Eye fundus anomaly interpretation method and interpretation system based on eye fundus image and server

Abstract

The invention discloses a fundus image-based interpretation method and interpretation system for fundus abnormality and a server. The interpretation method comprises the following steps: dividing a fundus image into a negative result image and a positive result image based on a first judgment result of the fundus image obtained by utilizing a machine learning method; sampling the negative result image to obtain a sampled negative image; judging the quality control image comprising the sampling negative image and / or the positive result image to obtain a second judgment result or serve as a first to-be-reread image; sampling the positive result image to obtain a sampled positive image; judging the sampling positive image to obtain a third judgment result of the sampling positive image or taking the third judgment result as a second to-be-reread image; judging a first to-be-reread image and a second to-be-reread image which serve as to-be-reread images to obtain a fourth judgment result; and performing total detection on the eye fundus image based on each judgment result. Therefore, the accuracy of interpreting fundus anomalies in the fundus image can be improved.

Description

[0001] This application is a divisional application of a patent application with an application date of 2021-03-24, an application number of 2021103163443, and an invention name of a system and method for interpreting fundus abnormalities based on fundus images. Technical Field

[0002] The present disclosure generally relates to a method for interpreting fundus abnormalities based on fundus images, and a system and server for interpreting fundus abnormalities. Background Art

[0003] Medical images often contain many details of body structures or tissues. In modern hospitals, most treatment information comes from medical images such as fundus images. In clinical practice, understanding these details in medical images can help doctors identify related diseases. Medical images have developed into the main method for clinical disease identification. However, traditional disease information identification based on medical images mainly relies on professional physicians to make judgments based on experience. In this case, developing a system that can assist doctors in identifying related diseases has become a hot topic in the field of medical imaging. With the development of artificial intelligence technology, disease identification systems based on computer vision and artificial intelligence such as machine learning have been developed and applied in medical image recognition.

[0004] For example, patent document 1 (CN105513077A) discloses a system for screening diabetic retinopathy, which includes: a fundus image acquisition device, an image processing and screening device, and a report output device. The fundus image acquisition device is used to collect or receive fundus images of the person being examined; the image processing and screening device is used to process the fundus image and detect whether there are lesions therein, and then transmit the detection results to the report output device; the report output device outputs a corresponding detection report based on the detection results.

[0005] However, in actual clinical applications, due to the diversity of fundus images, the accuracy of different fundus images on the screening system described in Patent Document 1 is often inconsistent, resulting in a decrease in the accuracy of interpreting fundus abnormalities in fundus images. Summary of the invention

[0006] The present disclosure is proposed in view of the above-mentioned situation, and its object is to provide a fundus abnormality interpretation system and interpretation method based on fundus images that can improve the accuracy of interpreting fundus abnormalities in fundus images.

[0007] To this end, the first aspect of the present disclosure provides a fundus abnormality interpretation system based on fundus images, including an acquisition module, a first interpretation module, a first sampling module, a second interpretation module, a second sampling module, a third interpretation module, a fourth interpretation module and a general inspection and release module; the acquisition module is used to acquire fundus images; the first interpretation module judges whether there is a fundus abnormality in the fundus image from the acquisition module based on a deep learning method and outputs a first judgment result, the first judgment result including a negative result or a positive result, and the fundus image with the first judgment result being a negative result is used as a negative result image, and the fundus image with the first judgment result being a positive result is used as a positive result image; the first sampling module is used to sample the negative result image and use the extracted negative result image as a sampled negative image; the second interpretation module is used to receive the positive result image from the first interpretation module and / or the positive result image from the first interpretation module A sampling module samples a negative image and uses it as a quality control image, judges the quality control image to obtain a second judgment result of the quality control image or uses the quality control image as a first image to be reread, wherein the first image to be reread includes a quality control image that meets the major positive condition; the second sampling module is used to sample the positive result image and use the sampled positive result image as a sampled positive image; the third judgment module is used to receive the sampled positive image from the second sampling module and judge it to obtain a third judgment result of the sampled positive image or use the sampled positive image as a second image to be reread, wherein the second image to be reread includes the sampled positive image that meets the major positive condition; the fourth judgment module is used to receive the first image to be reread from the second judgment module and the second image to be reread from the third judgment module as images to be reread, and judge the images to be reread to obtain a fourth judgment result;And the general inspection and release module performs a general inspection on the fundus image based on the first judgment result, the second judgment result, the third judgment result or the fourth judgment result of the fundus image to release the final judgment result of the fundus image or re-judge the fundus image. The general inspection and release module includes a general inspection module, a rereading module and a release module. The general inspection module obtains the second judgment result of the negative result image within a preset time period and determines whether the first judgment result of the negative result image of the at least one dimension is qualified according to at least one dimension. If qualified, the first judgment result of the negative result image of the at least one dimension is used as the final judgment result. Otherwise, the rereading module is used to reset the negative result image of the at least one dimension so that the negative result image is qualified. The reading system re-judges the negative result image of the at least one dimension, the general inspection module also obtains the third judgment result of the positive result image within the preset time period and determines whether the second judgment result of the positive result image of the at least one dimension is qualified according to the at least one dimension. If qualified, the second judgment result of the positive result image of the at least one dimension is used as the final judgment result, otherwise, the positive result image of the at least one dimension is reset by the re-reading module so that the reading system re-judges the positive result image of the at least one dimension, the general inspection module also uses the third judgment result or the fourth judgment result as the final judgment result, and the release module is used to release the final judgment result. ;

[0008] In the present disclosure, the fundus image is divided into a negative result image and a positive result image by using the first judgment result of the first judgment module, the negative result image is sampled and the second judgment result of the extracted negative result image is obtained by using the second judgment module, and then the first judgment result of the negative result image is determined by the general inspection and release module, and the positive result image is first determined by using the second judgment module to obtain the second judgment result, and then the positive result image is sampled and the third judgment result of the extracted positive result image is obtained by using the third judgment module, and then the second judgment result of the positive result image is determined by using the general inspection and release module. The fundus images that meet the major positive conditions obtained by the second and third judgment modules are uniformly judged by the fourth judgment module. In this case, the fundus images are divided into three fundus images: negative result images, positive result images, and fundus images that meet the major positive conditions, and different judgment modules are used to perform targeted judgments on the three fundus images and sample the judgment modules. Thus, the accuracy of judging fundus abnormalities in fundus images can be improved.

[0009] In addition, in the interpretation system involved in the first aspect of the present disclosure, optionally, the interpretation system further includes a fifth interpretation module, the fifth interpretation module is used to confirm and judge the first judgment result of the first interpretation module, the second judgment result of the second interpretation module, the third judgment result of the third interpretation module, or the fourth judgment result of the fourth interpretation module to obtain a fifth judgment result, and the final judgment result is the fifth judgment result. In this way, the judgment results of each interpretation module can be uniformly confirmed.

[0010] In addition, in the interpretation system involved in the first aspect of the present disclosure, optionally, the interpretation system further includes an over-conversion module, the over-conversion module is used to convert the first judgment result of the fundus image whose first judgment result is a negative result into a positive result to match the requirement of the preset negative prediction rate; the over-conversion module is also used to set the first judgment result of the fundus image that meets the preset conditions as a positive result, and the preset conditions include at least one of the following conditions: the existence of a history of fundus disease, the corrected visual acuity is lower than the preset visual acuity, the history of diabetes or the course of diabetes is greater than or equal to the first preset number of years, the history of hypertension or the course of hypertension is greater than or equal to the second preset number of years, and recent visual acuity decline. In this case, by setting a higher preset negative prediction rate, the interpretation accuracy of the fundus image with a negative result can be improved, and by setting the first judgment result of the fundus image that meets the preset conditions as a positive result, it can be convenient to make subsequent judgments on the fundus image with a positive result.

