CN111374632A - Retinopathy detection method, device and computer-readable storage medium - Google Patents
Retinopathy detection method, device and computer-readable storage medium Download PDFInfo
- Publication number
- CN111374632A CN111374632A CN201811642037.9A CN201811642037A CN111374632A CN 111374632 A CN111374632 A CN 111374632A CN 201811642037 A CN201811642037 A CN 201811642037A CN 111374632 A CN111374632 A CN 111374632A
- Authority
- CN
- China
- Prior art keywords
- detected
- picture
- retinopathy
- fundus
- fundus picture
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
- A61B3/10—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
- A61B3/10—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions
- A61B3/12—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for looking at the eye fundus, e.g. ophthalmoscopes
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
- A61B3/10—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions
- A61B3/14—Arrangements specially adapted for eye photography
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Ophthalmology & Optometry (AREA)
- Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Eye Examination Apparatus (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the application provides a retinopathy detection method, a retinopathy detection device and a computer-readable storage medium. The retinopathy detection method comprises the following steps: acquiring a fundus picture to be detected; inputting the fundus picture to be detected into a first neural network model, and carrying out quality judgment on the fundus picture to be detected; and when the quality of the fundus picture to be detected is qualified, inputting the fundus picture to be detected with qualified quality into a second neural network model for detecting retinopathy. The retinopathy detection method provided by the embodiment of the application can improve the detection efficiency of retinopathy of prematurity and can improve the detection precision.
Description
Technical Field
The present invention relates to the field of image recognition technologies, and in particular, to a method and an apparatus for detecting retinopathy, and a computer-readable storage medium.
Background
Retinopathy Of Prematurity (ROP) Of Prematurity is a vascular proliferative retinal disease affecting premature infants and infants with low birth weight, and the ROP incidence rate Of infants with the birth weight less than or equal to 1500g in China is 26.0%. Early screening and timely intervention are key factors in preventing ROP blindness, and the therapeutic window for ROP is small. Currently, a common method for fundus screening is to acquire a set of fundus image data of a neonate using a professional device and then diagnose the set of image data by a professional ophthalmologist, however ROP screening has many obstacles due to imbalance of medical resources in different geographical regions. First, medical equipment and personnel are lacking to conduct ROP examinations. Second, the training of ophthalmologists is not standard enough and qualified ophthalmologists are few. Third, developing nations do not fully implement the ROP screening policy. In less developed areas, many premature infants are blinded due to lack of timely screening and early treatment. Therefore, a more intelligent ROP detection method is needed.
Disclosure of Invention
The application provides a retinopathy detection method, which comprises the following steps:
acquiring a fundus picture to be detected;
inputting the fundus picture to be detected into a first neural network model, and carrying out quality judgment on the fundus picture to be detected;
and when the quality of the fundus picture to be detected is qualified, inputting the fundus picture to be detected with qualified quality into a second neural network model for detecting retinopathy.
The retinopathy detection method provided by the application comprises the steps of firstly randomly acquiring a certain number of fundus pictures to be detected, then inputting the fundus pictures to be detected into a first neural network model, wherein the first neural network model is used for preliminarily screening the quality of the fundus pictures to be detected, when the quality of the fundus pictures to be detected is qualified in preliminary screening, the fundus pictures to be detected with qualified quality are input into a second neural network model, and the second neural network model is used for detecting retinopathy of the characteristics displayed in the fundus pictures to be detected. The retinopathy detection method provided by the embodiment of the application can improve the efficiency of retinopathy detection of premature infants and is beneficial to improving the accuracy of detection.
The present application also provides a retinopathy detection device, which includes:
the acquisition module is used for acquiring a fundus picture to be detected;
the input judging module is used for inputting the fundus picture to be detected into a first neural network model and judging the quality of the fundus picture to be detected;
and the first input detection module is used for inputting the fundus picture to be detected with qualified quality into the second neural network model to detect the retinopathy when the quality of the fundus picture to be detected is qualified.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flowchart of a first retinopathy detection method provided by the embodiment of the present application.
Fig. 2 is a flowchart of a second retinopathy detection method provided by the embodiment of the present application.
Fig. 3 is a flowchart of a third retinopathy detection method provided by the embodiments of the present application.
Fig. 4 is a flowchart of a fourth retinopathy detecting method provided by the embodiment of the present application.
Fig. 5 is a flowchart of a fifth retinopathy detection method provided by the embodiment of the present application.
Fig. 6 is a flowchart of a sixth retinopathy detection method provided by the embodiment of the present application.
Fig. 7 is a schematic structural diagram of a first retinopathy detecting device provided by an embodiment of the present application.
Fig. 8 is a schematic structural diagram of a second retinopathy detecting device provided by the embodiment of the present application.
Fig. 9 is a schematic structural diagram of a third retinopathy detecting device provided by the embodiment of the present application.
Fig. 10 is a schematic structural diagram of a fourth retinopathy detecting device provided by the embodiment of the present application.
