CN109658385A - Eye fundus image judgment method and equipment - Google Patents
Eye fundus image judgment method and equipment Download PDFInfo
- Publication number
- CN109658385A CN109658385A CN201811408311.6A CN201811408311A CN109658385A CN 109658385 A CN109658385 A CN 109658385A CN 201811408311 A CN201811408311 A CN 201811408311A CN 109658385 A CN109658385 A CN 109658385A
- Authority
- CN
- China
- Prior art keywords
- eye fundus
- fundus image
- image
- judged
- difference
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
- G06T7/0014—Biomedical image inspection using an image reference approach
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20224—Image subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
Abstract
The present invention provides a kind of eye fundus image judgment method and equipment, and the eye fundus image judgment method includes the eye fundus image to be judged for obtaining user；Acquisition is associated with the eye fundus image to be judged to refer to eye fundus image, wherein it is normal eye fundus image that the reference eye fundus image, which is known state,；Difference image is generated according to the eyeground figure to be judged and the reference eye fundus image；Using the difference image as the input data of machine learning model, make the machine learning model output to the judging result of the eye fundus image to be judged, wherein the machine learning model is trained using training difference image and corresponding label data.
Description
Technical field
The present invention relates to medical image process fields, and in particular to a kind of eye fundus image judgment method and equipment.
Background technique
As image obtains the high speed development with memory technology, deep learning, so that deep learning is deep into each of society
A field.In medical field, due to China's medical resource wretched insufficiency, by cutting edge technologies such as deep learnings to medical imaging
Detecting screening becomes the field that research is very popular now.Diabetes mellitus in China patient has broken through 1.3 hundred million, wherein diabetes view
About 30,000,000 people of film lesion patient numbers.Diabetic retinopathy causes the probability of blindness very high after deteriorating.If
Disease is netted to sugar in early stage and carries out screening, diagnosing and treating intervention, that blindness's relative risk can decline 94.4%.
It is existing mainly to pass through the eye fundus image for collecting a large amount of real scenes based on ophthalmoscopic image disease detecting system, lead to
The intersection mark for crossing multiple Medical Technologist, obtains the data largely marked, by training data, passes through the neural network of deep layer, instruction
Practise classification/detection model.The eye fundus image of a unknown state can then be judged using trained model, be obtained
Corresponding tag along sort and lesion.
Since there may be differences for everyone eye fundus image of normal condition, the eye fundus image of same disease is caused to change
It is bigger, however for certain distant genius morbis, such as slight sugared net, small glass-film wart, slight artery sclerosis etc.,
It can be because the feature suppression of these diseases be gone down in the diversity variation of eye fundus image.Make so existing based on eye fundus image
Detection scheme it is poor for the detection effect of some tiny characteristic points.
Summary of the invention
In view of this, the present invention provides a kind of eye fundus image judgment method, comprising:
Obtain the eye fundus image to be judged of user；
Acquisition is associated with the eye fundus image to be judged to refer to eye fundus image, wherein the reference eye fundus image is
Know that state is normal eye fundus image；
Difference image is generated according to the eyeground figure to be judged and the reference eye fundus image；
Using the difference image as the input data of machine learning model, make the machine learning model output to described
The judging result of eye fundus image to be judged, wherein the machine learning model is to utilize training difference image and corresponding number of tags
According to what is be trained.
Optionally, the eye fundus image to be judged and it is described with reference to eye fundus image be the eye fundus image from same people,
Described in reference to eye fundus image acquisition time earlier than the eye fundus image to be judged acquisition time.
Optionally, the eye fundus image to be judged and it is described with reference to eye fundus image be the eye fundus image from different people,
Wherein acquisition time of the acquisition time with reference to eye fundus image earlier than the eye fundus image to be judged.
Optionally, the acquisition is associated with the eye fundus image to be judged refers to eye fundus image, comprising:
Judge whether to prestore the reference eye fundus image of the user；
When there not being the reference eye fundus image of the user, obtain preset with reference to eye fundus image.
