Eye fundus image restoration method and system based on deep learning

Abstract

The invention provides an eye fundus image restoration method and system based on deep learning, and the method comprises the steps: carrying out the abnormal feature extraction of an eye fundus image data set, and obtaining an abnormal feature extraction result of the eye fundus image data set; classifying abnormal feature extraction results of the fundus image data set to obtain one or more fundus lesion image type directories; and obtaining an image debugging result bound with the debugging instruction in the fundus image data set by combining the abnormal feature extraction results contained in the one or more fundus lesion image type directories. And the image debugging result bound with the debugging instruction in the fundus image data set is determined by analyzing each abnormal feature extraction result in each fundus lesion image type catalog step by step, so that the credibility and the precision of the image debugging result are ensured.

Description

Technical Field

[0001] The present application relates to the field of image restoration technology, and more specifically, to a fundus image restoration method and system based on deep learning. Background Art

[0002] As the global population ages, eye diseases, particularly cataracts, have become a major cause of visual impairment. Traditional fundus examination methods, such as ophthalmoscopy, can effectively identify eye lesions, but their accuracy and efficiency are significantly reduced when images are poor or blurry. Fundus images are often affected by factors such as insufficient lighting, blurry images, or noise, making it difficult for doctors to extract sufficient information about the lesion during diagnosis.

[0003] Cataracts are one of the leading causes of vision loss. Their pathological changes often manifest as clouding of the lens, directly impacting fundus image clarity and visualization of lesions. Because early cataract symptoms are subtle, many patients only discover the problem after their condition worsens, missing the optimal time for treatment. Therefore, improving the quality of fundus images and enabling timely and accurate identification of cataracts and other fundus lesions is crucial for enhancing diagnostic efficiency and patient treatment outcomes.

[0004] The goal of this project is to develop an efficient fundus image restoration and lesion recognition system using deep learning techniques, combined with image restoration and automatic recognition algorithms. This system will not only improve the quality of fundus images but also automatically identify cataracts and related lesions, providing ophthalmologists with a more accurate and rapid auxiliary diagnostic tool. Summary of the Invention

[0005] In order to improve the technical problems existing in the related technologies, this application provides a fundus image restoration method and system based on deep learning.

[0006] In a first aspect, a fundus image restoration method based on deep learning is provided, the method comprising: performing abnormal feature extraction processing on a fundus image dataset to obtain abnormal feature extraction results of the fundus image dataset; classifying the abnormal feature extraction results of the fundus image dataset to obtain one or more fundus lesion image type directories; and combining the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories to obtain an image debugging result bound to a debugging instruction in the fundus image dataset.

[0007] In the present application, the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories are combined to obtain the image debugging results bound to the debugging instructions in the fundus image data set, including: analyzing the abnormal feature extraction results contained in the first fundus lesion image type directory in the one or more fundus lesion image type directories to obtain the matching between the one or more debugging instructions and the one or more image debugging results contained in the first fundus lesion image type directory; combining the matching between the one or more debugging instructions and the one or more image debugging results contained in the first fundus lesion image type directory, analyzing the abnormal feature extraction results contained in the next fundus lesion image type directory of the first fundus lesion image type directory.

[0008] In the present application, the abnormal feature extraction result of the first fundus lesion image type directory includes the abnormal feature extraction results of one or more important element sets in the first fundus lesion image type directory; the abnormal feature extraction results contained in the first fundus lesion image type directory in the one or more fundus lesion image type directories are analyzed to obtain the matching between the one or more debugging instructions contained in the first fundus lesion image type directory and the one or more image debugging results, including: obtaining the first debugging instruction contained in the abnormal feature extraction result of the first important element set in the one or more important element sets, and the one or more debugging instructions including the first debugging instruction. instruction; determining whether the image debugging result of the first debugging instruction is included in the abnormal feature extraction results of one or more second important element sets that are similar to the first important element set in the one or more important element sets; identifying the image debugging result of the first debugging instruction in the abnormal feature extraction results of one or more third important element sets on the basis that the image debugging result of the first debugging instruction does not exist in the abnormal feature extraction results of the one or more second important element sets, wherein the third important element set is a fuzzy image description set obtained by deleting the first important element set and the one or more second important element sets from the one or more important element sets.

