Automatic generation of semantic description of visual findings in medical images

1. A method comprising using at least one hardware processor for:

receiving a medical image;

applying a mapping function to the medical image, wherein the mapping function is an SVM (Support Vector Machine) mapping function trained on a training set of medical images using a CRF (Conditional Random Field) model of the relationships between multiple received semantic descriptions of multiple medical findings in the training set;

generating a semantic description of a visual finding in the medical image based on the application of the mapping function; and

storing the medical image in association with the generated semantic description.

2. The method according to claim 1 , wherein the semantic description comprises a qualitative text.

3. The method according to claim 1 , wherein the semantic description comprises a quantitative text.

4. The method according to claim 1 , wherein the semantic description comprises a medical lexicon term.

5. The method according to claim 4 , wherein the medical lexicon is RadLex.

6. The method according to claim 1 , wherein the semantic description comprises a name of a parameter and a value of the parameter.

7. The method according to claim 1 , wherein the visual finding comprises an image feature associated with a portion of the medical image.

8. The method according to claim 1 , wherein the medical image is selected from the group consisting of: an X-Ray image, an MRI (Magnetic Resonance Imaging) image, a CT (Computerized Tomography) image, an angiography image, an ultrasound image, a nuclear image, a thermographic image and an echocardiographic image.

9. A method comprising using at least one hardware processor for:

providing, to one or more medical experts, a training set comprising multiple medical images;

receiving, from the one or more medical experts, multiple semantic descriptions of multiple visual findings in the multiple medical images;

training an SVM (Support Vector Machine) algorithm based on the training set and on a CRF (Conditional Random Field) model of the relationships between the multiple semantic descriptions; and

producing a mapping function from the multiple visual findings to the multiple semantic descriptions.

10. The method according to claim 9 , wherein the semantic descriptions comprise qualitative texts.

11. The method according to claim 9 , wherein the semantic descriptions comprise quantitative texts.

12. The method according to claim 9 , wherein the semantic descriptions comprise medical lexicon terms.

13. The method according to claim 12 , wherein the medical lexicon is RadLex.

14. The method according to claim 9 , wherein the semantic descriptions each comprises a name of a parameter and a value of the parameter.

15. The method according to claim 9 , wherein the visual findings each comprises an image feature associated with a portion of one of the multiple medical images.

16. The method according to claim 9 , wherein the multiple medical images are each selected from the group consisting of: an X-Ray image, an MRI (Magnetic Resonance Imaging) image, a CT (Computerized Tomography) image, an angiography image, an ultrasound image, a nuclear image, a thermographic image and an echocardiographic image.

17. The method according to claim 9 , further comprising using said at least one hardware processor for applying the mapping function to a medical image under investigation, to generate a semantic description of a visual finding in the medical image under investigation.

18. A computer program product for semantic description of visual findings in medical images, the computer program product comprising a non-transitory computer-readable storage medium having program code embodied therewith, the program code executable by at least one hardware processor to:

apply a mapping function to a medical image, wherein the mapping function is an SVM (Support Vector Machine) mapping function trained on a training set of medical images using a CRF (Conditional Random Field) model of multiple relationships between multiple received semantic descriptions of multiple medical findings in the training set;

generate a semantic description of a visual finding in the medical image based on the application of the mapping function; and

store the medical image in association with the generated semantic description.