CN105513077B - A kind of system for diabetic retinopathy screening - Google Patents
A kind of system for diabetic retinopathy screening Download PDFInfo
- Publication number
- CN105513077B CN105513077B CN201510920921.4A CN201510920921A CN105513077B CN 105513077 B CN105513077 B CN 105513077B CN 201510920921 A CN201510920921 A CN 201510920921A CN 105513077 B CN105513077 B CN 105513077B
- Authority
- CN
- China
- Prior art keywords
- image
- module
- equipment
- screening
- optic disk
- 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.)
- Active
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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Eye Examination Apparatus (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of system for diabetic retinopathy screening, the system comprises: eye fundus image obtains equipment, image procossing and screening equipment and report output equipment, and the eye fundus image obtains the eye fundus image that equipment is used to acquire or receive tested personnel；Then described image processing and screening equipment will test result and are conveyed to the report output equipment for being handled the eye fundus image and being detected wherein with the presence or absence of lesion；The report output equipment is based on the testing result output phase and answers examining report.System of the invention determines the sensitivity of sugared net picture up to 93.8%, and specificity is comparable with artificial judgement precision, but detection time greatly reduces up to 94.5%, average out to 8.7 seconds, only the 1/30 of worker's scoring time.Therefore, for sugared net, screening cost can not only be substantially saved, and can accomplish early discovery early treatment, greatly reduces the pain of patient, there is good potential applicability in clinical practice and social benefit.
Description
Technical field
The present invention relates to field of medical device, and in particular to a kind of system for diabetic retinopathy screening.
Background technique
Diabetic retinopathy, referred to as sugared net (DR), is one of most common chronic complicating diseases of diabetic.According to
The World Health Organization (WHO) prediction, to the year two thousand thirty, global DR patient numbers will be added to 3.66 hundred million, it has also become four substantially blind holes
One of disease, sugar net disease prevention and treatment will become a serious worldwide problem.
Studies have shown that carrying out early diagnosis and therapy to DR patient can effectively prevent the loss and blindness of vision, and prevent
The key controlled is then by eye-ground photography inspection, and regular follow-up finds the progress of the state of an illness, carries out laser therapeutic intervention in time.But
Any type of examination of eyes is not received more than 50% diabetic in the world at present, the diabetes view based on ophthalmoscopic image
The screening of retinopathy visually observes progress substantially or by oculist.
But when facing extensive screening, the data volume for needing doctor to analyze and handle is very big, and artificial interpretation method was both
Time-consuming and laborious, artificial screening can not be implemented.And manually screening subjectivity is strong, and data analysis is complicated and is difficult to quantify, very
Difficulty accomplishes quantitative follow-up.
So the above problem has become the major obstacle for implementing extensive screening for diabetic retinopathy, clinically urgently
A kind of objective, accurately and efficiently method is needed, it is quick to assist oculist to carry out the result of diabetic's eye-ground photography
Analysis.
In recent years, with the development of computer-aided diagnosis technology, the relevant technologies based on computer vision are in liver
Dirty disease, respiratory disease diagnostic imaging in obtain development and application.
But it can be used in currently on the market there are no a kind of screening system of precise and high efficiency to diabetic retinal
Lesion carries out rapid screening.
Although it is proposed that whether judging patient by carrying out feature extraction to patient's picture using computerized algorithm
With diabetic retinopathy, still, corresponding equipment is produced there is presently no people also, existing algorithm is verified
The accuracy rate of screening is relatively low afterwards.
Summary of the invention
In view of the above-mentioned problems, the present invention is desirable to provide a kind of diabetic retinopathy screening system based on image,
Its extensive automatic screening that can be realized sugar net disease, and then the diagnosis efficiency of doctor is greatly improved, allow doctor to read from heavy
It frees in piece work, and the subjectivity of people can be removed, avoid causing Error Diagnostics because of personal knowledge and experience difference, more
It is accurate to add, and sugared net screening that is objective, being efficiently completed ophthalmology image saves medical treatment cost.
Specifically, the present invention provides a kind of system for diabetic retinopathy screening, which is characterized in that institute
The system of stating includes: that eye fundus image obtains equipment, image procossing and screening equipment and report output equipment,
The eye fundus image obtains the eye fundus image that equipment is used to acquire or receive tested personnel；
Described image processing and screening equipment are for handling the eye fundus image and being detected wherein with the presence or absence of disease
Become, then will test result and be conveyed to the report output equipment；
The report output equipment is based on the testing result output phase and answers examining report.
