Detecting abnormalities in an x-ray image

1 . A system for detecting one or more abnormalities in a medical image, wherein an abnormality of the one or more abnormalities is indicative of a pathology, a disease or a clinical finding present in the medical image, the system comprising:

a data interface for accessing model data, wherein the model data includes data representing an image classifier trained to use one or more image quality features to determine classification scores for the one or more abnormalities from the medical image, and wherein the model data further includes data representing one or more feature extractors configured to extract the respective image quality features from the medical image indicative of a suitability of the medical image for detection of the one or more abnormalities;

an input interface for obtaining the medical image; and

a processor subsystem configured to:

apply the one or more feature extractors to the medical image to determine the respective image quality features for the medical image;

apply the image classifier to the medical image to determine the classification scores for the one or more abnormalities, wherein the image classifier uses the determined image quality features to determine said classification scores; and

output a classification result based on the determined classification scores,

wherein the processor subsystem is configured to determine a classification score for an abnormality by:

applying a classification part of the image classifier to the medical image to obtain a preliminary classification score for the abnormality;

applying a correction part of the image classifier to at least the image quality features to determine a correction value for the preliminary classification score; and

applying the correction value to the preliminary classification score to obtain the classification score for the abnormality.

2 . The system of claim 1 , wherein an image quality feature represents a type of view of the medical image, a rotation degree of the medical image subject with respect to the plane of the medical image, a degree of inspiration indicating a presence of an adequately inspired volume in the medical image, or a degree of penetration indicating an adequacy of exposure of the medical image.

3 . The system of claim 1 , wherein applying the correction part comprises applying a rule-based model to at least the determined image quality features.

4 . The system of claim 1 , wherein the processor subsystem is configured to determine a classification score for an abnormality by applying a feature extraction part of the image classification model to the medical image to obtain image features of the medical image, and applying a neural network to the image features and the image quality features to obtain the classification score for the abnormality.

5 . The system of claim 1 , wherein the processor subsystem is configured to include correction data in the classification result indicating an effect of the image quality features on the classification score.

6 . The system of claim 1 , wherein the processor subsystem is further configured to determine, based on the one or more image quality features, a subset of abnormalities that may be undetected due to image quality, and to output said subset of abnormalities as part of the classification result.

7 . The system of claim 6 , wherein the processor subsystem is configured to determine said subset based on applying a rule-based model, the processor subsystem being further configured to include rules used by the rule-based model to determine that an abnormality may be undetected in the classification result.

8 . The system of claim 1 , further comprising an output interface for outputting the classification result in a sensory-perceptible manner to a user.

9 . The system of claim 1 , further comprising:

a second data interface for accessing training data comprising multiple training images, a training image being labelled with one or more image quality features and one or more abnormalities;

a second processor subsystem configured to train the image classifier to, given a training image and using the one or more image quality features that the training image is labelled with, detect the one or more abnormalities that the training image is labelled with.

10 . The system of claim 9 , wherein the image classifier comprises a correction part configured to determine a correction value for a preliminary classification score according to a rule-based model, the processor subsystem being configured to determine a threshold for a rule of the rule-based system based on the training data.

11 . A computer-implemented method of detecting one or more abnormalities in an medical image, wherein an abnormality of the one or more abnormalities is indicative of a pathology, a disease or a clinical finding present in the medical image, the method comprising:

accessing model data, wherein the model data includes data representing an image classifier trained to use one or more image quality features to determine classification scores for the one or more abnormalities from the medical image, and wherein the model data further includes data representing one or more feature extractors configured to extract the respective image quality features from the medical image indicative of a suitability of the medical image for detection of the one or more abnormalities;

obtaining the medical image;

applying the one or more feature extractors to the medical image to determine the respective image quality features for the medical image;

applying the image classifier to the medical image to determine the classification scores for the one or more abnormalities, wherein the image classifier uses the determined image quality features to determine said classification scores; and

outputting a classification result based on the determined classification scores,

wherein a classification score is determined for an abnormality by:

applying a classification part of the image classifier to the medical image to obtain a preliminary classification score for the abnormality;

applying a correction part of the image classifier to at least the image quality features to determine a correction value for the preliminary classification score; and

applying the correction value to the preliminary classification score to obtain the classification score for the abnormality.

12 . The method of claim 11 , further comprising training the image classifier to detect the one or more abnormalities in the medical image by:

accessing training data comprising multiple training images, a training image being labelled with one or more image quality features and one or more abnormalities; and

training the image classifier to, given a training image and using the one or more image quality features that the training image is labelled with, detect the one or more abnormalities that the training image is labelled with.

13 . A non-transitory computer readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to:

access model data, wherein the model data includes data representing an image classifier trained to use one or more image quality features to determine classification scores for one or more abnormalities from a medical image, and wherein the model data further includes data representing one or more feature extractors configured to extract respective image quality features from the medical image indicative of a suitability of the medical image for detection of the one or more abnormalities;

obtain the medical image;

apply the one or more feature extractors to the medical image to determine the respective image quality features for the medical image;

apply the image classifier to the medical image to determine the classification scores for the one or more abnormalities, wherein the image classifier uses the determined image quality features to determine said classification scores; and

output a classification result based on the determined classification scores,

wherein the instructions, when executed by one or more processors, cause the one or more processors to determine a classification score for an abnormality by:

applying a classification part of the image classifier to the medical image to obtain a preliminary classification score for the abnormality;

applying a correction part of the image classifier to at least the image quality features to determine a correction value for the preliminary classification score;

applying the correction value to the preliminary classification score to obtain the classification score for the abnormality.

14 . The non-transitory computer readable medium of claim 13 , wherein an image quality feature represents a type of view of the medical image, a rotation degree of the medical image subject with respect to the plane of the medical image, a degree of inspiration indicating a presence of an adequately inspired volume in the medical image, or a degree of penetration indicating an adequacy of exposure of the medical image.

15 . The non-transitory computer readable medium of claim 13 , wherein applying the correction part comprises applying a rule-based model to at least the determined image quality features.

16 . The non-transitory computer readable medium of claim 13 , wherein the instructions, when executed by the one or more processors, further cause the one or more processors to: determine a classification score for an abnormality by applying a feature extraction part of the image classification model to the medical image to obtain image features of the medical image, and applying a neural network to the image features and the image quality features to obtain the classification score for the abnormality.

17 . The non-transitory computer readable medium of claim 13 , wherein the instructions, when executed by the one or more processors, further cause the one or more processors to: include correction data in the classification result indicating an effect of the image quality features on the classification score.

18 . The non-transitory computer readable medium of claim 13 , wherein the instructions, when executed by the one or more processors, further cause the one or more processors to: determine, based on the one or more image quality features, a subset of abnormalities that may be undetected due to image quality, and to output said subset of abnormalities as part of the classification result.

19 . The non-transitory computer readable medium of claim 18 , wherein the instructions, when executed by the one or more processors, further cause the one or more processors to: determine the subset based on applying a rule-based model, the processor subsystem being further configured to include rules used by the rule-based model to determine that an abnormality may be undetected in the classification result.

20 . The non-transitory computer readable medium of claim 13 , wherein the image classifier comprises a correction part configured to determine a correction value for a preliminary classification score according to a rule-based model.