[0009] In the present application, the abnormal feature extraction results contained in the next fundus lesion image type directory of the first fundus lesion image type directory are analyzed in combination with the matching between one or more debugging instructions and one or more image debugging results contained in the first fundus lesion image type directory, including: if the image debugging result bound to the first debugging instruction does not exist in the abnormal feature extraction results of the first fundus lesion image type directory, identifying the image debugging result bound to the first debugging instruction in the abnormal feature extraction results of the next fundus lesion image type directory; and / or if the debugging instruction bound to the first image debugging result does not exist in the abnormal feature extraction results of the first fundus lesion image type directory, identifying the debugging instruction bound to the first image debugging result in the abnormal feature extraction results of the next fundus lesion image type directory.

[0010] In the present application, the abnormal feature extraction results contained in the next fundus lesion image type directory of the first fundus lesion image type directory are analyzed in combination with the matching between one or more debugging instructions and one or more image debugging results contained in the first fundus lesion image type directory, and also include: analyzing the abnormal feature extraction results of the next fundus lesion image type directory in combination with the recognition result to obtain the matching between the debugging instructions and the image debugging results contained in the next fundus lesion image type directory.

[0011] In the present application, after obtaining the image debugging result bound to the debugging instructions in the fundus image dataset, it also includes: combining the mapping relationship between multiple debugging instructions in the fundus image dataset to denoise the image debugging result bound to the multiple debugging instructions.

[0012] In the present application, the image debugging results bound by the multiple debugging instructions are denoised, including: on the basis that the multiple debugging instructions include the first debugging instruction, the second debugging instruction and the third debugging instruction belonging to the same fundus lesion image type directory, denoising the first data bound by the first debugging instruction and the second data bound by the second debugging instruction belonging to the same fundus lesion image type directory and performing weighted processing to determine whether they comply with the image quality information bound by the third debugging instruction; and / or on the basis that the multiple debugging instructions include one or more important debugging instructions and all debugging instructions belonging to the same list, denoising whether the splicing result of the image quality information bound by all important debugging instructions belonging to the same list complies with all debugging results bound by all debugging instructions.

[0013] In the present application, after denoising the image debugging results bound to the multiple debugging instructions, it also includes: debugging the image debugging results bound to the multiple debugging instructions on the basis that the denoising result indicates that the mapping relationship does not meet the specified conditions, until the debugging result enables the mapping relationship to be realized.

[0014] In the present application, before performing abnormal feature extraction processing on the fundus image dataset, the method also includes: determining whether the fundus image dataset meets the lesion analysis indication; based on determining that the fundus image dataset does not meet the lesion analysis indication, updating the fundus image dataset to obtain the updated fundus image dataset; the abnormal feature extraction processing on the fundus image dataset includes: performing abnormal feature extraction processing on the updated fundus image dataset.

[0015] In a second aspect, a fundus image restoration system based on deep learning is provided, comprising a processor and a memory that communicate with each other, wherein the processor is used to read a computer program from the memory and execute it to implement the above method.

[0016] The deep learning-based fundus image restoration method and system provided in the embodiments of the present application perform abnormal feature extraction processing on a fundus image dataset to obtain abnormal feature extraction results of the fundus image dataset; classify the abnormal feature extraction results of the fundus image dataset to obtain one or more fundus lesion image type directories; and combine the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories to obtain image debugging results bound to debugging instructions in the fundus image dataset. In this way, the image debugging results bound to debugging instructions in the fundus image dataset are determined by gradually analyzing each abnormal feature extraction result in each fundus lesion image type directory, thereby ensuring the credibility and accuracy of the image debugging results. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A flowchart of a fundus image restoration method based on deep learning provided in an embodiment of the present application.