In a kind of preferred implementation, it is colored fundus camera that the eye fundus image, which obtains equipment, alternatively, the eyeground
Image acquisition equipment is eye fundus image receiving device.
In a kind of preferred implementation, described image processing and screening equipment include: image pre-processing module, image calibration
Positive module, blood vessel segmentation module, optic disk locating module, red lesion detection module, brightness lesion detection module and classification mould
Block, it is fixed that the categorization module is based on the red lesion detection module, brightness lesion detection module, blood vessel segmentation module and optic disk
The output of position module determines whether patient corresponding to detected picture has diabetic retinopathy.
In a kind of preferred implementation, described image correction module is used to carry out color, brightness and exposure to image
The normalized of distribution, normalized detailed process include: the exposure distribution according to image, fit a quadratic surface,
Then counter-bonification is carried out according to Luminance Distribution of the curved surface of fitting to image；Then, the preferable image conduct of a width quality is selected
Reference picture calculates the grey level histogram peak value in its color and luminance channel, the tone of image processed, saturation degree and bright
The grey level histogram of degree information all normalizes in the reference value of reference picture.
In a kind of preferred implementation, the blood vessel segmentation module divides the mistake of optic disk for dividing blood vessel and optic disk
Journey includes: the characteristic pattern that different scale is generated based on the eye fundus image；Corresponding characteristic remarkable is established based on the characteristic pattern
Property description；Different feature significance descriptions is normalized, and is added acquisition Saliency maps；It is calculated using Local threshold segmentation
Method extracts resulting Saliency maps, exports position of the maximum region of brightness value as optic disk.
In a kind of preferred implementation, the blood vessel segmentation module is when carrying out blood vessel segmentation, the region of interest of image
Polar coordinate representation is switched to by rectangular co-ordinate, and uses sub-pixel precision extracting method, seeks the ash at each pixel first
The gradient of angle value acquires the initial profile of blood vessel using Canny operator.
The present invention assists doctor to differentiate and accurately extract the disease marker of image by the relevant technologies of computer vision,
The disease information of intellectual analysis patient.The present invention includes that all kinds of sugared net lesions such as bleeding, exudation, microaneurysm are special by extracting
Sign carries out automatic classification diagnosis to ophthalmoscopic image further according to the clinical criteria of DR, to realize the extensive of sugar net disease
Automatic screening.
Since computer can carry out accurate quantitative calculating using image information comprehensively, doctor can be not only greatly improved
Diagnosis efficiency, allow doctor to free from heavy diagosis work, and the subjectivity of people can be removed, avoid knowing because of individual
Know and experience difference causes Error Diagnostics, more accurately, sugared net screening that is objective, being efficiently completed ophthalmology image, significantly
Medical treatment cost is saved, while establishing objective and accurate extensive health data management system convenient for future.It is objective for establishing
Accurate clinical criteria is very valuable.
By the result of system screening of the invention with compared through the authenticated screening results of expert show it is of the invention
System can accurately and efficiently complete the sugared net screening by eye fundus image, medical treatment cost be greatly saved, while convenient for not
It establishes objective and accurate extensive health data management system, shows good potential applicability in clinical practice.
Detailed description of the invention
Fig. 1 shows the structural schematic diagram of the screening system in one embodiment of the invention；
Fig. 2 shows image acquisition equipments employed in one embodiment；
Fig. 3 shows output equipment employed in one embodiment；
Fig. 4 shows system of the invention at work, to the concrete processing procedure of image；
Fig. 5, which is shown, is treated in journey eye fundus image, and processing is preceding and handles the eye fundus image in each stage,
In, a is the original image of Messidor data, and b is the image after image preprocessing, and the enhanced image of c, d is the defeated of blood vessel segmentation
Out as a result, e is the output of optic disk positioning as a result, f is the extraction result of two kinds of lesions；
Fig. 6 shows the example of a testing result；
Fig. 7 is the Saliency maps of each feature of eye fundus image generated according to vision attention theory, wherein (a) is aobvious for brightness
Work property figure (b) is color Saliency maps, (c) is direction Saliency maps；
Fig. 8 is the result that optic disk is positioned using Saliency maps, wherein (a) is original image, (b) is notable figure figure,
(c) figure is extracted for FOA, is (d) positioning optic disk figure；
Fig. 9 is that blood vessel is erased effect contrast figure, wherein (a) is green channel original image, (b) to scheme after blood vessel of erasing；
Figure 10 is the comparison diagram of image polar coordinates conversion front and back, wherein (a) is before polar coordinates are converted, (b) to turn for polar coordinates
After changing；
Figure 11 is the schematic diagram of edge extracting process, wherein (a) is sub-pixel edge extraction, is (b) connection edge view
Disk.
