Image recognition method and device, electronic equipment and readable storage medium

Abstract

The invention discloses an image recognition method and device, electronic equipment and a readable storage medium, and particularly relates to the technical field of image processing. The image recognition method comprises the following steps: based on a single-channel image corresponding to a fundus image to be recognized, determining at least one limitation atrophy focus candidate region in the single-channel image; based on the to-be-recognized fundus image, the at least one limitation atrophy focus candidate area in the single-channel image and a preset gray condition, a recognition result of the to-be-recognized fundus image is determined, and the recognition result comprises whether the to-be-recognized fundus image contains the limitation atrophy focus area or not. According to the image recognition method provided by the embodiment of the invention, through the to-be-recognized eye fundus image, the at least one limitation atrophy focus candidate area in the single-channel image and the preset gray condition, the result of whether the to-be-recognized eye fundus image contains the limitation atrophy focus area can be obtained, so that a doctor is assisted to determine whether the limitation atrophy focus exists or not; the problems of missed diagnosis, misdiagnosis and the like caused by experience of doctors and influence on patients are solved.

Description

Technical Field

[0001] This disclosure belongs to the field of image processing technology, and specifically relates to an image recognition method and apparatus, electronic device and readable storage medium. Background Technology

[0002] Localized fundus atrophy is a fundus lesion that affects vision. As the disease progresses, especially when the atrophic lesions involve the macula, central vision deteriorates significantly. Furthermore, localized fundus atrophy may increase the risk of complications, further impairing vision and severely impacting a patient's quality of life, including the ability to read, drive, and perform daily activities. Therefore, timely diagnosis and treatment are crucial for patients with localized fundus atrophy to slow disease progression and protect vision.

[0003] However, the traditional identification and diagnosis of localized fundus atrophy lesions usually relies on the doctor's experience. The doctor determines whether a patient has localized atrophy lesions based on experience and fundus images. However, different doctors have different levels of experience, and doctors in some regions have limited experience and cannot accurately identify and diagnose localized atrophy lesions, which prevents patients from receiving timely diagnosis and delays in treatment.

[0004] Therefore, there is an urgent need for an image recognition method to solve the problem of patients not receiving timely diagnosis due to the reliance on doctors' experience to identify limited atrophic lesions. Summary of the Invention

[0005] In view of this, this disclosure provides an image recognition method to assist doctors in determining whether there are localized atrophic lesions, thereby solving the problem that patients cannot receive timely diagnosis due to reliance on doctors' experience to identify localized atrophic lesions.

[0006] In a first aspect, an embodiment of this disclosure provides an image recognition method, comprising: determining at least one candidate region of localized atrophic lesions in a single-channel image corresponding to a fundus image to be recognized; and determining a recognition result of the fundus image to be recognized based on the fundus image to be recognized, at least one candidate region of localized atrophic lesions in the single-channel image, and a preset grayscale condition, wherein the recognition result includes whether the fundus image to be recognized contains a localized atrophic lesion region.

[0007] In conjunction with the first aspect, in certain implementations of the first aspect, determining the recognition result of the fundus image to be identified based on at least one candidate region of a localized atrophic lesion in the fundus image to be identified, a single-channel image, and preset grayscale conditions includes: determining at least one candidate region of a localized atrophic lesion in the fundus image to be identified based on at least one candidate region of a localized atrophic lesion in the fundus image to be identified and a single-channel image; determining, based on at least one candidate region of a localized atrophic lesion, a candidate region of a first channel, a candidate region of a second channel, and a candidate region of a third channel corresponding to each of the at least one candidate region of a localized atrophic lesion; and determining the recognition result of the fundus image to be identified based on the candidate regions of the first channel, the candidate regions of the second channel, and the candidate regions of the third channel corresponding to each of the at least one candidate region of a localized atrophic lesion and preset grayscale conditions.

[0008] In conjunction with the first aspect, in certain implementations of the first aspect, before determining the recognition result of the fundus image to be recognized based on the candidate region images of the first channel, the second channel, and the third channel, and a preset grayscale condition, the recognition method further includes: determining the fundus image to be recognized in the first channel, the second channel, and the third channel, respectively, based on the fundus image to be recognized; determining the grayscale data of each of the fundus images to be recognized in the first channel, the second channel, and the third channel, respectively; determining multiple first candidate grayscale values ​​based on the grayscale data of the fundus image to be recognized in the first channel and a first preset range condition; determining the evaluation grayscale value of the first channel based on the largest grayscale value among the multiple first candidate grayscale values; determining multiple second candidate grayscale values ​​based on the grayscale data of the fundus image to be recognized in the second channel and a second preset range condition; determining the evaluation grayscale value of the second channel based on the largest grayscale value among the multiple second candidate grayscale values; and determining the evaluation grayscale value of the second channel based on the grayscale data of the fundus image to be recognized in the third channel and the first preset range condition. Three preset range conditions are used to determine multiple third candidate grayscale values; based on the largest grayscale value among the multiple third candidate grayscale values, the evaluation grayscale value of the third channel is determined; based on the evaluation grayscale values ​​of the first channel, the second channel, and the third channel, preset grayscale conditions are determined; based on the candidate regions of the first channel, the second channel, and the third channel corresponding to at least one localized atrophic lesion candidate region, and the preset grayscale conditions, the recognition result of the fundus image to be identified is determined, including: for each localized atrophic lesion candidate region, if the grayscale values ​​of the pixels in the candidate region of the first channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluation grayscale value of the first channel, the grayscale values ​​of the pixels in the candidate region of the second channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluation grayscale value of the second channel, and the grayscale values ​​of all pixels in the candidate region of the third channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluation grayscale value of the third channel, the localized atrophic lesion candidate region is determined to be a localized atrophic lesion region.

[0009] In conjunction with the first aspect, in some implementations of the first aspect, determining at least one candidate region for a localized atrophic lesion in a single-channel image based on the fundus image to be identified includes: segmenting the single-channel image corresponding to the fundus image to be identified using a preset grayscale threshold to determine at least one first initial candidate region for a localized atrophic lesion; determining at least one second initial candidate region for a localized atrophic lesion based on the at least one first initial candidate region for a localized atrophic lesion and a preset roundness threshold; and determining at least one candidate region for a localized atrophic lesion in the single-channel image based on the at least one second initial candidate region for a localized atrophic lesion.