[0011] In addition, in the interpretation system involved in the first aspect of the present disclosure, optionally, the general inspection module obtains the second judgment result of the negative result image within the preset time period and calculates the first sampling error rate according to the at least one dimension. If the first sampling error rate is greater than the first preset error rate, the first judgment result of the negative result image in the at least one dimension is determined to be unqualified; the general inspection module obtains the third judgment result of the positive result image within the preset time period and calculates the second sampling error rate according to the at least one dimension. If the second sampling error rate is greater than the second preset error rate, the second judgment result of the positive result image in the at least one dimension is determined to be unqualified. In this way, it is possible to determine whether the first judgment result is qualified based on the first sampling error rate and whether the second judgment result is qualified based on the second sampling error rate.

[0012] In addition, in the interpretation system involved in the first aspect of the present disclosure, optionally, the dimension includes at least one of the source of the fundus image, the model of the shooting device of the fundus image, the shooting person of the fundus image, the shooting quality of the fundus image, and the interpretation person information of the fundus image. In this case, the fundus images can be classified based on different dimensions and sampled separately for each category. Thereby, the reliability of sampling can be improved.

[0013] In addition, in the interpretation system involved in the first aspect of the present disclosure, optionally, in the sampling, the sampling scheme is adjusted according to the dimension or the release time of the interpretation system, thereby improving the reliability of the sampling.

[0014] In addition, in the interpretation system involved in the first aspect of the present disclosure, optionally, the significant positive condition is the presence of pathological features that affect or threaten vision. Thus, the significant positive condition can be determined and the image to be re-read can be acquired based on the significant positive condition.

[0015] In addition, in the interpretation system involved in the first aspect of the present disclosure, optionally, the lesion characteristics include a cup-disc ratio greater than or equal to a preset cup-disc ratio, anterior optic disc membrane, myelinated nerve fibers, other optic nerve abnormalities, cotton wool spots, hard exudates, retinal vascular sclerosis, 、 At least one of retinal vascular occlusion, leopard-shaped fundus, macular pigment abnormality, macular hole, drusen, hemorrhage, proliferative membrane, other retinal abnormalities, non-proliferative diabetic retinopathy, and proliferative diabetic retinopathy, and the lesion characteristics that affect vision or threaten vision include at least one of macular hole, retinal vascular occlusion, and proliferative diabetic retinopathy. Thus, it can be determined that there are lesion characteristics that affect vision or threaten vision.

[0016] In addition, in the interpretation system involved in the first aspect of the present disclosure, optionally, the interpretation system further includes a quality control module and a recall module. For the judgment result of the fundus image corresponding to the released final judgment result, the quality control module samples the first interpretation module, the second interpretation module, the third interpretation module or the fourth interpretation module based on the preset cycle and the at least one dimension to determine whether the judgment result of the at least one dimension is qualified, and the recall module is used to perform recall processing on the fundus image of the at least one dimension determined as unqualified by the quality control module. In this way, the accuracy of interpreting fundus abnormalities in fundus images can be further improved.

[0017] The second aspect of the present disclosure provides a method for interpreting fundus abnormalities based on fundus images, comprising: acquiring a fundus image; judging whether the fundus image has a fundus abnormality based on a deep learning method and outputting a first judgment result, wherein the first judgment result includes a negative result or a positive result, and taking the fundus image whose first judgment result is a negative result as a negative result image, and taking the fundus image whose first judgment result is a positive result as a positive result image; sampling the negative result image and taking the extracted negative result image as a sampled negative image; taking the positive result image and / or the sampled negative image as a quality control image, and testing the quality control image. The first image to be reread includes the quality control image that meets the major positive condition; the positive result image is sampled and the sampled positive result image is used as the sampled positive image; the sampled positive image is judged to obtain the third judgment result of the sampled positive image or the sampled positive image is used as the second image to be reread, the second image to be reread includes the sampled positive image that meets the major positive condition; the first image to be reread and the second image to be reread are used as the images to be reread, and the images to be reread are judged. to obtain a fourth judgment result; and based on the first judgment result, the second judgment result, the third judgment result or the fourth judgment result of the fundus image, a general inspection is performed on the fundus image to release the final judgment result of the fundus image or to re-judge the fundus image. In the general inspection, the second judgment result of the negative result image within the preset time period is obtained, and the first judgment result of the negative result image of the at least one dimension is determined to be qualified according to at least one dimension. If qualified, the first judgment result of the negative result image of the at least one dimension is used as the final judgment result, otherwise, the negative result image of the at least one dimension is reset to re-judge the negative result image of the at least one dimension. In the general inspection, the third judgment result of the positive result image within the preset time period is also obtained, and the second judgment result of the positive result image of the at least one dimension is determined to be qualified according to the at least one dimension. If qualified, the second judgment result of the positive result image of the at least one dimension is used as the final judgment result, otherwise, the positive result image of the at least one dimension is reset to re-judge the positive result image of the at least one dimension. In the general inspection, the third judgment result or the fourth judgment result is also used as the final judgment result.

[0018] In the present disclosure, the fundus image is divided into a negative result image and a positive result image using the first judgment result, the negative result image is sampled and the second judgment result of the extracted negative result image is obtained, and then the first judgment result of the negative result image is determined to be qualified, and the positive result image is first judged to obtain the second judgment result, and then the positive result image is sampled and the third judgment result of the extracted positive result image is obtained to determine whether the second judgment result of the positive result image is qualified, and the fundus images that meet the major positive conditions are uniformly judged. In this case, the fundus images are divided into three types of fundus images: negative result images, positive result images, and fundus images that meet the major positive conditions, and the three fundus images are targeted for judgment and sampling. In this way, the accuracy of judging fundus abnormalities in fundus images can be improved.

[0019] According to the present disclosure, it is possible to provide a system and method for judging fundus abnormalities based on fundus images, which improve the accuracy of judging fundus abnormalities in fundus images. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The present disclosure will now be explained in further detail by way of example only with reference to the accompanying drawings, in which:

[0021] Figure 1 is a diagram showing an application scenario of the method for interpreting fundus abnormalities based on fundus images involved in the example of the present disclosure.

[0022] Figure 2 is a block diagram showing a system for interpreting fundus abnormalities based on fundus images according to an example of the present disclosure.

[0023] Figure 3 is a schematic diagram showing a fundus image related to the example of the present disclosure.

[0024] Figure 4 is a block diagram showing a general inspection and release module for fundus abnormalities based on fundus images involved in the example of the present disclosure.

[0025] Figure 5 is a block diagram showing a system for interpreting fundus abnormalities based on fundus images according to an example of the present disclosure.

[0026] Figure 6 Detailed description is a flowchart showing a method for determining fundus abnormality based on fundus images according to an example of the present disclosure.

[0027] Figure 7 Detailed description is a flowchart showing a method for determining fundus abnormality based on fundus images according to an example of the present disclosure. DETAILED DESCRIPTION

[0028] Hereinafter, with reference to the accompanying drawings, the preferred embodiments of the present disclosure are described in detail. In the following description, the same symbols are given to the same parts, and repeated descriptions are omitted. In addition, the accompanying drawings are only schematic diagrams, and the ratio of the sizes of the parts to each other or the shapes of the parts, etc. may be different from the actual ones. It should be noted that the terms "including" and "having" in the present disclosure and any variations thereof, such as the processes, methods, systems, products or devices of a series of steps or units included or possessed are not necessarily limited to those steps or units clearly listed, but may include or have other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. All methods described in the present disclosure may be performed in any suitable order, unless otherwise indicated herein or clearly contradictory to the context.