Specific embodiment
Embodiment 1
As shown in Figure 1, in the present embodiment, screening system includes that a colored fundus camera, image procossing and screening are set
Standby (it can be realized by being equipped with the computer of sugared net intellectual analysis software, can also be realized by dedicated processor) with
And diagnosis report printer.Other hospitals or base's screening unit acquired image can also be transferred to image procossing and screening
Equipment.
Fig. 2 shows the structural schematic diagrams of colored fundus camera employed in the present embodiment；Fig. 3 shows the present embodiment
Employed in diagnosis report printer structural schematic diagram.Colored fundus camera is used to obtain the colored eye fundus image of patient,
Acquisition is operated by the doctor of profession and under the fixed route of image transmitting to computer.
Acquired image will directly or indirectly be transferred to image procossing and screening equipment, image procossing and screening equipment base
Image procossing is carried out in acquired image and final output testing result gives diagnosis report printer, and the latter is according to testing result
The corresponding report of printing.
Core of the invention is image procossing and screening equipment, below mainly to the image procossing of the present embodiment and screening
Equipment is described in more detail.
Image procossing and screening equipment include: image pre-processing module, image correction module, blood vessel segmentation module, optic disk
Locating module, red lesion detection module, brightness lesion detection module and categorization module.
Image pre-processing module is for pre-processing eye fundus image.Anatomical structure and disease in detection eye fundus image
Before change, the visual field (ROI) of eye fundus image needs to find in advance by the method for adaptive threshold fuzziness and template matching, makes figure
The black background of picture is removed, and image pre-processing module is exactly for realizing the function.
Image correction module is for compensating different images.In order to obtain stabilization from multifarious eye fundus image
Information, it is very necessary that color, the normalized of brightness and exposure distribution are carried out to image.Image correction module elder generation according to
The exposure distribution of image fits a brightness of image quadratic surface, then according to the curved surface of fitting to the Luminance Distribution of image into
Row counter-bonification makes bright decrease, while then dark region enhancing is marked to guarantee the uniform of exposure from international Messidor
Quasi- eyeground picture library selectes a preferable image of width quality as reference picture, calculates the ash in reference picture color and luminance channel
Histogram peak is spent, the grey level histogram of the tone of image processed, saturation degree and luminance information is all normalized to reference to figure
In the reference value of picture, to ensure that the stability of image processed.
Blood vessel segmentation module is used for from the blood vessel being partitioned into image in image.Vascular system is most important in retina
One of anatomical structure.The accurate of blood vessel is divided for differentiation capillary and other anatomical structures from red lesion, such as
The detection of optic disk is all very crucial.In the present embodiment, blood vessel segmentation module is by Intensity threshold Separation and is based on color characteristic
Supervised classification method be effectively partitioned into vessel borders.
Optic disk locating module is for positioning optic disk in image.Optic disk is another critically important dissection knot of eyeground
Structure.Being properly positioned for optic disk is highly useful for the misjudgement of later period reduction brightness lesion.Optic disk be by its brightness, shape and
What the features such as the direction of blood vessel were positioned.
Red lesion detection module is for detecting red lesion.Red lesion, including microaneurysm, bleeding and increasing
Blood vessel is grown, is the important feature of DR, therefore, their correct detection is extremely important for DR screening system.Small red disease
The positioning of stove obtains candidate region based on mathematical morphology, and the positioning of big red lesion then passes through the prison based on color characteristic
Classification is superintended and directed to complete.The spy of candidate region shape, structure, color and contrast is described using the classifier for having supervision and one group
Sign can calculate the probability that detected candidate region is a real red lesion.