[0010] In conjunction with the first aspect, in some implementations of the first aspect, determining at least one candidate region for localized atrophic lesions in a single-channel image based on at least one second initial localized atrophic lesion candidate region includes: performing gradient calculations on the boundaries of each of the at least one second initial localized atrophic lesion candidate regions to determine the gradient data corresponding to each of the at least one second initial localized atrophic lesion candidate regions; determining at least one third initial localized atrophic lesion candidate region based on the gradient data corresponding to each of the at least one second initial localized atrophic lesion candidate regions and a preset gradient threshold; and determining at least one third localized atrophic lesion candidate region as a localized atrophic lesion candidate region when the at least one third initial localized atrophic lesion candidate region contains a vascular region.

[0011] In conjunction with the first aspect, in some implementations of the first aspect, before segmenting the single-channel image corresponding to the fundus image to be identified using a preset grayscale threshold to determine at least one first initial limited atrophic lesion candidate region, the image recognition method further includes: determining the optic disc-retinal-choroidal atrophy arc region of the fundus image to be identified; determining the pericoronal retinal-choroidal atrophy arc region of the fundus image to be identified; determining the grayscale values ​​of the optic disc-retinal-choroidal atrophy arc region and the pericoronal retinal-choroidal atrophy arc region; and determining a preset grayscale threshold based on the grayscale values ​​of the optic disc-retinal-choroidal atrophy arc region and the pericoronal retinal-choroidal atrophy arc region.

[0012] In conjunction with the first aspect, in some implementations of the first aspect, before determining at least one candidate region of a limited atrophic lesion in a single-channel image based on the single-channel image corresponding to the fundus image to be identified, the method further includes: performing enhancement processing on the region of interest of the initial fundus image to be identified to determine the enhanced image corresponding to the initial fundus image to be identified; performing channel separation processing on the enhanced image to determine the single-channel image corresponding to the enhanced image; and determining the single-channel image corresponding to the fundus image to be identified based on the single-channel image corresponding to the enhanced image.

[0013] Secondly, an image recognition device provided in one embodiment of this disclosure includes: a determining module, configured to determine at least one candidate region of localized atrophic lesions in a single-channel image based on a single-channel image corresponding to a fundus image to be recognized; and a recognizing module, configured to determine a recognition result of the fundus image to be recognized based on the fundus image to be recognized, at least one candidate region of localized atrophic lesions in the single-channel image, and a preset grayscale condition, wherein the recognition result includes whether the fundus image to be recognized contains a localized atrophic lesion region.

[0014] Thirdly, an embodiment of this disclosure provides an electronic device comprising: a processor; and a memory for storing processor-executable instructions, wherein the processor is configured to perform the method mentioned in the first aspect above.

[0015] Fourthly, one embodiment of this disclosure provides a computer-readable storage medium storing a computer program for performing the methods mentioned in the first aspect above.

[0016] The image recognition method provided in this disclosure determines the recognition result of the fundus image to be identified by using at least one candidate region of localized atrophic lesions in the fundus image to be identified, a single-channel image, and preset grayscale conditions. It can obtain the result of whether the fundus image to be identified contains a localized atrophic lesion region, thereby assisting doctors in judging whether localized atrophic lesions exist. This solves the problem of patients not receiving timely diagnosis due to reliance on doctors' experience to identify localized atrophic lesions. Attached Figure Description

[0017] The above and other objects, features, and advantages of this disclosure will become more apparent from the more detailed description of the embodiments thereof in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this disclosure and form part of the specification. They are used together with the embodiments of this disclosure to explain the disclosure and do not constitute a limitation thereof.

[0018] Figure 1 The diagram shown is an application scenario illustration provided by an embodiment of this disclosure.

[0019] Figure 2 The diagram shown is a flowchart of an image recognition method provided in an embodiment of this disclosure.

[0020] Figure 3 The diagram shows a flowchart of an embodiment of the present disclosure, which describes how to determine the recognition result of a fundus image based on a candidate region of a limited atrophic lesion in a single-channel image and preset grayscale conditions.

[0021] Figure 4 The diagram shown is a flowchart of an image recognition method provided in another embodiment of this disclosure.

[0022] Figure 5 The diagram shown is a flowchart illustrating a process for determining at least one candidate region of a localized atrophic lesion in a single-channel image based on a single-channel image corresponding to a fundus image to be identified, according to an embodiment of the present disclosure.

[0023] Figure 6 The diagram shown is a flowchart illustrating a process for determining at least one candidate region of a limited atrophic lesion in a single-channel image based on at least one second initial candidate region of limited atrophic lesion, according to an embodiment of the present disclosure.

[0024] Figure 7 The diagram shown is a flowchart of an image recognition method provided in another embodiment of this disclosure.

[0025] Figure 8 The diagram shown is a flowchart of an image recognition method provided in another embodiment of this disclosure.

[0026] Figure 9 The diagram shown is a structural schematic of an image recognition device provided in an embodiment of this disclosure.

[0027] Figure 10 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0028] The technical solutions of the present disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present disclosure, and not all embodiments.

[0029] Localized fundus atrophy is a fundus lesion that affects vision. As the disease progresses, especially when the atrophy involves the macula, central vision deteriorates significantly. Furthermore, localized fundus atrophy may increase the risk of complications such as choroidal neovascularization and macular edema, which can further impair vision. Decreased vision and visual field defects severely impact a patient's quality of life, including their ability to read, drive, and perform daily activities. Therefore, timely diagnosis and treatment are crucial for patients with localized fundus atrophy to slow disease progression and protect vision.

[0030] However, the traditional identification and diagnosis of localized fundus atrophy lesions usually relies on the doctor's experience. The doctor determines whether a patient has localized atrophy lesions based on experience and fundus images. However, different doctors have different levels of experience, and in some areas with limited medical resources, such as those with limited medical resources, doctors may not be able to accurately identify and diagnose localized atrophy lesions. This can lead to patients not receiving timely diagnosis, delaying treatment, causing the disease to progress, and affecting the patient's life.

[0031] Therefore, there is an urgent need for an image recognition method to solve the problem of patients not receiving timely diagnosis due to the reliance on doctors' experience to identify limited atrophic lesions.

[0032] The following is combined Figure 1 A brief introduction will be given to an application scenario of one embodiment of this disclosure.

[0033] Figure 1 The diagram illustrates an application scenario of one embodiment of this disclosure. Figure 1 As shown, this scenario is an image evaluation scenario. Specifically, the image evaluation scenario includes a server 110 and a user terminal 120 communicatively connected to the server 110. The server 110 is used to execute the methods mentioned in the embodiments of this disclosure.