Fig. 11 is a schematic structural diagram of a fifth retinopathy detecting device provided in the embodiment of the present application.
Fig. 12 is a schematic structural diagram of a sixth retinopathy detecting device according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art without any inventive effort based on the embodiments in the present application are within the scope of protection of the present application.
Referring to fig. 1, fig. 1 is a flowchart illustrating a first retinopathy detecting method according to an embodiment of the present disclosure. In the present embodiment, the retinopathy detection method includes, but is not limited to, steps S100, S200, and S300, and the steps S100, S200, and S300 will be described in detail as follows.
S100: and acquiring a fundus picture to be detected.
The fundus refers to the posterior tissue in the eyeball, i.e., the inner membrane of the eyeball includes the retina, the optic papilla, the macula, and the central retinal artery and vein. The fundus picture is a picture including the characteristics of the inner membrane of the eyeball.
The fundus picture to be measured can be acquired in real time or can be shot in advance. The fundus picture to be detected can be one or more.
S200: and inputting the fundus picture to be detected into a first neural network model, and carrying out quality judgment on the fundus picture to be detected.
The basic idea of neural networks is to simulate a plurality of interconnected cells in the brain of a computer, so that the computer can learn from the environment and recognize different patterns. A basic neural network contains millions of artificial neurons called cells. The units are arranged in layers, each layer being interconnected. The unit is divided into three parts: the input unit is used for receiving information of an external environment; the hidden units are finally input to the output unit, and each hidden unit is a compressed linear function of the input of the hidden unit; an output unit, these signals representing how the network should respond to the most recently acquired information.
In the present embodiment, the input unit is configured to receive an input fundus picture to be measured. The output unit is used for responding to the fundus picture to be detected received by the input unit and responding to the neural network model. The first neural network model can be a VGG-16 convolutional neural network model and has a deep learning function.
When the neural network model is trained, or just starts running after training, different information patterns are fed into the network using different input units. This information will trigger the layers of the hidden group and then reach the output unit. This is known as a feed forward network and is one of the common designs. When a neural network is fully trained using a training model, it reaches a stage where it presents a completely new set of inputs that it does not encounter during the training stage and it can predict a satisfactory output.
The first neural network model is an image quality judging model and is used for judging whether the fundus picture to be detected meets the quality requirement. The quality requirement here mainly refers to whether it is convenient to judge retinopathy from the fundus picture to be measured. When the condition of retinopathy can be judged according to the fundus picture to be detected, the quality of the fundus picture to be detected is considered to meet the requirement. When the condition of retinopathy cannot be judged according to the fundus picture to be detected, the quality of the fundus picture to be detected is considered to be not in accordance with the requirement, and at this time, the picture may need to be processed and corrected or the fundus picture is considered to be replaced.
S300: and when the quality of the fundus picture to be detected is qualified, inputting the fundus picture to be detected with qualified quality into a second neural network model for detecting retinopathy.
Among them, Retinopathy (ROP) is classified into many types, and the most common types include retinal detachment, macular degeneration, eye trauma, diabetic retinopathy, endophthalmitis, intrabulbar foreign body, and congenital eye diseases such as neonatal retinopathy and intraocular parasites, and the most common retinal detachment is exemplified.
The second neural network model is a retinopathy prediction model and is used for predicting retinopathy of prematurity corresponding to the fundus pictures with qualified quality. The second neural network model can also be a VGG-16 convolutional neural network model and has a deep learning function.
When the quality of the fundus picture to be detected is qualified, the retinopathy condition can be judged according to the fundus picture to be detected, so that the fundus picture to be detected with qualified quality needs to be input into the second neural network model for detecting in order to predict whether the premature infant corresponding to the fundus picture has the retinopathy condition.
The protocol can free the pediatric ophthalmologist from a lengthy and time-consuming picture reading task for ROP screening. The first advantage is stability, since all pictures are analyzed in the same way, without involving subjectivity. While this work is the best job for pediatric ophthalmologists, humans are not always in an optimal state due to fatigue, mood and various subjective reasons. A second advantage is that once properly trained, no such errors occur. A third advantage is configurability. Based on different task requirements, corresponding classification strategies can be given, such as high-sensitivity configuration for ROP screening.
The retinopathy detection method provided by the application comprises the steps of firstly randomly acquiring a certain number of fundus pictures to be detected, then inputting the fundus pictures to be detected into a first neural network model, wherein the first neural network model is used for preliminarily screening the quality of the fundus pictures to be detected, when the quality of the fundus pictures to be detected is qualified in preliminary screening, the fundus pictures to be detected with qualified quality are input into a second neural network model, and the second neural network model is used for detecting retinopathy of the characteristics displayed in the fundus pictures to be detected. The retinopathy detection method provided by the embodiment of the application can improve the efficiency of retinopathy detection of premature infants and is beneficial to improving the accuracy of detection.
