Architecture for artificial intelligence-based magnetic resonance reconstruction

Patent metadata

Field	Value
Publication number	US-12106402-B2
Application number	US-202117450075-A
Country	US
Kind code	B2
Filing date	Oct 6, 2021
Priority date	Oct 6, 2021
Publication date	Oct 1, 2024
Grant date	Oct 1, 2024

Abstract

For magnetic resonance (MR) reconstruction using artificial intelligence (AI), the AI-based reconstruction for MR imaging systems is offloaded to one or more servers. A remote server performs AI-based reconstruction. A library of recent, old, custom, and/or publicly available AI-based reconstruction processes may be rapidly deployed and available to the server, which has the memory and processing resources for AI-based reconstruction. Load balancing of the data and/or between servers may improve performance.

First claim

What is claimed is:

1. A system for magnetic resonance (MR) reconstruction in medical imaging, the system comprising:
an MR scanner configured to scan a region of a patient, the scan providing k-space data;
a scanner processor of the MR scanner, the scanner processor configured to extract the k-space data from an imaging processing of the MR scanner, the scanner processor configured to format the k-space data for transmission over a computer network;
a server remote from the MR scanner, the server communicatively connected to the scanner processor by the computer network, the server configured to reconstruct a representation from the k-space data with a reconstruction using a first machine-learned model selected from a library, wherein the scanner processor is configured to receive the representation and inject the received representation into the imaging process; and
a display of the MR scanner configured to display an image from the representation.

2. The system of claim 1 wherein the scanner processor comprises a reconstruction processor of the MR scanner, the reconstruction processor being in an adjacent room to coils of the MR scanner, and wherein the server is outside of the adjacent room.

3. The system of claim 1 wherein the computer network comprises a local-area network, a wide-area network, or the Internet, and wherein the scanner processor is configured to format the k-space data pursuant to an International Society of Magnetic Resonance in Medicine Raw Data format.

4. The system of claim 1 wherein the server comprises a supercomputer.

5. The system of claim 1 wherein the server comprises a workstation having greater processing power than the scanner processor where the scanner processor comprises a reconstruction processor of the MR scanner.

6. The system of claim 1 wherein the server is configured to select the first machine-learned model from the library of multiple machine-learned models accessible to the server.

7. The system of claim 6 wherein the library includes a first set of the multiple machine-learned models in Python and a second set of the multiple machine-learned models in another language with a Python wrapper.

8. The system of claim 7 wherein the first set comprises publicly available machine-learned models and the second set comprises custom machine-learned models that are publicly unavailable.

9. The system of claim 1 wherein the server is configured to load balance with splitting of the k-space data for the scan of the patient.

10. The system of claim 1 wherein the server is one of multiple servers and the MR scanner is one of multiple MR scanners where k-space data from different ones of the MR scanners are load balanced between the multiple servers.

11. A method for magnetic resonance (MR) reconstruction in medical imaging, the method comprising:
receiving k-space data over a computer network at a server from a remote MR imager;
load balancing, by the server, the k-space data;
selecting a first reconstruction process from a library of reconstruction processes, the reconstruction processes of the library including different machine-learned models;
reconstructing a representation from the load-balanced k-space data with the selected first reconstruction process and corresponding machine-learned model; and
transmitting the representation or an image from the representation to the remote MR imager over the computer network.

12. The method of claim 11 wherein receiving and transmitting comprise receiving and transmitting in an International Society of Magnetic Resonance in Medicine Raw Data format, wherein the remote MR imager is in a different building than the server, and the server comprises a supercomputer.

13. The method of claim 11 wherein load balancing comprises separating the k-space data for a scan of a patient into separate processing blocks for different processing units of the server.

14. The method of claim 11 wherein load balancing comprises assigning a processor for the reconstructing.

15. The method of claim 11 wherein selecting comprises selecting where the library includes publicly available reconstruction processes and custom reconstruction processes.

16. A system for magnetic resonance (MR) reconstruction in medical imaging, the system comprising:
an interface configured to receive k-space data from a plurality of different MR imagers in one or more different facilities than the interface; and
a first server connected with the interface, the first server configured to load balance the k-space data to different processing units of the first server and to reconstruct representations from the k-space data, the representations reconstructed using one or more machine-learned models;
wherein the interface is configured to communicate the representations and/or images from the representations to the different MR imagers.

17. The system of claim 16 further comprising a second server, the second server configured to load balance the k-space data between a bank of servers including the first server.

18. The system of claim 16 wherein the first server is configured to select one of a plurality of the machine-learned models for reconstruction of each representation, the plurality of the machine-learned models including native machine-learned models and internal machine-learned models.

Assignees

• Siemens Healthineers Ag

Inventors

• Janardhanan Nirmal
• Lazar Laszlo
• Mailhe Boris
• Arberet Simon
• Nadar Mariappan S
• Comaniciu Dorin
• Chow Kelvin
• Bush Michael

Classifications

• G06T12/10 Primary

  Image preprocessing, e.g. calibration, positioning of sources or scatter correction · CPC title

• G06T7/0012

  Biomedical image inspection · CPC title

• G01R33/4818

  MR characterised by data acquisition along a specific k-space trajectory or by the temporal order of k-space coverage, e.g. centric or segmented coverage of k-space · CPC title

• G06T2207/10088

  Magnetic resonance imaging [MRI] · CPC title

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title