[0034] For example, in practical applications, a user issues an instruction to evaluate an image through a user terminal 120. Upon receiving the instruction, the server 110 determines at least one candidate region for localized atrophic lesions in the single-channel image corresponding to the fundus image to be identified. Based on the fundus image to be identified, the at least one candidate region for localized atrophic lesions in the single-channel image, and preset grayscale conditions, the server 110 determines the identification result of the fundus image to be identified, including whether the fundus image to be identified contains a localized atrophic lesion region. The server 110 sends the identification result of the fundus image to be identified to the user terminal 120 so that the user can view the identification result of the fundus image to be identified through the user terminal 120 and check whether the fundus image to be identified contains a localized atrophic lesion region.

[0035] For example, the aforementioned fundus images to be evaluated include, but are not limited to, all image data stored by the medical institution for the same patient during treatment, relevant fundus image data of the patient input by the user, and fundus image data taken by the patient as needed during their visit. The fundus images to be evaluated can be 45° fundus images, 60° fundus images, wide-angle fundus images, or fundus images of other fields of view, or even fundus images of other modalities. Fundus images can be taken with the optic disc as the center, or with the macula as the center, or images of other eye positions. For example, the server 110 can directly receive the fundus images to be evaluated, or it can obtain them from a data storage device.

[0036] For example, the users mentioned above may be doctors, researchers of related diseases, or other personnel who want to understand the location of changes in fundus structures. This disclosure does not further limit the specific types of users.

[0037] For example, the user terminal 120 mentioned above includes, but is not limited to, computer terminals such as desktop computers and laptops, and mobile terminals such as tablet computers and mobile phones.

[0038] The following is combined Figures 2 to 8 A brief introduction to the image recognition method provided in this disclosure is given.

[0039] Figure 2 The diagram shown is a flowchart illustrating an image recognition method provided in an embodiment of this disclosure. Figure 2 As shown, an embodiment of the image recognition method provided by this disclosure includes the following steps.

[0040] Step S210: Based on the single-channel image corresponding to the fundus image to be identified, determine at least one candidate region of localized atrophic lesion in the single-channel image.

[0041] For example, the fundus image to be identified is a color image composed of three color channels: red (R), green (G), and blue (B), each channel containing intensity information of the corresponding color in the image. The fundus image to be identified is processed to obtain a corresponding single-channel image (e.g., the blue (B) channel), and at least one candidate region of localized atrophic lesion is identified in the single-channel image.

[0042] For example, based on a single-channel image corresponding to the fundus image to be identified, a trained deep learning model is used to determine candidate regions of localized atrophic lesions in the single-channel image. Alternatively, a preset threshold is used to segment the single-channel image corresponding to the fundus image to be identified, thereby determining candidate regions of localized atrophic lesions in the single-channel image.

[0043] In some embodiments, the specific execution method of step S210 is as follows: Figure 5 As shown, it will not be elaborated further here.

[0044] Step S220: Based on the fundus image to be identified, at least one candidate region of a limited atrophic lesion in the single-channel image, and the preset grayscale conditions, determine the identification result of the fundus image to be identified.

[0045] The identification results include whether the fundus image to be identified contains a localized atrophic lesion area.

[0046] For example, based on the fundus image to be identified corresponding to the candidate region of localized atrophic lesions, and / or at least one candidate region of localized atrophic lesions in a single-channel image obtained after the above processing, a preset grayscale condition is used. For example, based on the single-channel images under the R, G, and B channels, a preset grayscale value is compared. When the grayscale value of at least one image under the single-channel images under the R, G, and B channels is greater than or less than the preset grayscale value, the identification result of the fundus image to be identified can be determined, that is, whether the fundus image to be identified contains a localized atrophic lesion region can be determined. It should be understood that the preset grayscale condition can be selected according to actual needs, and the embodiments of this application do not further limit the preset grayscale condition.

[0047] In some embodiments, step S220 is specifically executed as follows: Figure 3 As shown, it will not be elaborated further here.

[0048] This embodiment of the present disclosure determines the recognition result of the fundus image to be identified based on at least one candidate region of localized atrophic lesions in the fundus image to be identified, a single-channel image, and preset grayscale conditions. It can obtain the result of whether the fundus image to be identified contains a localized atrophic lesion region, thereby assisting doctors in judging whether localized atrophic lesions exist. This solves the problem of patients not receiving timely diagnosis due to reliance on doctors' experience to identify localized atrophic lesions.

[0049] Figure 3The diagram shown is a flowchart illustrating an image recognition method according to another embodiment of this disclosure. Figure 3 As shown, the recognition result of the fundus image to be identified is determined based on at least one candidate region of a limited atrophic lesion in the fundus image to be identified, a single-channel image, and preset grayscale conditions, including the following steps.

[0050] Step S310: Based on at least one candidate region of a localized atrophic lesion in the fundus image to be identified and the single-channel image, determine at least one candidate region of a localized atrophic lesion in the fundus image to be identified.

[0051] For example, based on at least one candidate region of a localized atrophic lesion in the fundus image to be identified and a single-channel image, the fundus image to be identified corresponding to the candidate region of the localized atrophic lesion is determined, and then at least one candidate region of a localized atrophic lesion in the fundus image to be identified is determined.

[0052] Step S320: Based on at least one localized atrophic lesion candidate region, determine the candidate region of the first channel, the candidate region of the second channel, and the candidate region of the third channel corresponding to each of the at least one localized atrophic lesion candidate regions.

[0053] For example, the first channel, the second channel, and the third channel are R, G, and B channels, respectively. That is, based on at least one candidate region of a localized atrophic lesion, at least one candidate region of a localized atrophic lesion in the fundus image to be identified is determined in the R, G, and B channels, respectively. In other words, based on at least one candidate region of a localized atrophic lesion, the candidate region of the first channel, the candidate region of the second channel, and the candidate region of the third channel corresponding to each of the at least one candidate region of a localized atrophic lesion are determined, respectively.

[0054] Step S330: Based on the candidate regions of the first channel, the second channel, and the third channel corresponding to each of the candidate regions of at least one limited atrophic lesion, and the preset grayscale conditions, determine the recognition result of the fundus image to be identified.

