Fundus image processing method, computer apparatus, and storage medium

1. A method of processing fundus image, the method comprising:

receiving a collected fundus image;

identifying the collected fundus image via a first neural network to generate a first feature set of the collected fundus image, wherein the first feature set indicates a lesion type attribute of the collected fundus image, the second feature set indicates a lesion level attribute of the collected fundus image;

identifying the collected fundus image via a second neural network to generate a second feature set of the collected fundus image, wherein the first feature set and the second feature set indicate different lesion attributes of the collected fundus image;

combining the first feature set with the second feature set to obtain a combined feature set of the collected fundus image; and

inputting the combined feature set into a classifier to obtain a classification result, wherein the inputting includes inputting the combined feature set containing the lesion type attribute and the lesion level attribute into a multi-level classifier composed of a plurality class II classifiers according to a preset classification logic to obtain a multi-level classification result of the collected fundus image.

2. The method according to claim 1 , wherein the identifying the collected fundus image via the first neural network to obtain the first feature set of the fundus image comprises:

quadrant-dividing the fundus image to generate a quadrant image group;

inputting each quadrant image of the quadrant image group into the first neural network to obtain a feature vector of each quadrant image; and

combining the feature vectors of each quadrant image to generate the first feature set of the fundus image.

3. The method according to claim 1 , wherein the received collected fundus image comprises a left eye fundus image and a right eye fundus image of the same patient;

wherein the inputting the combined feature set into the classifier to obtain the classification result comprises:

connecting a combined feature set of the left eye fundus image with a combined feature set of the right eye fundus image to generate a combined feature sequence of the collected fundus image; and

inputting the combined feature sequence into the classifier to obtain the classification result.

4. The method according to claim 1 , wherein the received collected fundus image comprises a first view field left eye fundus image, a second view field left eye fundus image, a first view field right eye fundus image, and a second view field right eye fundus image of the same patient;

wherein the inputting the combined feature set into the classifier to obtain the classification result comprises:

connecting a combined feature set of the first view field left eye fundus image, a combined feature set of the second view field left eye fundus image, a combined feature set of the first view field right eye fundus image, with a combined feature set of the second view field right eye fundus image to generate the combined feature sequence of the collected fundus image; and

inputting the combined feature sequence into the classifier to obtain the classification result.

5. The method according to claim 1 , wherein the identifying the collected fundus image via the second neural network to generate the second feature set of the collected fundus image comprises:

identifying a lesion level attribute of the collected fundus image via the second neural network, and outputting a lesion level vector of the fundus image, wherein, when a fundus lesion is set to comprise n levels of the lesion, the generated lesion level vector has a length of n−1, wherein, in the feature vector of an i-th level lesion, the i-th element and the elements prior to the i-th element are configured as 1 and the rest elements are configured as 0.

6. The method according to claim 1 , wherein after receiving the collected fundus image, the method further comprises:

cropping the collected fundus image according to an information area of the collected fundus image; and

preprocessing the cropped fundus image.

7. A computer apparatus comprising one or more processors and a memory storing computer-readable instructions, which, when executed by the one or more processors, causing the one or more processors to perform a method comprising:

receiving a collected fundus image;

identifying the collected fundus image via a first neural network to generate a first feature set of the fundus image, wherein the first feature set indicates a lesion type attribute of the collected fundus image, the second feature set indicates a lesion level attribute of the collected fundus image;

identifying the collected fundus image via a second neural network to generate a second feature set of the collected fundus image, wherein the first feature set and the second feature set indicate different lesion attributes of the collected fundus image;

combining the first feature set and the second feature set to obtain a combined feature set of the fundus image; and

inputting the combined feature set into a classifier to obtain a classification result, wherein the inputting includes inputting the combined feature set containing the lesion type attribute and the lesion level attribute into a multi-level classifier composed of a plurality class II classifiers according to a preset classification logic to obtain a multi-level classification result of the collected fundus image.

8. The computer apparatus according to claim 7 , wherein when executing the computer-readable instructions, the one or more processors further perform the following steps:

quadrant-dividing the collected fundus image to generate a quadrant image group;

inputting each quadrant image of the quadrant image group into the first neural network to obtain a feature vector of each quadrant image; and

combining the feature vectors of each quadrant image to generate the first feature set of the collected fundus image.