Optionally, the eyeground figure to be judged according to and the reference eye fundus image generate difference image, comprising:
Optic disk and/or macula lutea are identified in the eyeground figure to be judged and the reference eye fundus image respectively；
It by eyeground the to be judged figure and described is carried out pair with reference to eye fundus image according to the position of the optic disk and/or macula lutea
Together；
Based on eyeground figure to be judged described in after alignment and described Difference Calculation is carried out with reference to eye fundus image obtain the difference
Partial image.
The present invention also provides a kind of eye fundus image judgment models training methods, comprising:
Obtain multiple images pair being made of the first eye fundus image and the second eye fundus image, first eye fundus image and
The acquisition time of two eye fundus images is not identical, and the First view base map seems normal condition, and second eye fundus image is just
Normal state or abnormality；
Respectively according to each image to generation difference image；
Respectively using the difference image as sample data, using the state of second eye fundus image as the sample number
According to label initial machine learning model is trained so that the machine learning model can be known according to the difference image
The state of not described second eye fundus image.
Optionally, first eye fundus image and the second eye fundus image are the eye fundus images from same people, wherein described
Acquisition time of the acquisition time of first eye fundus image earlier than second eye fundus image.
It is optionally, described respectively according to each image to generation difference image, comprising:
Optic disk and/or macula lutea are identified in described image pair respectively；
Respectively according to the position of the optic disk and/or macula lutea by described image to being aligned；
Image after being based respectively on alignment obtains the difference image to Difference Calculation is carried out.
Correspondingly, the present invention also provides a kind of electronic equipment, comprising: at least one processor；And with described at least one
The memory of a processor communication connection；Wherein, the memory is stored with the instruction that can be executed by one processor, institute
It states instruction to be executed by least one described processor, so that at least one described processor executes above-mentioned eye fundus image judgement side
Method.
Correspondingly, the present invention also provides another electronic equipments, comprising: at least one processor；And with it is described at least
The memory of one processor communication connection；Wherein, the memory is stored with the instruction that can be executed by one processor,
Described instruction is executed by least one described processor, so that at least one described processor executes above-mentioned eye fundus image and judges mould
Type training method.
The eye fundus image judgment method and equipment provided according to the present invention utilizes eyeground figure and known health to be judged
Difference image is generated with reference to eye fundus image, tiny characteristic content can be embodied by this difference image, then using artificial
Intellectual technology is based on big data and judges difference image, the state of eyeground figure to be judged is determined with this, it is possible thereby to mention
The accuracy that height judges eye fundus image state can obtain preferable judgement especially for certain unconspicuous genius morbis
As a result.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, is also possible to obtain other drawings based on these drawings.
Fig. 1 is the flow chart of eye fundus image judgment models training method provided in an embodiment of the present invention；
Fig. 2 is the eye fundus image of a normal condition；
Fig. 3 is the eye fundus image of another normal condition；
Fig. 4 is the eye fundus image of third normal condition；
Fig. 5 is the eye fundus image of an abnormality；
Fig. 6 is the difference image generated using image shown in Fig. 4 and Fig. 5；
Fig. 7 is the flow chart of eye fundus image judgment method provided in an embodiment of the present invention.
Specific embodiment
Technical solution of the present invention is clearly and completely described below in conjunction with attached drawing, it is clear that described implementation
Example is a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill
Personnel's every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
(classification, detection) is judged in order to the state to eye fundus image, and the embodiment of the invention provides a kind of eyes
As judgment models training method, the model trained can be used for identifying eye fundus image base map.This method can be by computer
Or server executes, this method comprises the following steps as shown in Figure 1:
S11, obtains multiple images pair being made of the first eye fundus image and the second eye fundus image, the first eye fundus image and the
The acquisition time of two eye fundus images is not identical, and First view base map seems normal (health) state, and the second eye fundus image is normal
State or abnormal (disease) state.
First eye fundus image and the second eye fundus image can be the image as shown in Figure 2 acquired by fundus camera.
A large amount of eye fundus image can be acquired first, by being manually labeled to these eye fundus images, namely determine their state, example
It such as can only be divided into health and the two abnormal states, or detailed mark, example further can also be carried out to abnormality
Such as certain morbid state.