[0029] The present disclosure relates to a fundus abnormality interpretation system 200 based on fundus images and a method for interpretation, which can improve the accuracy of interpretation of fundus abnormalities in fundus images. The fundus abnormality interpretation system 200 based on fundus images can sometimes be referred to as the interpretation system 200. The fundus abnormality interpretation method based on fundus images can sometimes be referred to as the interpretation method.

[0030] Figure 1 is a diagram showing an application scenario of the method for interpreting fundus abnormalities based on fundus images involved in the example of the present disclosure.

[0031] In some examples, the interpretation method (described later) can be applied to Figure 1 In the application scenario 100 shown. In the application scenario 100, the operator 110 can collect the fundus image of the fundus of the human eye 140 by controlling the acquisition device 130 connected to the terminal 120. After the acquisition device 130 completes the fundus image acquisition, the terminal 120 can submit the fundus image to the server 150 through the computer network. The server 150 can implement the interpretation method by executing the computer program instructions stored on the server 150. The interpretation method can receive the fundus image and generate a result report of the fundus image. The server 150 can return the generated result report of the fundus image to the terminal 120. In some examples, the terminal 120 can display the result report. In other examples, the result report can be stored in the memory of the terminal 120 or the server 150 as an intermediate result. In some examples, the result report can be displayed on the user side for the user to view, and the user can be a patient, for example. In other examples, the fundus image received by the interpretation method can be a fundus image stored in the terminal 120 or the server 150.

[0032] In some examples, the operator 110 may be a professional such as an ophthalmologist. In other examples, the operator 110 may be an ordinary person who has been trained in interpretation. The interpretation training may include but is not limited to the operation of the acquisition device 130 and the operation of the terminal 120 involving the interpretation method. In some examples, the terminal 120 may include but is not limited to a laptop, a tablet computer, or a desktop computer. In some examples, the acquisition device 130 may be a camera. The camera may be, for example, a handheld fundus camera or a desktop fundus camera. In some examples, the acquisition device 130 may be connected to the terminal 120 via a serial port. In some examples, the acquisition device 130 may be integrated in the terminal 120.

[0033] In some examples, the server 150 may include one or more processors and one or more memories. The processor may include a central processing unit, a graphics processing unit, and any other electronic component capable of processing data, and may execute computer program instructions. The memory may be used to store computer program instructions. In some examples, the server 150 may implement the interpretation method by executing the computer program instructions on the memory. In some examples, the server 150 may also be a cloud server.

[0034] The following is a detailed description of the interpretation system 200 of the present disclosure in conjunction with the accompanying drawings. The interpretation system 200 of the present disclosure is used to implement the above interpretation method. Figure 2 is a block diagram showing a system 200 for interpreting fundus abnormalities based on fundus images according to an example of the present disclosure.

[0035] like Figure 2As shown, in some examples, the interpretation system 200 may include an acquisition module 210, a first interpretation module 220, a first sampling module 230, a second interpretation module 240, a second sampling module 250, a third interpretation module 260, a fourth interpretation module 270 and a general inspection and release module 280. The acquisition module 210 can be used to acquire a fundus image. The first interpretation module 220 can judge the fundus image based on a deep learning method to obtain a first judgment result and divide the fundus image into a negative result image and a positive result image based on the first judgment result. The first sampling module 230 can be used to extract a negative result image and use it as a sampled negative image. The second interpretation module 240 can be used to judge a quality control image including a sampled negative image and / or a positive result image to obtain a second judgment result of the quality control image or use the quality control image as a first image to be reread. The second sampling module 250 can be used to extract a positive result image and use it as a sampled positive image. The third interpretation module 260 can be used to judge the sampled positive image to obtain the third judgment result of the sampled positive image or to use the sampled positive image as the second image to be reread. The fourth interpretation module 270 can be used to judge the fundus image that meets the major positive condition to obtain the fourth judgment result. The general inspection and release module 280 can perform a general inspection on the fundus image based on the judgment results of the fundus image in each interpretation module to release the final judgment result of the fundus image or re-judge the fundus image. In this case, the fundus image is divided into three types of fundus images: negative result image, positive result image and fundus image that meets the major positive condition. Different interpretation modules are used to perform targeted judgments on the three fundus images and sample the interpretation modules. In this way, the accuracy of judging fundus abnormalities in fundus images can be improved.

[0036] Figure 3 is a schematic diagram showing a fundus image related to the example of the present disclosure.

[0037] In some examples, the acquisition module 210 can be used to acquire fundus images. In some examples, the fundus images may include fundus images from the user's left eye and right eye. In some examples, the fundus images may be color fundus images. Color fundus images can clearly present rich fundus information such as the optic disc, optic cup, macula, blood vessels, etc. In addition, the fundus image may be an image in RGB mode, CMYK mode, Lab mode, or grayscale mode, etc. In some examples, the fundus image may be acquired by the acquisition device 130. In other examples, the fundus image may be a fundus image stored in the terminal 120 or the server 150. As an example of a fundus image, for example Figure 3 The fundus image of the human eye 140 is shown.

[0038] In some examples, after the acquisition module 210 acquires the fundus image, the fundus image may be preprocessed. In some instances, the preprocessing may include cropping the fundus image. Generally speaking, since the fundus image acquired by the acquisition module 210 may have different image formats or sizes, it is necessary to crop the fundus image to convert the fundus image into an image in a fixed standard form. The fixed standard form may refer to images with the same format and size. For example, in some examples, the size of the fundus image after preprocessing can be unified into a fundus image of 256×256, 374×374, 512×512, 768×768 or 1024×1024 pixels.

[0039] In addition, in some examples, preprocessing may include normalizing the fundus image. In some examples, normalization may include performing operations such as coordinate centering and scaling normalization on the fundus image. Thus, the differences between different fundus images can be overcome and the performance of the interpretation system 200 can be improved. In addition, in some examples, preprocessing may include performing noise reduction, grayscale processing, etc. on the fundus image. Thus, the features of the fundus image can be highlighted. In some examples, the fundus image may not need to be preprocessed, and the fundus image may be directly judged later.

[0040] In some examples, the first interpretation module 220 can be used to determine whether there is a fundus abnormality in the fundus image from the acquisition module 210. In some examples, the first interpretation module 220 can output a first judgment result. In some examples, the first judgment result can include a negative result or a positive result. In some examples, the first judgment result can also include the presence of a fundus abnormality.

[0041] In some examples, fundus abnormalities may include at least one of suspected glaucoma, optic nerve developmental abnormalities, optic nerve diseases, retinal vascular lesions, retinal vascular sclerosis, retinal vascular occlusion, leopard-shaped fundus, macular area abnormalities, macular holes, and diabetic retinopathy. Thus, multiple fundus abnormalities can be evaluated. However, the interpretation system 200 of the present disclosure does not limit the types of fundus abnormalities and can be easily extended to the evaluation of other fundus abnormalities. In some examples, different fundus abnormalities may correspond to different pathological characteristics.

[0042] In some examples, the first interpretation module 220 may use a machine learning method to judge the fundus image to obtain a first judgment result. In some examples, the machine learning method may be at least one of a traditional machine learning method and a deep learning method.

[0043] In some examples, the deep learning-based method may be based on a convolutional neural network (CNN). In this case, since the convolutional neural network (CNN) has the characteristic of high efficiency in image feature recognition, the performance of the interpretation system 200 can be effectively improved.