9. The computer apparatus according to claim 7 , wherein the received fundus image comprises a left eye fundus image and a right eye fundus image of the same patient; when executing the computer-readable instructions, the one or more processors further performs the following steps:

connecting a combined feature set of the left eye fundus image with a combined feature set of the right eye fundus image to generate a combined feature sequence of the collected fundus image; and

inputting the combined feature sequence into the classifier to obtain the classification result.

10. The computer apparatus according to claim 7 , wherein the received fundus image comprises a first view field left eye fundus image, a second view field left eye fundus image, a first view field right eye fundus image, and a second view field right eye fundus image of the same patient; when executing the computer-readable instructions, the one or more processors further perform the following steps:

connecting a combined feature set of the first view field left eye fundus image, a combined feature set of the second view field left eye fundus image, a combined feature set of the first view field right eye fundus image, with a combined feature set of the second view field right eye fundus image to generate the combined feature sequence of the collected fundus image; and

inputting the combined feature sequence into the classifier to obtain the classification result.

11. The computer apparatus according to claim 7 , wherein when executing the computer-readable instructions, the one or more processors further performs the following steps: identifying a lesion level attribute of the collected fundus image via the second neural network, and outputting a lesion level vector of the fundus image, wherein, when a fundus lesion is set to comprise n levels of the lesion, the generated lesion level vector has a length of n−1, wherein, in the feature vector of an i-th level lesion, the i-th element and the elements prior to the i-th element are 1 and the numerals on the rest element are 0.

12. The computer apparatus according to claim 7 , wherein after performing the receiving the collected fundus image, the one or more processors further perform the following steps:

cropping the collected fundus image according to an information area of the collected fundus image; and

preprocessing the cropped fundus image.

13. At least one non-transitory computer-readable storage medium storing computer-readable instructions, which, when executed by one or more processors, causing the one or more processors to perform a method comprising:

receiving a collected fundus image;

identifying the fundus image via a first neural network to generate a first feature set of the collected fundus image, wherein the first feature set indicates a lesion type attribute of the collected fundus image, the second feature set indicates a lesion level attribute of the collected fundus image;

identifying the fundus image via a second neural network to generate a second feature set of the collected fundus image, wherein the first feature set and the second feature set indicate different lesion attributes of the collected fundus image;

combining the first feature set and the second feature set to obtain a combined feature set of the collected fundus image; and

inputting the combined feature set into a classifier to obtain a classification result, wherein the inputting includes inputting the combined feature set containing the lesion type attribute and the lesion level attribute into a multi-level classifier composed of a plurality class II classifiers according to a preset classification logic to obtain a multi-level classification result of the fundus image.

14. The storage medium of claim 13 , wherein the computer-readable instructions when executed by the one or more processors, further cause the one or more processors to perform the following steps:

quadrant-dividing the collected fundus image to generate a quadrant image group;

inputting each quadrant image of the quadrant image group into the first neural network to obtain a feature vector of each quadrant image; and

combining the feature vectors of each quadrant image to generate the first feature set of the collected fundus image.

15. The storage medium according to claim 13 , wherein the received fundus image comprises a left eye fundus image and a right eye fundus image of the same patient; the computer-readable instructions when executed by the one or more processors, further cause the one or more processors to perform the following steps:

connecting a combined feature set of the left eye fundus image with a combined feature set of the right eye fundus image to generate a combined feature sequence of the collected fundus image; and

inputting the combined feature sequence into the classifier to obtain the classification result.

16. The storage medium according to claim 13 , wherein the received fundus image comprises a first view field left eye fundus image, a second view field left eye fundus image, a first view field right eye fundus image, and a second view field right eye fundus image of the same patient; the computer-readable instructions when executed by the one or more processors, further cause the one or more processors to perform the following steps:

connecting a combined feature set of the first view field left eye fundus image, a combined feature set of the second view field left eye fundus image, a combined feature set of the first view field right eye fundus image, with a combined feature set of the second view field right eye fundus image to generate the combined feature sequence of the collected fundus image; and

inputting the combined feature sequence into the classifier to obtain the classification result.

17. The storage medium of claim 13 , wherein the computer-readable instructions when executed by the one or more processors, further cause the one or more processors to perform the following steps:

identifying a lesion level attribute of the fundus image via the second neural network and

outputting a lesion level vector of the fundus image, wherein when a fundus lesion is set to comprise n levels of the lesion, the generated lesion level vector has a length of n−1, wherein, in the feature vector of an i-th level lesion, the i-th element and the elements prior to the i-th element are 1 and the rest elements are 0.