Referring to fig. 2, fig. 2 is a flowchart illustrating a second retinopathy detecting method according to an embodiment of the present disclosure. The second retinopathy detecting method is basically the same as the first retinopathy detecting method, except that in the present embodiment, the "S200: inputting the fundus picture to be measured into the first neural network model, and performing quality discrimination on the fundus picture to be measured "includes, but is not limited to, steps S210, S220, and S230, and details regarding steps S210, S220, and S230 are described below.
S210: judging whether the fundus picture to be detected has endobulbar features, wherein the endobulbar features comprise: one or more of the retina, the optic papilla, the macula, and the central retinal artery and vein.
The fundus picture to be detected can only comprise one of retina, optic papilla, macula lutea and central retinal artery and vein, and can also comprise a plurality of retinas, optic papilla, macula lutea and central retinal artery and vein.
When the quality of the fundus picture to be detected is judged, whether the fundus picture to be detected contains the endobulbar features or not is judged, only when the fundus picture to be detected contains the endobulbar features, whether the premature infant corresponding to the fundus picture to be detected has retinopathy or not can be detected, otherwise, if the fundus picture to be detected does not contain the endobulbar features at all, the subsequent processing of the fundus picture to be detected is meaningless, therefore, only when the fundus picture to be detected contains the endobulbar features, a subsequent series of operations need to be carried out on the fundus picture to be detected, and the workload is saved.
S220: when the fundus picture to be detected has the characteristics of the inner membrane of the eyeball, judging whether the fundus picture to be detected meets the recognizable requirements or not, wherein the recognizable requirements comprise: definition, picture visual angle and the proportion of the characteristics of the inner eye membrane in the fundus picture to be detected.
Specifically, when it is ensured that the fundus picture to be measured has the characteristics of the intraocular lens, it is necessary to further determine whether the fundus picture to be measured satisfies the recognizable requirements. And only when the characteristics of the inner eyeball membrane in the fundus picture to be detected are convenient to identify, whether the premature infant corresponding to the fundus picture to be detected has retinopathy can be judged.
Further, when the definition of the fundus picture to be detected reaches the preset definition, the intraocular lens characteristics in the fundus picture to be detected are considered to be convenient for clear observation, and at the moment, whether the premature infant corresponding to the fundus picture to be detected has retinopathy or not can be accurately predicted.
In addition, when the picture visual angle of the fundus picture to be detected is positioned at the normal observation visual angle, the more comprehensive characteristics of the inner membrane of the eyeball can be conveniently acquired, and at the moment, the condition whether the premature infant corresponding to the fundus picture to be detected has retinopathy or not can be accurately predicted.
Furthermore, when the proportion of the intraocular intima features in the fundus picture to be detected is within the preset range, the intraocular intima features can be conveniently and better extracted, and at the moment, the condition that whether the premature infant corresponding to the fundus picture to be detected has retinopathy can be accurately predicted.
It is understood that in other embodiments, the identifiable requirement may include whether the intraocular lens features are occluded, smeared, etc. in addition to the definition, the picture view angle, and the proportion of the intraocular lens features in the fundus picture to be tested.
S230: and when the fundus picture to be detected meets the recognizable requirement, judging that the quality of the fundus picture to be detected is qualified.
And when the fundus picture to be detected also meets the identification requirement on the basis of having the characteristics of the inner eyeball membrane of the premature infant, the quality of the fundus picture to be detected is considered to be qualified. At this time, relevant prediction work of retinopathy can be performed using fundus pictures of acceptable quality.
Referring to fig. 3, fig. 3 is a flowchart of a third retinopathy detecting method according to the embodiment of the present application. The third retinopathy detecting method is substantially the same as the first retinopathy detecting method, except that in the present embodiment, when the quality of the fundus picture to be measured is not acceptable, the retinopathy detecting method further includes, but is not limited to, steps S400 and S500, and the details regarding steps S400 and S500 are described below.
S400: and preprocessing the fundus picture to be detected so as to enable the fundus picture to be detected to meet the quality requirement.
Specifically, when the quality of the obtained fundus picture to be measured is not qualified, in order to avoid the need of re-obtaining the picture, a relatively intelligent method is to preprocess the fundus picture with unqualified quality, so that the fundus picture to be measured meets the quality requirement. For example, when the intraocular lens features in the fundus picture to be detected are shielded by foreign matters or smeared, the fundus picture to be detected can be subjected to sharpening treatment by means of an image processing technology, so that the fundus picture to be detected meets the quality requirement, and the accuracy of the retinopathy judgment result is improved.
S500: and inputting the pretreated qualified fundus picture to be detected into a second neural network model to detect retinopathy.
And further, inputting the qualified fundus picture to be detected after preprocessing into a second neural network model, and predicting retinopathy according to the extracted intraocular membrane characteristics on the fundus picture.
