Method and deep quantitative susceptibility mapping (QSM)
AssigneeSiemens Healthcare GmbH
InventorsKieran O'Brien, Markus Barth, Steffen Bollmann
Life of this patent
Estimating the term…
Abstract
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.
Claims20
A method for qualitative susceptibility mapping (QSM) to extract a magnetic susceptibility of a tissue by sample-induced magnetization in a static magnetic field, comprising:
- accessing, via a deep convolutional neuronal network, a set of simulated susceptibility distributions;
- computing, via the deep convolutional neuronal network, a respective set of phase distributions from the set of simulated susceptibility distributions using a unique forward solution;
- training, via the deep convolutional neuronal network, a model using the set of simulated susceptibility distributions and the respective set of phase distributions, the model representing a learned solution to an ill-posed field-to-source inversion problem;
- receiving, via the deep convolutional neuronal network, magnetic resonance imaging (MRI) phase data associated with the tissue;
- and
- solving, via the deep convolutional neuronal network, the ill-posed field-to-source inversion problem in accordance with the learned model to predict an underlying magnetic susceptibility distribution of the tissue using the MRI phase data.
A deep convolutional neuronal network for qualitative susceptibility mapping (QSM) to extract a magnetic susceptibility of a tissue by sample-induced magnetization in a static magnetic field, comprising:
- a storage medium configured to store a set of simulated susceptibility distributions;
- and
- one or more processors configured to:
- access the set of simulated susceptibility distributions via the storage medium;
- compute a respective set of phase distributions from the set of simulated susceptibility distributions using a unique forward solution;
- train a model using the set of simulated susceptibility distributions and the respective set of phase distributions, the model representing a learned solution to an ill-posed field-to-source inversion problem;
- access magnetic resonance imaging (MRI) phase data associated with the tissue;
- and
- solve the ill-posed field-to-source inversion problem in accordance with the learned model to predict an underlying magnetic susceptibility distribution of the tissue using the MRI phase data.
A non-transitory computer-readable medium associated with a deep convolutional neuronal network for qualitative susceptibility mapping (QSM) to extract a magnetic susceptibility of a tissue by sample-induced magnetization in a static magnetic field, the non-transitory medium having instructions stored thereon that, when executed by one or more processors, cause the one or more processors to:
- access the set of simulated susceptibility distributions via the storage medium;
- compute a respective set of phase distributions from the set of simulated susceptibility distributions using a unique forward solution;
- train a model using the set of simulated susceptibility distributions and the respective set of phase distributions, the model representing a learned solution to an ill-posed field-to-source inversion problem;
- access magnetic resonance imaging (MRI) phase data associated with the tissue;
- and
- solve the ill-posed field-to-source inversion problem in accordance with the learned model to predict an underlying magnetic susceptibility distribution of the tissue using the MRI phase data.
Dependencies are inferred from the claim text.