[0019] Figure 2 A block diagram of a fundus image restoration device based on deep learning provided in an embodiment of the present application.

[0020] Figure 3 This is an architectural diagram of a deep learning-based fundus image restoration system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to better understand the above technical solution, the technical solution of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0022] See also Figure 1 , shows a fundus image restoration method based on deep learning, which may include the technical solutions described in steps 100 and 200.

[0023] Step 100: perform abnormal feature extraction processing on the fundus image dataset to obtain the abnormal feature extraction results of the fundus image dataset; classify the abnormal feature extraction results of the fundus image dataset to obtain one or more fundus lesion image type catalogs.

[0024] Step 200 : combining the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories to obtain an image debugging result bound to the debugging instruction in the fundus image dataset.

[0025] It should be understood that when executing the contents described in the above steps 100 and 200, the fundus image dataset is subjected to abnormal feature extraction processing to obtain the abnormal feature extraction results of the fundus image dataset; the abnormal feature extraction results of the fundus image dataset are classified to obtain one or more fundus lesion image type directories; and the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories are combined to obtain the image debugging results bound to the debugging instructions in the fundus image dataset. In this way, the image debugging results bound to the debugging instructions in the fundus image dataset are determined by gradually analyzing each abnormal feature extraction result in each fundus lesion image type directory, thereby ensuring the credibility and accuracy of the image debugging results.

[0026] In a possible embodiment, the inventors have discovered after long-term research that when combining the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories, there may be a problem of abnormal feature extraction abnormalities, making it difficult to accurately obtain the image debugging results bound to the debugging instructions in the fundus image data set. In order to improve the above technical problems, the step described in step 200 of combining the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories to obtain the image debugging results bound to the debugging instructions in the fundus image data set can specifically include the contents described in the following steps 210 and 220.

[0027] Step 210: Analyze the abnormal feature extraction results contained in the first fundus lesion image type directory in the one or more fundus lesion image type directories to obtain the matching status between the one or more debugging instructions contained in the first fundus lesion image type directory and the one or more image debugging results.

[0028] Step 220: In combination with the matching between one or more debugging instructions and one or more image debugging results contained in the first fundus lesion image type directory, analyze the abnormal feature extraction results contained in the next fundus lesion image type directory of the first fundus lesion image type directory.

[0029] It should be understood that when executing the contents described in the above steps 210 and 220, combined with the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories, the problem of abnormal feature extraction anomalies is improved, so that the image debugging results bound to the debugging instructions in the fundus image data set can be accurately obtained.

[0030] In a possible embodiment, the inventors discovered after long-term research that the abnormal feature extraction results of the first fundus lesion image type directory include abnormal feature extraction results of one or more important element sets in the first fundus lesion image type directory; when analyzing the abnormal feature extraction results contained in the first fundus lesion image type directory among the one or more fundus lesion image type directories, there may be an analysis problem of inaccurate analysis, making it difficult to accurately obtain the matching between the one or more debugging instructions contained in the first fundus lesion image type directory and the one or more image debugging results. In order to improve the above technical problems, the step described in step 210 of analyzing the abnormal feature extraction results contained in the first fundus lesion image type directory among the one or more fundus lesion image type directories to obtain the matching between the one or more debugging instructions contained in the first fundus lesion image type directory and the one or more image debugging results can specifically include the contents described in the following steps 211-213.

[0031] Step 211: The abnormal feature extraction result of the first important element set in the one or more important element sets includes a first debugging instruction, and the one or more debugging instructions include the first debugging instruction.

[0032] Step 212: Determine whether the abnormal feature extraction results of one or more second important element sets similar to the first important element set in the one or more important element sets include the image debugging result of the first debugging instruction.

[0033] Step 213, based on the fact that the image debugging result of the first debugging instruction does not exist in the abnormal feature extraction results of the one or more second important element sets, identify the image debugging result of the first debugging instruction in the abnormal feature extraction results of one or more third important element sets, wherein the third important element set is a fuzzy image description set excluding the first important element set and the one or more second important element sets from the one or more important element sets.

