A quality control method and device for fundus images and a computer readable medium

Abstract

The embodiment of the present application provides a kind of quality control method of fundus image, this method includes: first, identify the region of interest in target fundus image;Second, extract eye feature and image feature from the region of interest;Finally, based on the preset scoring rule and the preset weight, the score result corresponding to eye feature and the score result corresponding to image feature are weighted, and image quality score is generated.The eye position judgment result, eye judgment result, average definition, average exposure and image resolution judgment result of the region of interest are combined for the evaluation of fundus image quality in the embodiment;Thus, the quality of fundus image can be more comprehensively and more accurately evaluated;The image feature score result and eye feature score result are weighted in the embodiment, so that the fundus image can be evaluated in view of the characteristics of fundus itself, the fineness of fundus image quality evaluation is improved, and it is beneficial to the detection of fundus disease in later period.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and particularly relates to a method, apparatus and computer-readable medium for quality control of fundus images. Background Technology

[0002] The fundus is the only part of the human body where capillary changes can be observed simply, non-invasively, and clearly. According to incomplete statistics, over 60 types of fundus diseases or systemic chronic diseases can cause corresponding lesions in the fundus; therefore, fundus images are often a clinical diagnostic reference for these diseases. In recent years, the urgent need for early and effective disease detection among a large population has spurred the development of computer-aided tools. In particular, automated computer-aided diagnostic and analysis technologies have rapidly developed from the detection of individual lesions and clues to image-level disease detection, thanks to the attention of policymakers, academics, clinicians, and industry. Due to the necessity of evaluating the performance of algorithms and technologies for disease and lesion detection, clean, clearly discernible, and sharp images are typically selected for technology development, training, and testing. However, in practice, such clean, high-quality images are often difficult to obtain. This may be due to imaging artifacts arising during image capture from misalignment of the acquisition optics, accidental movement of the subject, or defocus. When automated analysis systems encounter such images, the generated results may be unreliable.

[0003] In existing technologies, the evaluation of fundus image quality only considers factors related to image dimensions and factors that cause fundus lesions (e.g., image dimension factors include image contrast, blurriness, and entropy measurement; factors causing lesions include vascular density). However, it does not fully incorporate the structural features of fundus images for targeted analysis. Therefore, existing fundus image quality evaluation is only applicable to image quality evaluation in a broad sense and cannot effectively evaluate fundus images in a precise manner. Consequently, misjudgments often occur when using computer-aided tools for fundus disease detection. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, embodiments of the present invention provide a method, apparatus, and computer-readable medium for quality control of fundus images. This method utilizes quantitative and visual computation techniques to effectively and precisely assess the quality of fundus images, achieving automated assessment and improving assessment efficiency.

[0005] According to a first aspect of the present invention, a quality control method for fundus images is provided. The method includes: identifying a region of interest in a target fundus image; extracting eye features and image features from the region of interest; wherein the eye features include at least an eye position determination result and an eye identification result; the image features include at least an average sharpness, an average exposure, and an image resolution determination result; and weighting the scoring results corresponding to the eye features and the scoring results corresponding to the image features based on preset scoring rules and preset weights to generate an image quality score.

[0006] Optionally, the method further includes: if the image quality score does not meet a preset threshold, then determining that the target fundus image is unqualified; obtaining preset weights corresponding to eye features and image features in the region of interest respectively; and based on the preset weights, sorting the eye features and / or image features that cause the target fundus image to be unqualified according to priority, and using them as the reasons for the analysis of the evaluation results.

[0007] Optionally, when the eye features include eye position determination results and eye identification results, the step of extracting eye features from the region of interest includes: using a model to identify the eye structure features of the region of interest to obtain identification results; wherein, the eye structure features are used to indicate the optic disc, macula, and blood vessels; if the identification results include at least two of the eye structure features, then eye identification is performed based on any two of the eye structure features, and an eye identification result is output; if the identification results include at least two of the eye structure features, including the optic disc, then eye position determination is performed based on the optic disc and other eye structure features in the identification results, and an eye position determination result is output.

[0008] Optionally, the step of weighting the scoring results corresponding to the eye features and the scoring results corresponding to the image features based on preset scoring rules to generate an image scoring result includes: determining the eye position score corresponding to the eye position judgment result, the eye type score corresponding to the eye type judgment result, the sharpness score corresponding to the average sharpness, the exposure score corresponding to the average exposure, and the resolution score corresponding to the image resolution judgment result based on preset scoring rules; and weighting the eye position score, the eye type score, the sharpness score, the exposure score, and the resolution score based on preset weights to generate an image scoring result.

