Fundus image processing method and apparatus

Abstract

The present application relates to a fundus image processing method and apparatus. The method comprises: acquiring a retinal optical-coherence-tomography image of a target eye and a corresponding retinal angiography image; and inputting the retinal optical-coherence-tomography image and the corresponding retinal angiography image into a retinal lesion detection model, so as to obtain a retinal lesion detection result of the target eye. In the method, a retinal lesion detection result of a target eye is determined by introducing a retinal angiography image and also using a retinal optical-coherence-tomography image, and the retinal lesion detection result obtained in this manner is more accurate, thereby improving the accuracy of fundus lesion detection.

Description

Fundus image processing method and device

[0001] Related applications:

[0002] This application claims priority to Chinese patent application number 202310973303.0, filed on August 2, 2023, entitled “Fundus Image Processing Method and Device,” the entire text of which is hereby incorporated by reference. Technical Field

[0003] The present application relates to the field of artificial intelligence technology, and in particular to a fundus image processing method and device. Background Art

[0004] With the rapid development of artificial intelligence technology, artificial intelligence has gradually penetrated into different industries and fields, such as medical care, finance, education, e-commerce, transportation, aviation and tourism.

[0005] Taking the field of ophthalmology as an example, deep learning methods are usually used in related technologies to automatically detect fundus lesions to assist in the diagnosis of fundus diseases.

[0006] However, the fundus lesion detection method in the related art has the problem of insufficient detection accuracy.

[0007] Summary of the Invention

[0008] Based on this, it is necessary to provide a fundus image processing method and device to address the above technical problems. By introducing retinal blood flow images and combining them with retinal optical coherence tomography images, the retinal lesion detection results of the target eye are determined. The retinal lesion detection results obtained in this way are more accurate, thereby improving the accuracy of fundus lesion detection.

[0009] In a first aspect, an embodiment of the present application provides a method for processing fundus images. The method includes:

[0010] Acquiring a retinal optical coherence tomography image and a corresponding retinal blood flow image of the target eye;

[0011] The retinal optical coherence tomography image and the corresponding retinal blood flow image are input into the retinal lesion detection model to obtain the retinal lesion detection results of the target eye.

[0012] In one embodiment, obtaining a retinal optical coherence tomography image and a retinal blood flow image of a target eye includes:

[0013] Acquire three-dimensional retinal structure images and three-dimensional retinal blood flow images of the target eye;

[0014] Acquiring a plurality of first-direction retinal optical coherence tomography images and a plurality of second-direction retinal optical coherence tomography images based on the three-dimensional retinal structure image; determining each of the first-direction retinal optical coherence tomography images and each of the second-direction retinal optical coherence tomography images as a retinal optical coherence tomography image;

[0015] According to the three-dimensional retinal blood flow image, a plurality of first-direction retinal blood flow images and a plurality of second-direction retinal blood flow images are acquired; and each first-direction retinal blood flow image and each second-direction retinal blood flow image are determined as a retinal blood flow image.

[0016] In one embodiment, the retinal optical coherence tomography image and the corresponding retinal blood flow image are input into a retinal lesion detection model to obtain a retinal lesion detection result of the target eye, including:

[0017] Inputting the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a first lesion detection result and a first confidence level of the first lesion detection result;

[0018] A retinal lesion detection result of the target eye is determined according to the first lesion detection result and a first confidence level of the first lesion detection result.

[0019] In one embodiment, determining a retinal lesion detection result of a target eye based on a first lesion detection result and a first confidence level of the first lesion detection result includes:

[0020] Determining a second lesion detection result according to the first lesion detection result, the first confidence level, and a first preset confidence threshold;

[0021] Obtain retinal layering results from retinal optical coherence tomography;

[0022] The retinal lesion detection results were determined based on the retinal stratification results and the second lesion detection results.

[0023] In one embodiment, determining the retinal lesion detection result according to the retinal layering result and the second lesion detection result includes:

[0024] Determine the third lesion detection result based on the retinal layering result and the second lesion detection result;

[0025] The third lesion detection results are clustered to obtain retinal lesion detection results.

[0026] In one embodiment, determining a third lesion detection result based on the retinal layering result and the second lesion detection result includes:

[0027] Obtaining lesion type and lesion location information from the second lesion detection result;

[0028] The third lesion detection result is determined based on the lesion type, lesion location information and retinal stratification results.

[0029] In one embodiment, determining a third lesion detection result based on the lesion type, lesion location information, and retinal layering results includes:

[0030] Matching the lesion location information with the retinal layering result to obtain a matching result; the matching result indicates that the lesion is located at a corresponding position in the retinal layering result;

[0031] The third lesion detection result is determined based on the matching result and the lesion type.

[0032] In one embodiment, clustering the third lesion detection result to obtain a retinal lesion detection result includes:

[0033] Clustering the third lesion detection results to obtain clustered lesion detection results and a second confidence level of the clustered lesion detection results; the second confidence level is determined based on the first confidence level of the first lesion detection results included in the clustered lesion detection results;

[0034] The retinal lesion detection result is determined according to the clustered lesion detection result, the second confidence level and the second preset confidence threshold.

[0035] In one embodiment, the method further comprises:

[0036] acquiring a fundus projection map and / or fundus image of a target eye;

[0037] The retinal lesion detection result is projected onto the fundus projection map and / or the fundus image to obtain a fundus lesion distribution map of the target eye.

[0038] In a second aspect, an embodiment of the present application further provides a fundus image processing device. The device includes:

[0039] An image acquisition module, used to acquire a retinal optical coherence tomography image and a corresponding retinal blood flow image of a target eye;

[0040] The result determination module is used to input the retinal optical coherence tomography image and the corresponding retinal blood flow image into the retinal lesion detection model to obtain the retinal lesion detection result of the target eye. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0042] FIG1 is a diagram illustrating an application environment of a fundus image processing method according to an embodiment;

[0043] FIG2 is a schematic flow chart of a fundus image processing method according to an embodiment;

[0044] FIG3 is a schematic diagram of a retinal blood flow image according to an embodiment;

[0045] FIG4 is a schematic diagram of a retinal optical coherence tomography image according to one embodiment;

[0046] FIG5 is a schematic diagram of a process for obtaining retinal optical coherence tomography images and retinal blood flow images in one embodiment;

[0047] FIG6 is a schematic diagram of a process for determining retinal lesion detection results in one embodiment;