[0034] It should be understood that when executing the contents described in the above steps 211 to 213, when analyzing the abnormal feature extraction results contained in the first fundus lesion image type directory in the one or more fundus lesion image type directories, the problem of inaccurate analysis is improved, so that the matching between the one or more debugging instructions contained in the first fundus lesion image type directory and the one or more image debugging results can be accurately obtained.

[0035] In a possible embodiment, the inventors have discovered after long-term research that when analyzing the abnormal feature extraction results contained in the next fundus lesion image type directory of the first fundus lesion image type directory in combination with the matching between one or more debugging instructions contained in the first fundus lesion image type directory and one or more image debugging results, there may be a problem of inaccurate analysis. In order to improve the above technical problem, the step described in step 220 of analyzing the abnormal feature extraction results contained in the next fundus lesion image type directory of the first fundus lesion image type directory in combination with the matching between one or more debugging instructions contained in the first fundus lesion image type directory and one or more image debugging results can specifically include the contents described in the following steps 221 and 222.

[0036] Step 221: If the image debugging result bound to the first debugging instruction does not exist in the abnormal feature extraction result of the first fundus lesion image type directory, identify the image debugging result bound to the first debugging instruction in the abnormal feature extraction result of the next fundus lesion image type directory.

[0037] Step 222, and / or if the debugging instruction bound to the first image debugging result does not exist in the abnormal feature extraction result of the first fundus lesion image type directory, identify the debugging instruction bound to the first image debugging result in the abnormal feature extraction result of the next fundus lesion image type directory.

[0038] It should be understood that when executing the contents described in the above steps 221 and 222, combined with the matching between one or more debugging instructions contained in the first fundus lesion image type directory and one or more image debugging results, when analyzing the abnormal feature extraction results contained in the next fundus lesion image type directory of the first fundus lesion image type directory, the problem of inaccurate analysis can be improved, thereby ensuring the accuracy of the analysis.

[0039] Based on the above foundation, the analysis of the abnormal feature extraction results contained in the next fundus lesion image type directory of the first fundus lesion image type directory in combination with the matching between one or more debugging instructions and one or more image debugging results contained in the first fundus lesion image type directory may also include the following content: in combination with the recognition result, the analysis of the abnormal feature extraction results of the next fundus lesion image type directory is performed to obtain the matching between the debugging instructions and the image debugging results contained in the next fundus lesion image type directory.

[0040] It should be understood that the above specific implementation steps can effectively improve the accuracy of the analysis.

[0041] Based on the above, after obtaining the image debugging results bound to the debugging instructions in the fundus image dataset, the following contents may also be included: combining the mapping relationship between multiple debugging instructions in the fundus image dataset, denoising the image debugging results bound to the multiple debugging instructions.

[0042] It should be understood that, through the above specific implementation steps, the accuracy of denoising the image debugging result can be effectively improved.

[0043] In a possible embodiment, the inventors have discovered through long-term research that when denoising the image debugging results bound to the multiple debugging instructions, there may be a problem of inaccurate denoising, which may result in the inability to accurately obtain all debugging results. In order to improve the above technical problem, the step of denoising the image debugging results bound to the multiple debugging instructions may specifically include the contents described in the following steps a1 and a2.

[0044] Step a1, on the basis that the multiple debugging instructions include the first debugging instruction, the second debugging instruction and the third debugging instruction belonging to the same fundus lesion image type directory, denoise the first data bound to the first debugging instruction and the second data bound to the second debugging instruction belonging to the same fundus lesion image type directory and perform weighted processing to determine whether they meet the image quality information bound to the third debugging instruction.

[0045] Step a2, and / or on the basis that the multiple debugging instructions include one or more important debugging instructions and all debugging instructions belonging to the same list, whether the splicing result of denoising the image quality information bound to all important debugging instructions belonging to the same list conforms to the total debugging results bound to all debugging instructions.