[0009] Optionally, the step of determining eye position based on the optic disc and other ocular structural features in the recognition result, and outputting the eye position determination result, includes: if the recognition result contains a macula, then obtaining the position information of the macula and the optic disc in the region of interest; and based on the position information, determining a first distance between the center point of the macula and the region of interest, and a second distance between the optic disc and the center point of the region of interest; if the first distance is less than the second distance, then determining the eye position of the region of interest to be centered on the macula; if the first distance is not less than the second distance, then determining the eye position of the region of interest to be centered on the optic disc; if the recognition result does not include the macula, then identifying the blood vessels... The main vascular arch is identified, and the position information of the main vascular arch detection frame and the optic disc within the region of interest is obtained. Based on this position information, a third distance is determined between the center point of the main vascular arch detection frame and the center point of the region of interest, and a second distance is determined between the center point of the optic disc and the center point of the region of interest. If the third distance is less than the second distance and the main vascular arch and the optic disc are on the same side of the center point of the region of interest, the eye position of the region of interest is determined to be centered on the macula. If the third distance is within a preset threshold range, the eye position of the region of interest is determined to be centered on the macula. If the second distance is less than the preset threshold, the eye position of the region of interest is determined to be centered on the optic disc.

[0010] Optionally, the method further includes: determining the intersection of the trajectory circle and the main blood vessel as the first vertex of the vascular arch detection frame; wherein, the trajectory circle is used to indicate the arc formed with the center point of the optic disc as the center and a preset distance as the radius; drawing a perpendicular line along the Y-coordinate direction of the spatial coordinate system where the first vertex is located, and determining the intersection of the perpendicular line and the main blood vessel as the second vertex; and determining the position information of the main vascular arch detection frame in the region of interest based on the first vertex, the second vertex and the center point of the optic disc.

[0011] Optionally, the method further includes: if the eye identification result indicates that the region of interest is the right eye or the left eye, then outputting a first eye identification score based on a preset eye identification scoring rule; if the eye identification result indicates that the eye identification of the region of interest is unknown, then outputting a second eye identification score based on a preset eye identification scoring rule; wherein the second eye identification score is less than the first eye identification score; if the eye position identification result indicates that the region of interest is the optic disc or the macula, then outputting a first eye position score based on a preset eye position scoring rule; if the eye position identification result indicates that the eye position of the region of interest is other categories, then outputting a second eye position score based on a preset eye position scoring rule; wherein the second eye position score is less than the first eye position score.

[0012] Optionally, identifying the region of interest in the target fundus image includes: acquiring a normal color fundus image; performing recognition processing on the normal color fundus image to obtain a recognition result; if the recognition result indicates that the normal color fundus image is the target fundus image, then identifying the region of interest in the target fundus image.

[0013] Optionally, the method further includes: if the identification result includes at most one eye structure feature, the eye position determination result is unknown; if the identification result includes at most one eye structure feature or only the macula and blood vessels, the eye position determination result is other classification.

[0014] Optionally, when the image features include average sharpness, average exposure, and image resolution judgment results, the step of extracting image features from the region of interest includes: obtaining a sharpness distribution map, an exposure distribution map, and an image resolution corresponding to the region of interest; determining the average sharpness of the region of interest based on the sharpness distribution map; determining the average exposure of the region of interest based on the exposure distribution map; and determining the image resolution judgment result based on the image resolution and a preset threshold.

[0015] According to a second aspect of the present invention, a quality control device for fundus images is also provided. The device includes: a recognition module for recognizing a region of interest in a target fundus image; an extraction module for extracting eye features and image features from the region of interest; wherein the eye features include at least an eye position determination result and an eye identification result; and the image features include at least an average sharpness, an average exposure, and an image resolution determination result; and a generation module for weighting the scoring results corresponding to the eye features and the scoring results corresponding to the image features based on preset scoring rules and preset weights to generate an image quality score.

[0016] According to a third aspect of the present invention, a computer-readable medium is also provided, on which a computer program is stored, wherein the program, when executed by a processor, implements the method described in the first aspect.

[0017] This invention provides a quality control method for fundus images. The method includes: first, identifying a region of interest (ROI) in a target fundus image; second, extracting eye features and image features from the ROI; wherein the eye features include at least eye position determination results and eye type determination results; and the image features include at least average sharpness, average exposure, and image resolution determination results; finally, based on preset scoring rules and preset weights, weighting the scoring results corresponding to the eye features and the scoring results corresponding to the image features to generate an image quality score. This embodiment applies visual computing technology to quantitatively analyze fundus images, outputting eye position and eye type determination results, and digitally outputting image sharpness, exposure, and image resolution. Combining these results digitally describes the quality of the fundus image; thus, it can more comprehensively and precisely evaluate the quality of fundus images. Moreover, this embodiment also weights the image feature scoring results and the eye feature scoring results, thereby enabling targeted evaluation of fundus images based on the characteristics of the fundus itself, improving the precision of fundus image quality evaluation, and facilitating subsequent detection of fundus diseases. Attached Figure Description

[0018] The following sections will describe some specific embodiments of the invention in detail by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0019] Figure 1 This is a flowchart illustrating a quality control method for fundus images provided in an embodiment of the present invention.