[0044] However, the examples disclosed herein are not limited thereto. In other examples, the machine learning method of the first interpretation module 220 may be a traditional machine learning method. In some examples, the traditional machine learning method may include, but is not limited to, a linear regression algorithm, a logistic regression algorithm, a decision tree algorithm, a support vector machine algorithm, or a Bayesian algorithm. In this case, the fundus features in the fundus image may be first extracted using an image processing algorithm, and then the fundus features may be input into a judgment model established based on a traditional machine learning method to realize judgment of the fundus image.

[0045] In some examples, the first interpretation module 220 can establish a judgment model based on a machine learning method, and the judgment model can include multiple sub-judgment models. Each sub-judgment model can be for each fundus abnormality. Each sub-judgment model can receive a fundus image and obtain a sub-judgment result. In some examples, the first interpretation module 220 can obtain a first judgment result based on multiple sub-judgment results. Thus, the first judgment result can be obtained based on multiple sub-judgment models.

[0046] Specifically, different sub-judgment models can be established for each fundus abnormality and trained to obtain a sub-judgment result (e.g., a negative result or a positive result) of whether there is a fundus abnormality and the confidence level corresponding to the sub-judgment result, and then the first judgment result can be obtained according to each sub-judgment result. For example, if there is no fundus abnormality, it is a negative result, otherwise it is a positive result.

[0047] In some examples, the first interpretation module 220 may use the fundus image whose first judgment result is a negative result as a negative result image, and use the fundus image whose first judgment result is a positive result as a positive result image.

[0048] In some examples, the interpretation system 200 may include an over-conversion module (not shown). In some examples, the first interpretation module 220 may use the over-conversion module to convert the first judgment result of the fundus image whose first judgment result is a negative result into a positive result, and divide the fundus image into a negative result image and a positive result image based on the converted first judgment result.

[0049] In some examples, the over-conversion module can be used to convert the first judgment result of the fundus image with a negative result into a positive result to match the requirement of the preset negative prediction rate. Generally speaking, the higher the preset negative prediction rate, the more sensitive the first interpretation module 220 is to the negative result, and the first judgment result of the fundus image with an uncertain negative result will be set to a positive result. In this case, by setting a higher preset negative prediction rate, the accuracy of the interpretation of the fundus image with a negative result can be improved. In some examples, the preset negative prediction rate can be 95% to 99%. For example, the preset negative prediction rate can be 95%, 96%, 97%, 98% or 99%, etc.

[0050] Generally speaking, the division of negative results and positive results by the over-conversion module can be based on confidence, so the negative prediction rate can be adjusted to match the requirements of the preset negative prediction rate by setting the confidence threshold. Specifically, the confidence threshold can be reversely solved according to the preset negative prediction rate. Since on the gold standard data, the sensitivity, specificity, positive prediction rate (positive prediction rate can be the number of true positives / (the number of true positives + the number of false positives)) and negative prediction rate (negative prediction rate can be the number of true negatives / (the number of true negatives + the number of false negatives)) and other performance indicators corresponding to each confidence threshold are determined. In some examples, all confidence thresholds can be traversed according to the preset step size and the relevant performance indicators can be solved. For example, the preset step size can be 0.01, 0.001 or 0.0001, etc. In this case, a table recording the performance indicators corresponding to each confidence threshold can be obtained, in which the confidence threshold corresponding to the preset negative prediction rate is the confidence threshold. In some examples, when calculating the confidence threshold, the corresponding relationship between the confidence threshold and the preset negative prediction rate can be calculated and a curve graph can be formed, and the confidence threshold corresponding to the preset negative prediction rate can be obtained according to the curve graph.

[0051] In some examples, the confidence threshold can be for each sub-judgment model. That is, based on the confidence threshold, the sub-judgment result of the fundus image with a negative sub-judgment result is converted to a positive result. For example, assuming that the sub-judgment result of a fundus image is that glaucoma does not exist (that is, a negative result) and the confidence is 90%, if the confidence threshold is 92%, since the confidence of the sub-judgment result is less than 92%, the sub-judgment result of the fundus image can be set to the presence of glaucoma. In this case, the confidence threshold can be applied to the sub-judgment results of each sub-judgment model to affect the first judgment result.

[0052] In some examples, the over-conversion module is also used to set the first judgment result of the fundus image that meets the preset conditions as a positive result. In some examples, the preset conditions may include at least one of the following conditions: the presence of a history of fundus disease, the corrected visual acuity is lower than the preset visual acuity, the history of diabetes or the course of diabetes is greater than or equal to the first preset number of years, the history of hypertension or the course of hypertension is greater than or equal to the second preset number of years, and recent visual acuity decline. Thus, it is convenient to make subsequent judgments on fundus images with positive results. In some examples, the preset visual acuity may be 0.6. In some examples, the first preset number of years may be 5 to 10 years. In some examples, the second preset number of years may be 5 to 10 years. In some examples, recent visual acuity decline may be a recent visual condition provided to the user, such as whether there is obvious blurred vision in the near future.

[0053] In some examples, the first sampling module 230 can be used to sample negative result images and use the extracted negative result images as sampled negative images. Specifically, the first sampling module 230 can sample negative result images according to a preset time (e.g., every hour) or a preset number of sheets (e.g., 100 sheets). In some examples, the extracted negative result images can be marked to distinguish the extracted negative result images and use the extracted negative result images as sampled negative images. In some examples, the negative result images that have not been extracted can be input into the general inspection and release module 280 (described later) for general inspection.

[0054] In some examples, fundus images can be sampled based on a sampling scheme. In some examples, the sampling scheme can include uniformly sampling fundus images. Specifically, a sampling method can be performed on all fundus images. In some examples, the sampling method can include, but is not limited to, random sampling, stratified sampling, overall sampling, or systematic sampling. In some examples, the sampling method can be based on national standards for sampling inspection. In some examples, national standards for sampling inspection can include, but are not limited to, GB / T2828.1-2012, GB / T2828.5-2011, or GB / T8052-2002.

[0055] In addition, in some examples, the sampling scheme may include classifying the fundus images according to one or more different dimensions, and sampling separately for each category (for example, several dimensions can be used as a category). In some examples, the dimensions may include at least one of the source of the fundus image, the model of the shooting device of the fundus image, the shooting personnel of the fundus image, the shooting quality of the fundus image, and the interpretation personnel information of the fundus image. In some examples, the judgment model established based on the machine learning method to judge the fundus image can be regarded as the interpretation personnel, and if there are multiple judgment models, multiple interpretation personnel can be corresponding. In this case, the fundus images can be classified based on different dimensions and sampled separately for each category. Thereby, the reliability of sampling can be improved.

[0056] In some examples, in sampling, the sampling scheme can be adjusted according to the release time of the dimension or the interpretation system 200. In some examples, the sampling scheme can have different degrees of sampling. In some examples, the degree of sampling can include general, special, strict or relaxed, etc. In some examples, different degrees of sampling can correspond to different sampling methods or different sampling parameters. In some examples, the sampling parameters can include but are not limited to sampling ratios or AQL (Acceptable Quality Limit) values, etc.

[0057] In some examples, a preliminary sampling ratio may be obtained, and the preliminary sampling ratio may be adjusted based on the degree of sampling to determine the sampling ratio. For example, if the degree of sampling is relaxed, the preliminary sampling ratio may be reduced to determine the sampling ratio.

[0058] Generally speaking, AQL refers to the worst average process quality level allowed when a series of batches are submitted for acceptance sampling. If the quality level is expressed as a percentage of unqualified products, the AQL value should not exceed 10%. If the quality level is expressed as the number of unqualified products per 100 units of fundus images to be sampled, the AQL value should not exceed 10. Different degrees of sampling can correspond to different AQL values.