Then the eye fundus image of these known states can be matched, so that each image pair at least one is just
The eye fundus image of normal state.Obtained image is to may be two kinds of situations: two normal eye fundus images, a normal conditions
Eye fundus image and an abnormality eye fundus image.
Fig. 2, it is shown in Fig. 3 be normal eye fundus image, it is opposite that the two images can form a figure；It is shown in Fig. 4
Be normal eye fundus image figure, it is shown in fig. 5 be abnormal fundus image, it is opposite that the two images can form another figure.Fig. 5
In include an abnormal area 51, dotted line frame therein is intended merely to clearly demonstrate made aid mark, in actual conditions
In actual conditions and this dotted line frame is not included.
S12, respectively according to each image to generation difference image.It is poor for two eye fundus images of a specific user
Point calculate can only focus on optic disk, macula lutea, train of thought blood vessel feature difference, and ignore it is certain due to external environment, acquisition equipment
Caused difference (such as brightness).
The process of Difference Calculation, which can be, to be first passed through optic disk and macula lutea positioning two opens one's eyes base map alignment, determines that two open one's eyes bottom
The discrepancy of image or region, then by these discrepancys or zone marker in the second eye fundus image.In order to enhance difference
As a result, the second colored eye fundus image can be converted to gray level image, by discrepancy or zone marker in gray level image.
Thus obtained difference image can embody the difference of this two images, these differences can be subtle region
Even pixel is also possible to apparent region.Since Fig. 2 and Fig. 3 are normal eye fundus image, the difference of the two is very
Small, the difference content marked in Fig. 3 is some pixels, and human eye even can not observe these contents；And Fig. 4 and figure
Their difference compared to there are apparent difference contents, is marked the difference image shown in available Fig. 6 in Fig. 5, wherein wrapping by 5
Include diff area 61.The size and shape of diff area 61 and abnormal area 51 may be not fully identical.
S13, respectively using difference image as sample data, using the state of the second eye fundus image as the label of sample data
Initial machine learning model is trained, so that machine learning model can identify the second eye fundus image according to difference image
State.Such as using the state of image shown in Fig. 5 as the label data of difference image shown in Fig. 6.
First eye fundus image of image pair must be normal condition, and the second eye fundus image may be normal condition or
Abnormality, or specific abnormal conditions (such as disease category).The each training number inputted to initial machine learning model
It needs to include sample data and corresponding label data in, the present embodiment, should by generating using difference image as sample data
The state of second image used in difference image is as label training initial machine learning model, according to the training condition of setting
Obtain eye fundus image judgment models.
The eye fundus image judgment models that training obtains can will judge the shape of a wherein image according to eyeground difference image
State (or being interpreted as classifying with output label information to difference image).
Machine learning model for example can be support vector machines, neural network model, deep learning model, such as can adopt
With inception resnet model.
The eye fundus image judgment models training method provided according to embodiments of the present invention, utilizes the First view by known state
The image of base map picture and the second eye fundus image composition can embody small spy by this difference image to difference image is generated
Content is levied, then initial machine learning model is trained using a large amount of difference image, makes machine learning to a large amount of numbers
According to study is carried out to obtain eye fundus image judgment models, thus obtained model has the judgement of eye fundus image state higher
Accuracy, preferable judging result can be obtained especially for certain unconspicuous genius morbis.
Although the eye fundus image of different health (normal) people is generally there is no larger difference, machine is carrying out difference
Still some contents unrelated with abnormal (lesion) may be judged as difference content when calculating and be embodied in difference image, this can
It can reduce the recognition accuracy of machine learning model.As a preferred embodiment, scheme in opposite work generating,
It is preferable to use two eye fundus images from same people, and the acquisition time of the first eye fundus image therein is earlier than the second eyeground
The acquisition time of image, such as acquisition time can differ some months or several years.
Due to for same people, difference caused by non-disease (exception) reason will very little, this to be subsequently generated
There is a possibility that obvious advantage, the difference content embodied in difference image is lesion bigger in the operation of difference image.