[0020] Figure 2 A flowchart illustrating a quality control method for fundus images provided in another embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of the process for extracting eye features from a region of interest in one embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram of the process for extracting image features from a region of interest in one embodiment of the present invention.

[0023] Figure 5 This is a schematic diagram of the structure of a quality control device for fundus images provided in an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] like Figure 1 The diagram shown is a flowchart illustrating a quality control method for fundus images provided in an embodiment of the present invention.

[0026] A quality control method for fundus images, the method comprising at least the following steps:

[0027] S101, Identify the region of interest in the target fundus image;

[0028] S102, extract eye features and image features from the region of interest, respectively; wherein, the eye features include at least the eye position determination result and the eye identification result; the image features include at least the average sharpness, average exposure and image resolution determination results;

[0029] S103, based on preset scoring rules and preset weights, performs weighted processing on the scoring results corresponding to eye features and the scoring results corresponding to image features to generate an image quality score.

[0030] In S101, the target fundus image is a fundus image that needs to be quality evaluated; the region of interest in the target fundus image is identified based on the trained model.

[0031] In S102, eye features and image features in the region of interest are extracted based on the pre-trained model; the eye position judgment results are divided into three categories: optic disc, macula, and others; the eye identification results are divided into three categories: left eye, right eye, and unknown; the average sharpness and average exposure are both quantized results; the image resolution judgment results include two categories: qualified and unqualified.

[0032] In S103, there are corresponding scoring rules for five evaluation indicators: eye position, eye type, average sharpness, average exposure, and image resolution. For any of the five evaluation indicators: a scoring rule is constructed for that indicator; based on the mapping relationship in the scoring rule, the judgment result corresponding to that evaluation indicator is scored to generate the corresponding score result. For example: the eye position scoring rule is: the score for the optic disc is 10 points, the score for the macula is 10 points, and the score for other categories is 3 points; based on the eye position judgment result of the optic disc, the eye position score is determined to be 10 points. The average sharpness scoring rule is: average sharpness ≤ 30%, the corresponding score is 2 points; 30% < average sharpness ≤ 50%, the corresponding score is 4 points; 50% < average sharpness ≤ 70%, the corresponding score is 7 points; 70% < average sharpness, the corresponding score is 9 points; based on the average image sharpness of 80%, the average sharpness score of the region of interest is determined to be 9 points.

[0033] Here, the scoring rules for the evaluation indicators are established based on the judgment results of the evaluation indicators. Other classifications in the eye position judgment results are used to indicate eye position classifications not defined in this application, other than optic disc and macular eye position.

[0034] The preset weights of evaluation indicators vary in different scenarios; within the same scenario, the preset weights of evaluation indicators are fixed and do not change; however, the preset weights of evaluation indicators change depending on the scenario. For example, in one scenario, the weight of eye position is greater than the weight of eye type; while in another scenario, the weight of eye position is less than the weight of eye type.

[0035] This embodiment combines eye position, eye type, average sharpness, average exposure, and image resolution in the region of interest for evaluating fundus image quality. This allows for a more comprehensive, accurate, and detailed assessment of fundus image quality. Furthermore, this embodiment applies weights to each evaluation indicator and obtains an image quality score through weighted processing. This enables a more targeted evaluation of fundus images based on the characteristics of the fundus itself, improving the precision of fundus image quality assessment and facilitating the detection of fundus diseases in the future.

[0036] Furthermore, the quality control method in this embodiment is based on algorithms and mathematical models for evaluation, which can be integrated with automated systems to achieve automated fundus image quality evaluation and screening. As a result, not only is production efficiency and quality control improved, but the evaluation results are also unaffected by subjective factors and can provide objective evaluation results.

[0037] In a preferred embodiment, the image quality score can also be displayed in an intuitive and visual way, such as a bar chart or radar chart, so that users can easily understand the quality of the fundus image.

[0038] like Figure 2The diagram shown is a flowchart illustrating a quality control method for fundus images provided in another embodiment of the present invention.

[0039] A quality control method for fundus images includes at least the following steps:

[0040] S201, acquire a standard color fundus image;

[0041] S202, using a binary classification model to process ordinary color fundus images for recognition, and obtaining recognition results;

[0042] S203, if the recognition result indicates that the ordinary color fundus image is the target fundus image, then identify the region of interest in the target fundus image;

[0043] S204, extract eye features and image features from the region of interest, respectively; wherein, the eye features include at least the eye position determination result and the eye identification result; the image features include at least the average sharpness, average exposure and image resolution determination results;

[0044] S205, based on preset scoring rules and preset weights, weights the scoring results corresponding to eye features and the scoring results corresponding to image features to generate an image quality score;

[0045] S206, determine whether the image quality score is greater than the preset threshold; if yes, proceed to step S208; if no, proceed to step S207.