[0046] It should be understood that when executing the contents described in steps a1 and a2 above, when denoising the image debugging results bound to the multiple debugging instructions, the problem of inaccurate denoising is improved, so that all debugging results can be accurately obtained.

[0047] Based on the above, after denoising the image debugging results bound to the multiple debugging instructions, the following content may also be included: on the basis that the denoising result indicates that the mapping relationship does not meet the specified conditions, debugging the image debugging results bound to the multiple debugging instructions until the debugging result enables the mapping relationship to be realized.

[0048] It should be understood that by debugging the image debugging results bound to multiple debugging instructions, the debugging results can be effectively improved so that an accurate mapping relationship can be achieved.

[0049] Based on the above, before performing abnormal feature extraction processing on the fundus image dataset, the following contents may also be included: determining whether the fundus image dataset meets the lesion analysis indication; and on the basis of determining that the fundus image dataset does not meet the lesion analysis indication, updating the fundus image dataset to obtain an updated fundus image dataset.

[0050] It should be understood that by continuously updating the fundus image dataset, the fundus image dataset can be accurately acquired.

[0051] In a possible embodiment, the inventors have discovered after long-term research that when performing abnormal feature extraction processing on the fundus image dataset, there may be a problem of inaccurate abnormal feature extraction, so further processing is required. In order to improve the above technical problems, a further implementation method is as follows, which performs abnormal feature extraction processing on the updated fundus image dataset.

[0052] Based on the above, please refer to Figure 2, provides a fundus image restoration device 200 based on deep learning, which is applied to a fundus image restoration system based on deep learning, and the device includes: The abnormal feature extraction module 210 is configured to perform abnormal feature extraction processing on the fundus image dataset to obtain abnormal feature extraction results of the fundus image dataset; classify the abnormal feature extraction results of the fundus image dataset to obtain one or more fundus lesion image type catalogs; The result debugging module 220 is configured to obtain an image debugging result bound to the debugging instruction in the fundus image dataset by combining the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories.

[0053] Based on the above, please refer to Figure 3 , shows a fundus image restoration system 300 based on deep learning, including a processor 310 and a memory 320 that communicate with each other, and the processor 310 is used to read and execute a computer program from the memory 320 to implement the above method.

[0054] Based on the above, a computer-readable storage medium is also provided, on which a computer program stored implements the above method when running.

[0055] In summary, based on the above scheme, the fundus image dataset is subjected to abnormal feature extraction processing to obtain the abnormal feature extraction results of the fundus image dataset; the abnormal feature extraction results of the fundus image dataset are classified to obtain one or more fundus lesion image type directories; and the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories are combined to obtain the image debugging results bound to the debugging instructions in the fundus image dataset. In this way, the image debugging results bound to the debugging instructions in the fundus image dataset are determined by gradually analyzing each abnormal feature extraction result in each fundus lesion image type directory, thereby ensuring the credibility and accuracy of the image debugging results.

[0056] It should be understood that the system and its modules shown above can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated hardware. It should be understood by those skilled in the art that the above methods and systems can be implemented using computer-executable instructions and / or contained in processor control code, for example, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the system and its modules of the present application. Not only can hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc. be implemented, they can also be implemented using software executed by various types of processors, and can also be implemented by a combination of the above hardware circuits and software (for example, firmware).

[0057] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other possible beneficial effects.

[0058] The basic concepts have been described above. It will be apparent to those skilled in the art that the detailed disclosure above is merely illustrative and does not limit the present application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and adjustments to the present application. Such modifications, improvements, and adjustments are suggested in the present application and remain within the spirit and scope of the exemplary embodiments of the present application.

[0059] At the same time, this application uses specific terms to describe the embodiments of this application. For example, "one embodiment," "an embodiment," and / or "some embodiments" refer to a certain feature, structure, or characteristic related to at least one embodiment of this application. Therefore, it should be emphasized and noted that "one embodiment," "an embodiment," or "an alternative embodiment" mentioned twice or multiple times in different locations in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this application may be appropriately combined.