The recognition accuracy of machine learning model will be further increased when using these training datas.
Nevertheless, the effect for still thering are some factors to will affect production difference image, such as eye fundus image common at present
Acquisition equipment be by manual operation, meanwhile, people repeatedly shooting eye fundus image when may not use identical equipment, these
Factor offsets from each other the position that will lead to two eye fundus images.In response to this, can specifically take in step s 12 as
Lower processing mode:
S121 identifies optic disk and/or macula lutea in image pair respectively, the step for the mode of artificial intelligence also can be used
It is identified, can also be identified by the way of machine vision based on the linear feature in image；
S122, respectively according to the position of optic disk and/or macula lutea by image to being aligned；
S123, the image after being based respectively on alignment obtain difference image to Difference Calculation is carried out.
Above-mentioned preferred embodiment, to alignment operation is carried out, avoids leading since eye fundus image location of content deviates for all images
Thus the Difference Calculation mistake of cause improves the quality of generated difference image, and then the identification for improving machine learning model is accurate
Property.
Correspondingly, the embodiment of the invention also provides a kind of electronic equipment, comprising: at least one processor；And with institute
State the memory of at least one processor communication connection；Wherein, the memory, which is stored with, to be executed by one processor
Instruction, described instruction executed by least one described processor, so that at least one described processor executes above-mentioned eyeground figure
As judgment models training method.
The embodiment of the invention also provides a kind of eye fundus image judgment models training device, which can be hardware composition
Device or the corresponding virtual bench of computer program, be also possible to program and device that hardware combines, which includes:
Module is obtained, for obtaining multiple images pair being made of the first eye fundus image and the second eye fundus image, described the
The acquisition time of one eye fundus image and the second eye fundus image is not identical, and the First view base map seems normal condition, and described
Two eye fundus images are normal condition or abnormality；
Difference block, for respectively according to each image to generation difference image；
Training module, for respectively using the difference image as sample data, by the state of second eye fundus image
Label as the sample data is trained initial machine learning model, so that the machine learning model being capable of basis
The difference image identifies the state of second eye fundus image.
Preferably, the difference block includes:
Identification module, for identifying optic disk and/or macula lutea in described image pair respectively；
Alignment module, for respectively according to the position of the optic disk and/or macula lutea by described image to being aligned；
Computing module obtains the difference image to Difference Calculation is carried out for being based respectively on the image after being aligned.
The present invention also provides a kind of eye fundus image judgment method, this method can be based on the machine that the above method is trained
Learning model judges that eye fundus image, this method can be executed by computer or server.This method as shown in Figure 7
Include:
S21, obtains the eye fundus image to be judged of user, which is the unknown state figure acquired by fundus camera
Picture, e.g. such as Fig. 2, Fig. 3, Fig. 4 or image shown in fig. 5, their state is unknown in the present embodiment；
S22, acquisition is associated with eye fundus image to be judged to refer to eye fundus image, wherein being known shape with reference to eye fundus image
State is normal eye fundus image, can be the image of he or she with reference to eye fundus image, is also possible to the eye fundus image of other people, only
It need to determine its state (health), such as can be image as shown in Figure 2, Figure 3, Figure 4.
S23 generates difference image according to eyeground figure to be judged and with reference to eye fundus image, and there are many difference algorithms, the present invention
It can use any existing algorithm and carry out Difference Calculation.Difference Calculation can only focus on optic disk, macula lutea, train of thought blood vessel feature
Difference, and ignore it is certain due to external environment, acquisition equipment caused by differences (such as brightness).
The process of Difference Calculation, which can be, to be first passed through optic disk and macula lutea positioning two opens one's eyes base map alignment, determines that two open one's eyes bottom
The discrepancy of image or region, then by these discrepancys or zone marker wait judge in eye fundus image.In order to enhance difference
As a result, colored eye fundus image to be judged can be converted to gray level image, by discrepancy or zone marker in gray level image
In.
Obtained difference image can embody the difference of this two images, these differences can be subtle region very
To being pixel, it is also possible to apparent region.Such as image as shown in FIG. 6 can be generated.