[0046] S207: Obtain the preset weights corresponding to the eye features and image features in the region of interest, respectively; based on the preset weights, sort the eye features and / or image features that cause the target fundus image to be unqualified according to priority, and use them as the reasons for the evaluation results; then execute S209.

[0047] S208, Determine the target fundus image as a qualified image;

[0048] S209, output the evaluation results; the evaluation results include image quality score, pass or fail, analysis of reasons, and sharpness distribution map and exposure distribution map.

[0049] Specifically, after using a binary classification model to identify ordinary color fundus images, if the identification result indicates that the ordinary color fundus image is not the target fundus image, no processing is performed, and the result without an image quality score is directly output. If the identification result indicates that the ordinary color fundus image is the target fundus image, the target fundus image needs to be evaluated to generate an image quality score; if the image quality score is unsatisfactory, the reasons for the unsatisfactory image quality need to be analyzed based on five evaluation indicators, and the reasons for the unsatisfactory image quality need to be output in order of importance. If the image quality score is satisfactory, the results of the five evaluation indicators can also be output sequentially.

[0050] Generally, the output evaluation results include three parts: 1. Whether the image is qualified and the image quality score; 2. The reason for the image being unqualified; 3. Sharpness distribution map and exposure distribution map. The sharpness distribution map and exposure distribution map can intuitively show where the image is not clear and where it is overexposed. For example, unclear parts may be displayed semi-transparently.

[0051] It should be noted that ordinary fundus images include, but are not limited to, all image data stored by medical institutions during the treatment of the same or different patients. Ordinary fundus images can be 45° fundus images, or 60°, 80°, 120° fundus images, or wide-angle fundus images or fundus images of other fields of view, or even fundus images of other modalities. Ordinary 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.

[0052] This embodiment's method, when determining that the target fundus image is substandard, can also analyze the reasons for the substandardness and output the analyzed reasons. This enables a more refined assessment of the target fundus image quality, providing more detailed information to better understand the specific problems with the fundus image quality, thus achieving fine-grained evaluation. Furthermore, the image quality score provided in this embodiment is a specific numerical score, which quantifies image quality, making the evaluation results easier to understand and compare.

[0053] like Figure 3 The diagram shown is a flowchart illustrating the process of extracting eye features from the region of interest in one embodiment of the present invention.

[0054] Extracting eye features from the region of interest includes at least the following steps:

[0055] S301, the model is used to identify the eye structure features of the region of interest and obtain the identification results; among them, the eye structure features are used to indicate the optic disc, macula and main vascular arch;

[0056] S302, if the recognition result includes at least two of the aforementioned eye structure features, then eye identification is performed based on any two eye structure features, and the eye identification result is output;

[0057] S303, if the recognition result includes at least two of the aforementioned eye structure features, including the optic disc, then the eye position is determined based on the optic disc and other eye structure features in the recognition result, and the eye position determination result is output.

[0058] S304. If the recognition result includes at most one eye structure feature, the eye identification result is unknown.

[0059] S305. If the recognition result includes at most one eye structure feature or only the macula and blood vessels, the eye position judgment result is other classification.

[0060] Here, eye identification and eye position determination can be achieved based on a trained model or on a software program.

[0061] Therefore, this embodiment can determine eye position and eye type based on the eye structure features of the region of interest, which not only realizes automated recognition and reduces the influence of human subjective factors, but also improves the accuracy of eye position and eye type recognition, thereby improving the precision of fundus image quality assessment.

[0062] In a preferred embodiment, first determine whether the eye position is centered on the macula; if not, then determine whether it is centered on the optic disc. The eye position determination based on the optic disc and other eye structure features in the recognition result, and the output of the eye position determination result, includes:

[0063] If the recognition result contains a macula, then the position information of the macula and the optic disc in the region of interest is obtained respectively; and based on the position information, a first distance between the center point of the macula and the region of interest, and a second distance between the optic disc and the center point of the region of interest are determined; if the first distance is less than the second distance, then the eye position of the region of interest is determined to be centered on the macula; if the first distance is not less than the second distance, then the eye position of the region of interest is determined to be centered on the optic disc.

[0064] If the recognition result does not include the macula, then the main vascular arch in the blood vessel is identified, and the position information of the main vascular arch detection frame and the optic disc in the region of interest is obtained respectively. Based on this position information, a third distance between the center point of the main vascular arch detection frame and the center point of the region of interest, and a second distance between the center point of the optic disc and the center point of the region of interest are determined. If the third distance is less than the second distance and the main vascular arch and the optic disc are on the same side of the center point of the region of interest, then the eye position of the region of interest is determined to be centered on the macula. If the third distance is within a preset threshold range, then the eye position of the region of interest is determined to be centered on the macula. If the second distance is less than the preset threshold, then the eye position of the region of interest is determined to be centered on the optic disc. For example, the third distance is greater than 1.5PD and less than 3PD, or the third distance is greater than 2PD and less than 3PD, or the third distance is greater than 1.5PD and less than 3.5PD, etc., which can be adjusted by those skilled in the art according to the actual camera model.