[0055] For example, based on the candidate regions of the first channel, the candidate regions of the second channel, and the candidate regions of the third channel corresponding to each of the at least one candidate region of a localized atrophic lesion, the grayscale condition is selected as the grayscale value of the candidate region image being greater than a preset value. If the candidate regions of the first channel, the candidate regions of the second channel, and the candidate regions of the third channel corresponding to each of the at least one candidate region of a localized atrophic lesion satisfy the grayscale condition, then the candidate regions of the first channel, the candidate regions of the second channel, and the candidate regions of the third channel corresponding to each of the at least one candidate region of a localized atrophic lesion can be determined as localized atrophic lesions. That is, the fundus image to be identified contains a localized atrophic lesion region, thereby determining the recognition result of the fundus image to be identified.

[0056] This embodiment of the disclosure determines the recognition result of the fundus image to be identified based on the candidate regions of the first, second, and third channels corresponding to at least one candidate region of a localized atrophic lesion, as well as preset grayscale conditions. By using preset grayscale conditions to determine the recognition result of the fundus image to be identified, it is possible to quickly and accurately determine whether the fundus image to be identified contains a localized atrophic lesion region. Furthermore, the preset grayscale conditions can be set according to requirements, adapting to more application scenarios.

[0057] Figure 4 The diagram shown is a flowchart illustrating an image recognition method provided in another embodiment of this disclosure. Figure 3 Extending from the illustrated embodiment Figure 4 The illustrated embodiment will be described in detail below. Figure 4 The illustrated embodiments and Figure 3 The differences between the embodiments shown are not repeated here, and the similarities are not repeated here.

[0058] like Figure 4 As shown, another embodiment of this disclosure provides another image recognition method that, before determining the recognition result of the fundus image to be recognized based on the candidate region image of the first channel, the candidate region image of the second channel, the candidate region image of the third channel, and preset grayscale conditions, further includes the following steps.

[0059] Step S410: Based on the fundus image to be identified, determine the fundus image to be identified in the first channel, the fundus image to be identified in the second channel, and the fundus image to be identified in the third channel.

[0060] For example, the RGB channels of the fundus image to be identified are separated. Information from the red component is extracted to generate a grayscale image, thus obtaining the first channel of the fundus image to be identified. Information from the green component is extracted to generate another grayscale image, thus obtaining the second channel of the fundus image to be identified. Information from the blue component is extracted to generate another grayscale image, thus obtaining the third channel of the fundus image to be identified.

[0061] Step S420: Determine the grayscale data of the fundus image to be identified in the first channel, the fundus image to be identified in the second channel, and the fundus image to be identified in the third channel, respectively.

[0062] For example, the brightness information of each pixel in the fundus image to be identified in the first channel, the brightness information of each pixel in the fundus image to be identified in the second channel, and the brightness information of each pixel in the fundus image to be identified in the third channel are determined respectively, that is, the brightness of the pixels in the image is represented by numerical values.

[0063] Step S430: Based on the grayscale data of the fundus image to be identified in the first channel and the first preset range condition, determine a plurality of first candidate grayscale values.

[0064] For example, a grayscale histogram is generated based on the grayscale data of the fundus image to be identified in the first channel. The first preset range condition is selected as the top 10% of grayscale values, thereby determining multiple first candidate grayscale values, that is, selecting the top 10% of grayscale values.

[0065] Step S440: Determine the evaluation gray value of the first channel based on the largest gray value among multiple first candidate gray values.

[0066] For example, based on the top 10% of grayscale values, the largest grayscale value is determined, and thus the maximum value is determined as the evaluation grayscale value of the first channel.

[0067] Step S450: Based on the grayscale data of the fundus image to be identified in the second channel and the second preset range condition, determine a plurality of second candidate grayscale values.

[0068] For example, a grayscale histogram is generated based on the grayscale data of the fundus image to be identified in the second channel. The top 10% of grayscale values ​​with the largest proportion are selected as the second preset range condition, thereby determining multiple second candidate grayscale values, that is, selecting the top 10% of grayscale values ​​with the largest proportion.

[0069] Step S460: Determine the evaluation gray value of the second channel based on the largest gray value among multiple second candidate gray values.

[0070] For example, the largest gray value among the top 10% of gray values ​​is determined as the evaluation gray value of the second channel.

[0071] Step S470: Based on the grayscale data of the fundus image to be identified in the third channel and the third preset range condition, determine multiple third candidate grayscale values.

[0072] For example, a grayscale histogram is generated based on the grayscale data of the fundus image to be identified in the third channel. The top 10% of grayscale values ​​with the largest proportion are selected as the third preset range condition, thereby determining multiple third candidate grayscale values, that is, selecting the top 10% of grayscale values ​​with the largest proportion.

[0073] Step S480: Determine the evaluation gray value of the third channel based on the largest gray value among multiple third candidate gray values.

[0074] For example, the largest gray value among the top 10% of gray values ​​is determined as the evaluation gray value of the third channel.

[0075] Step S490: Determine preset grayscale conditions based on the evaluation grayscale values ​​of the first channel, the second channel, and the third channel.

[0076] For example, based on the evaluation grayscale values ​​of the first channel, the second channel, and the third channel, a preset grayscale condition is determined to be that all grayscale values ​​are greater than the evaluation grayscale values ​​of the first channel, the second channel, and the third channel. It should be understood that the preset grayscale condition can also be other conditions related to the evaluation grayscale values ​​of the first channel, the second channel, and the third channel, such as being greater than or equal to the evaluation grayscale values ​​of the first channel, the second channel, and the third channel.

[0077] In the embodiments provided in this disclosure, the recognition result of the fundus image to be identified is determined based on the candidate regions of the first channel, the candidate regions of the second channel, and the candidate regions of the third channel corresponding to each of the at least one localized atrophic lesion candidate regions and preset grayscale conditions. This includes: for each localized atrophic lesion candidate region among the at least one localized atrophic lesion candidate regions, if the grayscale values ​​of the pixels in the candidate regions of the first channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluation grayscale value of the first channel, the grayscale values ​​of the pixels in the candidate regions of the second channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluation grayscale value of the second channel, and the grayscale values ​​of all pixels in the candidate regions of the third channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluation grayscale value of the third channel, then the localized atrophic lesion candidate region is determined to be a localized atrophic lesion region.

[0078] For example, based on the preset condition that all gray values ​​are greater than the evaluation gray values ​​of the first channel, the second channel, and the third channel, the recognition result of the fundus image to be identified is determined based on the candidate regions of the first channel, the second channel, and the third channel corresponding to at least one localized atrophic lesion candidate region and the preset gray value condition. Specifically, if the gray values ​​of the pixels in the candidate regions of the first channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluation gray value of the first channel, the gray values ​​of the pixels in the candidate regions of the second channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluation gray value of the second channel, and the gray values ​​of all pixels in the candidate regions of the third channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluation gray value of the third channel, then the localized atrophic lesion candidate region is determined to be a localized atrophic lesion region.