[0060] In addition, it should be understood by those skilled in the art that the various aspects of the present application can be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, the various aspects of the present application can be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as a "data block", "module", "engine", "unit", "component" or "system". In addition, the various aspects of the present application may be manifested as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0061] A computer storage medium may include a propagated data signal embodying the computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may be in a variety of forms, including electromagnetic, optical, or any suitable combination thereof. A computer storage medium may be any computer-readable medium other than a computer-readable storage medium that can be connected to an instruction execution system, apparatus, or device to communicate, propagate, or transfer the program for use. The program code on the computer storage medium may be transmitted via any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of these.

[0062] The computer program code required for the operation of the various parts of this application can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C, Visual Basic, Fortran 2003, Perl, COBOL 2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy, or other programming languages. The program code can be executed entirely on the user's computer, or as a stand-alone software package on the user's computer, or partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer via any network, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0063] In addition, unless expressly stated in the claims, the order of the processing elements and sequences described in this application, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this application. Although the above disclosure discusses some of the invention embodiments currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all debugging and equivalent combinations that are consistent with the essence and scope of the embodiments of this application. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.

[0064] Similarly, it should be noted that, in order to simplify the presentation of this application and thus facilitate understanding of one or more embodiments of the invention, the foregoing descriptions of the embodiments of this application sometimes combine multiple features into a single embodiment, figure, or description thereof. However, this disclosure method does not mean that the subject matter of this application requires more features than those recited in the claims. In fact, an embodiment may have fewer features than all of the features of a single embodiment disclosed above.

[0065] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise stated, "about", "approximately" or "substantially" indicate that the numbers allow adaptive changes. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which can be changed according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of the present application are approximate values, in specific embodiments, the settings of such numerical values ​​are as accurate as possible within the feasible range.

[0066] Each patent, patent application, patent application disclosure, and other materials, such as articles, books, specifications, publications, documents, etc., cited in this application is hereby incorporated by reference in its entirety. This includes application history documents that are inconsistent with or conflict with the content of this application, as well as documents (currently or subsequently attached to this application) that limit the broadest scope of the claims of this application. It should be noted that if the descriptions, definitions, and / or use of terms in the accompanying materials of this application are inconsistent or conflicting with the content of this application, the descriptions, definitions, and / or use of terms in this application shall prevail.

[0067] Finally, it should be understood that the embodiments described in this application are merely illustrative of the principles of the embodiments of this application. Other variations may also fall within the scope of this application. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this application may be considered consistent with the teachings of this application. Accordingly, the embodiments of this application are not limited to the embodiments explicitly introduced and described in this application.

[0068] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A fundus image restoration method based on deep learning, characterized in that: The method comprises: performing abnormal feature extraction processing on the fundus image dataset to obtain abnormal feature extraction results of the fundus image dataset; classifying the abnormal feature extraction results of the fundus image dataset to obtain one or more fundus lesion image type catalogs; The image debugging result bound to the debugging instruction in the fundus image dataset is obtained by combining the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories.

2. The method according to claim 1, characterized in that The step of combining the abnormal feature extraction results contained in each of the one or more fundus lesion image type directories to obtain the image debugging result bound to the debugging instruction in the fundus image dataset includes: Analyzing abnormal feature extraction results contained in a first fundus lesion image type directory among the one or more fundus lesion image type directories to obtain a matching condition between one or more debugging instructions contained in the first fundus lesion image type directory and one or more image debugging results; In combination with the matching between one or more debugging instructions contained in the first fundus lesion image type directory and one or more image debugging results, the abnormal feature extraction results contained in the next fundus lesion image type directory of the first fundus lesion image type directory are analyzed.