[0065] The position information of the main vascular arch detection frame in the region of interest is obtained by the following method: the intersection of the trajectory circle and the main vascular vessel is determined as the first vertex of the vascular arch detection frame; wherein, the trajectory circle is used to indicate the arc formed with the optic disc center point as the center and a preset distance as the radius; a perpendicular line is drawn along the Y-coordinate direction of the spatial coordinate system where the first vertex is located, and the intersection of the perpendicular line and the main vascular vessel is determined as the second vertex; based on the first vertex, the second vertex, and the optic disc center point, the position information of the main vascular arch detection frame in the region of interest is determined. In a preferred embodiment, the step of determining the eye based on any two of the stated eye structural features and outputting the eye determination result includes: acquiring the position information of the macula, optic disc, and aortic arch within the region of interest; if the position information indicates that the optic disc is to the left of the macula, then the region of interest is determined to be the left eye; if the position information indicates that the optic disc is to the right of the macula, then the region of interest is determined to be the right eye; if the position information indicates that the optic disc is to the left of the aortic arch, then the region of interest is determined to be the left eye; if the position information indicates that the optic disc is to the right of the aortic arch, then the region of interest is determined to be the right eye; if the position information indicates that the macula is to the right of the aortic arch, then the region of interest is determined to be the left eye; if the position information indicates that the macula is to the left of the aortic arch, then the region of interest is determined to be the right eye.

[0066] In a preferred embodiment, the step of weighting the scoring results corresponding to the eye features and the scoring results corresponding to the image features based on preset scoring rules to generate an image scoring result further includes: if the eye identification result indicates that the region of interest is the right eye or the left eye, then outputting a first eye identification score based on the preset eye identification scoring rules; if the eye identification result indicates that the eye identification of the region of interest is unknown, then outputting a second eye identification score based on the preset eye identification scoring rules; wherein, the second eye identification score is less than the first eye identification score; if the eye position identification result indicates that the region of interest is the optic disc or the macula, then outputting a first eye position score based on the preset eye position scoring rules; if the eye position identification result indicates that the eye position of the region of interest is other categories, then outputting a second eye position score based on the preset eye position scoring rules; wherein, the second eye position score is less than the first eye position score.

[0067] It should be noted that the optic disc, macula, and aortic arch are important features of the fundus; if one or more of these features are missing, it indicates that the image quality is poor or that the region of interest is too small to obtain more useful information.

[0068] like Figure 4 The diagram shown is a flowchart illustrating the process of extracting image features from a region of interest in one embodiment of the present invention.

[0069] Extracting image features from a region of interest includes at least the following steps:

[0070] S401, Obtain the sharpness distribution map, exposure distribution map, and image resolution corresponding to the region of interest;

[0071] S402, determine the average sharpness of the region of interest based on the sharpness distribution map;

[0072] S403, determine the average exposure of the region of interest based on the exposure distribution map;

[0073] S404, determine the image resolution judgment result based on the image resolution and a preset threshold.

[0074] Specifically, all sharpness values ​​are obtained from the sharpness distribution map, and the average of all sharpness values ​​is calculated to obtain the average sharpness; all exposure values ​​are obtained from the exposure distribution map, and the average of all exposure values ​​is calculated to obtain the average exposure; it is determined whether the image resolution is greater than a preset threshold; if yes, the image resolution is deemed acceptable; if no, the image resolution is deemed unacceptable.

[0075] Therefore, this embodiment can perform a refined evaluation of the quality of the target fundus image from the image dimension by judging the results based on average exposure, average sharpness, and image resolution; thus further improving the precision of fundus image quality evaluation.

[0076] The quality control method for fundus images provided in this embodiment will be described in detail below with reference to specific application scenarios.

[0077] A quality control method for fundus images includes at least the following steps:

[0078] S1, acquire a normal color fundus image; perform recognition processing on the normal color fundus image to obtain a recognition result; if the recognition result indicates that the normal color fundus image is a target fundus image, then identify the region of interest in the target fundus image;

[0079] S2, using a model to identify the eye structure features of the region of interest, obtaining an identification result; wherein, the eye structure features are used to indicate the optic disc, macula, and main vascular arch; if the identification result includes at least two of the eye structure features, then an eye classification judgment is performed based on any two of the eye structure features, and the eye classification judgment result is output; if the identification result includes at least two of the eye structure features, including the optic disc, then an eye position judgment is performed based on the optic disc and other eye structure features in the identification result, and the eye position judgment result is output; if the identification result includes at most one eye structure feature, then the eye classification judgment result is unknown; if the identification result includes at most one eye structure feature or only includes the macula and blood vessels, then the eye position judgment result is other classification.

