Deep learning medical systems and methods for image reconstruction and quality evaluation

Patent metadata

Field	Value
Publication number	US-10896352-B2
Application number	US-201916697904-A
Country	US
Kind code	B2
Filing date	Nov 27, 2019
Priority date	Nov 23, 2016
Publication date	Jan 19, 2021
Grant date	Jan 19, 2021

Abstract

Methods and apparatus to automatically generate an image quality metric for an image are provided. An example method includes automatically processing a first medical image using a deployed learning network model to generate an image quality metric for the first medical image, the deployed learning network model generated from a digital learning and improvement factory including a training network, wherein the training network is tuned using a set of labeled reference medical images of a plurality of image types, and wherein a label associated with each of the labeled reference medical images indicates a central tendency metric associated with image quality of the image. The example method includes computing the image quality metric associated with the first medical image using the deployed learning network model by leveraging labels and associated central tendency metrics to determine the associated image quality metric for the first medical image.

First claim

What is claimed is:

1. An apparatus comprising:
memory including instructions; and
at least one processor to execute the instructions to:
extract features from an image;
train an image quality learning network in which nodes represent features and weights are associated with at least a portion of the nodes in the image quality learning network, the image quality learning network trained to generate an image quality index;
process the image using a deployed model of the trained image quality learning network to generate the image quality index for the image; and
trigger a change in at least one of image acquisition or image reconstruction when the image quality index is less than a threshold.

2. The apparatus of claim 1 , wherein the at least one processor is to process the image by processing a portion of the image to generate the image quality index for the portion of the image.

3. The apparatus of claim 1 , wherein the image quality index includes at least one metric associated with spatial resolution, noise level, or detectability extracted from the image.

4. The apparatus of claim 1 , wherein the image quality learning network is trained using second features extracted from a set of labeled reference medical images, and wherein a label associated with each of the labeled reference medical images indicates an image quality index for the respective reference medical image, the second features associated with a target value for the image quality index.

5. The apparatus of claim 4 , wherein the at least one processor is to compare the image quality metric to the target value to determine a reliability of the deployed model.

6. The apparatus of claim 4 , wherein the wherein the trained image quality learning network is tested using one or more unlabeled images before the deployed model is generated from the trained image quality learning network.

7. The apparatus of claim 1 , wherein the image quality index is defined on a 5-point scale, wherein a low score on the 5-point scale indicates a low confidence to make a decision based on the image and wherein a high score on the 5-point scale indicates a high confidence to make the decision based on the image.

8. The apparatus of claim 1 , wherein the features are calculated from image intensity values including at least one of mean, standard deviation, skewness, kurtosis, energy, contract, moment, or entropy.

9. The apparatus of claim 1 , wherein the image quality learning network is trained using a reference medical image and at least one blurred variant of the reference medical image.

10. At least one computer-readable storage medium comprising instructions that, when executed, cause at least one processor to:
extract features from an image;
train an image quality learning network in which nodes represent features and weights are associated with at least a portion of the nodes in the image quality learning network, the image quality learning network trained to generate an image quality index;
process the image using a deployed model of the trained image quality learning network to generate the image quality index for the image; and
trigger a change in at least one of image acquisition or image reconstruction when the image quality index is less than a threshold.

11. The at least one computer-readable storage medium of claim 10 , wherein the instructions, when executed, cause the at least one processor to process the image by processing a portion of the image to generate the image quality index for the portion of the image.

12. The at least one computer-readable storage medium of claim 10 , wherein the image quality index includes at least one metric associated with spatial resolution, noise level, or detectability extracted from the image.

13. The at least one computer-readable storage medium of claim 10 , wherein the instructions, when executed, cause the at least one processor to train the image quality learning network using second features extracted from a set of labeled reference medical images, and wherein a label associated with each of the labeled reference medical images indicates an image quality index for the respective reference medical image, the second features associated with a target value for the image quality index.

14. The at least one computer-readable storage medium of claim 13 , wherein the instructions, when executed, cause the at least one processor to compare the image quality metric to the target value to determine a reliability of the deployed model.

15. The at least one computer-readable storage medium of claim 13 , wherein the instructions, when executed, cause the at least one processor to test the trained image quality learning network using one or more unlabeled images before the deployed model is generated from the trained image quality learning network.

16. A method to extract an image quality index from an image, the method comprising:
extracting features from an image;
training an image quality learning network in which nodes represent features and weights are associated with at least a portion of the nodes in the image quality learning network, the image quality learning network trained to generate an image quality index;
processing the image using a deployed model of the trained image quality learning network to generate the image quality index for the image; and
triggering a change in at least one of image acquisition or image reconstruction when the image quality index is less than a threshold.

17. The method of claim 16 , wherein processing the image includes processing a portion of the image to generate the image quality index for the portion of the image.

18. The method of claim 16 , wherein training the image quality learning network includes training the image quality learning network using second features extracted from a set of labeled reference medical images, wherein a label associated with each of the labeled reference medical images indicates an image quality index for the respective reference medical image, the second features associated with a target value for the image quality index.

19. The method of claim 18 , further including comparing the image quality metric to the target value to determine a reliability of the deployed model.

20. The method of claim 16 , further including testing the trained image quality learning network using one or more unlabeled images before the deployed model is generated from the trained image quality learning network.

Assignees

• Gen Electric

Inventors

• Hsieh Jiang
• Avinash Gopal
• Sirohey Saad
• Wang Xin
• Yin Zhye
• De Man Bruno

Classifications

• G06N3/084 Primary

  Backpropagation, e.g. using gradient descent · CPC title

• G06F18/2193

  based on specific statistical tests · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06V10/454

  Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN] · CPC title

• G06V10/993

  Evaluation of the quality of the acquired pattern · CPC title