Machine learning from noisy labels for abnormality assessment in medical imaging

Patent metadata

Field	Value
Publication number	US-11776117-B2
Application number	US-202017072424-A
Country	US
Kind code	B2
Filing date	Oct 16, 2020
Priority date	Jul 22, 2020
Publication date	Oct 3, 2023
Grant date	Oct 3, 2023

Abstract

For machine learning for abnormality assessment in medical imaging and application of a machine-learned model, the machine learning uses regularization of the loss, such as regularization being used for training for abnormality classification in chest radiographs. The regularization may be a noise and/or correlation regularization directed to the noisy ground truth labels of the training data. The resulting machine-learned model may better classify abnormalities in medical images due to the use of the noise and/or correlation regularization in the training.

First claim

We claim:

1. A method for machine learning abnormality assessment in medical imaging by a machine, the method comprising:
obtaining training data comprising medical images and ground truth labels for the medical images, the ground truth labels designating an abnormality represented by the medical images;
machine training, by the machine, a model from the training data, the machine training using a loss function, the loss function including a regularization, the regularization comprising a noise regularization, where the noise regularization includes a predetermined function of prior label noise probabilities; and
storing the model resulting from the machine training in a memory.

2. The method of claim 1 wherein machine training comprises machine training with the loss function comprising a cross-entropy function comparing a classification of abnormality output of the model with the ground truth labels.

3. The method of claim 1 where machine training comprises machine training with the ground truth labels comprising binary labels for absence or presence of the abnormality and the loss function being weighted as a function of number of positive and number of negative instances of the abnormality in the medical images of the training data.

4. The method of claim 1 further comprising measuring a noise level of the ground truth labels, and wherein machine training comprises machine training with the noise regularization being a function of the noise level.

5. The method of claim 4 wherein the noise level comprises a specificity and a sensitivity of the ground truth labels for the abnormality, and wherein the noise regularization comprises a first weight that is a function of the specificity and a second weight that is a function of the sensitivity.

6. The method of claim 1 wherein the noise regularization comprises an inverse binary cross-entropy function.

7. The method of claim 1 wherein the ground truth labels designate at least first and second types of abnormalities, and wherein machine training comprises machine training with the regularization of the loss function further comprises a correlation regularization, the correlation regularization correlating the ground truth labels for the first type of abnormality to the ground truth labels for the second type of abnormality.

8. The method of claim 7 wherein the correlation regularization comprises a covariance.

9. The method of claim 8 wherein the at least first and second types of abnormalities comprise at least four types of abnormalities, and wherein machine training comprises machine training with the correlation regularization as a sum of the covariance between all of the at least four types of abnormalities.

10. The method of claim 1 wherein machine training comprises machine training with the regularization comprising both the noise regularization and a correlation regularization.

11. The method of claim 1 wherein obtaining comprises obtaining the medical images of the training data as chest radiographs and wherein the abnormality comprises effusion, cardiomegaly, consolidation, atelectasis, and/or mass.

12. The method of claim 1 further comprising applying the model resulting from the machine training to a patient image for a patient, the applying outputting a classification of the patient image has having or not having the abnormality.

13. A system for abnormality detection in medical imaging, the system comprising:
a medical imaging system configured to generate an image of a patient;
a processor configured to apply a machine-learned model to the image of the patient, the machine-learned model having been trained with a noise regularization to detect an abnormality in the image, where the noise regularization includes a predetermined function of prior label noise probabilities; and
a display configured to display a classification of the patient as having or not having the abnormality based on the abnormality detection.

14. The system of claim 13 wherein the noise regularization accounting for noise in ground truth labels used in machine training.

15. The system of claim 13 wherein the processor is configured to apply the machine-learned model having been trained with a correlation regularization accounting for mischaracterization between different types of abnormalities.

16. The system of claim 13 wherein the medical imaging system comprises an x-ray system and wherein the noise regularization is for ground truth labels for abnormalities in x-ray images.

17. A system for machine training for abnormality classification, the system comprising:
a memory configured to store training data including images of anatomy and ground truth classifications for the images and to store a machine-learned classifier; and
a processor configured to machine train from the training data, the machine training including calculation of loss with a noise regularization, label noise probabilities, and the machine training with the calculation of loss resulting in the machine-learned classifier.

18. The system of claim 17 wherein the processor is further configured to machine train with a correlation regularization.

19. The system of claim 17 wherein the processor is further configured to:
measure a noise level of the ground truth labels; and
machine train with the noise regularization being a function of the noise level.

20. The system of claim 17 wherein the noise regularization comprises an inverse binary cross-entropy function.

Assignees

• Siemens Healthcare Gmbh

Inventors

• Guendel Sebastian
• Adiyoso Arnaud Arindra
• Ghesu Florin-Cristian
• Grbic Sasa
• Georgescu Bogdan
• Comaniciu Dorin

Classifications

• G06T7/0012 Primary

  Biomedical image inspection · CPC title

• G06T5/002

  Physics · mapped topic

• G06T2207/10116

  X-ray image · CPC title

• G06T2207/20081

  Training; Learning · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title