[0079] The image recognition method provided in this disclosure determines preset grayscale conditions using fundus images from a first channel, a second channel, and a third channel. This makes the grayscale data more accurate, and the determined preset grayscale conditions better meet the usage requirements, thus obtaining more precise results.

[0080] Figure 5 The diagram illustrates a process for determining at least one candidate region for a localized atrophic lesion in a single-channel image based on an embodiment of the present disclosure. Figure 5 As shown, determining at least one candidate region for a localized atrophic lesion in a single-channel image based on the fundus image to be identified includes the following steps.

[0081] Step S510: Using a preset grayscale threshold, the single-channel image corresponding to the fundus image to be identified is segmented to determine at least one first initial limited atrophic lesion candidate region.

[0082] For example, the preset grayscale threshold can be selected as needed. For instance, in the fundus image to be identified, the grayscale value A corresponding to the peripapillary retinal choroidal atrophy arc region in channel B corresponds to a grayscale value A in channel B. The preset grayscale threshold is selected to be between grayscale value A and 20. Thus, using the preset grayscale threshold, the single-channel image corresponding to the fundus image to be identified is segmented to determine at least one first initial limited atrophic lesion candidate region. It should be understood that the preset grayscale threshold can be selected as needed.

[0083] Step S520: Based on at least one first initial limited atrophic lesion candidate region and a preset roundness threshold, determine at least one second initial limited atrophic lesion candidate region.

[0084] For example, the preset roundness threshold can be selected as 0.3. Based on at least one first initial limited atrophic lesion candidate region, a region greater than the preset roundness threshold of 0.3 is determined, thereby determining at least one second initial limited atrophic lesion candidate region.

[0085] Step S530: Based on at least one second initial limited atrophic lesion candidate region, determine at least one limited atrophic lesion candidate region in the single-channel image.

[0086] For example, based on at least one second initial limited atrophic lesion candidate region, and utilizing the boundary of the second initial limited atrophic lesion candidate region, at least one limited atrophic lesion candidate region in a single-channel image is determined using a gradient image and corresponding preset gradient conditions. Alternatively, based on at least one second initial limited atrophic lesion candidate region, at least one limited atrophic lesion candidate region in a single-channel image is determined using a trained model.

[0087] In some embodiments, the specific implementation steps of step S530 are as follows: Figure 6 As shown, it will not be elaborated further here.

[0088] This embodiment utilizes a preset grayscale threshold to determine at least one first initial limited atrophic lesion candidate region, and based on the at least one first initial limited atrophic lesion candidate region and a preset roundness threshold, determines at least one second initial limited atrophic lesion candidate region. Then, based on the at least one second initial limited atrophic lesion candidate region, at least one limited atrophic lesion candidate region in a single-channel image is determined. By using the preset grayscale threshold and the preset roundness threshold, at least one limited atrophic lesion candidate region in a single-channel image can be determined more accurately, thereby improving the accuracy and robustness of the image recognition method.

[0089] Figure 6 The diagram illustrates a flowchart of an embodiment of this disclosure, illustrating the process of determining at least one candidate region for a localized atrophic lesion in a single-channel image based on at least one second initial candidate region for localized atrophic lesions. Figure 6 As shown, determining at least one candidate region for a localized atrophic lesion in a single-channel image based on at least one second initial localized atrophic lesion candidate region includes the following steps.

[0090] Step S610: Perform gradient calculation on the boundaries of at least one second initial limited atrophic lesion candidate region to determine the gradient data corresponding to each of the at least one second initial limited atrophic lesion candidate regions.

[0091] For example, gradient operators (such as Sobel, Prewitt, or Scharr operators) are used to compute the gradient of the image. The gradient includes gradient values ​​in the horizontal direction (x-direction) and the vertical direction (y-direction). Gradients are computed for the boundaries of at least one second initial limited atrophic foci candidate region, and the gradient magnitude of each pixel is calculated by combining the horizontal and vertical gradients, thereby determining the gradient data corresponding to each of the at least one second initial limited atrophic foci candidate region.

[0092] Step S620: Based on the gradient data corresponding to each of the at least one second initial limited atrophic lesion candidate regions and the preset gradient threshold, determine at least one third initial limited atrophic lesion candidate region.

[0093] For example, the gradient threshold is selected to be greater than 100. Based on the gradient data corresponding to each of the at least one second initial limited atrophic lesion candidate regions, the region corresponding to the data greater than 100 in the gradient data corresponding to each of the at least one second initial limited atrophic lesion candidate regions is determined, thereby determining at least one third initial limited atrophic lesion candidate region.

[0094] Step S630: Determine whether at least one third initial localized atrophic lesion candidate region contains a vascular region.

[0095] For example, step S640 is executed when at least one third initial localized atrophic lesion candidate region contains a vascular region. When at least one third initial localized atrophic lesion candidate region contains a vascular region, step S610 may be repeated until the end.

[0096] Step S640: Determine at least one third localized atrophic lesion candidate region as a localized atrophic lesion candidate region.

[0097] For example, when at least one third initial localized atrophic lesion candidate region contains a vascular region, at least one third localized atrophic lesion candidate region is determined as a localized atrophic lesion candidate region.

[0098] The embodiments of this disclosure determine a third initial localized atrophic lesion candidate region by using a second initial localized atrophic lesion candidate region and a preset gradient threshold. When at least one third initial localized atrophic lesion candidate region contains a vascular region, at least one third localized atrophic lesion candidate region is determined as a localized atrophic lesion candidate region. This can further improve the accuracy of the results of determining localized atrophic lesion candidate regions, thereby improving the accuracy of image recognition results.

[0099] Figure 7 The diagram shown is a flowchart illustrating an image recognition method provided in another embodiment of this disclosure. Figure 5 Extending from the illustrated embodiment Figure 7 The illustrated embodiment will be described in detail below. Figure 5 The illustrated embodiments and Figure 7 The differences between the embodiments shown are not repeated here, and the similarities are not repeated here.

[0100] like Figure 7 As shown, another embodiment of this disclosure provides another image recognition method that, before segmenting the single-channel image corresponding to the fundus image to be identified using a preset grayscale threshold to determine at least one first initial limited atrophic lesion candidate region, further includes the following steps.

[0101] Step S710: Determine the optic disc-retinal-choroid atrophy arc region in the fundus image to be identified.