Referring to fig. 4, fig. 4 is a flowchart illustrating a fourth method for detecting retinopathy according to an embodiment of the present disclosure. The fourth retinopathy detecting method is substantially the same as the third retinopathy detecting method, except that in the present embodiment, the "S400: the preprocessing is performed on the fundus picture to be measured so that the fundus picture to be measured meets the quality requirement "includes, but is not limited to, steps S410, S420, and S430, and the details regarding steps S410, S420, and S430 are described below.
S410: and amplifying the fundus picture to be detected by a preset multiple.
Specifically, in order to facilitate viewing of the characteristics of the intraocular membrane in the fundus picture, the characteristics of the intraocular membrane in the fundus picture may be enlarged by a preset factor, so as to facilitate extraction and analysis of the characteristics of the intraocular membrane by the second neural network model.
S420: and correcting the characteristics of the inner eyeball membrane in the fundus picture to be detected after the preset magnification is amplified so as to enable the characteristics of the inner eyeball membrane to be in proper positions and angles in the fundus picture to be detected.
Further, in a preferred embodiment, in order to make the features of the inner eye membrane more completely and clearly present in the fundus picture to be tested and be in proper positions and angles, the features of the inner eye membrane in the fundus picture to be tested need to be corrected, so that the second neural network model can extract and analyze the features of the inner eye membrane.
S430: and removing the background image without the characteristics of the inner membrane of the eyeball from the fundus picture to be detected.
In order to improve the judgment of the second neural network model on whether the premature infant corresponding to the fundus picture to be detected has retinopathy, the background image from which the endobulbar features are removed from the fundus picture to be detected can be removed in advance, so that the second neural network model can rapidly extract the endobulbar features from the fundus picture, the solution time of the second neural network model for retinopathy is saved, and the efficiency of the second neural network model is improved.
Referring to fig. 5, fig. 5 is a flowchart illustrating a fifth retinopathy detecting method according to an embodiment of the present disclosure. The fifth retinopathy detecting method is basically the same as the first retinopathy detecting method, except that in the present embodiment, in the "S200: inputting the fundus picture to be detected into a first neural network model, and performing quality judgment on the fundus picture to be detected and the step S300: before inputting the fundus picture to be tested with qualified quality into the second neural network model for detecting retinopathy when the quality of the fundus picture to be tested is qualified, the retinopathy detecting method further comprises but is not limited to the steps S140, S160 and S180, and the detailed description about the steps S140, S160 and S180 is as follows.
S140: inputting a plurality of randomly acquired pictures into a first neural network model to train the first neural network model, so that the first neural network model has the capability of identifying the picture quality.
This is a process of training the first neural network model, in which the randomly acquired several pictures include fundus pictures of satisfactory quality and fundus pictures of unsatisfactory quality. And training a first neural network model by using a training set consisting of fundus pictures with satisfactory quality and fundus pictures with unsatisfactory quality, so that the first neural network model has the capability of judging the image quality.
S160: when the first neural network judges that the target picture meets the quality requirement, marking the target picture, wherein the marking comprises that the target picture has retinopathy and the target picture does not have retinopathy.
The process of supervised learning refers to a process of adjusting parameters of a classifier by using a group of samples of known classes to enable the classifier to achieve required performance, and is also called supervised training or teacher learning.
A supervised model is a form of learning that infers a particular function from previously labeled training data. It uses a supervised learning algorithm that contains a set of inputs with corresponding labeled correct outputs. The input of the tag is compared with the output of the tag. In view of the difference between the two, you can calculate an error value and then use an algorithm to learn the mapping between the input and the output. The final goal here is to approximate the mapping function, and if new input data is received, accurate output data can be predicted. Similar to the teacher supervising the learning process, the learning process stops when the algorithm achieves satisfactory performance or accuracy.
Specifically, a training set of tens of thousands of fundus pictures is preprocessed through a first neural network model, and pictures with unsatisfactory quality are removed. And then, the processed picture is marked for the ophthalmologist of the child, and finally, the second neural network model is trained through the marked picture, so that whether the second neural network model is accurate or not can be verified.
S180: inputting the marked target picture into a second neural network model to train the second neural network model, so that the second neural network model has the capability of judging whether the picture has retinopathy.
Again, this is the process of training the second neural network model. The icon picture is a picture which meets the quality requirement after being screened by the first neural network model, and the picture comprises marks added by children ophthalmologists.
Referring to fig. 6, fig. 6 is a flowchart illustrating a sixth method for detecting retinopathy according to an embodiment of the present disclosure. The sixth retinopathy detecting method is substantially the same as the first retinopathy detecting method, except that in the present embodiment, the "S300: when the quality of the fundus picture to be tested is qualified, inputting the fundus picture to be tested into the second neural network model for detecting the retinopathy includes, but is not limited to, the steps S310, S320, S330 and S340, and the details about the steps S310, S320, S330 and S340 are described below.
S310: and extracting the intraocular lens features in the fundus picture to be detected with qualified quality.
S320: and characterizing the characteristics of the inner eyeball membrane by adopting a vector to obtain a characterization vector.