[0048] FIG7 is a schematic diagram of a process for determining retinal lesion detection results in another embodiment;

[0049] FIG8 is a schematic diagram of a process for determining retinal lesion detection results in another embodiment;

[0050] FIG9 is a schematic diagram of a process for determining a third lesion detection result in one embodiment;

[0051] FIG10 is a schematic diagram of a process for determining retinal lesion detection results in another embodiment;

[0052] FIG11 is a schematic flow chart of a method for determining a fundus lesion distribution map according to one embodiment;

[0053] FIG12 is a schematic diagram of a fundus lesion distribution map based on a fundus projection map in one embodiment;

[0054] FIG13 is a schematic diagram of a fundus lesion distribution map based on a fundus image in one embodiment;

[0055] FIG14 is a schematic flow chart of a fundus image processing method according to another embodiment;

[0056] FIG15 is a schematic diagram of the structure of a fundus image processing device according to one embodiment;

[0057] FIG16 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0059] The fundus image processing method provided in the embodiment of the present application can be applied to the application environment shown in Figure 1. Therein, the optical coherence tomography scanner 102 communicates with the server 104 via a network. The data storage system can store data that the server 104 needs to process. Optionally, the optical coherence tomography scanner 102 is used to obtain retinal optical coherence tomography images and corresponding retinal blood flow images. The data storage system can be integrated on the optical coherence tomography scanner 102, or it can be placed on the server 104 or on the cloud or other network servers. The server 104 can be implemented as an independent server or a server cluster consisting of multiple servers.

[0060] In one embodiment, as shown in FIG2 , a fundus image processing method is provided. The method is described by taking the server 104 in FIG1 as an example, and includes the following steps:

[0061] S201, obtaining a retinal optical coherence tomography image and a corresponding retinal blood flow image of a target eye.

[0062] Optionally, an optical coherence tomography (OCT) scanner is used to scan the retina of the target eye to obtain a retinal optical coherence tomography image. Retinal blood flow images are generated using optical coherence tomography angiography (OCTA). Figure 3 shows a retinal blood flow image; the black dots in region ① represent blood flow signals in the image. Figure 4 shows a retinal optical coherence tomography image.

[0063] S202: Input the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a retinal lesion detection result of the target eye.

[0064] In an embodiment of the present application, the pre-trained retinal lesion detection model can detect at least 22 types of lesions, including epiretinal membrane, intraretinal hyperreflective material, retinal atrophy, retinal pigment epithelial disorder, retinal hole, drusen, serous retinal pigment epithelial detachment, non-serous retinal pigment epithelial detachment, subretinal fluid, intraretinal fluid, choroidal mass, choroidal pit, retinoschisis, preretinal hemorrhage, retinal middle cell edema, retinal inner cell edema, intraretinal hemorrhage, subretinal hyperreflective material, posterior vitreous detachment, vitreomacular traction, focal ellipsoidal band loss and outer retinal tube formation, etc. Among them, the retinal lesion detection model can be a deep neural network model, such as YOLOv7 model, YOLOv5 model, DETR model, YOLOX model, YOLOR model, EfficientDet model, Faster-RCNN model, YOLOv6 model, YOLOv8 model and PPYOLOE model.

[0065] The retinal lesion detection results may include information such as lesion type, lesion location, lesion area and lesion diameter.

[0066] It should be noted that the retinal lesion detection model in the related art can detect about 10 types of lesions. Therefore, the retinal lesion detection model introduced in the embodiment of the present application solves the problem that the related art can only detect a few types of lesions.

[0067] In the fundus image processing method provided in the embodiments of the present application, a retinal optical coherence tomography image and a corresponding retinal blood flow image of the target eye are obtained, and then the retinal optical coherence tomography image and the corresponding retinal blood flow image are input into a retinal lesion detection model to obtain a retinal lesion detection result for the target eye. By introducing the retinal blood flow image and combining it with the retinal optical coherence tomography image to determine the retinal lesion detection result for the target eye, the retinal lesion detection result obtained in this manner is more accurate, thereby improving the accuracy of fundus lesion detection.

[0068] By acquiring retinal optical coherence tomography images and corresponding retinal blood flow images from different directions, the two sets of data from different directions can be used to complement each other, preventing the omission of lesions and improving the accuracy of the detection results. Based on this, in one embodiment, an optional method for acquiring retinal optical coherence tomography images and retinal blood flow images is provided. As shown in Figure 5, the following steps may be included:

[0069] S301, acquiring a three-dimensional retinal structure image and a three-dimensional retinal blood flow image of a target eye.

[0070] S302, acquiring a plurality of first-direction retinal optical coherence tomography images and a plurality of second-direction retinal optical coherence tomography images based on the three-dimensional retinal structure image; and determining each first-direction retinal optical coherence tomography image and each second-direction retinal optical coherence tomography image as a retinal optical coherence tomography image.

[0071] S303: Acquire a plurality of first-direction retinal blood flow images and a plurality of second-direction retinal blood flow images according to the three-dimensional retinal blood flow image; and determine each first-direction retinal blood flow image and each second-direction retinal blood flow image as a retinal blood flow image.

[0072] In an embodiment of the present application, the first-direction retinal optical coherence tomography image can be a transverse retinal optical coherence tomography image, and the second-direction retinal optical coherence tomography image can be a longitudinal retinal optical coherence tomography image; similarly, the first-direction retinal blood flow image can be a transverse retinal blood flow image, and the second-direction retinal blood flow image can be a longitudinal retinal blood flow image.

[0073] Optionally, for three-dimensional retinal structure images, sampling can be performed from two different directions, namely the horizontal and vertical directions, to obtain multiple first-direction retinal optical coherence tomography images and multiple second-direction retinal optical coherence tomography images, and then each first-direction retinal optical coherence tomography image and each second-direction retinal optical coherence tomography image can be determined as a retinal optical coherence tomography image.

[0074] Similarly, for three-dimensional retinal blood flow images, sampling can be performed from two different directions, horizontal and vertical, to obtain multiple first-direction retinal blood flow images and multiple second-direction retinal blood flow images, and then each first-direction retinal blood flow image and each second-direction retinal blood flow image can be determined as a retinal blood flow image.

[0075] The above describes the process of obtaining a 3D retinal structure image and a 3D retinal blood flow image of a target eye through 3D scanning, and then obtaining a retinal optical coherence tomography image and a corresponding retinal blood flow image based on the 3D retinal structure image and the 3D retinal blood flow image. In this embodiment, a retinal optical coherence tomography image and a corresponding retinal blood flow image can also be obtained by directly performing a 2D scan on the target object.

