Method and image processor for evaluating a contrast agent-enhanced magnetic resonance slice image of a heart

Patent metadata

Field	Value
Publication number	US-10521906-B2
Application number	US-201815907592-A
Country	US
Kind code	B2
Filing date	Feb 28, 2018
Priority date	Mar 1, 2017
Publication date	Dec 31, 2019
Grant date	Dec 31, 2019

Abstract

In a method and processor for evaluating a contrast agent-enhanced two-dimensional magnetic resonance slice image of a heart of a patient in order to determine picture elements revealing contrast agent deposits in the myocardium, an endocardium contour in the magnetic resonance slice image, taking into consideration deposition information describing picture elements potentially revealing contrast agent deposits and determined by image analysis on the basis of a shape assumption for the heart structure that is to be examined, in particular the left ventricle, such that picture elements potentially revealing contrast agent deposits are avoided as much as possible as a contour component. An epicardium contour enclosing the endocardium contour is then determined. Picture elements are marked that indicate contrast agent enhancement in the myocardium lying between the epicardium contour and the endocardium contour as contrast agent deposit.

First claim

The invention claimed is:

1. A method for evaluating a contrast agent-enhanced two-dimensional magnetic resonance slice image of a heart of a patient in order to determine picture elements in said slice image revealing contrast agent deposits in the myocardium of the heart, said method comprising:
providing a computer with a contrast agent-enhanced two-dimensional magnetic resonance slice image of a heart, said slice image being comprised of picture elements;
in said computer, determining an endocardium contour of the endocardium of the heart in the magnetic resonance slice image, using deposition information describing picture elements that are candidates for revealing contrast agent deposits, by executing an image analysis based on an assumption, provided to the computer, of a shape for a structure of the heart in the slice image and, in said image analysis, avoiding picture elements as components of said endocardium contour that are said candidates for revealing contrast agent deposits;
in said computer, determining an epicardium contour that encloses the endocardium contour;
in said computer, marking picture elements that indicate contrast agent enhancement in the myocardium lying between the epicardium contour and the endocardium contour, as contrast agent deposit; and
making the slice image, with the picture elements marked as contrast agent deposit, available from the computer in electronic form, as a data file.

2. A method as claimed in claim 1 comprising, in said image analysis in said computer, using a shape assumption for the left ventricle of the heart, as said assumption of the shape of said heart structure.

3. A method as claimed in claim 1 comprising determining said endocardium contour and said epicardium contour by localizing said heart structure using background information describing said structure as having a circular shape, as said assumption of said shape.

4. A method as claimed in claim 3 comprising detecting said structure in said image analysis using at least one of circular Hough transforms and a circularity measure of said circular shape.

5. A method as claimed in claim 1 comprising determining said endocardium contour by first determining an estimated endocardium contour by localization of blood within said structure, and using said estimated contour to determine said deposition information.

6. A method as claimed in claim 5 comprising determining said estimated endocardium contour by executing a morphological active evolution algorithm for a contour curve.

7. A method as claimed in claim 6 comprising executing said morphological active evolution algorithm for a contour curve without taking edges into consideration.

8. A method as claimed in claim 5 comprising determining said deposition information by first defining an inner region within said estimated endocardium contour and an outer region surrounding the estimated endocardium contour, and marking picture elements in the outer region, as representing contrast agent deposit, which exceed a threshold value determined by statistical evaluation of values of the picture elements in said inner region.

9. A method as claimed in claim 5 comprising identifying candidate picture elements, from said estimated endocardium contour, for inclusion as contour points in said endocardium contour, by equidistant point selection along radial rays with respect to said structure, and applying an endocardium classifier to each candidate picture element that produces an output, for each candidate picture element applied to said endocardium classifier, comprising at least one of a value describing whether the respective candidate picture elements represents contrast agent deposit, and a value representing a probability that the respective candidate picture element belongs said endocardium contour.

10. A method as claimed in claim 9 comprising operating said endocardium classifier according to a classification algorithm trained by machine learning, using training slice images as training data that contain fundamental truths for said classification algorithm.

11. A method as claimed in claim 10 comprising using a random forest algorithm as said classification algorithm.

12. A method as claimed in claim 9 comprising operating said endocardium classifier to evaluate at least one property of each candidate picture element selected from the group consisting of an intensity of the respective candidate picture element, a gradient of the respective candidate picture element, and an environment in said slice image that encompasses the candidate picture element.

13. A method as claimed in claim 1 comprising determining said endocardium contour by:
applying an endocardium classifier to candidate picture elements for inclusion as contour points of said endocardium contour, that produces an output value, for each candidate picture element, describing a probability of the respective candidate picture element being a contour point of said endocardium contour; and
determining a cost map that designates candidate picture elements having a probability value above a predetermined value and avoiding candidate picture elements, according to said deposition information, that represent contrast agent deposits, and using said cost map to define picture elements, among said candidate picture elements for inclusion in said endocardium contour.

14. A method as claimed in claim 13 comprising determining said cost map by one of:
executing a pathfinding algorithm with path length weighting in order to identify a minimum cost path as said cost path; and
using a polar coordinate system referenced to a centerpoint of said structure in order to search for a path along a polar angle that is shortest.

15. A method as claimed in claim 1 comprising:
in said computer, determining edge information describing a position of edges in said magnetic resonance slice image; and
determining said endocardium contour by applying a cost map to said edge information in order to identify path preferring edges, while avoiding picture elements representing contrast agent deposits, in order to identify picture elements for inclusion in said endocardium contour.

16. A method as claimed in claim 15 comprising determining said cost map by one of:
executing a pathfinding algorithm with path length weighting in order to identify a minimum cost path as said cost path; and
using a polar coordinate system referenced to a centerpoint of said structure in order to search for a path along a polar angle that is shortest.

17. A method as claimed in claim 1 wherein providing said computer with said contrast agent-enhanced two-dimensional magnetic resonance slice image comprises one of providing said computer with a short-axis magnetic resonance slice image, and providing said computer with a plurality of magnetic resonance slice images of a slice image stack covering the heart from the base to the apex.

18. A method as claimed in claim 1 comprising, at a display in communication with said computer, displaying said magnetic resonance slice image with said picture elements marked as contrast agent deposit having a different visual appearance than other picture elements in the displayed magnetic resonance slice image.

19. An image processing computer comprising:
a processor;
said processor having an input that receives a contrast agent-enhanced two-dimensional magnetic resonance slice image of a heart, said slice image being comprised of picture elements;
said processor being configured to determine an endocardium contour of the

Assignees

• Siemens Healthcare Gmbh

Inventors

• Forman Christoph
• Kurzendorfer Tanja
• Brost Alexander

Classifications

• G06T7/155 Primary

  involving morphological operators · CPC title

• G06T7/0012 Primary

  Biomedical image inspection · CPC title

• G06F18/2111

  by using evolutionary computational techniques, e.g. genetic algorithms · CPC title

• G06F18/241

  relating to the classification model, e.g. parametric or non-parametric approaches · CPC title

• G06T15/08

  Volume rendering · CPC title