[0059] In some examples, during sampling, the sampling plan can be adjusted according to the release time of the interpretation system 200. For example, the sampling degree of the interpretation system 200 that has just been released can be tightened, and after a period of operation, it is determined that the interpretation accuracy of the interpretation system 200 meets the requirements, the sampling degree can be adjusted to be relaxed. In some examples, during sampling, the sampling plan can be adjusted according to the dimension. For example, the sampling degree of fundus images interpreted by readers with lower seniority can be tightened.

[0060] In some examples, the second judgment module 240 can be used to receive the positive result image from the first judgment module 220 and the sampled negative image from the first sampling module 230. That is, the positive result image of the first judgment module 220 and the sampled negative image of the first sampling module 230 can be input into the second judgment module 240. In some examples, the positive result image of the first judgment module 220 and the sampled negative image of the first sampling module 230 can be used as quality control images.

[0061] In some examples, the second interpretation module 240 can judge the quality control image to obtain a second judgment result of the quality control image. The second judgment result can include a negative result or a positive result. In some examples, the second judgment result can also include the presence of fundus abnormalities. In some examples, the second interpretation module 240 can judge the quality control image to use the quality control image as the first image to be reread.

[0062] In some examples, the first image to be reread may include a quality control image that meets the major positive condition. In some examples, the major positive condition may be the presence of a lesion feature that affects vision or threatens vision. Thus, the major positive condition can be determined and the first image to be reread can be acquired based on the major positive condition. In some examples, the lesion features may include a cup-to-disc ratio greater than or equal to a preset cup-to-disc ratio, anterior optic disc membrane, myelinated nerve fibers, other optic nerve abnormalities, cotton wool spots, hard exudates, retinal vascular sclerosis, 、 At least one of retinal vascular occlusion, leopard-shaped fundus, macular pigment abnormality, macular hole, drusen, hemorrhage, proliferative membrane, other retinal abnormalities, non-proliferative diabetic retinopathy, and proliferative diabetic retinopathy. In some examples, the preset cup-to-disc ratio can be 0.6. In some examples, the pathological features that affect vision or threaten vision can include at least one of macular hole, retinal vascular occlusion, and proliferative diabetic retinopathy. Thus, it can be determined that there are pathological features that affect vision or threaten vision. However, the examples disclosed in the present invention are not limited to this. In other examples, it is possible to define compliance with major positive conditions according to actual conditions.

[0063] Specifically, the second interpretation module 240 can obtain and display the basic information of the quality control image, and the basic information may include but is not limited to the eye type (e.g., left eye or right eye), creation time, affiliated medical institution, shooting model, and current interpretation module (e.g., second interpretation module 240). After the interpretation personnel judge the quality control image based on the basic information and generate judgment information, the second interpretation module 240 can receive the judgment information of the interpretation personnel. If the quality control image does not meet the major positive condition, the second interpretation module 240 can generate a second judgment result of the quality control image based on the judgment information, otherwise the quality control image can be marked as the first image to be reread. In some examples, the basic information may also include medical history, vision information, gender, and age.

[0064] In some examples, the judgment information may include but is not limited to the existence of fundus abnormalities, image quality level of fundus images, existence of lesion characteristics, etc. In some examples, the image quality level may be five levels: very good, good, average, poor, and extremely poor.

[0065] In some examples, the second sampling module 250 can be used to sample the positive result images and use the extracted positive result images as sampled positive images. Specifically, the second sampling module 250 can sample the positive result images according to a preset time (e.g., every hour) or a preset number of sheets (e.g., 100 sheets). In some examples, the extracted positive result images can be marked to distinguish the extracted positive result images and use the extracted positive result images as sampled positive images. In some examples, the positive result images that have not been extracted can be input into the general inspection release module 280 (described later) for general inspection.

[0066] In some examples, the fundus images may be sampled based on the sampling scheme. For details, see the description of the sampling scheme in the first sampling module 230. In some examples, the second sampling module 250 may be used to sample the positive result images with the second judgment result and use the sampled positive result images as sampled positive images (i.e., positive result images that meet the major positive conditions are not sampled).

[0067] In some examples, the third interpretation module 260 can be used to receive the sampled positive image from the second sampling module 250. In some examples, the sampled positive image can be judged to obtain a third judgment result of the sampled positive image. The third judgment result may include a negative result or a positive result. In some examples, the third judgment result may also include the existence of fundus abnormalities. In some examples, the sampled positive image can be judged to use the sampled positive image as the second image to be reread. In some examples, the second image to be reread may include a sampled positive image that meets the major positive condition. In some examples, the seniority of the interpretation personnel of the third interpretation module 260 may be higher than the seniority of the interpretation personnel of the second interpretation module 240.

[0068] Specifically, the third judgment module 260 can obtain the above-mentioned basic information corresponding to the sampled positive image, wherein the current judgment module in the basic information can be the third judgment module 260. After the judgment personnel judge the sampled positive image based on the basic information and generate judgment information, the third judgment module 260 can receive the judgment information of the judgment personnel, and if the sampled positive image does not meet the major positive condition, the third judgment module 260 can generate a third judgment result of the sampled positive image based on the judgment information, otherwise the sampled positive image can be marked as the second image to be re-read.

[0069] In some examples, the fourth interpretation module 270 can be used to judge the fundus image that meets the major positive condition to obtain a fourth judgment result. In some examples, the fundus image that meets the major positive condition can be the first image to be reread from the second interpretation module 240 and the second image to be reread from the third interpretation module 260. In some examples, the fundus image that meets the major positive condition can be used as the image to be reread and uniformly judged by the fourth interpretation module 270. In some examples, the fourth interpretation module 270 can judge the image to be reread to obtain a fourth judgment result. In some examples, the fourth judgment result can include a negative result or a positive result. In some examples, the fourth judgment result can also include the existence of fundus abnormalities. In some examples, the seniority of the interpretation personnel of the fourth interpretation module 270 can be higher than the seniority of the interpretation personnel of the third interpretation module 260.

[0070] Specifically, the fourth interpretation module 270 may obtain the above basic information corresponding to the image to be reread, wherein the current interpretation module in the basic information may be the fourth interpretation module 270. After the interpretation personnel judge the image to be reread based on the basic information and generate judgment information, the fourth interpretation module 270 may receive the judgment information of the interpretation personnel and generate a fourth judgment result of the image to be reread based on the judgment information.

[0071] In some examples, the general inspection and release module 280 may perform a general inspection on the fundus image based on the judgment results of the fundus image in each interpretation module to release the final judgment result of the fundus image. In some examples, the general inspection and release module 280 may perform a general inspection on the fundus image based on the judgment results of the fundus image in each interpretation module to re-judge the fundus image. In some examples, the judgment results of each interpretation module may include the first judgment result of the first interpretation module 220, the second judgment result of the second interpretation module 240, the third judgment result of the third interpretation module 260, or the fourth judgment result of the fourth interpretation module 270.

[0072] Figure 4 is a block diagram showing a general inspection and release module 280 for fundus abnormalities based on fundus images involved in an example of the present disclosure.

[0073] like Figure 4 As shown, in some examples, the general inspection and release module 280 may include a general inspection module 281. The general inspection module 281 may be used to determine whether the determination result of the fundus image is qualified.

[0074] In some examples, the general inspection module 281 can be used to determine whether the first judgment result of the negative result image is qualified. As described above, the second interpretation module 240 can use the positive result image of the first interpretation module 220 and the sampled negative image of the first sampling module 230 as quality control images, and can judge the quality control image to obtain the second judgment result of the quality control image. Therefore, the sampled negative image has the second judgment result (that is, the extracted negative result image has the second judgment result). Specifically, the general inspection module 281 can obtain the second judgment result of the negative result image within a preset time period (for example, the same day, morning or afternoon), and determine whether the first judgment result of the negative result image of at least one dimension is qualified according to at least one dimension (for example, the information of the interpretation personnel). If qualified, the first judgment result of the negative result image of at least one dimension is used as the final judgment result, otherwise the negative result image of at least one dimension is reset so that the interpretation system 200 re-judges the negative result image of at least one dimension. In some examples, the negative result image of at least one dimension can be reset using the rereading module 282 (described later).