[0076] Optionally, a plurality of first-direction retinal optical coherence tomography images and a plurality of second-direction retinal optical coherence tomography images are acquired; each first-direction retinal optical coherence tomography image and each second-direction retinal optical coherence tomography image are determined as retinal optical coherence tomography images; a plurality of first-direction retinal blood flow images and a plurality of second-direction retinal blood flow images are acquired; each first-direction retinal blood flow image and each second-direction retinal blood flow image are determined as retinal blood flow images.

[0077] For the target eye of the target object, image acquisition is performed at different positions of the target eye in the first direction and the second direction, so that a plurality of first-direction retinal optical coherence tomography images and corresponding plurality of first-direction retinal blood flow images, as well as a plurality of second-direction retinal optical coherence tomography images and corresponding plurality of second-direction retinal blood flow images can be obtained.

[0078] It should be noted that when inputting the retinal optical coherence tomography image and the corresponding retinal blood flow image into the retinal lesion detection model to obtain the retinal lesion detection result of the target eye, a first-direction retinal optical coherence tomography image and a corresponding first-direction retinal blood flow image can be used as a group, and each first-direction retinal optical coherence tomography image and each first-direction retinal blood flow image can be input into the retinal lesion detection model in batches. Alternatively, a second-direction retinal optical coherence tomography image and a corresponding second-direction retinal blood flow image can be used as a group, and each second-direction retinal optical coherence tomography image and each second-direction retinal blood flow image can be input into the retinal lesion detection model in batches. It should be noted that a group of first-direction retinal optical coherence tomography images and first-direction retinal blood flow images correspond to the same imaging position of the target eye. In the embodiment of the present application, by using image data in two directions to obtain the lesion detection result, the detection results obtained from data in different directions can provide data support to each other, so as to obtain the retinal lesion detection result faster and more accurately.

[0079] Alternatively, multiple retinal optical coherence tomography images in the first direction and corresponding multiple retinal blood flow images in the first direction, as well as multiple retinal optical coherence tomography images in the second direction and corresponding multiple retinal blood flow images in the second direction, may be simultaneously input into the retinal lesion detection model to obtain retinal lesion detection results for the target eye. In the embodiment of the present application, by simultaneously inputting all retinal optical coherence tomography images and corresponding retinal blood flow images in different directions into the retinal detection model to obtain retinal lesion detection results, more reference quantities are obtained during the detection process, thereby obtaining more accurate retinal lesion detection results.

[0080] In addition, in the follow-up scenario of the target object, the retinal optical coherence tomography image and the corresponding retinal blood flow image are input into the retinal lesion detection model to obtain the retinal lesion detection result of the target eye. The method can be to input a first-direction retinal optical coherence tomography image and a corresponding first-direction retinal blood flow image into the retinal lesion detection model to obtain the retinal lesion detection result of the target eye, or to input a second-direction retinal optical coherence tomography image and a corresponding second-direction retinal blood flow image into the retinal lesion detection model to obtain the retinal lesion detection result of the target eye. In an embodiment of the present application, in the follow-up scenario, the target object is directly scanned in two dimensions to obtain a first-direction retinal optical coherence tomography image and a corresponding first-direction retinal blood flow image, or directly obtain a second-direction retinal optical coherence tomography image and a corresponding second-direction retinal blood flow image, and input them into the retinal lesion detection model to quickly obtain the retinal lesion detection result of the target eye, providing a method suitable for quickly obtaining lesion detection results in a follow-up scenario.

[0081] In an embodiment of the present application, a method is provided for sampling three-dimensional retinal structure images and three-dimensional retinal blood flow images from different directions, which can obtain multiple sets of retinal optical coherence tomography images and retinal blood flow images, providing data support for subsequent determination of retinal lesion detection results.

[0082] When detecting fundus lesions, one can usually choose to train a network model and obtain the lesion detection type in the input image based on the trained model. To avoid false positive results, an optional method for determining retinal lesion detection results is provided in one embodiment. As shown in Figure 6, the following steps may be included:

[0083] S401: Inputting a retinal optical coherence tomography image and a corresponding retinal blood flow image into a retinal lesion detection model to obtain a first lesion detection result and a first confidence level of the first lesion detection result.

[0084] S402: Determine a retinal lesion detection result of the target eye according to the first lesion detection result and a first confidence level of the first lesion detection result.

[0085] Optionally, the retinal optical coherence tomography image and the corresponding retinal blood flow image are input into a pre-trained retinal lesion detection model, and the retinal lesion detection model outputs a first lesion detection result and a first confidence level of the first lesion detection result, and then the retinal lesion detection result of the target eye is determined based on the first lesion detection result and the first confidence level of the first lesion detection result.

[0086] In an embodiment of the present application, the retinal lesion detection result of the target eye can be determined based on the first lesion detection result and the first confidence level of the first lesion detection result, thereby providing an optional method for quickly determining the retinal lesion detection result.

[0087] When removing false positive results based on the confidence level output by the retinal lesion detection model, a method of introducing a preset confidence threshold can generally be used to filter out false positive detection results. Based on this, in one embodiment, another optional method for retinal lesion detection results is provided. As shown in Figure 7, the following steps may be included:

[0088] S501 : Determine a second lesion detection result according to the first lesion detection result, a first confidence level, and a first preset confidence threshold.

[0089] S502, obtaining retinal layering results of retinal optical coherence tomography images.

[0090] S503: Determine the retinal lesion detection result according to the retinal stratification result and the second lesion detection result.

[0091] In the embodiment of the present application, different lesion types are set with different first preset confidence thresholds, which can usually be determined based on the application scenario (physical examination / diagnosis) and the degree of impact of the disease type on the patient's health.

[0092] Optionally, for any first lesion detection result, the first confidence level of the lesion detection result is compared with a corresponding first preset confidence threshold. If the first confidence level of the lesion detection result is greater than the corresponding first preset confidence threshold, the first lesion detection result is considered the second lesion detection result. Furthermore, the retinal optical coherence tomography image can be input into a pre-trained retinal layering model to obtain a retinal layering result of the retinal optical coherence tomography image. The retinal lesion detection result can then be determined based on the retinal layering result and the second lesion detection result.