[0102] For example, the fundus image to be identified undergoes some preprocessing, such as grayscale conversion, contrast enhancement, and noise reduction. Taking advantage of the optic disc's typically circular shape and high brightness, template matching, Hough circle detection, or morphological methods are used to initially locate the optic disc region. Threshold segmentation or region growing algorithms are then used to accurately extract the optic disc's boundaries. Edge detection (such as the Canny operator) or texture analysis (such as the Gabor filter) is used to extract potential regions, and threshold segmentation is performed to determine the optic disc-retinal-choroidal atrophy arc region. It should be understood that a trained model can be used to determine the optic disc-retinal-choroidal atrophy arc region, and this disclosure does not further limit the specific method for determining the optic disc-retinal-choroidal atrophy arc region.

[0103] Step S720: Determine the pericoronary choroidal atrophy arc region in the fundus image to be identified.

[0104] For example, after image preprocessing of the fundus image to be identified and locating the optic disc region, the peripapillary region is defined, for example, a ring-shaped area within a certain range centered on the optic disc, such as 1 to 2 times the diameter of the optic disc. Threshold segmentation is then used to determine the peripapillary retinal-choroidal atrophy arc region of the fundus image to be identified.

[0105] Step S730: Determine the gray values ​​of the optic disc retinal choroidal atrophy arc region and the peripapillary retinal choroidal atrophy arc region.

[0106] For example, the grayscale values ​​of all pixels are extracted from the optic disc atrophy arc region and the peripapillary retinal atrophy arc region obtained above. Statistical characteristics of the grayscale values ​​in each region, such as mean, median, and standard deviation, are calculated to determine the grayscale values ​​of the optic disc-retinal-choroidal atrophy arc region and the peripapillary retinal-choroidal atrophy arc region.

[0107] Step S740: Determine a preset grayscale threshold based on the grayscale values ​​of the optic disc retinal choroidal atrophy arc region and the peripapillary retinal choroidal atrophy arc region.

[0108] For example, based on the gray values ​​of the optic disc-retinal choroidal atrophy region and the pericoronal retinal choroidal atrophy region, a preset gray value threshold is determined as required, either as the gray value of the optic disc-retinal choroidal atrophy region and the pericoronal retinal choroidal atrophy region, or as a range of gray values ​​including the gray values ​​of the optic disc-retinal choroidal atrophy region and the pericoronal retinal choroidal atrophy region.

[0109] This embodiment of the present disclosure determines a preset grayscale threshold based on the grayscale values ​​of the optic disc-retinal choroidal atrophy arc region and the peripapillary retinal choroidal atrophy arc region. It fully considers the relationship between the grayscale values ​​of the optic disc-retinal choroidal atrophy arc region and the peripapillary retinal choroidal atrophy arc region and the grayscale values ​​of the candidate region image, thereby improving the accuracy of the results in determining the candidate region of the initial localized atrophic lesion, and further improving the accuracy of image evaluation.

[0110] Figure 8 The diagram shown is a flowchart illustrating an image recognition method provided in another embodiment of this disclosure. Figure 2 Extending from the illustrated embodiment Figure 8 The illustrated embodiment will be described in detail below. Figure 8 The illustrated embodiments and Figure 2 The differences between the embodiments shown are not repeated here, and the similarities are not repeated here.

[0111] like Figure 8 As shown, another embodiment of this disclosure provides another image recognition method, which further includes the following steps before determining at least one candidate region of a localized atrophic lesion in a single-channel image based on a single-channel image corresponding to the fundus image to be identified.

[0112] Step S810: Enhance the region of interest in the initial fundus image to be identified and determine the enhanced image corresponding to the initial fundus image to be identified.

[0113] For example, deep learning models (such as SRCNN and ESRGAN) can be used to process the initial fundus image to be identified. Alternatively, morphological processing can be employed, such as dilation, erosion, opening, and closing operations. Furthermore, the region of interest (ROI) of the initial fundus image can be enhanced through grayscale transformation, histogram processing, image smoothing, and image sharpening to determine the enhanced image corresponding to the initial fundus image. It should be understood that in practical applications, other methods can be used to enhance the ROI of the initial fundus image, as long as the enhanced image corresponding to the initial fundus image meets the image recognition requirements.

[0114] Step S820: Perform channel separation processing on the enhanced image to determine the single-channel image corresponding to the enhanced image.

[0115] For example, channel separation processing is performed on the enhanced image, and the enhanced images under different channels (the enhanced images may include HSV images or RGB images) are used to determine the single-channel images corresponding to the enhanced images.

[0116] Step S830: Based on the single-channel image corresponding to the enhanced image, determine the single-channel image corresponding to the fundus image to be identified.

[0117] For example, the single-channel image corresponding to the enhanced image is determined as the single-channel image corresponding to the fundus image to be identified.

[0118] The embodiments of this disclosure determine the single-channel image corresponding to the fundus image to be identified based on the enhanced image, which can highlight important information in the image, reduce noise in the image, improve image quality, and adapt to different needs, thereby improving the quality and accuracy of the fundus image to be identified.

[0119] Figure 9 The diagram shown is a structural schematic of an image recognition device provided in an embodiment of this disclosure. Figure 9 As shown, an embodiment of the present disclosure provides an image recognition device, including: a determining module 910 and a recognizing module 920; the determining module 910 is used to determine at least one candidate region of localized atrophic lesions in a single-channel image based on a single-channel image corresponding to a fundus image to be recognized; the recognizing module 920 is used to determine the recognition result of the fundus image to be recognized based on the fundus image to be recognized, at least one candidate region of localized atrophic lesions in the single-channel image, and a preset grayscale condition, wherein the recognition result includes whether the fundus image to be recognized contains a localized atrophic lesion region.

[0120] In some embodiments, the recognition module 920 is further configured to: determine at least one candidate region of a localized atrophic lesion in the fundus image to be recognized based on at least one candidate region of a localized atrophic lesion in the fundus image to be recognized and the single-channel image; determine, based on the at least one candidate region of a localized atrophic lesion, a candidate region of a first channel, a candidate region of a second channel, and a candidate region of a third channel corresponding to each of the at least one candidate region of a localized atrophic lesion; and determine the recognition result of the fundus image to be recognized based on the candidate regions of the first channel, the candidate regions of the second channel, and the candidate regions of the third channel corresponding to each of the at least one candidate region of a localized atrophic lesion and a preset grayscale condition.