And vectorizing the intraocular lens features to obtain a characterization vector for characterizing the intraocular lens features.
S330: comparing the characterization vector to a set of vectors in the second neural network model.
S340: and judging the retinopathy of the fundus picture to be detected by calculating the association degree between the characterization vector and the vector set.
Referring to fig. 7, fig. 7 is a schematic structural diagram of a first retinopathy detecting device according to an embodiment of the present disclosure. In the present embodiment, the retinopathy detecting device 10 includes, but is not limited to, an acquiring module 100, an input discriminating module 200, and a first input detecting module 310, and the following description is provided for the acquiring module 100, the input discriminating module 200, and the first input detecting module 310.
The acquisition module 100 is used for acquiring a fundus picture to be detected.
And the input judging module 200 is used for inputting the fundus picture to be detected into the first neural network model and judging the quality of the fundus picture to be detected.
The first input detection module 310 is configured to, when the quality of the fundus picture to be detected is qualified, input the fundus picture to be detected with qualified quality into the second neural network model for detecting retinopathy.
Referring to fig. 8, fig. 8 is a schematic structural diagram of a second retinopathy detecting device according to an embodiment of the present application. The structure of the second retinopathy detecting device 10 is substantially the same as that of the first retinopathy detecting device 10, except that in the present embodiment, the input discriminating module 200 includes, but is not limited to, a first judging module 210, a second judging module 220 and a judging module 230, and the first judging module 210, the second judging module 220 and the judging module 230 are described as follows.
A first determining module 210, configured to determine whether the fundus picture to be detected has an endobulbar feature, where the endobulbar feature includes: one or more of the retina, the optic papilla, the macula, and the central retinal artery and vein.
A second determining module 220, configured to determine whether the fundus picture to be detected meets identifiable requirements when the fundus picture to be detected has an intraocular membrane feature, where the identifiable requirements include: definition, picture visual angle and the proportion of the characteristics of the inner eye membrane in the fundus picture to be detected.
The determination module 230 is configured to determine that the quality of the fundus picture to be detected is qualified when the fundus picture to be detected meets the identifiable requirement.
Referring to fig. 9, fig. 9 is a schematic structural diagram of a third retinopathy detecting device according to an embodiment of the present application. The structure of the third retinopathy detecting device 10 is substantially the same as that of the first retinopathy detecting device 10, except that in the present embodiment, the retinopathy detecting device 10 further includes, but is not limited to, a preprocessing module 400 and a second input detecting module 320, which are described below with respect to the preprocessing module 400 and the second input detecting module 320.
The preprocessing module 400 is configured to preprocess the fundus picture to be detected, so that the fundus picture to be detected meets the quality requirement.
And the second input detection module 320 is used for inputting the preprocessed qualified fundus picture to be detected into the second neural network model to detect retinopathy.
Referring to fig. 10, fig. 10 is a schematic structural diagram of a fourth retinopathy detecting device according to an embodiment of the present application. The fourth retinopathy detecting device 10 has a structure substantially the same as that of the third retinopathy detecting device 10, except that, in the present embodiment, the preprocessing module 400 includes, but is not limited to, a magnifying module 410, a correcting module 420, and a removing module 430, which are described below with respect to the magnifying module 410, the correcting module 420, and the removing module 430.
And the amplifying module 410 is used for amplifying the fundus picture to be detected by a preset multiple.
A correcting module 420, configured to correct the endobulbar feature in the fundus picture to be detected after the preset magnification is amplified, so that the endobulbar feature is in a proper position and angle in the fundus picture to be detected.
And a removing module 430, configured to remove the background image with the endobulbar features removed from the fundus picture to be detected.
Referring to fig. 11, fig. 11 is a schematic structural diagram of a fifth retinopathy detecting device according to an embodiment of the present application. The fifth retinopathy detecting device 10 has a structure substantially the same as that of the first retinopathy detecting device 10, except that in the present embodiment, the retinopathy detecting device 10 further includes a first training module 510, a labeling module 520, and a second training module 530, which are described below with respect to the first training module 510, the labeling module 520, and the second training module 530.
The first training module 510 is configured to input a plurality of randomly acquired pictures into a first neural network model to train the first neural network model, so that the first neural network model has a capability of recognizing picture quality.
A marking module 520, configured to mark the target picture when the first neural network determines that the target picture meets the quality requirement, where the marking includes that the target picture has retinopathy and that the target picture does not have retinopathy.
The second training module 530 is configured to input the marked target picture into a second neural network model to train the second neural network model, so that the second neural network model has a capability of determining whether the picture has retinopathy.
Referring to fig. 12, fig. 12 is a schematic structural diagram of a sixth retinopathy detecting device according to an embodiment of the present application. The structure of the sixth retinopathy detecting device 10 is substantially the same as that of the first retinopathy detecting device 10, except that in the present embodiment, the first input detecting module 310 includes, but is not limited to, an extracting module 311, a characterizing module 312, an input comparing module 313 and a calculating and judging module 314, and the extracting module 311, the characterizing module 312, the input comparing module 313 and the calculating and judging module 314 are described as follows.
