Automatic method and system for vessel refine segmentation in biomedical images using tree structure based deep learning model

• US 10,573,005 B2
• Filed: 08/01/2019
• Issued: 02/25/2020
• Est. Priority Date: 04/24/2018
• Status: Active Grant

• Current Assignee: Shenzhen Keya Medical Technology Corporation
• Original Assignee: Shenzhen Keya Medical Technology Corporation
• Inventors: Wang, Xin, Yin, Youbing, Bai, Junjie, Lu, Yi, Song, Qi, Cao, Kunlin
• Primary Examiner(s): Bayat, Ali
• Application Number:
  US16/529,769
• Publication Number:
  US 20190355120A1
• Time in Patent Office: 208 Days
• Field of Search: 382173
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
• CPC Class Codes: G06T 2207/10072 Tomographic images
• CPC Class Codes: G06T 2207/10116 X-ray image
• CPC Class Codes: G06T 2207/10132 Ultrasound image
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/20084 Artificial neural networks [ANN]
• CPC Class Codes: G06T 2207/30101 Blood vessel; Artery; Vein; Vascular
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/174 involving the use of two or more images

Abstract

Embodiments of the disclosure provide systems and methods for analyzing a biomedical image including at least one tree structure object. The system includes a communication interface configured to receive a learning model and a plurality of model inputs derived from the biomedical image. The biomedical image is acquired by an image acquisition device. The system further includes at least one processor configured to apply the learning model to the plurality of model inputs to analyze the biomedical image. The learning model includes a first network configured to process the plurality of model inputs to construct respective feature maps and a second network configured to process the feature maps collectively. The second network is a tree structure network that models a spatial constraint of the tree structure object.

Claims

1. A system for analyzing a biomedical image including at least one tree structure object, comprising:

a communication interface configured to receive a learning model and a plurality of model inputs derived from the biomedical image, the biomedical image being acquired by an image acquisition device; and

at least one processor, configured to apply the learning model to the plurality of model inputs to analyze the biomedical image, wherein the learning model includes a first network configured to process the plurality of model inputs to construct respective feature maps and a second network configured to process the feature maps collectively, wherein the second network is a tree structure network that models a spatial constraint of the tree structure object.

2. The system of claim 1, wherein the model inputs are image patches extracted along a centerline of the tree structure object in the biomedical image.

3. The system of claim 1, wherein the first network includes a plurality of convolutional networks each configured to process a model input to construct a feature map.

4. The system of claim 3, wherein the tree structure network includes a plurality of nodes each receiving at least one feature map constructed by at least one convolutional network.

5. The system of claim 1, wherein the tree structure network includes a plurality of nodes arranged in a structure consistent with that of the tree structure object.

6. The system of claim 5, wherein the plurality of nodes include a root and a plurality of terminals, wherein the tree structure network is implemented to allow information flow in a single direction either from the root to the terminals or from the terminals to the root.

7. The system of claim 5, wherein the tree structure network is implemented to allow information flow in two directions among the plurality of nodes.

8. The system of claim 1, wherein the tree structure network is a tree structure convolution recurrent neural network.

9. The system of claim 2, wherein biomedical image is a 2D image, wherein the spatial constraint is determined based on a spatial relationship between two neighboring image patches along the centerline extracted within the biomedical image.

10. The system of claim 2, wherein biomedical image is a 3D image including a plurality of image slices, wherein the spatial constraint is determined based on a spatial relationship between corresponding image patches extracted from two neighboring image slices.

11. The system of claim 1, wherein the tree structure object is a blood vessel.

12. The system of claim 1, wherein the tree structure object includes at least one trunk and a plurality of branches, wherein the model inputs include at least one model input derived from each branch.

13. A method for analyzing a biomedical image including at least one tree structure object, comprising:

receiving a learning model and a plurality of model inputs derived from the biomedical image, the biomedical image being acquired by an image acquisition device; and

applying, by at least one processor, the learning model to the plurality of model inputs to analyze the biomedical image, wherein the learning model includes a first network configured to process the plurality of model inputs to construct respective feature maps and a second network configured to process the feature maps collectively, wherein the second network is a tree structure network that models a spatial constraint of the tree structure object.

14. The method of claim 13, wherein the tree structure object includes at least one trunk and a plurality of branches, wherein the model inputs include at least one model input derived from each branch.

15. The method of claim 13, wherein the first network includes a plurality of convolutional networks each configured to process a model input to construct a feature map, wherein the tree structure network includes a plurality of nodes each receiving at least one feature map constructed by at least one convolutional network.

16. The method of claim 13, wherein the tree structure network includes a plurality of nodes arranged in a structure consistently with that of the tree structure object.

17. The method of claim 16, wherein the plurality of nodes include a root and a plurality of terminals, wherein the tree structure network is implemented to allow information flow in a single direction either from the root to the terminals or from the terminals to the root.

18. The method of claim 16, wherein the tree structure network is implemented to allow information flow in two directions among the plurality of nodes.

19. The method of claim 13, wherein the tree structure object is a blood vessel.

20. A non-transitory computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by at least one processor, performs a method for analyzing a biomedical image including at least one tree structure object, the method comprising:

receiving a learning model and a plurality of model inputs derived from the biomedical image, the biomedical image being acquired by an image acquisition device; and

applying the learning model to the plurality of model inputs to analyze the biomedical image, wherein the learning model includes a first network configured to process the plurality of model inputs to construct respective feature maps and a second network configured to process the feature maps collectively, wherein the second network is a tree structure network that models a spatial constraint of the tree structure object.