[0075] In some examples, it is determined whether the first judgment result of the negative result image of at least one dimension is qualified based on the first sampling error rate. Specifically, the general inspection module 281 can obtain the second judgment result of the negative result image within a preset time period, and calculate the first sampling error rate according to at least one dimension. If the first sampling error rate is greater than the first preset error rate, the first judgment result of the negative result image of at least one dimension is determined to be unqualified. Thus, it is possible to determine whether the first judgment result is qualified based on the first sampling error rate. In some examples, the first preset error rate can be set according to actual conditions.

[0076] In some examples, if the first judgment result of the negative result image is inconsistent with the second judgment result, the first judgment result of the negative result image is an erroneous judgment result.

[0077] In some examples, the general inspection module 281 can be used to determine whether the second judgment result of the positive result image is qualified. As described above, the sampled positive image can be judged to obtain the third judgment result of the sampled positive image. Therefore, the sampled positive image has the third judgment result (that is, the extracted positive result image has the third judgment result). Specifically, the general inspection module 281 can obtain the third judgment result of the positive result image within a preset time period (for example, the same day, morning or afternoon), and determine whether the second judgment result of the positive result image of at least one dimension is qualified according to at least one dimension (for example, the information of the judge), if qualified, then the second judgment result of the positive result image of at least one dimension is used as the final judgment result, otherwise the positive result image of at least one dimension is reset so that the reading system 200 re-judges the positive result image of at least one dimension. In some examples, the rereading module 282 (described later) can be used to reset the positive result image of at least one dimension.

[0078] In some examples, it is determined whether the second judgment result of the positive result image of at least one dimension is qualified based on the second sampling error rate. Specifically, the general inspection module 281 can obtain the third judgment result of the positive result image within the preset time period and calculate the second sampling error rate according to at least one dimension. If the second sampling error rate is greater than the second preset error rate, the second judgment result of the positive result image of at least one dimension is determined to be unqualified. Thus, it is possible to determine whether the second judgment result is qualified based on the second sampling error rate. In some examples, the second preset error rate can be set according to actual conditions.

[0079] In some examples, if the second judgment result of the positive result image is inconsistent with the third judgment result, the second judgment result of the positive result image is an erroneous judgment result.

[0080] In some examples, the first sampling error rate or the second sampling error rate may be the number of sampling errors divided by the total number of samples. As described above, the sampling scheme may include classifying fundus images according to one or more different dimensions and sampling separately for each category. In this case, the general inspection module 281 may obtain the total number of fundus images of each category, the total number of samples, and the number of sampling errors. The total number of samples may be the number of fundus images extracted from the fundus images of each category, and the number of sampling errors may be the number of fundus images corresponding to the erroneous judgment results in the fundus images of each category.

[0081] In some examples, the sampling scheme may be adjusted based on the sampling error rate. For example, if the sampling error rate corresponding to a certain dimension is low, the degree of sampling corresponding to the sampling scheme may be adjusted to be relaxed.

[0082] As described above, the general inspection module 281 can determine whether the first judgment result or the second judgment result is qualified and obtain the final judgment result. In some examples, the general inspection module 281 can also use the third judgment result as the final judgment result. In some examples, the general inspection module 281 can also use the fourth judgment result as the final judgment result.

[0083] In some examples, the general inspection and release module 280 may also include a re-read module 282 (see Figure 4 ). The rereading module 282 can be used to reset the fundus image corresponding to the unqualified judgment result so that the interpretation system 200 re-judges the fundus image corresponding to the unqualified judgment result. In some examples, resetting the fundus image corresponding to the unqualified judgment result can be resetting the relevant information of the unqualified fundus image (here refers to the fundus image corresponding to the judgment result of the unqualified fundus image) so that the acquisition module 210 can acquire the fundus image. For example, if the acquisition module 210 acquires a fundus image with a processing state to be interpreted, the processing state of the unqualified fundus image can be reset to to be interpreted. For another example, if the acquisition module 210 acquires the fundus image from a specific data set, the unqualified fundus image can be reinserted into the data set. For another example, if the acquisition module 210 acquires the fundus image by message, a message can be resent to notify the acquisition module 210 to acquire the unqualified fundus image.

[0084] In some examples, the general inspection release module 280 may also include a release module 283 (see Figure 4 ). In some examples, the release module 283 can be used to release the final judgment result. In some examples, the release module 283 can be used to display the final judgment result and release the result report.

[0085] Specifically, the release module 283 can obtain the above basic information of the fundus image, the final judgment result, the interpretation personnel information corresponding to the final judgment result, the result report and the interpretation details, and receive the release command from the release personnel to release the result report. In some examples, the interpretation details can include the judgment results of each interpretation module.

[0086] In some examples, the result report may include the final judgment result. In some examples, the released result report may be displayed on the user side for the user to view. In some examples, the result report may include user information, medical history, fundus image, image quality level of fundus image, comprehensive evaluation result, comprehensive suggestion, auditor information and details of fundus abnormality based on the final judgment result. Thus, the result report can be output. In some examples, the user information may include but is not limited to name, gender and age, etc. In addition, in some examples, the medical history may include but is not limited to history of diabetes, history of hypertension and history of eye diseases, etc. In addition, in some examples, the fundus image in the result report may include fundus image of the left eye and fundus image of the right eye. In addition, in some examples, the comprehensive evaluation result may include evaluation of the left eye and evaluation of the right eye. As an example of the comprehensive evaluation result, for example, the comprehensive evaluation result may be that there is no obvious abnormality in the fundus of the left eye, and the fundus of the right eye is visible. In some examples, the details of the fundus abnormality based on the final judgment result may include at least one of medical explanation, common causes and guidance suggestions.

[0087] In some examples, the interpretation system 200 also includes a fifth interpretation module (not shown). The fifth interpretation module can be used to confirm and judge each judgment result to obtain a fifth judgment result. In some examples, each judgment result may include a first judgment result of the first interpretation module 220, a second judgment result of the second interpretation module 240, a third judgment result of the third interpretation module 260, or a fourth judgment result of the fourth interpretation module 270. Specifically, the fifth interpretation module can be used to check each judgment result and confirm whether to retain the judgment result of the fundus image in the last interpretation module (for example, the last interpretation module of the negative result image that has not been extracted is the first interpretation module 220) and use it as the fifth judgment result. If not retained, re-judge to obtain the fifth judgment result. Thus, the judgment results of each interpretation module can be uniformly confirmed. In some examples, the fifth judgment result may include a negative result or a positive result. In some examples, the fifth judgment result may also include an existing fundus abnormality.

[0088] In some examples, the final judgment result may be the fifth judgment result. Specifically, if the general inspection module 281 determines that the first judgment result of the fundus image or the second judgment result of the fundus image is qualified, the general inspection module 281 may use the fifth judgment result corresponding to the qualified fundus image as the final judgment result. In some examples, the general inspection module 281 may directly use the fifth judgment result as the final judgment result.

[0089] Figure 5 is a block diagram showing a system 200 for interpreting fundus abnormalities based on fundus images according to an example of the present disclosure.