[0093] Among them, the retinal stratification results may include the inner limiting membrane, nerve fiber layer, ganglion cell layer, inner plexiform layer, inner nuclear layer, outer plexiform layer, outer nuclear layer, retinal pigment epithelium, external limiting membrane, choroid layer, etc.

[0094] In an embodiment of the present application, an optional method for quickly determining the retinal lesion detection results of the target eye is provided.

[0095] In order to ensure the accuracy and reliability of the fundus lesion detection results, the second lesion detection results can be optimized again according to the retinal layering results. Based on this, in one embodiment, an optional method for determining the retinal lesion detection results is provided. As shown in Figure 8, the following steps may be included:

[0096] S601, determining a third lesion detection result according to the retinal stratification result and the second lesion detection result.

[0097] S602: Cluster the third lesion detection result to obtain a retinal lesion detection result.

[0098] In one implementation, the lesion type and lesion location information in the second lesion detection result may be obtained; and the third lesion detection result may be determined based on the lesion type, lesion location information, and retinal stratification result.

[0099] Optionally, the detected lesion type and lesion location information can be obtained from the second lesion detection result, and then the lesion type, lesion location information and retinal stratification results can be input into a pre-trained model, and the model outputs a third lesion detection result.

[0100] In the embodiment of the present application, the accuracy and reliability of the fundus lesion detection results are further improved by further optimizing the second lesion detection results according to the retinal stratification results.

[0101] Because some lesions occur at specific locations in the retinal layering results, false positive test results can be excluded based on this consideration. Based on this, in one embodiment, an optional method for determining the third lesion detection result is provided. As shown in Figure 9, the following steps may be included:

[0102] S701, matching the lesion location information with the retinal layering result to obtain a matching result.

[0103] S702: Determine a third lesion detection result based on the matching result and the lesion type.

[0104] In the embodiment of the present application, the matching result indicates that the lesion is located at a corresponding position in the retinal layering result.

[0105] Optionally, the lesion location information is matched with the retinal stratification results to determine which layer or layers of the retinal stratification results the lesion location information is located in, and then, based on the matching results and the lesion type, it is determined whether the location of the lesion corresponding to the lesion type in the retinal optical coherence tomography image is consistent with the location where the lesion type should actually occur. If so, the second lesion detection result is determined to be the third lesion detection result. For example, posterior vitreous detachment occurs at the junction of the vitreous and the internal limiting membrane. If the posterior vitreous detachment lesion is detected in the second lesion detection result, the lesion location information is first matched with the retinal stratification results to determine which layer of the lesion is located in the retinal stratification results, and then determine whether the layer is at the junction of the vitreous and the internal limiting membrane; if so, the second lesion detection result is determined to be the third lesion detection result; if not, the second lesion detection result is determined to be a false positive, and the lesion detection result is removed.

[0106] In an embodiment of the present application, an optional method for determining the third lesion detection result is provided.

[0107] When clustering the third lesion detection results to obtain the retinal lesion detection results of the target eye, in order to further improve the accuracy of the fundus lesion detection results, another filtering can be performed after clustering. Based on this, in one embodiment, another optional method for determining the retinal lesion detection results is provided. As shown in Figure 10, the following steps may be included:

[0108] S801 , clustering the third lesion detection result to obtain a clustered lesion detection result and a second confidence level of the clustered lesion detection result.

[0109] S802 : Determine a retinal lesion detection result according to the clustered lesion detection result, the second confidence level, and the second preset confidence threshold.

[0110] In the embodiment of the present application, the second confidence level is determined based on the first confidence level of the first lesion detection result included in the clustered lesion detection results; for example, the second confidence level can be obtained by taking the average, median, and maximum values ​​of the first confidence levels of the included first lesion detection results. Similarly, different second preset confidence thresholds are set for different lesion types.

[0111] Optionally, the third lesion detection result is input into the spatial clustering algorithm for cluster analysis to obtain the cluster lesion detection result and the second confidence of the cluster lesion detection result. Then, for any cluster lesion detection result, the second confidence of the cluster lesion detection result is compared with the corresponding second preset confidence threshold. If the second confidence of the cluster lesion detection result is greater than the corresponding second preset confidence threshold, the cluster lesion detection result is determined to be a retinal lesion detection result.

[0112] It should be noted that the second confidence level needs to be set to be greater than the corresponding first confidence level, which can not only prevent the omission of lesions, but also make the final retinal lesion detection result relatively accurate.

[0113] In the embodiment of the present application, the second confidence level is introduced to filter the clustered lesion detection results to obtain retinal lesion detection results, thereby further improving the accuracy of the fundus lesion detection results.

[0114] After obtaining the retinal lesion detection results of the target eye, the retinal lesion detection results can be projected onto a fundus projection diagram or fundus image to obtain a fundus lesion distribution map, so as to quickly understand the lesion distribution of the target eye. Based on this, in one embodiment, an optional method for determining the fundus lesion distribution map is provided. As shown in Figure 11, the following steps may be included:

[0115] S901: Acquire a fundus projection image and / or fundus image of a target eye.

[0116] S902 : Projecting the retinal lesion detection result into a fundus projection diagram and / or a fundus image to obtain a fundus lesion distribution diagram of the target eye.

[0117] The fundus image refers to an image obtained by using any one of a fundus camera, a confocal scanning system ophthalmoscope, a confocal laser scanning ophthalmoscope, and a scanning laser ophthalmoscope.

[0118] Optionally, the pixel coordinate values ​​of each pixel point in the retinal lesion detection result projected along the axial direction of the eye, as well as the fundus projection map and fundus image of the target eye, can be obtained. Then, based on the pixel coordinate values ​​of each pixel point projected along the axial direction of the eye, the retinal lesion detection result is projected onto the fundus projection map and / or fundus image to obtain a fundus lesion distribution map of the target eye. As shown in Figure 12, a fundus lesion distribution map obtained based on the fundus projection map is shown; wherein, areas ②, ③, ④, and ⑤ respectively represent retinal lesion detection results corresponding to different types; as shown in Figure 13, a fundus lesion distribution map obtained based on the fundus image is shown; wherein, area ⑥ indicates the area where lesions are present.

[0119] In an embodiment of the present application, an optional method for obtaining a fundus lesion distribution map of a target eye is provided.

[0120] In addition, in one embodiment, the present application also provides an optional example of a fundus image processing method. As shown in FIG14 , the method includes:

[0121] S1001, acquiring a three-dimensional retinal structure image and a three-dimensional retinal blood flow image of a target eye.