And the extraction module 311 is configured to extract the endobulbar features in the fundus picture to be detected, which are qualified in quality.
A characterization module 312, configured to characterize the intraocular lens features by using the vector to obtain a characterization vector.
An input alignment module 313 for aligning the characterization vector with a set of vectors in the second neural network model.
And the calculation and judgment module 314 is configured to judge retinopathy of the fundus picture to be detected by calculating the association degree between the characterization vector and the vector set.
The utility model provides a retinopathy detection device at first acquires the eyeground picture that awaits measuring of a certain amount at random, then inputs these eyeground pictures that await measuring into first neural network model, first neural network model is arranged in carrying out preliminary screening to the quality of the eyeground picture that awaits measuring, when the preliminary screening of the quality of the eyeground picture that awaits measuring is qualified, again with the qualified eyeground picture that awaits measuring of quality input to second neural network model in, second neural network model is arranged in carrying out the retinopathy to the characteristic that appears in the eyeground picture that awaits measuring and detects. The retinopathy detection method provided by the embodiment of the application can improve the efficiency of retinopathy detection of premature infants and is beneficial to improving the accuracy of detection.
The present invention also provides a computer-readable storage medium storing a computer program for retinopathy detection, wherein the computer program for retinopathy detection, when executed, performs the retinopathy detection method provided by any of the embodiments above.
Embodiments of the present application also provide a computer program product comprising a non-transitory computer readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps of any of the methods as set forth in the above-described cardiopulmonary resuscitation guidance method embodiments. The computer program product may be a software installation package and the computer comprises the retinopathy detecting means.
It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present application is not limited by the order of acts described, as some steps may occur in other orders or concurrently depending on the application. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required in this application.
In the foregoing embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus may be implemented in other manners. For example, the above-described embodiments of the apparatus are merely illustrative, and for example, a division of a unit is merely a logical division, and an actual implementation may have another division, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of some interfaces, devices or units, and may be an electric or other form.
Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable memory. Based on such understanding, the technical solution of the present application may be substantially implemented or a part of or all or part of the technical solution contributing to the prior art may be embodied in the form of a software product stored in a memory, and including several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method of the embodiments of the present application. And the aforementioned memory comprises: various media capable of storing program codes, such as a usb disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic disk, or an optical disk.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable memory, which may include: flash disks, Read-Only memories (ROMs), Random Access Memories (RAMs), magnetic or optical disks, and the like.
The foregoing detailed description of the embodiments of the present application has been presented to illustrate the principles and implementations of the present application, and the above description of the embodiments is only provided to help understand the method and the core concept of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Claims (13)
1. A retinopathy detection method characterized by comprising:
acquiring a fundus picture to be detected;
inputting the fundus picture to be detected into a first neural network model, and carrying out quality judgment on the fundus picture to be detected;
and when the quality of the fundus picture to be detected is qualified, inputting the fundus picture to be detected with qualified quality into a second neural network model for detecting retinopathy.
2. The retinopathy detection method according to claim 1, wherein the "inputting the fundus picture to be measured into the first neural network model, and performing quality discrimination on the fundus picture to be measured" includes:
judging whether the fundus picture to be detected has endobulbar features, wherein the endobulbar features comprise: one or more of the retina, the optic papilla, the macula, and the central retinal artery and vein;
when the fundus picture to be detected has the characteristics of the inner membrane of the eyeball, judging whether the fundus picture to be detected meets the recognizable requirements or not, wherein the recognizable requirements comprise: definition, picture visual angle and the proportion of the endobulbar features in the fundus picture to be detected;
and when the fundus picture to be detected meets the recognizable requirement, judging that the quality of the fundus picture to be detected is qualified.
3. The retinopathy detection method according to claim 1, wherein when the fundus picture to be measured is not of good quality, the retinopathy detection method further comprises:
preprocessing the fundus picture to be detected so as to enable the fundus picture to be detected to meet the quality requirement;
and inputting the pretreated qualified fundus picture to be detected into a second neural network model to detect retinopathy.
4. The retinopathy detection method of claim 3, wherein the "preprocessing the fundus picture to be detected so that the fundus picture to be detected meets quality requirements" includes:
amplifying the fundus picture to be detected by a preset multiple;
correcting the endobulbar features in the fundus picture to be detected after the preset times are amplified so as to enable the endobulbar features to be in proper positions and angles in the fundus picture to be detected;
and removing the background image without the characteristics of the inner membrane of the eyeball from the fundus picture to be detected.