[0121] In some embodiments, the recognition module 920 is further configured to: determine, based on the fundus image to be recognized, a fundus image to be recognized in a first channel, a fundus image to be recognized in a second channel, and a fundus image to be recognized in a third channel; determine the grayscale data of each of the fundus images to be recognized in the first channel, the second channel, and the third channel; determine a plurality of first candidate grayscale values ​​based on the grayscale data of the fundus image to be recognized in the first channel and a first preset range condition; and determine the evaluation grayscale of the first channel based on the largest grayscale value among the plurality of first candidate grayscale values. The evaluation grayscale value of the second channel is determined based on the grayscale data of the fundus image to be identified in the second channel and the second preset range condition. Based on the largest grayscale value among the multiple second candidate grayscale values, the evaluation grayscale value of the second channel is determined. Based on the grayscale data of the fundus image to be identified in the third channel and the third preset range condition, multiple third candidate grayscale values ​​are determined. Based on the largest grayscale value among the multiple third candidate grayscale values, the evaluation grayscale value of the third channel is determined. Based on the evaluation grayscale value of the first channel, the evaluation grayscale value of the second channel, and the evaluation grayscale value of the third channel, the preset grayscale condition is determined. The recognition module 920 is further configured to determine the recognition result of the fundus image to be recognized based on the candidate regions of the first channel, the candidate regions of the second channel, and the candidate regions of the third channel corresponding to each of the at least one localized atrophic lesion candidate regions, and preset grayscale conditions. This includes: for each localized atrophic lesion candidate region, if the grayscale values ​​of the pixels in the candidate regions of the first channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluated grayscale value of the first channel, the grayscale values ​​of the pixels in the candidate regions of the second channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluated grayscale value of the second channel, and the grayscale values ​​of all pixels in the candidate regions of the third channel corresponding to the localized atrophic lesion candidate region are all greater than the evaluated grayscale value of the third channel, then the localized atrophic lesion candidate region is determined to be a localized atrophic lesion region. In some embodiments, the determining module 910 is further configured to: segment the single-channel image corresponding to the fundus image to be identified using a preset grayscale threshold to determine at least one first initial localized atrophic lesion candidate region; determine at least one second initial localized atrophic lesion candidate region based on at least one first initial localized atrophic lesion candidate region and a preset roundness threshold; and determine at least one localized atrophic lesion candidate region in the single-channel image based on at least one second initial localized atrophic lesion candidate region.

[0122] In some embodiments, the determining module 910 is further configured to: perform gradient calculation on the boundaries of at least one second initial localized atrophic lesion candidate region respectively, and determine the gradient data corresponding to each of the at least one second initial localized atrophic lesion candidate regions; determine at least one third initial localized atrophic lesion candidate region based on the gradient data corresponding to each of the at least one second initial localized atrophic lesion candidate regions and a preset gradient threshold; and determine at least one third localized atrophic lesion candidate region as a localized atrophic lesion candidate region when at least one third initial localized atrophic lesion candidate region contains a vascular region.

[0123] In some embodiments, the determining module 910 is further configured to: determine the optic disc-retinal-choroidal atrophy arc region of the fundus image to be identified; determine the pericoronal retinal-choroidal atrophy arc region of the fundus image to be identified; determine the gray values ​​of the optic disc-retinal-choroidal atrophy arc region and the pericoronal retinal-choroidal atrophy arc region; and determine a preset gray value threshold based on the gray values ​​of the optic disc-retinal-choroidal atrophy arc region and the pericoronal retinal-choroidal atrophy arc region.

[0124] In some embodiments, the determining module 910 is further configured to: perform enhancement processing on the region of interest of the initial fundus image to be identified, and determine the enhanced image corresponding to the initial fundus image to be identified; perform channel separation processing on the enhanced image, and determine the single-channel image corresponding to the enhanced image; and determine the single-channel image corresponding to the fundus image to be identified based on the single-channel image corresponding to the enhanced image.

[0125] Figure 10 The diagram shown is a schematic representation of the structure of an electronic device provided in an exemplary embodiment of this disclosure. The electronic device 1000 (specifically, it may be a computer device) includes a memory 1001, a processor 1002, a communication interface 1003, and a bus 1004. The memory 1001, processor 1002, and communication interface 1003 are interconnected via the bus 1004.

[0126] The memory 1001 may be a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1001 may store a program, and when the program stored in the memory 1001 is executed by the processor 1002, the processor 1002 and the communication interface 1003 are used to execute the various steps in the image recognition method of the embodiments of this disclosure.

[0127] The processor 1002 may be a general-purpose central processing unit (CPU), microprocessor, application specific integrated circuit (ASIC), graphics processing unit (GPU), or one or more integrated circuits, used to execute relevant programs to achieve the functions required by each unit of the image recognition device of this disclosure embodiment.

[0128] The processor 1002 can also be an integrated circuit chip with signal processing capabilities. In implementation, each step of the image processing method or fundus image comparison method of this disclosure can be completed by the integrated logic circuitry in the hardware of the processor 1002 or by instructions in software form. The processor 1002 described above can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this disclosure can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory 1001. The processor 1002 reads the information in the memory 1001 and, in conjunction with its hardware, performs the functions required by the units included in the image recognition device of this disclosure, or executes the image recognition method of this disclosure.

[0129] The communication interface 1003 uses a transceiver device, such as, but not limited to, a transceiver, to enable communication between the electronic device 1000 and other devices or communication networks. For example, an image of the fundus to be identified can be acquired through the communication interface 1003.

[0130] Bus 1004 may include a pathway for transmitting information between various components of electronic device 1000 (e.g., memory 1001, processor 1002, communication interface 1003).

[0131] It should be noted that, although Figure 10The illustrated electronic device 1000 only shows the memory, processor, and communication interface. However, those skilled in the art should understand that in specific implementations, the electronic device 1000 may also include other devices necessary for normal operation. Furthermore, depending on specific needs, those skilled in the art should understand that the electronic device 1000 may also include hardware devices for implementing other additional functions. Moreover, those skilled in the art should understand that the electronic device 1000 may only include the devices necessary for implementing the embodiments of this disclosure, and may not necessarily include... Figure 10 All the devices shown.

[0132] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0133] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0134] In the embodiments provided in this disclosure, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0135] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0136] In addition, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0137] Embodiments of this disclosure can also be computer-readable storage media storing computer program instructions that, when executed by a processor, cause the processor to perform the steps of the methods described above according to various embodiments of this disclosure. If the functionality is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks. The computer-readable storage medium can be any combination of one or more readable media. A readable medium can be a readable signal medium or a readable storage medium. A readable storage medium can be, for example, including but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof.

