Determining characteristics of muscle structures using artificial neural network

Patent metadata

Field	Value
Publication number	US-12354730-B2
Application number	US-202217662475-A
Country	US
Kind code	B2
Filing date	May 9, 2022
Priority date	Jun 1, 2021
Publication date	Jul 8, 2025
Grant date	Jul 8, 2025

Abstract

Techniques of determining a quantification of at least one characteristic of a muscle structure comprising at least one muscle and at least one tendon are disclosed. The quantification of the at least one characteristic of the rotator cuff may be determined by using at least one artificial neural network and based on one or more medical images depicting the muscle structure of a patient.

First claim

The invention claimed is:

1. A computer-implemented method, comprising:
obtaining one or more medical images, the one or more medical images depicting a muscle structure of a patient, wherein the muscle structure comprises at least one muscle and at least one tendon,
determining a pre-segmentation of a region of interest associated with the muscle structure,
using a first one of at least one artificial neural network, determining a segmentation of the region of interest associated with the muscle structure in the one or more medical images, wherein the segmentation is determined based on the pre-segmentation, and
determining a quantification of at least one characteristic of the muscle structure using a second one of the at least one artificial neural network based on the segmentation of the region of interest and the one or more medical images.

2. The computer-implemented method of claim 1 , wherein the pre-segmentation is determined using a convolutional neural network, the convolutional neural network providing, as an output, a pixel probability map, wherein the pre-segmentation is obtained as a pixel mask by applying a threshold comparison to probability values of the pixel probability map and selecting a largest contiguous area.

3. The computer-implemented method of claim 2 , wherein the segmentation comprises a bounding box which is determined based on the pixel mask.

4. The computer-implemented method of claim 1 , wherein determining a quantification of at least one characteristic of the muscle structure using a second one of the at least one artificial neural network based on the segmentation of the region of interest and the one or more medical images comprises:
determining a value indicative of the quantification of the at least one characteristic.

5. The computer-implemented method of claim 1 ,
wherein the one or more medical images are multi-slice images,
wherein the method further comprises:
determining a reference point of the region of interest based on the segmentation,
selecting a given slice from multiple slices of the one or more medical images based on the reference point,
wherein the quantification of the at least one characteristic is determined based on an appearance of the muscle structure in the given slice.

6. The computer-implemented method of claim 1 ,
wherein said obtaining of the one or more medical images comprises pre-processing the one or medical images based on a landmark detection algorithm configured to detect a region of interest in the one or more medical images, the region of interest being associated with the muscle structure.

7. The computer-implemented method of claim 1 ,
wherein the at least one characteristic comprises multiple characteristics,
wherein the second one of the at least one artificial neural network comprises multiple decoder branches configured to determine the quantification of the multiple characteristics based on a shared set of latent features determined based on the one or more medical images.

8. The method of claim 1 , wherein the quantification of the at least one characteristic comprises at least one of a length, width, thickness, or musculotendinous junction position of a tendon tear.

9. The method of claim 8 , wherein the second one of the at least one artificial neural network comprises a first artificial neural network configured to determine at least one of 1) a presence or absence of the tendon tear in the at least one tendon or 2) a type of the tendon tear in the at least one tendon.

10. The method of claim 9 ,
wherein the second one of the at least one artificial neural network comprises a second artificial neural network configured to determine the at least one of the length, width, thickness, or the musculotendinous junction position of the tendon tear,
wherein the second artificial neural network obtains, as an input, an output of the first artificial neural network.

11. The method of claim 1 , wherein the at least one characteristic comprises a muscle atrophy of the at least one muscle of the muscle structure.

12. The method of claim 11 , wherein the second one of the at least one artificial neural network is configured to determine the quantification of the muscle atrophy.

13. The method of claim 1 , wherein the at least one characteristic comprises a fat infiltration of the at least one muscle of the muscle structure.

14. The method of claim 13 , wherein the second one of the at least one artificial neural network is configured to determine the fat infiltration.

15. The method of claim 1 , further comprising:
determining a classification of the at least one characteristic or at least one further characteristic of the muscle structure using the at least one artificial neural network.

16. A system comprising:
at least one processor; and
at least one memory, wherein upon loading and executing program code from the at least one memory, the at least one processor is configured to:
obtain one or more medical images, the one or more medical images depicting a muscle structure of a patient, wherein the muscle structure comprises at least one muscle and at least one tendon,
determine a pre-segmentation of a region of interest associated with the muscle structure,
using a first one of at least one artificial neural network, determine a segmentation of the region of interest associated with the muscle structure in the one or more medical images, wherein the segmentation is determined based on the pre-segmentation, and
determine a quantification of at least one characteristic of the muscle structure using a second one of the at least one artificial neural network based on the segmentation of the region of interest and the one or more medical images.

17. A medical imaging scanner, the medical imaging scanner comprising the system of claim 16 .

18. A non-transitory computer readable medium storing computer program instructions, the computer program instructions when executed by a processor cause the processor to perform operations comprising:
obtaining one or more medical images, the one or more medical images depicting a muscle structure of a patient, wherein the muscle structure comprises at least one muscle and at least one tendon,
determining a pre-segmentation of a region of interest associated with the muscle structure,
using a first one of at least one artificial neural network, determining a segmentation of the region of interest associated with the muscle structure in the one or more medical images, wherein the segmentation is determined based on the pre-segmentation, and
determining a quantification of at least one characteristic of the muscle structure using a second one of the at least one artificial neural network based on the segmentation of the region of interest and the one or more medical images.

Assignees

• Siemens Healthineers Ag
• Univ New York

Inventors

• Schwier Michael
• Geiger Bernhard
• Grbic Sasa
• Raithel Esther
• Lin Dana
• Chabin Guillaume

Classifications

• G06N3/0895

  Weakly supervised learning, e.g. semi-supervised or self-supervised learning · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/092

  Reinforcement learning · CPC title

• G06N3/09

  Supervised learning · CPC title