5. The retinopathy detection method according to claim 1, wherein before the "inputting the fundus picture to be detected into a first neural network model and performing quality discrimination on the fundus picture to be detected" and the "inputting the fundus picture to be detected of qualified quality into a second neural network model and performing retinopathy detection when the quality of the fundus picture to be detected is qualified", the retinopathy detection method further comprises:
inputting a plurality of randomly acquired pictures into a first neural network model to train the first neural network model, so that the first neural network model has the capacity of identifying the picture quality;
when the first neural network judges that a target picture meets the quality requirement, marking the target picture, wherein the marking comprises that the target picture has retinopathy and the target picture does not have retinopathy;
inputting the marked target picture into a second neural network model to train the second neural network model, so that the second neural network model has the capability of judging whether the picture has retinopathy.
6. The retinopathy detection method according to claim 1, wherein the "inputting the fundus picture to be detected of qualified quality into a second neural network model for retinopathy detection when the quality of the fundus picture to be detected is qualified" includes:
extracting the intraocular lens features in the fundus picture to be detected with qualified quality;
characterizing the characteristics of the inner eyeball membrane by adopting vectors to obtain a characterization vector;
comparing the characterization vector to a set of vectors in the second neural network model;
and judging the retinopathy of the fundus picture to be detected by calculating the association degree between the characterization vector and the vector set.
7. A retinopathy detection device characterized by comprising:
the acquisition module is used for acquiring a fundus picture to be detected;
the input judging module is used for inputting the fundus picture to be detected into a first neural network model and judging the quality of the fundus picture to be detected;
and the first input detection module is used for inputting the fundus picture to be detected with qualified quality into the second neural network model to detect the retinopathy when the quality of the fundus picture to be detected is qualified.
8. The retinopathy detection device of claim 7, wherein the input discrimination module includes:
the first judgment module is used for judging whether the fundus picture to be detected has the endobulbar features, and the endobulbar features comprise: one or more of the retina, the optic papilla, the macula, and the central retinal artery and vein;
the second judging module is used for judging whether the fundus picture to be detected meets the identifiable requirements or not when the fundus picture to be detected has the characteristics of the inner eyeball membrane, and the identifiable requirements comprise: definition, picture visual angle and the proportion of the endobulbar features in the fundus picture to be detected;
and the judging module is used for judging that the quality of the fundus picture to be detected is qualified when the fundus picture to be detected meets the identifiable requirements.
9. The retinopathy detection device of claim 7, further comprising:
the preprocessing module is used for preprocessing the fundus picture to be detected so as to enable the fundus picture to be detected to meet the quality requirement;
and the second input detection module is used for inputting the preprocessed qualified fundus picture to be detected into the second neural network model to detect retinopathy.
10. The retinopathy detection device of claim 9, wherein the preprocessing module comprises:
the amplifying module is used for amplifying the fundus picture to be detected by a preset multiple;
the correction module is used for correcting the endobulbar features in the fundus picture to be detected after the preset times of amplification so as to enable the endobulbar features to be in proper positions and angles in the fundus picture to be detected;
and the removing module is used for removing the background image with the inner eyeball membrane characteristics in the fundus picture to be detected.
11. The retinopathy detection device of claim 7, further comprising:
the first training module is used for inputting a plurality of randomly acquired pictures into a first neural network model to train the first neural network model so that the first neural network model has the capacity of identifying the picture quality;
the marking module is used for marking a target picture when the first neural network judges that the target picture meets the quality requirement, wherein the marking comprises that the target picture has retinopathy and the target picture does not have retinopathy;
and the second training module is used for inputting the marked target picture into a second neural network model to train the second neural network model, so that the second neural network model has the capability of judging whether the picture has retinopathy.
12. The retinopathy detection device of claim 7, wherein the first input detection module comprises:
the extraction module is used for extracting the intraocular lens features in the fundus picture to be detected with qualified quality;
the characterization module is used for characterizing the intraocular lens features by adopting vectors to obtain characterization vectors;
an input comparison module for comparing the characterization vector with a set of vectors in the second neural network model;
and the calculation and judgment module is used for judging the retinopathy of the fundus picture to be detected by calculating the association degree between the characterization vector and the vector set.
