Reconstruction module and method for reconstruction of medical images
AssigneeSiemens Healthineers AG
InventorsUvo Hölscher, Ralf Kartäusch
Life of this patent
Estimating the term…
Abstract
A reconstruction module may include at least one parameter interface to receive a set of measurement sequence parameters and a memory interface to a memory having a trained mathematical model stored in the memory. The trained mathematical model determines a results dataset containing at least one iterative denoising parameter and at least one edge enhancement parameter for at least one received measurement sequence parameter. A control of the reconstruction module controls a reconstruction of the medical images using a reconstruction algorithm based on the determined parameters for parameterizing an iterative denoising function and an edge enhancement function.
Claims15
A reconstruction computer for a magnetic resonance (MR) tomography scanner for the reconstruction of medical images, comprising:
- a first memory storing a first reconstruction algorithm;
- and
- processing circuitry configured to:
- receive a set of measurement sequence parameters;
- interface with a second memory having a trained mathematical model stored therein, the trained mathematical model being configured to determine a results dataset containing at least one iterative denoising (ID) parameter and at least one edge enhancement (EE) parameter for at least one received measurement sequence parameter, wherein the measurement sequence parameters are read out directly from an MR protocol and/or indirectly from an at least partially reconstructed image, the at least partially reconstructed image being generated using a second reconstruction algorithm that applies no ID function and no EE function;
- and
- execute the first reconstruction algorithm to control a reconstruction of the medical images, based on the determined at least one ID parameter and the at least one EE parameter, to parameterize an iterative denoising function and an edge enhancement function.
A computer-implemented method for reconstruction of medical images, comprising:
- receiving a set of measurement sequence parameters, wherein the measurement sequence parameters are read out directly from a magnetic resonance protocol and/or indirectly from an at least partially reconstructed image;
- and
- determining a results dataset using a trained mathematical model and based on the received set of measurement sequence parameters, the results dataset containing at least one iterative denoising (ID) parameter and at least one edge enhancement (EE) parameter, wherein the trained mathematical model is adapted for a reconstruction algorithm configured to reconstruct the medical images, the at least partially reconstructed image being generated using another reconstruction algorithm that applies no ID function and no EE function.
A system for the reconstruction of medical images, comprising:
- a memory storing a trained mathematical model;
- and
- a magnetic resonance tomography scanner configured to acquire detector signals for the reconstruction of medical images, the magnetic resonance tomography scanner including a computer configured to engage in data interchange with the memory and having at least one processor configured to execute a first reconstruction algorithm to:
- determine, using the trained mathematical model and based on measurement sequence parameters, a results dataset containing at least one iterative denoising (ID) parameter and at least one edge enhancement (EE) parameter for at least one received measurement sequence parameter, wherein the measurement sequence parameters are determined directly from a magnetic resonance protocol and/or indirectly from an at least partially reconstructed image generated using a second reconstruction algorithm that applies no ID function and no EE function;
- and
- reconstruct the medical images, based on the determined at least one ID parameter and the at least one EE parameter, to parameterize an iterative denoising function and an edge enhancement function.
Dependencies are inferred from the claim text.