[0080] S3, obtain the sharpness distribution map, exposure distribution map, and image resolution corresponding to the region of interest; determine the average sharpness of the region of interest based on the sharpness distribution map; determine the average exposure of the region of interest based on the exposure distribution map; determine the image resolution judgment result based on the image resolution and a preset threshold.

[0081] S4. Based on the preset scoring rules, determine the eye position score corresponding to the eye position judgment result, the eye type score corresponding to the eye type judgment result, the sharpness score corresponding to the average sharpness, the exposure score corresponding to the average exposure, and the resolution score corresponding to the image resolution judgment result.

[0082] S5. Based on preset weights, the eye position score, the eye-specific score, the sharpness score, the exposure score, and the resolution score are weighted to generate an image score result.

[0083] S6, determine whether the image quality score is greater than a preset threshold; if yes, proceed to step S8; if no, proceed to step S7.

[0084] S7, obtain the preset weights corresponding to the eye features and image features in the region of interest respectively; based on the preset weights, sort the eye features and / or image features that cause the target fundus image to be unqualified according to priority, and use them as the reasons for the evaluation results; then execute S9;

[0085] S8, determine that the target fundus image is a qualified image;

[0086] S9, output the evaluation results; wherein, the evaluation results include image quality score, pass or fail, analysis of reasons, and sharpness distribution map and exposure distribution map.

[0087] The method described in this embodiment can quickly evaluate a large number of fundus images, improve evaluation efficiency, and perform well in real-time applications.

[0088] It should be noted that the method in this embodiment can be customized and extended as needed to meet the requirements of specific application scenarios. For example, if you want to make the fundus image quality assessment more lenient, you can adjust the parameters; if you want to make the assessment more precise, you can also adjust the parameters to support different fundus image quality assessment standards.

[0089] like Figure 5 The diagram shown is a structural schematic of a quality control device for fundus images provided in an embodiment of the present invention.

[0090] A quality control device for fundus images, the device 500 comprising at least: a recognition module 501 for recognizing a region of interest in a target fundus image; an extraction module 502 for extracting eye features and image features from the region of interest; wherein the eye features include at least eye position determination results and eye identification results; and the image features include at least average sharpness, average exposure, and image resolution determination results; and a generation module 503 for weighting the scoring results corresponding to the eye features and the scoring results corresponding to the image features based on preset scoring rules and preset weights to generate an image quality score.

[0091] In a preferred embodiment, the device further includes: a determining module, configured to determine that the target fundus image is unqualified if the image quality score does not meet a preset threshold; and an analysis module, configured to obtain preset weights corresponding to eye features and image features in the region of interest, respectively; and based on the preset weights, to sort the eye features and / or image features that cause the target fundus image to be unqualified according to priority, and use them as the analysis reasons for the evaluation results.

[0092] In a preferred embodiment, the extraction module includes: a recognition unit, configured to use a model to recognize eye structure features of the region of interest and obtain a recognition result; wherein the eye structure features are used to indicate the optic disc, macula, and blood vessels; an eye identification unit, configured to perform eye identification based on any two of the eye structure features if the recognition result includes at least two of the eye structure features, and output an eye identification result; and an eye position determination unit, configured to perform eye position determination based on the optic disc and other eye structure features in the recognition result if the recognition result includes at least two of the eye structure features, including the optic disc, and output an eye position determination result.

[0093] In a preferred embodiment, the eye-specific determination unit includes: an acquisition subunit, configured to acquire position information of the macula, optic disc, and main vascular arch within the region of interest; a first determination subunit, configured to determine the region of interest as the left eye if the position information indicates that the optic disc is to the left of the macula, and as the right eye if the position information indicates that the optic disc is to the right of the macula; a second determination subunit, configured to determine the region of interest as the left eye if the position information indicates that the optic disc is to the left of the main vascular arch, and as the right eye if the position information indicates that the optic disc is to the right of the main vascular arch; and a third determination subunit, configured to determine the region of interest as the left eye if the position information indicates that the macula is to the right of the main vascular arch, and as the right eye if the position information indicates that the macula is to the left of the main vascular arch.

[0094] In a preferred embodiment, the eye position determination unit includes: a first eye position determination subunit, configured to, if the recognition result includes a macula, acquire position information of the macula and the optic disc in the region of interest; and, based on the position information, determine a first distance between the center point of the macula and the region of interest, and a second distance between the optic disc and the center point of the region of interest; if the first distance is less than the second distance, determine that the eye position of the region of interest is centered on the macula; if the first distance is not less than the second distance, determine that the eye position of the region of interest is centered on the optic disc; and a second eye position determination subunit, configured to, if the recognition result does not include the macula, identify the main vein in the blood vessel. The vascular arch is detected, and the position information of the main vascular arch detection frame and the optic disc in the region of interest is obtained respectively. Based on the position information, a third distance between the center point of the main vascular arch detection frame and the center point of the region of interest, and a second distance between the center point of the optic disc and the center point of the region of interest are determined. If the third distance is less than the second distance and the main vascular arch and the optic disc are on the same side of the center point of the region of interest, the eye position of the region of interest is determined to be centered on the macula. If the third distance is within a preset threshold range, the eye position of the region of interest is determined to be centered on the macula. If the second distance is less than the preset threshold, the eye position of the region of interest is determined to be centered on the optic disc.