S24 makes machine learning model output treat judgement eye using difference image as the input data of machine learning model
The judging result of base map picture, wherein machine learning model is trained using training difference image and corresponding label data
It arrives.Machine learning model therein can be the machine learning mould obtained according to above-mentioned eye fundus image judgment models training method
Type.
About judging result, can be normal (health) or extremely both as a result, being also possible to specific exception class
Type (disease type), this depends on the content of used label data when training machine learning model.
The eye fundus image judgment method provided according to embodiments of the present invention utilizes eyeground figure and known health to be judged
Difference image is generated with reference to eye fundus image, tiny characteristic content can be embodied by this difference image, then using artificial
Intellectual technology is based on big data and judges difference image, the state of eyeground figure to be judged is determined with this, it is possible thereby to mention
The accuracy that height judges eye fundus image state can obtain preferable judgement especially for certain unconspicuous genius morbis
As a result.
Although the eye fundus image of different health (normal) people is generally there is no larger difference, machine is carrying out difference
Still some contents unrelated with abnormal (lesion) may be judged as difference content when calculating and be embodied in difference image, this can
It can reduce the recognition accuracy of machine learning model.As a preferred embodiment, in the operation for obtaining reference picture
In, the present embodiment preferably using the eye fundus image to be judged from same people and refers to eye fundus image, wherein referring to eye fundus image
Acquisition time earlier than eye fundus image to be judged acquisition time.
This requires detected user that the eye fundus image under its health status has been provided previously before this detection.
Due to for same people, difference caused by non-disease (exception) reason will very little, this has the difference image generated
A possibility that obvious advantage, the difference embodied in difference image becomes apparent from, is lesion when a discrepancy exists, is bigger, thus
The identification accuracy of machine learning model can be improved.
This method when detected user did not provide the eye fundus image under its health status before this detection
It can also successfully carry out, eye fundus image to be judged can be learnt according to the principle of this method and be also possible to reference to eye fundus image
Eye fundus image from different people, wherein referring to the acquisition time of eye fundus image earlier than when judging the acquisition of eye fundus image
Between.
When being detected using this method, user can provide unique user information, and system can decide whether to deposit
In the reference eye fundus image of the user, and if so, difference image is generated using the image, if there is no then available
Preset to refer to eye fundus image, which can be from other people's.
Image alignment operation can be executed when detecting eye fundus image, specifically, step S23 may include:
S231, respectively in eyeground figure to be judged and with reference to identifying optic disk and/or macula lutea in eye fundus image, the step for can also
It is identified, can also be carried out by the way of machine vision based on the linear feature in image in a manner of making manually intelligence
Identification；
S232 eyeground figure to be judged and will be aligned with reference to eye fundus image according to the position of optic disk and/or macula lutea；
S233 obtains difference image based on the eyeground figure to be judged after alignment and with reference to eye fundus image progress Difference Calculation.
Above-mentioned preferred embodiment, to alignment operation is carried out, is avoided since eye fundus image location of content is inclined for two eye fundus images
Thus Difference Calculation mistake caused by moving improves the quality of generated difference image, and then improves the identification of machine learning model
Accuracy.
Correspondingly, the embodiment of the invention also provides a kind of electronic equipment, comprising: at least one processor；And with institute
State the memory of at least one processor communication connection；Wherein, the memory, which is stored with, to be executed by one processor
Instruction, described instruction executed by least one described processor, so that at least one described processor executes above-mentioned eyeground figure
As judgment method.
The embodiment of the invention provides a kind of eye fundus image judgment means, the device can be hardware composition device or
The corresponding virtual bench of computer program, is also possible to program and device that hardware combines, which includes:
First obtains module, for obtaining the eye fundus image to be judged of user；
Second obtains module, associated with the eye fundus image to be judged with reference to eye fundus image for obtaining, wherein institute
State with reference to eye fundus image be known state be normal eye fundus image；
Difference block, for generating difference image according to the eyeground figure to be judged and the reference eye fundus image；
Judgment module, for making the machine learning using the difference image as the input data of machine learning model
Model exports the judging result to the eye fundus image to be judged, wherein the machine learning model is to utilize training difference image
It is trained with corresponding label data.