[0122] S1002, acquiring a plurality of first-direction retinal optical coherence tomography images and a plurality of second-direction retinal optical coherence tomography images based on the three-dimensional retinal structure image; and determining each first-direction retinal optical coherence tomography image and each second-direction retinal optical coherence tomography image as a retinal optical coherence tomography image.

[0123] S1003, acquiring a plurality of first-direction retinal blood flow images and a plurality of second-direction retinal blood flow images based on the three-dimensional retinal blood flow image; and determining each first-direction retinal blood flow image and each second-direction retinal blood flow image as a retinal blood flow image.

[0124] S1004: Input the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a first lesion detection result and a first confidence level of the first lesion detection result.

[0125] S1005 , determining a second lesion detection result according to the first lesion detection result, the first confidence level, and a first preset confidence threshold.

[0126] S1006, obtaining retinal layering results of retinal optical coherence tomography images.

[0127] S1007: Obtain lesion type and lesion location information in the second lesion detection result.

[0128] S1008, matching the lesion location information with the retinal layering result to obtain a matching result.

[0129] S1009: Determine a third lesion detection result based on the matching result and the lesion type.

[0130] S1010 , clustering the third lesion detection result to obtain a clustered lesion detection result and a second confidence level of the clustered lesion detection result.

[0131] S1011 , determining a retinal lesion detection result based on the clustered lesion detection result, the second confidence level, and the second preset confidence threshold.

[0132] The above processes of S1001-S1011 can be referred to the description of the above method embodiment, and the implementation principles and technical effects are similar, so they will not be repeated here.

[0133] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0134] Based on the same inventive concept, embodiments of the present application also provide a fundus image processing device for implementing the fundus image processing method described above. The solution provided by this device is similar to the solution described in the method described above. Therefore, the specific limitations of one or more fundus image processing device embodiments provided below can be found in the limitations of the fundus image processing method described above and will not be repeated here.

[0135] In one embodiment, as shown in FIG15 , a fundus image processing device 1 is provided, comprising: an image acquisition module 10 and a result determination module 20 , wherein:

[0136] An image acquisition module 10 is used to acquire a retinal optical coherence tomography image and a corresponding retinal blood flow image of a target eye;

[0137] The result determination module 20 is used to input the retinal optical coherence tomography image and the corresponding retinal blood flow image into the retinal lesion detection model to obtain the retinal lesion detection result of the target eye.

[0138] In one embodiment, the image acquisition module 10 may be used to:

[0139] Acquire a three-dimensional retinal structure image and a three-dimensional retinal blood flow image of the target eye; acquire a plurality of first-direction retinal optical coherence tomography images and a plurality of second-direction retinal optical coherence tomography images based on the three-dimensional retinal structure image; determine each first-direction retinal optical coherence tomography image and each second-direction retinal optical coherence tomography image as a retinal optical coherence tomography image; acquire a plurality of first-direction retinal blood flow images and a plurality of second-direction retinal blood flow images based on the three-dimensional retinal blood flow image; determine each first-direction retinal blood flow image and each second-direction retinal blood flow image as a retinal blood flow image.

[0140] In one embodiment, the result determination module 20 may be used to:

[0141] The retinal optical coherence tomography image and the corresponding retinal blood flow image are input into the retinal lesion detection model to obtain a first lesion detection result and a first confidence level of the first lesion detection result; based on the first lesion detection result and the first confidence level of the first lesion detection result, the retinal lesion detection result of the target eye is determined.

[0142] In one embodiment, the result determination module 20 is further configured to:

[0143] Determine a second lesion detection result based on the first lesion detection result, the first confidence level and the first preset confidence threshold; obtain a retinal stratification result of the retinal optical coherence tomography image; and determine a retinal lesion detection result based on the retinal stratification result and the second lesion detection result.

[0144] In one embodiment, the result determination module 20 is further configured to:

[0145] The third lesion detection result is determined according to the retinal stratification result and the second lesion detection result; the third lesion detection result is clustered to obtain the retinal lesion detection result.

[0146] In one embodiment, the result determination module 20 is further configured to:

[0147] Obtain the lesion type and lesion location information in the second lesion detection result; and determine the third lesion detection result based on the lesion type, lesion location information, and retinal stratification results.

[0148] In one embodiment, the result determination module 20 is further configured to:

[0149] The lesion location information is matched with the retinal stratification result to obtain a matching result; the matching result indicates that the lesion is located at a corresponding position in the retinal stratification result; and a third lesion detection result is determined based on the matching result and the lesion type.

[0150] In one embodiment, the result determination module 20 is further configured to:

[0151] The third lesion detection result is clustered to obtain a clustered lesion detection result and a second confidence level of the clustered lesion detection result; the second confidence level is determined based on the first confidence level of the first lesion detection result included in the clustered lesion detection result; and the retinal lesion detection result is determined based on the clustered lesion detection result, the second confidence level, and a second preset confidence threshold.

[0152] In one embodiment, the fundus image processing device 1 is further configured to:

[0153] Obtain a fundus projection map and / or fundus image of the target eye; project the retinal lesion detection results onto the fundus projection map and / or fundus image to obtain a fundus lesion distribution map of the target eye.

[0154] Each module in the fundus image processing device described above can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0155] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in FIG16 . The computer device includes a processor, a memory, and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store fundus image processing data. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a fundus image processing method is implemented.

[0156] Those skilled in the art will understand that the structure shown in Figure 16 is merely a block diagram of a portion of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0157] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0158] Acquiring a retinal optical coherence tomography image and a corresponding retinal blood flow image of the target eye;

[0159] The retinal optical coherence tomography image and the corresponding retinal blood flow image are input into the retinal lesion detection model to obtain the retinal lesion detection results of the target eye.

[0160] In one embodiment, when a processor executes logic in a computer program for acquiring a retinal optical coherence tomography image and a retinal blood flow image of a target eye, the following steps may be implemented:

[0161] Acquire a three-dimensional retinal structure image and a three-dimensional retinal blood flow image of the target eye; acquire a plurality of first-direction retinal optical coherence tomography images and a plurality of second-direction retinal optical coherence tomography images based on the three-dimensional retinal structure image; determine each first-direction retinal optical coherence tomography image and each second-direction retinal optical coherence tomography image as a retinal optical coherence tomography image; acquire a plurality of first-direction retinal blood flow images and a plurality of second-direction retinal blood flow images based on the three-dimensional retinal blood flow image; determine each first-direction retinal blood flow image and each second-direction retinal blood flow image as a retinal blood flow image.