3. The method according to claim 2, characterized in that The abnormal feature extraction result of the first fundus lesion image type directory includes abnormal feature extraction results of one or more important element sets in the first fundus lesion image type directory; the analyzing the abnormal feature extraction results contained in the first fundus lesion image type directory in the one or more fundus lesion image type directories to obtain a matching situation between one or more debugging instructions contained in the first fundus lesion image type directory and one or more image debugging results includes: Obtaining that a result of extracting abnormal features of a first important element set in the one or more important element sets includes a first debugging instruction, and the one or more debugging instructions include the first debugging instruction; determining whether an abnormal feature extraction result of one or more second important element sets similar to the first important element set in the one or more important element sets includes an image debugging result of the first debugging instruction; Based on the fact that the image debugging result of the first debugging instruction does not exist in the abnormal feature extraction results of the one or more second important element sets, the image debugging result of the first debugging instruction is identified in the abnormal feature extraction results of one or more third important element sets, wherein the third important element set is a fuzzy image description set except for the first important element set and the one or more second important element sets deleted from the one or more important element sets.

4. The method according to claim 2, characterized in that The analyzing, based on a match between one or more debugging instructions and one or more image debugging results included in the first fundus lesion image type directory, an abnormal feature extraction result included in a next fundus lesion image type directory of the first fundus lesion image type directory, includes: If the image debugging result bound to the first debugging instruction does not exist in the abnormal feature extraction results of the first fundus lesion image type directory, identifying the image debugging result bound to the first debugging instruction in the abnormal feature extraction results of the next fundus lesion image type directory; And / or if the debugging instruction bound to the first image debugging result does not exist in the abnormal feature extraction result of the first fundus lesion image type directory, the debugging instruction bound to the first image debugging result is identified in the abnormal feature extraction result of the next fundus lesion image type directory.

5. The method according to claim 4, characterized in that The method of analyzing the abnormal feature extraction results contained in the next fundus lesion image type directory of the first fundus lesion image type directory in combination with the matching between one or more debugging instructions and one or more image debugging results contained in the first fundus lesion image type directory also includes: analyzing the abnormal feature extraction results of the next fundus lesion image type directory in combination with the recognition result to obtain the matching between the debugging instructions and the image debugging results contained in the next fundus lesion image type directory.

6. The method according to claim 5, characterized in that After obtaining the image debugging result bound to the debugging instruction in the fundus image dataset, the method further includes: denoising the image debugging result bound to the multiple debugging instructions in combination with the mapping relationship between the multiple debugging instructions in the fundus image dataset.

7. The method according to claim 6, characterized in that Denoising the image debugging results bound to the multiple debugging instructions, including: On the basis that the multiple debugging instructions include a first debugging instruction, a second debugging instruction, and a third debugging instruction belonging to the same fundus lesion image type directory, performing weighted processing on first data bound to the first debugging instruction and second data bound to the second debugging instruction, which belong to the same fundus lesion image type directory, to determine whether they meet image quality information bound to the third debugging instruction; And / or on the basis that the multiple debugging instructions include one or more important debugging instructions and all debugging instructions belonging to the same list, whether the splicing result of denoising the image quality information bound to all important debugging instructions belonging to the same list conforms to the total debugging results bound to all debugging instructions.

8. The method according to claim 6, characterized in that After denoising the image debugging results bound to the multiple debugging instructions, the method further includes: debugging the image debugging results bound to the multiple debugging instructions on the basis that the denoising result indicates that the mapping relationship does not meet the specified condition, until the debugging result enables the mapping relationship to be realized.

9. The method according to claim 5, characterized in that Before performing abnormal feature extraction processing on the fundus image dataset, the method also includes: determining whether the fundus image dataset meets the lesion analysis indication; based on determining that the fundus image dataset does not meet the lesion analysis indication, updating the fundus image dataset to obtain an updated fundus image dataset; the abnormal feature extraction processing on the fundus image dataset includes: performing abnormal feature extraction processing on the updated fundus image dataset.

10. A fundus image restoration system based on deep learning, characterized in that: The invention comprises a processor and a memory communicating with each other, wherein the processor is used to read a computer program from the memory and execute the computer program to implement the method according to any one of claims 1 to 9.

Images