13. A computer-readable storage medium storing a computer program for retinopathy detection, wherein the computer program for retinopathy detection when executed performs the retinopathy detection method of any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811642037.9A CN111374632B (en) | 2018-12-29 | 2018-12-29 | Retinopathy detection method, device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811642037.9A CN111374632B (en) | 2018-12-29 | 2018-12-29 | Retinopathy detection method, device and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111374632A true CN111374632A (en) | 2020-07-07 |
| CN111374632B CN111374632B (en) | 2023-05-26 |
Family
ID=71220602
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811642037.9A Active CN111374632B (en) | 2018-12-29 | 2018-12-29 | Retinopathy detection method, device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111374632B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112381821A (en) * | 2020-12-08 | 2021-02-19 | 北京青燕祥云科技有限公司 | Intelligent handheld fundus camera and image analysis method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104751186A (en) * | 2015-04-10 | 2015-07-01 | 山东师范大学 | Iris image quality classification method based on BP (back propagation) network and wavelet transformation |
| CN106530295A (en) * | 2016-11-07 | 2017-03-22 | 首都医科大学 | Fundus image classification method and device of retinopathy |
| CN106934798A (en) * | 2017-02-20 | 2017-07-07 | 苏州体素信息科技有限公司 | Diabetic retinopathy classification stage division based on deep learning |
| CN108231194A (en) * | 2018-04-04 | 2018-06-29 | 苏州医云健康管理有限公司 | A kind of disease diagnosing system |
| CN108470359A (en) * | 2018-02-11 | 2018-08-31 | 艾视医疗科技成都有限公司 | A kind of diabetic retinal eye fundus image lesion detection method |
| CN108577803A (en) * | 2018-04-26 | 2018-09-28 | 上海鹰瞳医疗科技有限公司 | Eye fundus image detection method based on machine learning, apparatus and system |
| CN108876775A (en) * | 2018-06-12 | 2018-11-23 | 广州图灵人工智能技术有限公司 | The rapid detection method of diabetic retinopathy |
-
2018
- 2018-12-29 CN CN201811642037.9A patent/CN111374632B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104751186A (en) * | 2015-04-10 | 2015-07-01 | 山东师范大学 | Iris image quality classification method based on BP (back propagation) network and wavelet transformation |
| CN106530295A (en) * | 2016-11-07 | 2017-03-22 | 首都医科大学 | Fundus image classification method and device of retinopathy |
| CN106934798A (en) * | 2017-02-20 | 2017-07-07 | 苏州体素信息科技有限公司 | Diabetic retinopathy classification stage division based on deep learning |
| CN108470359A (en) * | 2018-02-11 | 2018-08-31 | 艾视医疗科技成都有限公司 | A kind of diabetic retinal eye fundus image lesion detection method |
| CN108231194A (en) * | 2018-04-04 | 2018-06-29 | 苏州医云健康管理有限公司 | A kind of disease diagnosing system |
| CN108577803A (en) * | 2018-04-26 | 2018-09-28 | 上海鹰瞳医疗科技有限公司 | Eye fundus image detection method based on machine learning, apparatus and system |
| CN108876775A (en) * | 2018-06-12 | 2018-11-23 | 广州图灵人工智能技术有限公司 | The rapid detection method of diabetic retinopathy |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112381821A (en) * | 2020-12-08 | 2021-02-19 | 北京青燕祥云科技有限公司 | Intelligent handheld fundus camera and image analysis method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111374632B (en) | 2023-05-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3373798B1 (en) | Method and system for classifying optic nerve head | |
| Arbelaez et al. | Use of a support vector machine for keratoconus and subclinical keratoconus detection by topographic and tomographic data | |
| Kauppi et al. | The diaretdb1 diabetic retinopathy database and evaluation protocol. | |
| Chutatape | A model-based approach for automated feature extraction in fundus images | |
| JP4767570B2 (en) | Corneal shape analysis system | |
| Kälviäinen et al. | DIARETDB1 diabetic retinopathy database and evaluation protocol | |
| Huang et al. | Glaucoma detection using adaptive neuro-fuzzy inference system | |
| Aquino | Establishing the macular grading grid by means of fovea centre detection using anatomical-based and visual-based features | |
| Kauppi | Eye fundus image analysis for automatic detection of diabetic retinopathy | |
| Lesmana et al. | Abnormal condition detection of pancreatic beta-cells as the cause of diabetes mellitus based on iris image | |
| Chin et al. | Automatic fovea location in retinal images using anatomical priors and vessel density | |
| CN112869697A (en) | Judgment method for simultaneously identifying stage and pathological change characteristics of diabetic retinopathy | |
| CN113768461A (en) | Fundus image analysis method and system and electronic equipment | |
| CN111179258A (en) | Artificial intelligence method and system for identifying retinal hemorrhage image | |
| CN113066066A (en) | Retinal abnormality analysis method and device | |
| KR102220573B1 (en) | Method, apparatus and computer program for calculating quality score of fundus image data using artificial intelligence | |
| Kabir et al. | Multi-classification based Alzheimer's disease detection with comparative analysis from brain MRI scans using deep learning | |
| CN111374632B (en) | Retinopathy detection method, device and computer readable storage medium | |
| Santhosh et al. | Retinal Glaucoma Detection from Digital Fundus Images using Deep Learning Approach | |
| CN115661101A (en) | Premature infant retinopathy detection system based on random sampling and deep learning | |
| Purwanithami et al. | Hemorrhage diabetic retinopathy detection based on fundus image using neural network and FCM segmentation | |
| Quellec et al. | Instant automatic diagnosis of diabetic retinopathy | |
| US20240062378A1 (en) | Quality control method and quality control system for data annotation on fundus image | |
| Grałek et al. | Application of artificial intelligence in pediatric ophthalmic practice | |
| Anggraeni et al. | Detection of the emergence of exudate on the image of retina using extreme learning machine method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |