techniques are disclosed to leverage the use of convolutional neural networks or similar machine learning algorithms to predict an underlying susceptibility distribution from MRI phase data, thereby solving the ill-posed inverse problem. These techniques include the use of Deep Quantitative Susceptibility 'DeepQSM' mapping, which uses a large amount of simulated susceptibility distributions and computes phase distribution using a unique forward solution. These examples are then used to train a deep convolutional neuronal network to invert the ill-posed problem.

Method and deep quantitative susceptibility mapping (QSM)

In a magnetic resonance method and apparatus, deficiencies in conventional masking in quantitative susceptibility mapping (QSM) are addressed by the inclusion of an additional step in the conventional QSM post-processing pipeline. In this additional step, atlas-based segmentation techniques, which have been developed for morphological applications such as T1w MPRAGE are used in order to provide the mask. This mask is then fed to the remainder of the QSM post-processing pipeline.

Magnetic resonance method and apparatus using atlas-based masking for quantitative susceptibility mapping

An imaging system includes determination of a first range of values of an imaging parameter, determination of a cost function expressing a difference between a first pulse profile and a second pulse profile, the second pulse profile generated based on respective values of each of a set of pulse parameters, identification of first coefficient values of each function of a set of functions which substantially minimize the cost function over the first range of values of the imaging parameter, where each of the set of functions determines a value of a respective one of the set of pulse parameters based on a value of the imaging parameter, and storage of the first coefficient values of each function of the set of functions in association with the first range of values.

RF pulse generation for magnetic resonance imaging

In a method and magnetic resonance (MR) apparatus for producing an MR image of a subject, MR signals are acquired respectively with multiple MR signal reception channels, and with an ultra-high basic magnetic field in the MR data acquisition scanner. Each of the acquired MR signals has a phase and exhibiting phase noise, and the acquired MR signals from the multiple MR signal reception channels are entered into a computer, as raw data in which said phase noise is preserved. The computer calculates the respective phase noise of each MR signal reception channel relative to the phase noise in each other MR signal reception channel, and calculates a phase noise map from the relative phase noise calculation for each of said reception channels, the phase noise map representing a spatial distribution of phase noise over the multiple MR signal reception channels. The phase noise map is applied to said raw data and raw data are selected that, dependent on the phase noise map, have a phase noise that is below a predetermined threshold. MR image data are reconstructed only from the selected raw data.

Growing community of inventors

Highgate Hill, Australia

Kieran O'Brien