[0090] like Figure 5 As shown, in some examples, the interpretation system 200 may further include a quality control module 290 and a recall module 300. In some examples, for the judgment result of the fundus image corresponding to the final judgment result released by the release module 283, the quality control module 290 may sample each interpretation module (e.g., the first interpretation module 220, the second interpretation module 240, the third interpretation module 260, or the fourth interpretation module 270) based on a preset period (e.g., every month or every quarter) and at least one dimension to determine whether the judgment result of at least one dimension is qualified. The recall module 300 may be used to perform a recall process on the fundus image of at least one dimension that is determined to be unqualified by the quality control module 290. In some examples, the recalled fundus images may be fully inspected, the sampling ratio of the dimension corresponding to the recalled fundus images may be increased, and at least one of the software and hardware of the interpretation system 200 may be repaired. Thus, the accuracy of interpreting fundus abnormalities in fundus images can be further improved.

[0091] The following, combined Figure 6 The method for interpreting fundus abnormalities based on fundus images of the present disclosure is described in detail. The method for interpreting fundus abnormalities based on fundus images of the present disclosure may sometimes be referred to as an interpretation method. The interpretation method is applied to the above-mentioned interpretation system 200. Figure 6 Detailed description is a flowchart showing a method for determining fundus abnormality based on fundus images according to an example of the present disclosure.

[0092] In some examples, such as Figure 6As shown, the interpretation method may include acquiring a fundus image (step S110), judging the fundus image based on a deep learning method to obtain a first judgment result and dividing the fundus image into a negative result image and a positive result image based on the first judgment result (step S120), extracting the negative result image and using it as a sampled negative image (step S130), judging a quality control image including the sampled negative image and / or the positive result image to obtain a second judgment result of the quality control image or using the quality control image as a first image to be reread (step S140), extracting the positive result image and using it as a sampled positive image (step S150), judging the sampled positive image to obtain a third judgment result of the sampled positive image or using the sampled positive image as a second image to be reread (step S160), judging the fundus image that meets the major positive condition to obtain a fourth judgment result (step S170) and conducting a general inspection of the fundus image based on the various judgment results of the fundus image to release the final judgment result of the fundus image or re-judge the fundus image (step S180). In this case, the fundus images are divided into three types: negative result images, positive result images, and fundus images that meet the requirements of major positive results, and the three types of fundus images are judged and sampled in a targeted manner, thereby improving the accuracy of judging fundus abnormalities in fundus images.

[0093] In some examples, in step S110, a fundus image may be acquired. For a detailed description, see the related description of the acquisition module 210.

[0094] In some examples, in step S120, it is possible to judge whether there is a fundus abnormality in the fundus image based on a deep learning method and output a first judgment result. In some examples, the first judgment result may include a negative result or a positive result. In some examples, the fundus image with a negative first judgment result may be used as a negative result image, and the fundus image with a positive first judgment result may be used as a positive result image. For a specific description, see the relevant description of the first interpretation module 220.

[0095] In some examples, the first judgment result can be over-converted. In some examples, in the over-conversion, the first judgment result of the fundus image with a negative first judgment result can be converted to a positive result, and the fundus image can be divided into a negative result image and a positive result image based on the converted first judgment result. In some examples, the first judgment result of the fundus image with a negative first judgment result can be converted to a positive result to match the requirement of the preset negative prediction rate. In this case, by setting a higher preset negative prediction rate, the accuracy of the interpretation of the fundus image of the negative result can be improved. In some examples, in the over-conversion, the first judgment result of the fundus image that meets the preset conditions can be set as a positive result. In some examples, the preset conditions can include at least one of the conditions of having a history of fundus disease, corrected vision lower than the preset vision, having a history of diabetes or a history of diabetes greater than or equal to a first preset period of years, having a history of hypertension or a history of hypertension greater than or equal to a second preset period of years, and recent vision loss. Thus, it is convenient to make subsequent judgments on the fundus image of the positive result. For a specific description, see the relevant description of the over-conversion module.

[0096] In some examples, in step S130, negative result images can be sampled and the extracted negative result images can be used as sampled negative images. In some examples, the negative result images that have not been extracted can be transferred to step S180 (described later) for processing. In some examples, in sampling, the sampling scheme can be adjusted according to the release duration of the dimension or the interpretation system 200. Thus, the reliability of sampling can be improved. In some examples, the dimension can include at least one of the source of the fundus image, the shooting device model of the fundus image, the shooting personnel of the fundus image, the shooting quality of the fundus image, and the interpretation personnel information of the fundus image. In this case, the fundus image can be classified based on different dimensions and sampled separately for each category. Thus, the reliability of sampling can be improved. For a specific description, see the relevant description of the first sampling module 230.

[0097] In some examples, in step S140, the positive result image and the sampled negative image can be used as quality control images. In some examples, the quality control image can be judged to obtain a second judgment result of the quality control image or the quality control image can be used as the first image to be reread. In some examples, the first image to be reread may include a quality control image that meets a major positive condition. In some examples, the major positive condition may be the presence of pathological features that affect vision or threaten vision. Thus, it is possible to determine the major positive condition and obtain the image to be reread based on the major positive condition. In some examples, the pathological features may include a cup-to-disc ratio greater than or equal to a preset cup-to-disc ratio, anterior optic disc membrane, myelinated nerve fibers, other abnormalities of the optic nerve, cotton wool spots, hard exudates, retinal vascular sclerosis, etc. 、At least one of retinal vascular occlusion, leopard-shaped fundus, macular pigment abnormality, macular hole, drusen, hemorrhage, proliferative membrane, other retinal abnormalities, non-proliferative diabetic retinopathy, and proliferative diabetic retinopathy. In some examples, the lesion characteristics that affect vision or threaten vision may include at least one of macular hole, retinal vascular occlusion, and proliferative diabetic retinopathy. Thus, it can be determined that there are lesion characteristics that affect vision or threaten vision. For a specific description, see the relevant description of the second interpretation module 240.

[0098] In some examples, in step S150, the positive result image can be sampled and the extracted positive result image can be used as the sampled positive image. In some examples, the positive result image that has not been extracted can be transferred to step S180 (described later) for processing. For specific description, see the relevant description of the second sampling module 250.

[0099] In some examples, in step S160, the sampled positive image can be judged to obtain a third judgment result of the sampled positive image or the sampled positive image can be used as the second image to be reread. In some examples, the second image to be reread can include a sampled positive image that meets the major positive condition. For a specific description, see the relevant description of the third judgment module 260.

[0100] In some examples, in step S170, the first image to be reread and the second image to be reread may be used as the image to be reread. In some examples, the image to be reread may be judged to obtain a fourth judgment result. For a detailed description, see the related description of the fourth judgment module 270.

[0101] In some examples, in step S180, the fundus image may be subjected to a general inspection based on each of the judgment results of the fundus image to release the final judgment result of the fundus image. In some examples, the fundus image may be subjected to a general inspection based on each of the judgment results of the fundus image to re-judge the fundus image. In some examples, each of the judgment results may include a first judgment result, a second judgment result, a third judgment result, or a fourth judgment result.

[0102] In some examples, in the general inspection, the second judgment result of the negative result image within the preset time period can be obtained and the first judgment result of the negative result image of at least one dimension can be determined according to at least one dimension. If qualified, the first judgment result of the negative result image of at least one dimension can be used as the final judgment result, otherwise the negative result image of at least one dimension can be reset to re-judge the negative result image of at least one dimension. In some examples, whether the first judgment result of the negative result image of at least one dimension is qualified can be determined based on the first sampling error rate. Specifically, the second judgment result of the negative result image within the preset time period can be obtained, and the first sampling error rate can be calculated according to at least one dimension. If the first sampling error rate is greater than the first preset error rate, the first judgment result of the negative result image of at least one dimension can be determined to be unqualified. Thus, it is possible to determine whether the first judgment result is qualified based on the first sampling error rate. For a specific description, see the relevant description of the general inspection release module 280.

