Methods, systems and use for detecting irregularities in medical images by means of a machine learning model

1. A method for training a computer-implemented machine learning model for detecting irregularities in medical images, the method including:

identifying at least one predetermined type of body region ( 14 ) depicted in a medical image ( 10 ), said body region ( 14 ) having a depicted irregularity ( 12 );

defining a plurality of image segments ( 20 ) each including at least part of the depicted body region ( 14 ), wherein a resolution of the image segments ( 20 ) is maintained or not reduced by more than 20% compared to the medical image ( 10 ); and

using said image segments ( 20 ) to train a machine learning model to detect similar irregularities ( 12 ) in other medical images ( 10 );

wherein the machine learning model comprises a neural network; wherein the neural network uses the image information of medical images ( 10 ) at an input layer and/or provides the detected irregularity at an output layer;

further including:

discarding or replacing at least part of the image information of the medical image ( 10 ) which does not depict the identified body region ( 14 );

wherein at least portions of at least two image segments of the said plurality of image segments ( 20 ) that include at least part of the depicted body region and which are used to train the machine learning model to detect similar irregularities overlap one another; and

wherein an overlapping portion of the said at least two image segments ( 20 ) includes at least part of the depicted irregularity ( 12 ).

2. The method according to claim 1 , wherein the machine learning model comprises a neural network.

3. The method according to claim 1 , wherein the neural network uses the image information of medical images ( 10 ) at an input layer and/or provides the detected irregularity at an output layer.

4. The method according to claim 1 , wherein the image segments ( 20 ) are defined so that the depicted irregularity ( 12 ) is included in at least one image segment ( 20 ) as a whole.

5. The method according to claim 1 , wherein the body region ( 14 ) is the lungs.

6. The method according to claim 5 , wherein each left and right lung ( 11 , 13 ) is segmented by means of the at least two image segments ( 20 ).

7. The method according to claim 6 , wherein an image segment ( 20 ) does not include more than one lung ( 11 , 13 ).

8. The method according to claim 7 , wherein at least one further depicted body region ( 40 ) is identified in the medical image ( 10 ), at least one further image segment of the said plurality of image segments ( 20 ) is defined that includes said further depicted body region ( 40 ) and said further image segment ( 20 ) is also used for training the machine learning model.

9. The method according to claim 8 , wherein the further depicted body region ( 40 ) is the heart or a region in which the heart overlaps with the lungs.

10. The method according to claim 5 , wherein an image segment ( 20 ) does not include more than one lung ( 11 , 13 ).

11. The method according to claim 1 , wherein at least one further depicted body region ( 40 ) is identified in the medical image ( 10 ), at least one further image segment of said plurality of image segments ( 20 ) is defined that includes said further depicted body region ( 40 ) and said further image segment ( 20 ) is also used for training the machine learning model.

12. The method according to claim 11 , wherein the further depicted body region ( 40 ) is the heart or a region in which the heart overlaps with the lungs.

13. Use of a machine learning model for detecting an irregularity in a medical image, wherein the machine learning model has been trained by means of a method according to one of the previous claims.

14. A system ( 1 ) for detecting irregularities in medical images ( 10 ), the system including at least one computer device ( 2 ), the computer device ( 2 ) being configured to detect irregularities ( 12 ) in medical images ( 10 ) by means of a machine learning model that has been trained by means of a method according to claim 1 .

15. A system ( 1 ) for training a machine learning model for detecting irregularities ( 12 ) in medical images ( 10 ), the system ( 1 ) including at least one computer device ( 2 ) and being configured to:

identify at least one predetermined type of body region ( 14 ) depicted in a medical image ( 10 ), said body region ( 14 ) having a depicted irregularity ( 12 );

define a plurality of different image segments ( 20 ) each including at least part of the depicted body region ( 14 ), wherein a resolution of the image segments ( 20 ) is not changed or not reduced by more than 20% compared to the medical image ( 10 ); and

use said image segments ( 20 ) to train the machine learning model to detect similar irregularities ( 12 ) in other medical images ( 10 );

wherein the machine learning model comprises a neural network; wherein the neural network uses the image information of medical images ( 10 ) at an input layer and/or provides the detected irregularity at an output layer;

the at least one computer device ( 2 ) and being further configured to:

discard or replace at least part of the image information of the medical image ( 10 ) which does not depict the identified body region ( 14 );

wherein at least portions of at least two image segments of the said plurality of different image segments ( 20 ) that each include at least part of the depicted body region and which are used to train the machine learning model to detect similar irregularities in other medical images overlap one another; and

wherein an overlapping portion of the said at least two image segments ( 20 ) includes at least part of the depicted irregularity ( 12 ).