[0138] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. An image recognition method, characterized in that, include: Based on the single-channel image corresponding to the fundus image to be identified, at least one candidate region of localized atrophic lesion in the single-channel image is determined; Based on the fundus image to be identified, at least one candidate region in the single-channel image, and a preset grayscale condition, the identification result of the fundus image to be identified is determined, and the identification result includes whether the fundus image to be identified contains a localized atrophic lesion region.

2. The image recognition method according to claim 1, characterized in that, The determination of the recognition result of the fundus image to be identified based on the fundus image to be identified, at least one candidate region of localized atrophic lesions in the single-channel image, and preset grayscale conditions includes: Based on at least one candidate region of a localized atrophic lesion in the fundus image to be identified and the single-channel image, at least one candidate region of a localized atrophic lesion in the fundus image to be identified is determined; Based on the at least one limited atrophic lesion candidate region, the candidate region of the first channel, the candidate region of the second channel, and the candidate region of the third channel corresponding to each of the at least one limited atrophic lesion candidate regions are determined respectively. Based on the candidate regions of the first channel, the second channel, and the third channel corresponding to each of the at least one limited atrophic lesion candidate regions, and the preset grayscale conditions, the recognition result of the fundus image to be identified is determined.

3. The image recognition method according to claim 2, characterized in that, Before determining the recognition result of the fundus image to be identified based on the candidate region image of the first channel, the candidate region image of the second channel, the candidate region image of the third channel, and the preset grayscale conditions, the method further includes: Based on the fundus image to be identified, the fundus image to be identified in the first channel, the fundus image to be identified in the second channel, and the fundus image to be identified in the third channel are determined respectively; The grayscale data of the fundus image to be identified in the first channel, the fundus image to be identified in the second channel, and the fundus image to be identified in the third channel are determined respectively. Based on the grayscale data of the fundus image to be identified in the first channel and the first preset range condition, a number of first candidate grayscale values ​​are determined. The evaluation gray value of the first channel is determined based on the largest gray value among the plurality of first candidate gray values. Based on the grayscale data of the fundus image to be identified in the second channel and the second preset range condition, multiple second candidate grayscale values ​​are determined. The evaluation gray value of the second channel is determined based on the largest gray value among the plurality of second candidate gray values. Based on the grayscale data of the fundus image to be identified in the third channel and the third preset range condition, multiple third candidate grayscale values ​​are determined. The evaluation gray value of the third channel is determined based on the largest gray value among the plurality of third candidate gray values. Based on the evaluation grayscale value of the first channel, the evaluation grayscale value of the second channel, and the evaluation grayscale value of the third channel, the preset grayscale condition is determined; The determination of the recognition result of the fundus image to be identified based on the candidate regions of the first channel, the candidate regions of the second channel, and the candidate regions of the third channel corresponding to each of the at least one limited atrophic lesion candidate regions, and the preset grayscale conditions, includes: For each of the at least one limited atrophic lesion candidate regions, if the gray values ​​of all pixels in the candidate region of the first channel corresponding to the limited atrophic lesion candidate region are greater than the evaluation gray value of the first channel, the gray values ​​of all pixels in the candidate region of the second channel corresponding to the limited atrophic lesion candidate region are greater than the evaluation gray value of the second channel, and the gray values ​​of all pixels in the candidate region of the third channel corresponding to the limited atrophic lesion candidate region are greater than the evaluation gray value of the third channel, then the limited atrophic lesion candidate region is determined to be a limited atrophic lesion region.

4. The image recognition method according to claim 1, characterized in that, The step of determining at least one candidate region for a localized atrophic lesion in a single-channel image corresponding to the fundus image to be identified includes: Using a preset grayscale threshold, the single-channel image corresponding to the fundus image to be identified is segmented to determine at least one first initial limited atrophic lesion candidate region; Based on the at least one first initial limited atrophic lesion candidate region and the preset roundness threshold, at least one second initial limited atrophic lesion candidate region is determined; At least one candidate region for a limited atrophic lesion in the single-channel image is determined based on at least one second initial limited atrophic lesion candidate region.

5. The image recognition method according to claim 4, characterized in that, The step of determining at least one candidate region for a localized atrophic lesion in the single-channel image based on at least one second initial localized atrophic lesion candidate region includes: Gradient calculations are performed on the boundaries of the at least one second initial limited atrophic lesion candidate region to determine the gradient data corresponding to each of the at least one second initial limited atrophic lesion candidate regions. Based on the gradient data and preset gradient threshold corresponding to each of the at least one second initial limited atrophic lesion candidate regions, at least one third initial limited atrophic lesion candidate region is determined. If the at least one third initial localized atrophic lesion candidate region contains a vascular region, the at least one third localized atrophic lesion candidate region is determined as the localized atrophic lesion candidate region.

6. The image recognition method according to claim 4, characterized in that, Before segmenting the single-channel image corresponding to the fundus image to be identified using a preset grayscale threshold to determine at least one first initial localized atrophic lesion candidate region, the method further includes: Determine the optic disc-retinal-choroidal atrophy arc region in the fundus image to be identified; Identify the pericoronary choroidal atrophy arc region in the fundus image to be identified; Determine the grayscale values ​​of the optic disc retinal choroidal atrophy arc region and the peripapillary retinal choroidal atrophy arc region; The preset grayscale threshold is determined based on the grayscale values ​​of the optic disc retinal choroid atrophy arc region and the peripapillary retinal choroid atrophy arc region.

7. The image recognition method according to any one of claims 1 to 6, characterized in that, Before determining at least one candidate region of localized atrophic lesions in the single-channel image corresponding to the fundus image to be identified, the method further includes: The region of interest in the initial fundus image to be identified is enhanced to determine the enhanced image corresponding to the initial fundus image to be identified; The enhanced image is subjected to channel separation processing to determine the single-channel image corresponding to the enhanced image; Based on the single-channel image corresponding to the enhanced image, the single-channel image corresponding to the fundus image to be identified is determined.

8. An image recognition device, characterized in that, include: The determination module is used to determine at least one candidate region of localized atrophic lesions in the single-channel image based on the single-channel image corresponding to the fundus image to be identified; The recognition module is used to determine the recognition result of the fundus image to be recognized based on the fundus image to be recognized, at least one localized atrophic lesion candidate region in the single-channel image, and preset grayscale conditions. The recognition result includes whether the fundus image to be recognized contains a localized atrophic lesion region.

9. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions. The processor is used to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program for performing the method described in any one of claims 1 to 7.

Images