[0162] In one embodiment, when a processor executes the logic of a computer program for inputting a retinal optical coherence tomography image and a corresponding retinal blood flow image into a retinal lesion detection model to obtain a retinal lesion detection result of a target eye, the following steps may be implemented:

[0163] The retinal optical coherence tomography image and the corresponding retinal blood flow image are input into the retinal lesion detection model to obtain a first lesion detection result and a first confidence level of the first lesion detection result; based on the first lesion detection result and the first confidence level of the first lesion detection result, the retinal lesion detection result of the target eye is determined.

[0164] In one embodiment, when a processor executes logic in a computer program for determining a retinal lesion detection result of a target eye based on a first lesion detection result and a first confidence level of the first lesion detection result, the following steps may be implemented:

[0165] Determine a second lesion detection result based on the first lesion detection result, the first confidence level and the first preset confidence threshold; obtain a retinal stratification result of the retinal optical coherence tomography image; and determine a retinal lesion detection result based on the retinal stratification result and the second lesion detection result.

[0166] In one embodiment, when a processor executes logic in a computer program to determine a retinal lesion detection result based on a retinal layering result and a second lesion detection result, the following steps may be implemented:

[0167] The third lesion detection result is determined according to the retinal stratification result and the second lesion detection result; the third lesion detection result is clustered to obtain the retinal lesion detection result.

[0168] In one embodiment, when a processor executes logic in a computer program to determine a third lesion detection result based on a retinal layering result and a second lesion detection result, the following steps may be implemented:

[0169] Obtain the lesion type and lesion location information in the second lesion detection result; and determine the third lesion detection result based on the lesion type, lesion location information, and retinal stratification results.

[0170] In one embodiment, when the processor executes the logic in the computer program for determining the third lesion detection result based on the lesion type, lesion location information, and retinal layering results, the following steps may be implemented:

[0171] The lesion location information is matched with the retinal stratification result to obtain a matching result; the matching result indicates that the lesion is located at a corresponding position in the retinal stratification result; and a third lesion detection result is determined based on the matching result and the lesion type.

[0172] In one embodiment, when the processor executes the logic of clustering the third lesion detection result to obtain the retinal lesion detection result in the computer program, the following steps may be implemented:

[0173] The third lesion detection result is clustered to obtain a clustered lesion detection result and a second confidence level of the clustered lesion detection result; the second confidence level is determined based on the first confidence level of the first lesion detection result included in the clustered lesion detection result; and the retinal lesion detection result is determined based on the clustered lesion detection result, the second confidence level, and a second preset confidence threshold.

[0174] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0175] Obtain a fundus projection map and / or fundus image of the target eye; project the retinal lesion detection results onto the fundus projection map and / or fundus image to obtain a fundus lesion distribution map of the target eye.

[0176] The principles and processes of the computer device provided above in implementing each embodiment can be found in the description of the fundus image processing method in the aforementioned embodiment, and will not be repeated here.

[0177] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0178] Obtain a retinal optical coherence tomography image and a corresponding retinal blood flow image of the target eye; input the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a retinal lesion detection result of the target eye.

[0179] In one embodiment, when the logic of acquiring the retinal optical coherence tomography image and the retinal blood flow image of the target eye in the computer program is executed by a processor, the following steps may be implemented:

[0180] Acquire a three-dimensional retinal structure image and a three-dimensional retinal blood flow image of the target eye; acquire a plurality of first-direction retinal optical coherence tomography images and a plurality of second-direction retinal optical coherence tomography images based on the three-dimensional retinal structure image; determine each first-direction retinal optical coherence tomography image and each second-direction retinal optical coherence tomography image as a retinal optical coherence tomography image; acquire a plurality of first-direction retinal blood flow images and a plurality of second-direction retinal blood flow images based on the three-dimensional retinal blood flow image; determine each first-direction retinal blood flow image and each second-direction retinal blood flow image as a retinal blood flow image.

[0181] In one embodiment, when the logic of a computer program for inputting a retinal optical coherence tomography image and a corresponding retinal blood flow image into a retinal lesion detection model and obtaining a retinal lesion detection result for a target eye is executed by a processor, the following steps may be implemented:

[0182] The retinal optical coherence tomography image and the corresponding retinal blood flow image are input into the retinal lesion detection model to obtain a first lesion detection result and a first confidence level of the first lesion detection result; based on the first lesion detection result and the first confidence level of the first lesion detection result, the retinal lesion detection result of the target eye is determined.

[0183] In one embodiment, when the logic in the computer program for determining the retinal lesion detection result of the target eye based on the first lesion detection result and the first confidence level of the first lesion detection result is executed by a processor, the following steps may be implemented:

[0184] Determine a second lesion detection result based on the first lesion detection result, the first confidence level and the first preset confidence threshold; obtain a retinal stratification result of the retinal optical coherence tomography image; and determine a retinal lesion detection result based on the retinal stratification result and the second lesion detection result.

[0185] In one embodiment, when the logic of determining the retinal lesion detection result based on the retinal layering result and the second lesion detection result in the computer program is executed by a processor, the following steps may be implemented:

[0186] The third lesion detection result is determined according to the retinal stratification result and the second lesion detection result; the third lesion detection result is clustered to obtain the retinal lesion detection result.

[0187] In one embodiment, when the logic of determining the third lesion detection result based on the retinal layering result and the second lesion detection result in the computer program is executed by a processor, the following steps may be implemented:

[0188] Obtain the lesion type and lesion location information in the second lesion detection result; and determine the third lesion detection result based on the lesion type, lesion location information, and retinal stratification results.

[0189] In one embodiment, when the logic in the computer program for determining the third lesion detection result based on the lesion type, lesion location information, and retinal layering results is executed by a processor, the following steps may be implemented:

[0190] The lesion location information is matched with the retinal stratification result to obtain a matching result; the matching result indicates that the lesion is located at a corresponding position in the retinal stratification result; and a third lesion detection result is determined based on the matching result and the lesion type.

[0191] In one embodiment, when the logic of clustering the third lesion detection result to obtain the retinal lesion detection result in the computer program is executed by the processor, the following steps may be implemented:

[0192] The third lesion detection result is clustered to obtain a clustered lesion detection result and a second confidence level of the clustered lesion detection result; the second confidence level is determined based on the first confidence level of the first lesion detection result included in the clustered lesion detection result; and the retinal lesion detection result is determined based on the clustered lesion detection result, the second confidence level, and a second preset confidence threshold.

[0193] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0194] Obtain a fundus projection map and / or fundus image of the target eye; project the retinal lesion detection results onto the fundus projection map and / or fundus image to obtain a fundus lesion distribution map of the target eye.

[0195] The principles and processes of the computer-readable storage medium provided above in implementing each embodiment can be found in the description of the fundus image processing method in the aforementioned embodiment, and will not be repeated here.

[0196] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0197] Obtain a retinal optical coherence tomography image and a corresponding retinal blood flow image of the target eye; input the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a retinal lesion detection result of the target eye.

[0198] In one embodiment, when the logic of acquiring the retinal optical coherence tomography image and the retinal blood flow image of the target eye in the computer program is executed by a processor, the following steps may be implemented:

[0199] Acquire a three-dimensional retinal structure image and a three-dimensional retinal blood flow image of the target eye; acquire a plurality of first-direction retinal optical coherence tomography images and a plurality of second-direction retinal optical coherence tomography images based on the three-dimensional retinal structure image; determine each first-direction retinal optical coherence tomography image and each second-direction retinal optical coherence tomography image as a retinal optical coherence tomography image; acquire a plurality of first-direction retinal blood flow images and a plurality of second-direction retinal blood flow images based on the three-dimensional retinal blood flow image; determine each first-direction retinal blood flow image and each second-direction retinal blood flow image as a retinal blood flow image.

[0200] In one embodiment, when the logic of a computer program for inputting a retinal optical coherence tomography image and a corresponding retinal blood flow image into a retinal lesion detection model and obtaining a retinal lesion detection result for a target eye is executed by a processor, the following steps may be implemented:

[0201] The retinal optical coherence tomography image and the corresponding retinal blood flow image are input into the retinal lesion detection model to obtain a first lesion detection result and a first confidence level of the first lesion detection result; based on the first lesion detection result and the first confidence level of the first lesion detection result, the retinal lesion detection result of the target eye is determined.

[0202] In one embodiment, when the logic in the computer program for determining the retinal lesion detection result of the target eye based on the first lesion detection result and the first confidence level of the first lesion detection result is executed by a processor, the following steps may be implemented:

[0203] Determine a second lesion detection result based on the first lesion detection result, the first confidence level and the first preset confidence threshold; obtain a retinal stratification result of the retinal optical coherence tomography image; and determine a retinal lesion detection result based on the retinal stratification result and the second lesion detection result.

[0204] In one embodiment, when the logic of determining the retinal lesion detection result based on the retinal layering result and the second lesion detection result in the computer program is executed by a processor, the following steps may be implemented:

[0205] The third lesion detection result is determined according to the retinal stratification result and the second lesion detection result; the third lesion detection result is clustered to obtain the retinal lesion detection result.

[0206] In one embodiment, when the logic of determining the third lesion detection result based on the retinal layering result and the second lesion detection result in the computer program is executed by a processor, the following steps may be implemented:

[0207] Obtain the lesion type and lesion location information in the second lesion detection result; and determine the third lesion detection result based on the lesion type, lesion location information, and retinal stratification results.

[0208] In one embodiment, when the logic in the computer program for determining the third lesion detection result based on the lesion type, lesion location information, and retinal layering results is executed by a processor, the following steps may be implemented:

[0209] The lesion location information is matched with the retinal stratification result to obtain a matching result; the matching result indicates that the lesion is located at a corresponding position in the retinal stratification result; and a third lesion detection result is determined based on the matching result and the lesion type.

[0210] In one embodiment, when the logic of clustering the third lesion detection result to obtain the retinal lesion detection result in the computer program is executed by the processor, the following steps may be implemented:

[0211] The third lesion detection result is clustered to obtain a clustered lesion detection result and a second confidence level of the clustered lesion detection result; the second confidence level is determined based on the first confidence level of the first lesion detection result included in the clustered lesion detection result; and the retinal lesion detection result is determined based on the clustered lesion detection result, the second confidence level, and a second preset confidence threshold.

[0212] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0213] Obtain a fundus projection map and / or fundus image of the target eye; project the retinal lesion detection results onto the fundus projection map and / or fundus image to obtain a fundus lesion distribution map of the target eye.

[0214] The principles and processes of the computer program products provided above in implementing the various embodiments can be found in the description of the fundus image processing method in the aforementioned embodiments, and will not be repeated here.

[0215] It should be noted that the data involved in this application (including but not limited to data used for analysis, stored data, displayed data, etc.) are all information and data that are authorized or fully authorized by all parties.

[0216] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0217] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0218] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A fundus image processing method, characterized in that: The method comprises: Acquire retinal optical coherence tomography images and corresponding retinal blood flow images of the target eye; The retinal optical coherence tomography image and the corresponding retinal blood flow image are input into a retinal lesion detection model to obtain the retinal lesion detection result of the target eye.

2. The method according to claim 1, characterized in that The step of obtaining a retinal optical coherence tomography image and a corresponding retinal blood flow image of a target eye includes: Acquire a plurality of first-direction retinal optical coherence tomography images and a plurality of second-direction retinal optical coherence tomography images; determine each of the first-direction retinal optical coherence tomography images and each of the second-direction retinal optical coherence tomography images as the retinal optical coherence tomography images; Acquire a plurality of first-direction retinal blood flow images and a plurality of second-direction retinal blood flow images; and determine each of the first-direction retinal blood flow images and each of the second-direction retinal blood flow images as the retinal blood flow images.

3. The method according to claim 2, characterized in that The method of acquiring a plurality of retinal optical coherence tomography images in the first direction and a plurality of retinal optical coherence tomography images in the second direction, and acquiring a plurality of retinal blood flow images in the first direction and a plurality of retinal blood flow images in the second direction, comprises: Acquire a three-dimensional retinal structure image of the target eye and a three-dimensional retinal blood flow image of the target eye; According to the three-dimensional retinal structure image, the three-dimensional retinal structure image is sampled horizontally to obtain retinal coherence tomography images in the first direction; the three-dimensional retinal structure image is sampled vertically to obtain retinal optical coherence tomography images in the second direction; According to the three-dimensional retinal blood flow image, the three-dimensional retinal blood flow image is sampled horizontally to obtain each retinal blood flow image in the first direction; and the three-dimensional retinal blood flow image is sampled longitudinally to obtain each retinal blood flow image in the second direction.

