Granted Patent B2

US 12,361,547 · App. 18/075,314 · Granted Jul 15, 2025

Inventors: Dong Wei (Shenzhen, CN); Yuexiang Li (Shenzhen, CN); Yi Lin (Shenzhen, CN); Kai Ma (Shenzhen, CN); Yefeng Zheng (Shenzhen, CN)

Assignee: TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED

G06T7/0012G06V10/774G06V40/14G06T2207/20081G06T2207/20221G06T2207/30101

Abstract

Claims (69)

1. A method, performed by a computer device, comprising:

acquiring an initial sample angiography image set, the initial sample angiography image set comprising a first sample angiography image corresponding to a first blood vessel at a target site, and a first label of the first sample angiography image, wherein the first label indicates that the first sample angiography image corresponds to the first blood vessel at the target site;

performing data expansion processing on the first sample angiography image based on physical characteristics of blood vessels at the target site to obtain a processed sample angiography image;

performing label conversion processing on the first label based on the physical characteristics of the blood vessels at the target site to obtain a second label of the processed sample angiography image; and

adding the processed sample angiography image and the second label to a target sample angiography image set, the second label indicating that the processed sample angiography image corresponds to a second blood vessel at the target site, wherein a region between the first blood vessel and the second blood vessel satisfies the physical characteristics of the blood vessels at the target site; and

training an angiography image recognition model using the initial sample angiography image set and the target sample angiography image set to obtain a trained angiography image recognition model, the trained angiography image recognition model identifying a target blood vessel at the target site in an angiography image.

2. The method according to claim 1 , wherein the physical characteristics of the blood vessels at the target site comprise a symmetry; and performing data expansion processing on the first sample angiography image comprises:

forming a mirror image of the first sample angiography image based on the symmetry of the blood vessels at the target site to obtain the processed sample angiography image, wherein the processed sample angiography image is symmetrical to the first sample angiography image.

3. The method according to claim 2 , wherein a label of an angiography image is a vector comprising a plurality of elements, each element characterizing a possibility that a respective blood vessel is in the angiography image; and

in the first label of the first sample angiography image, an element value corresponding to the first blood vessel has a first value, and an element value corresponding to blood vessels other than the first blood vessel has a second element value.

4. The method according to claim 3 , wherein performing label conversion processing comprises:

adjusting the element value corresponding to the first blood vessel in the first label from the first value to the second value; and adjusting the element value corresponding to the second blood vessel in the first label from the second value to the first value; and

using the adjusted first label as the second label of the processed sample angiography image.

5. The method according to claim 1 , wherein the initial sample angiography image set further comprises a second sample angiography image corresponding to a third blood vessel, distinct from the first blood vessel, at the target site, and wherein the physical characteristics of the blood vessels at the target site comprise a fusibility; and

performing data expansion processing on the first sample angiography image comprises:

performing fusion processing on the first sample angiography image and the second sample angiography image based on the fusibility to obtain the processed sample angiography image.

6. The method according to claim 5 , wherein performing label conversion processing on the first label comprises:

adjusting the element value corresponding to the third intracranial blood vessel in the first label from a second element to a first element value; and

setting the adjusted first label as the second label of the processed sample angiography image.

7. The method according to claim 1 , wherein training the angiography image recognition model using the initial sample angiography image set and the target sample angiography image set to obtain a trained angiography image recognition model comprises:

performing recognition processing, using the angiography image recognition model, on a plurality of sample angiography images in the initial sample angiography image set to obtain a first recognition result corresponding to each of the plurality of sample angiography images in the initial sample angiography image set;

performing recognition processing, using the angiography image recognition model, on a plurality of sample angiography images in the target sample angiography image set to obtain a second recognition result corresponding to each of the plurality of sample angiography images in the target sample angiography image set;

determining a first loss function of the angiography image recognition model based on a difference between a plurality of the first recognition results and the first labels corresponding sample angiography images, and a difference between a plurality of the second recognition results and the second labels corresponding to sample angiography images; and

updating parameters of the angiography image recognition model based on the first loss function to train the angiography image recognition model.

8. The method according to claim 7 , wherein updating parameters of the angiography image recognition model based on the first loss function comprises:

acquiring an annotated sample angiography image set comprising a plurality of annotated sample angiography images and corresponding annotation information, the annotation information of the annotated sample angiography image indicating rupture information of a target object in a fourth blood vessel at the target site in the annotated sample angiography image, the rupture information comprising at least one of rupture risk information of the target object and rupture time information of the target object;

performing prediction processing, using the angiography image recognition model, on a plurality of annotated sample angiography images in the annotated sample angiography image set to obtain predicted rupture information of the target object in the plurality of annotated sample angiography images;

determining a second loss function of the angiography image recognition model according to a difference between the predicted rupture information of the target object in each annotated sample angiography image and the annotation information of the annotated sample angiography image; and

optimizing the angiography image recognition model based on the first loss function and the second loss function to obtain the trained angiography image recognition model, the trained angiography image recognition model identifying a target blood vessel at the target site in an angiography image and predicting rupture information of a target object in the target blood vessel in the angiography image.

9. The method according to claim 8 , wherein optimizing the angiography image recognition model based on the first loss function and the second loss function comprises:

performing weighting processing on the first loss function and the second loss function to obtain a third loss function; and

updating the parameters of the angiography image recognition model based on the third loss function.

10. The method according to claim 8 , wherein the method further comprises:

acquiring an angiography image to be analyzed; and

performing recognition processing, using the trained angiography image recognition model or the target angiography image recognition model, on the angiography image to obtain rupture information of a target object in a target blood vessel at the target site in the angiography image.