[0095] In a preferred embodiment, the second eye position determination subunit includes: a first determining unit, used to determine the intersection of the trajectory circle and the main blood vessel as the first vertex of the vascular arch detection frame; wherein, the trajectory circle is used to indicate an arc formed with the center point of the optic disc as the center and a preset distance as the radius; a second determining unit, used to draw a perpendicular line along the Y-coordinate direction of the spatial coordinate system where the first vertex is located, and determine the intersection of the perpendicular line and the main blood vessel as the second vertex; a third determining unit, used to determine the position information of the main vascular arch detection frame in the region of interest based on the two vertices and the center point of the optic disc.

[0096] In a preferred embodiment, the generation module includes: a scoring unit, configured to determine, based on preset scoring rules, an eye position score corresponding to the eye position judgment result, an eye type score corresponding to the eye type judgment result, a sharpness score corresponding to the average sharpness, an exposure score corresponding to the average exposure, and a resolution score corresponding to the image resolution judgment result; and a weighting unit, configured to perform weighted processing on the eye position score, the eye type score, the sharpness score, the exposure score, and the resolution score based on preset weights to generate an image scoring result.

[0097] In a preferred embodiment, the scoring unit includes: an eye-specific scoring subunit, configured to output a first eye-specific score based on a preset eye-specific scoring rule if the eye-specific judgment result indicates that the region of interest is the right eye or the left eye; and to output a second eye-specific score based on a preset eye-specific scoring rule if the eye-specific judgment result indicates that the eye-specificity of the region of interest is unknown; wherein the second eye-specific score is less than the first eye-specific score; and an eye position scoring subunit, configured to output a first eye position score based on a preset eye position scoring rule if the eye position judgment result indicates that the region of interest is the optic disc or the macula; and to output a second eye position score based on a preset eye position scoring rule if the eye position judgment result indicates that the eye position of the region of interest is other categories; wherein the second eye position score is less than the first eye position score.

[0098] In a preferred embodiment, the recognition module includes: an acquisition unit for acquiring a normal color fundus image; a first recognition unit for performing recognition processing on the normal color fundus image to obtain a recognition result; and a second recognition unit for recognizing a region of interest in the target fundus image if the recognition result indicates that the normal color fundus image is a target fundus image.

[0099] In a preferred embodiment, the extraction module further includes an unknown judgment unit, which is used to determine the eye type as unknown if the recognition result includes at most one eye structure feature, or only the macula and blood vessels if the recognition result includes at most one eye structure feature, or only the macula and blood vessels.

[0100] In a preferred embodiment, the extraction module includes: an acquisition unit, configured to acquire a sharpness distribution map, an exposure distribution map, and an image resolution corresponding to the region of interest; determine the average sharpness of the region of interest based on the sharpness distribution map; a first determination unit, configured to determine the average exposure of the region of interest based on the exposure distribution map; and a second determination unit, configured to determine an image resolution judgment result based on the image resolution and a preset threshold.

[0101] The above-described apparatus can execute the fundus image quality control method provided in an embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the fundus image quality control method. Technical details not described in detail in this embodiment can be found in the fundus image quality control method provided in an embodiment of the present invention.

[0102] The present invention also provides an electronic device, comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the quality control method for fundus images according to the present invention.

[0103] In addition to the methods and apparatus described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of this application described in the "Exemplary Methods" section above.

[0104] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0105] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the methods according to the following embodiments of this application described in the "Exemplary Methods" section above.

[0106] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0107] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0108] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0109] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0110] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0111] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