4. The method according to claim 2, characterized in that: The step of inputting the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a retinal lesion detection result of the target eye includes: A first-direction retinal optical coherence tomography image and a corresponding first-direction retinal blood flow image are taken as a group, and each of the first-direction retinal optical coherence tomography images and each of the first-direction retinal blood flow images are input into the retinal lesion detection model in batches to obtain the retinal lesion detection result of the target eye.

5. The method according to claim 2, characterized in that: The step of inputting the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a retinal lesion detection result of the target eye includes: A second-direction retinal optical coherence tomography image and a corresponding second-direction retinal blood flow image are taken as a group, and each of the second-direction retinal optical coherence tomography images and each of the second-direction retinal blood flow images are input into the retinal lesion detection model in batches to obtain the retinal lesion detection result of the target eye.

6. The method according to claim 2, characterized in that The step of inputting the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a retinal lesion detection result of the target eye includes: The first-direction retinal optical coherence tomography images and the corresponding first-direction retinal blood flow images, as well as the second-direction retinal optical coherence tomography images and the corresponding second-direction retinal blood flow images are simultaneously input into the retinal lesion detection model to obtain the retinal lesion detection result of the target eye.

7. The method according to claim 1, characterized in that The step of inputting the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a retinal lesion detection result of the target eye includes: A first-direction retinal optical coherence tomography image and a corresponding first-direction retinal blood flow image are input into the retinal lesion detection model to obtain a retinal lesion detection result of the target eye.

8. The method according to claim 7, characterized in that The step of inputting the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a retinal lesion detection result of the target eye includes: A second-direction retinal optical coherence tomography image and a corresponding second-direction retinal blood flow image are input into the retinal lesion detection model to obtain a retinal lesion detection result of the target eye.

9. The method according to claim 1, characterized in that: The step of inputting the retinal optical coherence tomography image and the corresponding retinal blood flow image into a retinal lesion detection model to obtain a retinal lesion detection result of the target eye includes: Inputting the retinal optical coherence tomography image and the corresponding retinal blood flow image into the retinal lesion detection model to obtain a first lesion detection result and a first confidence level of the first lesion detection result; The retinal lesion detection result of the target eye is determined according to the first lesion detection result and a first confidence level of the first lesion detection result.

10. The method according to claim 9, characterized in that The step of determining the retinal lesion detection result of the target eye according to the first lesion detection result and the first confidence level of the first lesion detection result includes: Determining a second lesion detection result according to the first lesion detection result, the first confidence level, and a first preset confidence threshold; Obtaining retinal layering results of the retinal optical coherence tomography image; The retinal lesion detection result is determined according to the retinal stratification result and the second lesion detection result.

11. The method according to claim 10, characterized in that Determining the retinal lesion detection result according to the retinal stratification result and the second lesion detection result includes: Determining a third lesion detection result according to the retinal stratification result and the second lesion detection result; The third lesion detection result is clustered to obtain the retinal lesion detection result.

12. The method according to claim 11, characterized in that Determining a third lesion detection result according to the retinal stratification result and the second lesion detection result includes: Obtaining lesion type and lesion location information in the second lesion detection result; The third lesion detection result is determined according to the lesion type, the lesion location information and the retinal stratification result.

13. The method according to claim 12, characterized in that Determining the third lesion detection result according to the lesion type, the lesion location information and the retinal stratification result includes: Matching the lesion location information with the retinal stratification result to obtain a matching result; the matching result indicates that the lesion is located at a corresponding position in the retinal stratification result; The third lesion detection result is determined according to the matching result and the lesion type.

14. The method according to claim 11, characterized in that The clustering of the third lesion detection result to obtain the retinal lesion detection result includes: Clustering the third lesion detection result to obtain a clustered lesion detection result and a second confidence level of the clustered lesion detection result; the second confidence level is determined according to a first confidence level of a first lesion detection result included in the clustered lesion detection result; The retinal lesion detection result is determined according to the clustered lesion detection result, the second confidence level and a second preset confidence threshold.

15. The method according to any one of claims 1 to 14, characterized in that: The method further comprises: Acquiring a fundus projection map and / or a fundus image of the target eye; The retinal lesion detection result is projected onto the fundus projection map and / or the fundus image to obtain a fundus lesion distribution map of the target eye.

16. The method according to claim 15, characterized in that The step of projecting the retinal lesion detection result onto the fundus projection map and / or the fundus image to obtain a fundus lesion distribution map of the target eye includes: Obtaining the pixel coordinate value of each pixel point in the retinal lesion detection result projected along the eye axis direction, as well as the fundus projection map and fundus image of the target eye; According to the pixel coordinate value of each pixel point projected along the eye axis direction, the retinal lesion detection result is projected into the fundus projection map and / or the fundus image to obtain the fundus lesion distribution map of the target eye.

17. A fundus image processing device, characterized in that: The device comprises: An image acquisition module, used to acquire a retinal optical coherence tomography image and a corresponding retinal blood flow image of a target eye; The result determination module is used to input the retinal optical coherence tomography image and the corresponding retinal blood flow image into the retinal lesion detection model to obtain the retinal lesion detection result of the target eye.

18. The device according to claim 17, characterized in that The image acquisition module is also used for: Acquire a plurality of first-direction retinal optical coherence tomography images and a plurality of second-direction retinal optical coherence tomography images; determine each of the first-direction retinal optical coherence tomography images and each of the second-direction retinal optical coherence tomography images as the retinal optical coherence tomography images; Acquire a plurality of first-direction retinal blood flow images and a plurality of second-direction retinal blood flow images; and determine each of the first-direction retinal blood flow images and each of the second-direction retinal blood flow images as the retinal blood flow images.

19. The device according to claim 17, characterized in that The result determination module is also used for: A first-direction retinal optical coherence tomography image and a corresponding first-direction retinal blood flow image are input into the retinal lesion detection model to obtain a retinal lesion detection result of the target eye.

20. The device according to claim 17, characterized in that The result determination module is also used for: A second-direction retinal optical coherence tomography image and a corresponding second-direction retinal blood flow image are input into the retinal lesion detection model to obtain a retinal lesion detection result of the target eye.

Images