11. The method according to claim 1 , wherein after performing training on the angiography image recognition model, the method further comprises:

acquiring an annotated sample angiography image set, the annotated sample angiography image set comprising a plurality of annotated sample angiography images having respective associated annotation information, the respective associated annotation information of the annotated sample angiography image indicating rupture information of a target object in a fourth blood vessel at the target site in the annotated sample angiography image, the rupture information comprising at least one of rupture risk information of the target object and rupture time information of the target object; and

training the trained angiography image recognition model using the annotated sample angiography image set to obtain a target angiography image recognition model for identifying a target blood vessel at the target site in an angiography image and rupture information of a target object in the target blood vessel.

12. An electronic device, comprising:

one or more processors; and

memory storing one or more programs, the one or more programs comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:

acquiring an initial sample angiography image set, the initial sample angiography image set comprising a first sample angiography image corresponding to a first blood vessel at a target site, and a first label of the first sample angiography image, wherein the first label indicates that the first sample angiography image corresponds to the first blood vessel at the target site;

performing data expansion processing on the first sample angiography image based on physical characteristics of blood vessels at the target site to obtain a processed sample angiography image;

performing label conversion processing on the first label based on the physical characteristics of the blood vessels at the target site to obtain a second label of the processed sample angiography image; and

adding the processed sample angiography image and the second label to a target sample angiography image set, the second label indicating that the processed sample angiography image corresponds to a second blood vessel at the target site, wherein a region between the first blood vessel and the second blood vessel satisfies the physical characteristics of the blood vessels at the target site; and

training an angiography image recognition model using the initial sample angiography image set and the target sample angiography image set to obtain a trained angiography image recognition model, the trained angiography image recognition model identifying a target blood vessel at the target site in an angiography image.

13. The electronic device according to claim 12 , wherein the initial sample angiography image set further comprises a second sample angiography image corresponds to a third blood vessel, distinct from the first blood vessel, at the target site, and wherein the physical characteristics of the blood vessels at the target site comprise a fusibility; and

performing data expansion processing on the first sample angiography image comprises:

performing fusion processing on the first sample angiography image and the second sample angiography image based on the fusibility to obtain the processed sample angiography image.

14. The electronic device according to claim 12 , wherein the physical characteristics of the blood vessels at the target site comprise a symmetry; and performing data expansion processing on the first sample angiography image comprises:

forming a mirror image of the first sample angiography image based on the symmetry of the blood vessels at the target site to obtain the processed sample angiography image, wherein the processed sample angiography image is symmetrical to the first sample angiography image.

15. The electronic device according to claim 14 , wherein a label of an angiography image is a vector comprising a plurality of elements, each element characterizing a possibility that a respective blood vessel is in the angiography image; and

in the first label of the first sample angiography image, an element value corresponding to the first blood vessel has a first value, and an element value corresponding to blood vessels other than the first blood vessel has a second element value.

16. The electronic device according to claim 15 , wherein performing label conversion processing comprises:

adjusting the element value corresponding to the first blood vessel in the first label from the first value to the second value; and adjusting the element value corresponding to the second blood vessel in the first label from the second value to the first value; and

using the adjusted first label as the second label of the processed sample angiography image.

17. A non-transitory computer-readable storage medium, storing a computer program, the computer program, when executed by one or more processors of an electronic device, cause the one or more processors to perform operations comprising:

acquiring an initial sample angiography image set, the initial sample angiography image set comprising a first sample angiography image corresponding to a first blood vessel at a target site, and a first label of the first sample angiography image, wherein the first label indicates that the first sample angiography image corresponds to the first blood vessel at the target site;

performing data expansion processing on the first sample angiography image based on physical characteristics of blood vessels at the target site to obtain a processed sample angiography image;

performing label conversion processing on the first label based on the physical characteristics of the blood vessels at the target site to obtain a second label of the processed sample angiography image; and

adding the processed sample angiography image and the second label to a target sample angiography image set, the second label indicating that the processed sample angiography image corresponds to a second blood vessel at the target site, wherein a region between the first blood vessel and the second blood vessel satisfies the physical characteristics of the blood vessels at the target site; and

training an angiography image recognition model using the initial sample angiography image set and the target sample angiography image set to obtain a trained angiography image recognition model, the trained angiography image recognition model identifying a target blood vessel at the target site in an angiography image.

18. The non-transitory computer-readable storage medium according to claim 17 , wherein the physical characteristics of the blood vessels at the target site comprise a symmetry; and performing data expansion processing on the first sample angiography image comprises:

forming a mirror image of the first sample angiography image based on the symmetry of the blood vessels at the target site to obtain the processed sample angiography image, wherein the processed sample angiography image is symmetrical to the first sample angiography image.

19. The non-transitory computer-readable storage medium according to claim 18 , wherein a label of an angiography image is a vector comprising a plurality of elements, each element characterizing a possibility that a respective blood vessel is in the angiography image; and

in the first label of the first sample angiography image, an element value corresponding to the first blood vessel has a first value, and an element value corresponding to blood vessels other than the first blood vessel has a second element value.

20. The non-transitory computer-readable storage medium according to claim 19 , wherein performing label conversion processing comprises:

adjusting the element value corresponding to the first blood vessel in the first label from the first value to the second value; and adjusting the element value corresponding to the second blood vessel in the first label from the second value to the first value; and

using the adjusted first label as the second label of the processed sample angiography image.

Assignments (1)

Priority Claims (1)

Continuity (2)

References Cited (28)