[0112] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0113] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0114] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A quality control method of a fundus image, characterized by, include: Identify the region of interest in the target fundus image; Eye features and image features are extracted from the region of interest, respectively; wherein, the eye features include at least the eye position determination result and the eye identification result; the image features include at least the average sharpness, average exposure, and image resolution determination results; Based on preset scoring rules and preset weights, the scoring results corresponding to the eye features and the scoring results corresponding to the image features are weighted and processed to generate an image quality score. When the eye features include eye position determination results and eye identification results, the step of extracting eye features from the region of interest includes: using a model to identify the eye structure features of the region of interest to obtain identification results; wherein, the eye structure features are used to indicate the optic disc, macula, and blood vessels; if the identification results include at least two of the eye structure features, then eye identification is performed based on any two of the eye structure features, and an eye identification result is output; if the identification results include at least two of the eye structure features, including the optic disc, then eye position determination is performed based on the optic disc and other eye structure features in the identification results, and an eye position determination result is output. The step of determining eye position based on the optic disc and other eye structure features in the recognition result, and outputting the eye position determination result, includes: If the recognition result contains a macula, then the position information of the macula and the optic disc in the region of interest is obtained respectively; and based on the position information, a first distance between the center point of the macula and the region of interest, and a second distance between the optic disc and the center point of the region of interest are determined; if the first distance is less than the second distance, then the eye position of the region of interest is determined to be centered on the macula; if the first distance is not less than the second distance, then the eye position of the region of interest is determined to be centered on the optic disc. If the recognition result does not include the macula, then the main vascular arch in the blood vessel is identified, and the position information of the main vascular arch detection frame and the optic disc in the region of interest is obtained respectively; based on the position information, a third distance between the center point of the main vascular arch detection frame and the center point of the region of interest, and a second distance between the center point of the optic disc and the center point of the region of interest are determined; if the third distance is less than the second distance and the main vascular arch and the optic disc are on the same side of the center point of the region of interest, then the eye position of the region of interest is determined to be centered on the macula; if the third distance is within a preset threshold range, then the eye position of the region of interest is determined to be centered on the macula; if the second distance is less than the preset threshold, then the eye position of the region of interest is determined to be centered on the optic disc.

2. The method of claim 1, wherein, Also includes: If the image quality score does not meet the preset threshold, the target fundus image is determined to be unqualified. Preset weights are obtained for the eye features and image features in the region of interest, respectively; and based on the preset weights, the eye features and / or image features that cause the target fundus image to be unqualified are sorted by priority and used as the reasons for the evaluation results.

3. The method according to claim 1, characterized in that, The step of generating an image scoring result by weighting the scoring results corresponding to the eye features and the scoring results corresponding to the image features based on preset scoring rules includes: Based on the preset scoring rules, determine the eye position score corresponding to the eye position judgment result, the eye type score corresponding to the eye type judgment result, the sharpness score corresponding to the average sharpness, the exposure score corresponding to the average exposure, and the resolution score corresponding to the image resolution judgment result. Based on preset weights, the eye position score, the eye-specific score, the sharpness score, the exposure score, and the resolution score are weighted to generate an image score result.

4. The method according to claim 1, characterized in that, Also includes: The intersection of the trajectory circle and the main blood vessel is determined as the first vertex of the vascular arch detection frame; wherein, the trajectory circle is used to indicate the arc formed with the center point of the visual disc as the center and a preset distance as the radius; Draw a perpendicular line along the Y-coordinate direction of the spatial coordinate system where the first vertex is located, and determine the intersection of the perpendicular line and the main blood vessel as the second vertex; Based on the first vertex, the second vertex, and the center point of the optic disc, the position information of the main vascular arch detection frame in the region of interest is determined.

5. The method according to claim 1, characterized in that, Also includes: If the eye identification result indicates that the region of interest is the right eye or the left eye, then the first eye identification score is output based on the eye identification preset scoring rules; If the eye identification result indicates that the eye identification of the region of interest is unknown, then a second eye identification score is output based on the preset eye identification scoring rules; wherein, the second eye identification score is less than the first eye identification score; If the eye position determination result indicates that the region of interest is the optic disc or macula, then a first eye position score is output based on the preset eye position scoring rules; if the eye position determination result indicates that the eye position of the region of interest is other categories, then a second eye position score is output based on the preset eye position scoring rules; wherein, the second eye position score is less than the first eye position score.

6. The method according to claim 1, characterized in that, The identification of the region of interest in the target fundus image includes: Obtain a standard color fundus image; The ordinary color fundus image is processed to obtain the recognition result; If the recognition result indicates that the ordinary color fundus image is the target fundus image, then the region of interest in the target fundus image is identified.

7. The method according to claim 1, characterized in that, Also includes: If the recognition result includes at most one eye structure feature, then the eye identification result is unknown; If the recognition result includes at most one eye structure feature or only the macula and blood vessels, then the eye position determination result is classified as other.

8. The method according to claim 1, characterized in that, When the image features include average sharpness, average exposure, and image resolution determination results, the step of extracting image features from the region of interest includes: Obtain the sharpness distribution map, exposure distribution map, and image resolution corresponding to the region of interest; The average sharpness of the region of interest is determined based on the sharpness distribution map. The average exposure of the region of interest is determined based on the exposure distribution map. The image resolution judgment result is determined based on the image resolution and the preset threshold.

9. A quality control device applied to the quality control method according to any one of claims 1-8, characterized in that, include: The recognition module is used to identify the region of interest in the target fundus image; An extraction module is used to extract eye features and image features from the region of interest, respectively; wherein, the eye features include at least eye position determination results and eye identification results; the image features include at least average sharpness, average exposure, and image resolution determination results; The generation module is used to generate an image quality score by weighting the score results corresponding to the eye features and the score results corresponding to the image features based on preset scoring rules and preset weights.

10. A computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any one of claims 1-8.

Images