Preferably, the second acquisition module includes:
Historical data judgment module, for judging whether to prestore the reference eye fundus image of the user；
Pre-set image obtains module, for obtaining preset reference when there not being the reference eye fundus image of the user
Eye fundus image.
Preferably, the difference block includes:
Identification module, for respectively the eyeground figure to be judged and it is described with reference to identify in eye fundus image optic disk and/or
Macula lutea；
Alignment module, for according to the position of the optic disk and/or macula lutea by the eyeground figure to be judged and the reference
Eye fundus image is aligned；
Computing module, the eyeground figure to be judged described in after alignment and described carries out based on difference with reference to eye fundus image
Calculation obtains the difference image.
It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the present invention
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the present invention, which can be used in one or more,
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces
The form of product.
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Obviously, the above embodiments are merely examples for clarifying the description, and does not limit the embodiments.It is right
For those of ordinary skill in the art, can also make on the basis of the above description it is other it is various forms of variation or
It changes.There is no necessity and possibility to exhaust all the enbodiments.And it is extended from this it is obvious variation or
It changes still within the protection scope of the invention.
Claims (10)
1. a kind of eye fundus image judgment method characterized by comprising
Obtain the eye fundus image to be judged of user；
Acquisition is associated with the eye fundus image to be judged to refer to eye fundus image, wherein the reference eye fundus image is known shape
State is normal eye fundus image；
Difference image is generated according to the eyeground figure to be judged and the reference eye fundus image；
Using the difference image as the input data of machine learning model, make the machine learning model output to described wait sentence
The judging result of disconnected eye fundus image, wherein the machine learning model be using training difference image and corresponding label data into
Row training obtains.
2. the method according to claim 1, wherein the eye fundus image to be judged and it is described refer to eye fundus image
It is the eye fundus image from same people, wherein the acquisition time with reference to eye fundus image is earlier than the eye fundus image to be judged
Acquisition time.
3. the method according to claim 1, wherein the eye fundus image to be judged and it is described refer to eye fundus image
It is the eye fundus image from different people, wherein the acquisition time with reference to eye fundus image is earlier than the eye fundus image to be judged
Acquisition time.
4. according to the method described in claim 3, it is characterized in that, the acquisition is associated with the eye fundus image to be judged
With reference to eye fundus image, comprising:
Judge whether to prestore the reference eye fundus image of the user；
When there not being the reference eye fundus image of the user, obtain preset with reference to eye fundus image.
5. the method according to claim 1, wherein the eyeground figure to be judged according to refers to eye with described
Base map picture generates difference image, comprising:
Optic disk and/or macula lutea are identified in the eyeground figure to be judged and the reference eye fundus image respectively；
Eyeground the to be judged figure and the eye fundus image that refers to are aligned according to the position of the optic disk and/or macula lutea；
Based on eyeground figure to be judged described in after alignment and described Difference Calculation is carried out with reference to eye fundus image obtain the difference diagram
Picture.
6. a kind of eye fundus image judgment models training method characterized by comprising
Obtain multiple images pair being made of the first eye fundus image and the second eye fundus image, first eye fundus image and second
The acquisition time of base map picture is not identical, and the First view base map seems normal condition, and second eye fundus image is normal shape
State or abnormality；
Respectively according to each image to generation difference image；
Respectively using the difference image as sample data, using the state of second eye fundus image as the sample data
Label is trained initial machine learning model, so that the machine learning model can identify institute according to the difference image
State the state of the second eye fundus image.
7. according to the method described in claim 6, it is characterized in that, first eye fundus image and the second eye fundus image are to come from
The eye fundus image of same people, wherein the acquisition time of the First view base map picture earlier than second eye fundus image acquisition when
Between.
8. according to the method described in claim 6, it is characterized in that, it is described respectively according to each image to generate difference image,
Include:
Optic disk and/or macula lutea are identified in described image pair respectively；
Respectively according to the position of the optic disk and/or macula lutea by described image to being aligned；
Image after being based respectively on alignment obtains the difference image to Difference Calculation is carried out.