[0103] In some examples, in the general inspection, the third judgment result of the positive result image within the preset time period can also be obtained and the second judgment result of the positive result image of at least one dimension can be determined according to at least one dimension. If qualified, the second judgment result of the positive result image of at least one dimension can be used as the final judgment result, otherwise the positive result image of at least one dimension can be reset to re-judge the positive result image of at least one dimension. In some examples, in the general inspection, the third judgment result or the fourth judgment result can also be used as the final judgment result. In some examples, whether the second judgment result of the positive result image of at least one dimension is qualified can be based on the second sampling error rate. Specifically, the third judgment result of the positive result image within the preset time period can be obtained and the second sampling error rate can be calculated according to at least one dimension. If the second sampling error rate is greater than the second preset error rate, the second judgment result of the positive result image of at least one dimension can be determined to be unqualified. Thus, it is possible to determine whether the second judgment result is qualified based on the second sampling error rate. For a specific description, see the relevant description of the general inspection release module 280.

[0104] Figure 7 Detailed description is a flowchart showing a method for determining fundus abnormality based on fundus images according to an example of the present disclosure.

[0105] like Figure 7As shown, in some examples, the interpretation method may further include step S190. In step S190, the fundus image may be confirmed and judged based on each judgment result of the fundus image to obtain a fifth judgment result. In some examples, each judgment result may include a first judgment result, a second judgment result, a third judgment result, or a fourth judgment result. In some examples, the final judgment result may be the fifth judgment result. For a specific description, see the relevant description of the fifth interpretation module.

[0106] In some examples, for the judgment results of the fundus images corresponding to the final judgment results that have been released, the interpretation method can also sample each judgment result (e.g., the third judgment result or the fourth judgment result) based on a preset period (e.g., every month or every quarter) and at least one dimension to determine whether the judgment result of at least one dimension is qualified. In some examples, the interpretation method can also perform a recall process on the fundus images of at least one dimension that are judged to be unqualified. In some examples, the recalled fundus images can be fully inspected, the sampling ratio of the dimensions corresponding to the recalled fundus images can be increased, and at least one of the software and hardware of the interpretation system 200 can be repaired. In this way, the accuracy of interpreting fundus abnormalities in fundus images can be further improved. For a specific description, see the relevant description of the quality control module 290 and the recall module 300.

[0107] Although the present disclosure is specifically described above in conjunction with the accompanying drawings and examples, it is to be understood that the above description does not limit the present disclosure in any form. Those skilled in the art may modify and change the present disclosure as needed without departing from the essential spirit and scope of the present disclosure, and these modifications and changes all fall within the scope of the present disclosure.

Claims

1. A method for interpreting fundus abnormalities based on fundus images, characterized in that: include: Acquire fundus images; Judging the fundus image based on a machine learning method to obtain a first judgment result and dividing the fundus image into a negative result image and a positive result image based on the first judgment result; Sampling the negative result image and using the extracted negative result image as a sampled negative image; judging the quality control image including the sampled negative image and / or the positive result image to obtain a second judgment result of the quality control image or using the quality control image as a first image to be reread; Sampling the positive result image and using the sampled positive result image as a sampled positive image; The sampled positive image is judged to obtain a third judgment result of the sampled positive image or the sampled positive image is used as a second image to be reread; the first image to be reread and the second image to be reread are used as images to be reread, and the images to be reread are judged to obtain a fourth judgment result; and based on the first judgment result, the second judgment result, the third judgment result or the fourth judgment result of the fundus image, a general inspection of the fundus image is performed to release the final judgment result of the fundus image or to re-judge the fundus image.

2. The interpretation method according to claim 1, characterized in that: Also includes: The first judgment result of the fundus image with the first judgment result being a negative result is converted into a positive result to match the requirement of a preset negative predictive rate, and / or the first judgment result of the fundus image that meets preset conditions is set to a positive result, wherein the preset conditions include at least one of the following conditions: a history of fundus disease, a corrected visual acuity lower than a preset visual acuity, a history of diabetes or a history of diabetes with a duration greater than or equal to a first preset number of years, a history of hypertension or a history of hypertension with a duration greater than or equal to a second preset number of years, and a recent decrease in visual acuity.

3. The interpretation method according to claim 1, characterized in that: Also includes: The first judgment result, the second judgment result, the third judgment result or the fourth judgment result is checked to confirm whether the last judgment result of the fundus image is retained and used as the fifth judgment result. If not, a re-judgment is performed to obtain the fifth judgment result. The final judgment result is the fifth judgment result.

4. The interpretation method according to claim 1, characterized in that: The machine learning method is at least one of a traditional machine learning method and a deep learning method.

5. The interpretation method according to claim 1, characterized in that: Also includes: For the judgment result of the fundus image corresponding to the released final judgment result, the first judgment result, the second judgment result, the third judgment result or the fourth judgment result is sampled based on a preset period and at least one dimension to determine whether the judgment result of the at least one dimension is qualified, and recall processing is performed on the fundus image of the at least one dimension that is judged to be unqualified.

6. The interpretation method according to claim 1, characterized in that: At least one of the following processes is performed in the general inspection: obtaining a second judgment result of the negative result image within a preset time period and determining whether the first judgment result of the negative result image of the at least one dimension is qualified according to at least one dimension; if qualified, taking the first judgment result of the negative result image of the at least one dimension as the final judgment result; otherwise, resetting the negative result image of the at least one dimension to re-judge the negative result image of the at least one dimension; Obtaining a third judgment result of the positive result image within the preset time period and determining whether the second judgment result of the positive result image of the at least one dimension is qualified according to the at least one dimension; if qualified, taking the second judgment result of the positive result image of the at least one dimension as the final judgment result; otherwise, resetting the positive result image of the at least one dimension to re-judge the positive result image of the at least one dimension; The fourth judgment result is used as the final judgment result.

7. The interpretation method according to claim 1, characterized in that: The sampling inspection plan includes at least one of the following: unified sampling; classification according to one or more different dimensions, and separate sampling for each category.

8. The interpretation method according to claim 1, characterized in that: The first image to be reread includes a quality control image that meets the major positive condition; the second image to be reread includes a sampled positive image that meets the major positive condition; the major positive condition is the presence of pathological features that affect or threaten vision.

9. A fundus abnormality interpretation system based on fundus images, characterized in that: It comprises an acquisition module, a first judgment module, a first sampling module, a second judgment module, a second sampling module, a third judgment module, a fourth judgment module and a general inspection and release module; the acquisition module is used to acquire fundus images; the first judgment module is used to judge the fundus images from the acquisition module by using a machine learning method to obtain a first judgment result and divide the fundus images into negative result images and positive result images based on the first judgment result; The first sampling module is used to sample the negative result image and use the extracted negative result image as a sampled negative image; the second judgment module is used to judge the quality control image including the sampled negative image and / or the positive result image to obtain a second judgment result of the quality control image or use the quality control image as the first image to be reread; the second sampling module is used to sample the positive result image and use the extracted positive result image as a sampled positive image; The third judgment module is used to judge the sampled positive image to obtain a third judgment result of the sampled positive image or use the sampled positive image as a second image to be reread; the fourth judgment module is used to judge the first image to be reread as the image to be reread and the second image to be reread to obtain a fourth judgment result; The general inspection and release module performs a general inspection on the fundus image based on the first judgment result, the second judgment result, the third judgment result or the fourth judgment result of the fundus image to release the final judgment result of the fundus image or re-judge the fundus image.

10. A server, characterized in that: Computer program instructions are stored, and when the computer program instructions are executed, the judgment method according to any one of claims 1 to 8 is implemented.

Images