9. a kind of electronic equipment characterized by comprising at least one processor；And it is logical at least one described processor
Believe the memory of connection；Wherein, the memory is stored with the instruction that can be executed by one processor, and described instruction is by institute
The execution of at least one processor is stated, so that at least one described processor is executed as described in any one of claim 1-5
Eye fundus image judgment method.
10. a kind of electronic equipment characterized by comprising at least one processor；And it is logical at least one described processor
Believe the memory of connection；Wherein, the memory is stored with the instruction that can be executed by one processor, and described instruction is by institute
The execution of at least one processor is stated, so that at least one described processor is executed as described in any one of claim 6-8
Eye fundus image judgment models training method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811408311.6A CN109658385A (en) | 2018-11-23 | 2018-11-23 | Eye fundus image judgment method and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811408311.6A CN109658385A (en) | 2018-11-23 | 2018-11-23 | Eye fundus image judgment method and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109658385A true CN109658385A (en) | 2019-04-19 |
Family
ID=66112226
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811408311.6A Pending CN109658385A (en) | 2018-11-23 | 2018-11-23 | Eye fundus image judgment method and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109658385A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298850A (en) * | 2019-07-02 | 2019-10-01 | 北京百度网讯科技有限公司 | The dividing method and device of eye fundus image |
| CN110751636A (en) * | 2019-10-12 | 2020-02-04 | 天津工业大学 | Fundus image retinal arteriosclerosis detection method based on improved coding and decoding network |
| CN111834004A (en) * | 2020-05-25 | 2020-10-27 | 杭州深睿博联科技有限公司 | Unknown disease category identification method and device based on centralized space learning |
| WO2020215682A1 (en) * | 2019-09-17 | 2020-10-29 | 平安科技（深圳）有限公司 | Fundus image sample expansion method and apparatus, electronic device, and computer non-volatile readable storage medium |
| CN114847871A (en) * | 2022-07-06 | 2022-08-05 | 北京鹰瞳科技发展股份有限公司 | Method, system and related product for analyzing fundus variation trend of subject |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104463830A (en) * | 2013-09-18 | 2015-03-25 | 通用电气公司 | System and method for detecting plaques in blood vessels |
| CN105431077A (en) * | 2013-07-31 | 2016-03-23 | 佳能株式会社 | Image processing apparatus and image processing method |
| US20160227999A1 (en) * | 2015-02-05 | 2016-08-11 | Carl Zeiss Meditec, Inc. | Acquisition and analysis techniques for improved outcomes in optical coherence tomography angiography |
| CN106846293A (en) * | 2016-12-14 | 2017-06-13 | 海纳医信（北京）软件科技有限责任公司 | Image processing method and device |
| CN107045720A (en) * | 2017-05-04 | 2017-08-15 | 深圳硅基智能科技有限公司 | Artificial neural network and system for recognizing eye fundus image lesion |
| CN107423571A (en) * | 2017-05-04 | 2017-12-01 | 深圳硅基仿生科技有限公司 | Diabetic retinopathy identifying system based on eye fundus image |
-
2018
- 2018-11-23 CN CN201811408311.6A patent/CN109658385A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105431077A (en) * | 2013-07-31 | 2016-03-23 | 佳能株式会社 | Image processing apparatus and image processing method |
| CN104463830A (en) * | 2013-09-18 | 2015-03-25 | 通用电气公司 | System and method for detecting plaques in blood vessels |
| US20160227999A1 (en) * | 2015-02-05 | 2016-08-11 | Carl Zeiss Meditec, Inc. | Acquisition and analysis techniques for improved outcomes in optical coherence tomography angiography |
| CN106846293A (en) * | 2016-12-14 | 2017-06-13 | 海纳医信（北京）软件科技有限责任公司 | Image processing method and device |
| CN107045720A (en) * | 2017-05-04 | 2017-08-15 | 深圳硅基智能科技有限公司 | Artificial neural network and system for recognizing eye fundus image lesion |
| CN107423571A (en) * | 2017-05-04 | 2017-12-01 | 深圳硅基仿生科技有限公司 | Diabetic retinopathy identifying system based on eye fundus image |
Non-Patent Citations (1)
| Title |
|---|
| 庞浩等: "用于糖尿病视网膜病变检测的深度学习模型", 《软件学报》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298850A (en) * | 2019-07-02 | 2019-10-01 | 北京百度网讯科技有限公司 | The dividing method and device of eye fundus image |
| CN110298850B (en) * | 2019-07-02 | 2022-03-15 | 北京百度网讯科技有限公司 | Segmentation method and device for fundus image |
| WO2020215682A1 (en) * | 2019-09-17 | 2020-10-29 | 平安科技（深圳）有限公司 | Fundus image sample expansion method and apparatus, electronic device, and computer non-volatile readable storage medium |
| CN110751636A (en) * | 2019-10-12 | 2020-02-04 | 天津工业大学 | Fundus image retinal arteriosclerosis detection method based on improved coding and decoding network |
| CN110751636B (en) * | 2019-10-12 | 2023-04-21 | 天津工业大学 | Fundus image retinal arteriosclerosis detection method based on improved coding and decoding network |
| CN111834004A (en) * | 2020-05-25 | 2020-10-27 | 杭州深睿博联科技有限公司 | Unknown disease category identification method and device based on centralized space learning |
| CN114847871A (en) * | 2022-07-06 | 2022-08-05 | 北京鹰瞳科技发展股份有限公司 | Method, system and related product for analyzing fundus variation trend of subject |
| CN114847871B (en) * | 2022-07-06 | 2022-10-18 | 北京鹰瞳科技发展股份有限公司 | Method, system and related product for analyzing fundus variation trend of subject |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109658385A (en) | Eye fundus image judgment method and equipment | |
| Al-Bander et al. | Multiscale sequential convolutional neural networks for simultaneous detection of fovea and optic disc | |
| US9779492B1 (en) | Retinal image quality assessment, error identification and automatic quality correction | |
| KR20200005409A (en) | Fundus image management device and method for determining suitability of fundus image | |
| CN110327013A (en) | Eye fundus image detection method, device and equipment and storage medium | |
| CN105431078B (en) | System and method for the tracking of coaxial eye gaze | |
| CN107145845A (en) | The pedestrian detection method merged based on deep learning and multi-characteristic points | |
| TW200938162A (en) | Method and device for finding and tracking pairs of eyes | |
| CN107003737A (en) | The indicator projection inputted for natural user | |
| CN108717868A (en) | Glaucoma eye fundus image screening method based on deep learning and system | |
| CN109697719A (en) | A kind of image quality measure method, apparatus and computer readable storage medium | |
| CN109961848A (en) | Macula lutea image classification method and equipment | |
| CN109635669B (en) | Image classification method and device and classification model training method and device | |
| CN109658393A (en) | Eye fundus image joining method and system | |
| CN108629378A (en) | Image-recognizing method and equipment | |
| CN102567734A (en) | Specific value based retina thin blood vessel segmentation method | |
| Sedai et al. | Multi-stage segmentation of the fovea in retinal fundus images using fully convolutional neural networks | |
| CN110188613A (en) | Image classification method and equipment | |
| Niu et al. | Automatic localization of optic disc based on deep learning in fundus images | |
| Farooq et al. | Improved automatic localization of optic disc in Retinal Fundus using image enhancement techniques and SVM | |
| Zhang et al. | A human-in-the-loop deep learning paradigm for synergic visual evaluation in children | |
| Ramachandran et al. | U-COSFIRE filters for vessel tortuosity quantification with application to automated diagnosis of retinopathy of prematurity | |
| Jemima Jebaseeli et al. | Retinal blood vessel segmentation from depigmented diabetic retinopathy images | |
| Ponnibala et al. | Proliferative diabetic retinopathy diagnostic investigation using retinal blood vessels mining technique | |
| El Hossi et al. | Applied CNN for automatic diabetic retinopathy assessment using fundus images |
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 |