Brightness lesion detection module is used to detect the brightness lesion in image.Brightness lesion, such as exudation, cotton-wool patches or glass
Glass film wart, frequently appears in DR Mass screening.Only first two lesion is related to DR.It is similar with the detection of red lesion,
The present embodiment is used to obtain candidate brightness lesion region using the Supervised classification of Pixel-level.Each candidate region is really bright
The probability for spending lesion is obtained by using the supervised classification average value of one group of candidate region feature, these features include shape, right
Than degree, color and at a distance from nearest red lesion.
Blood vessel that categorization module is used to obtain based on above-mentioned each processing and detection module, optic disk, red lesion and bright
Lesion etc. is spent to carry out detection classification to eye fundus image.The output of above-mentioned these different processing and detection module must be combined
Together to obtain the final result checked about patient.In order to realize this target, it is necessary to calculate above-mentioned module
The feature of each output.In the present embodiment, categorization module using supervised classifier come a possibility that determining lesion, thus for into one
The diagnostic criteria of step provides foundation, realizes the automatic screening to DR.Categorization module is according to Pathological Information and international sugar net classification
Standard separates different grades, and provides diagnostic recommendations.
Experiments verify that sugared net screening system of the invention determines that the sensitivity of sugared net picture is 93.8%, specificity is
94.5%, be comparable with artificial judgement precision, but detection time greatly reduces, average out to 8.7 seconds, only worker's diagosis when
Between 1/30, these description of test show that the automatic screening system of sugar net developed based on computer vision algorithms make can be accurate, efficiently
Completion ophthalmology image sugared net screening, and computer another big advantage be can be with 24 hours uninterrupted works
Make, this is that diagosis doctor can not accomplish, the screening for million picture ranks of the following processing, this automatic screening it is huge
Advantage just embodies.Increasingly raising of the future with aging of population and people to health demand, the health sieve of large area
A kind of necessity will be become by looking into, and be assisted screening using the technology of computer, can not only substantially be saved medical treatment cost, and can accomplish
Early discovery early treatment, greatly reduces the pain of patient, has good potential applicability in clinical practice and social benefit.In addition, computer
The automatic screening of sugar net of realization will bring glad tidings for the diabetic of those remote districts for lacking medical resource.Because
The low cost of computer and convenient and efficient, patients of these remote districts can be carried out the base to routinize inspection, effectively anti-
Only disease to it is very serious when the interrogation of Cai Qu hospital the case where.
Following table is the contrast table using system of the invention compared with artificial detection, as can be seen from the table using the present invention
The advantage of system.
1 artificial detection of table and automatic running duration contrast table
Embodiment 2
In the present embodiment, screening system includes eye fundus image receiver, (it is soft that it is equipped with sugared net intellectual analysis to computer
Part) and diagnosis report display equipment.
In the present embodiment, eye fundus image receiver remotely receives the eyeground figure that other fundus cameras non-indigenous are sent
Picture.Computer includes processor, memory, and sugared net intellectual analysis software is equipped in computer.
Sugared net intellectual analysis software can execute each process step shown in Fig. 4.As shown, receiving first at a distance
The eye fundus image that fundus camera or other equipment obtain, then carries out image preprocessing.
Followed by image quality correction, segmentation blood vessel, optic disk positioning, red lesion are extracted, brightness lesion is extracted.So
Afterwards, the characteristic informations such as blood vessel, optic disk, the lesion obtained using above steps, are analyzed, and determine whether image has sugared net
Lesion.
Fig. 5 shows output result of the eye fundus image example after each step process.Fig. 6 shows detection report
The example of announcement.
Embodiment 3
In the present embodiment, the construction of system is same as Example 1, and only optic disk locating module uses a kind of special
Localization method.
Specifically, optic disk locating module is put forward for the first time base by means of some progress of calculating neurology in the present embodiment
Optic disk is positioned in the method for human visual attention model.Principle is as follows: firstly generating the characteristic pattern of multiple and different scales.It is specific and
Speech, by gaussian filtering and continuous down-sampling, extracts image in the Gauss gold word of multiple and different scales after image preprocessing
Tower, after establishing gaussian pyramid, the extraction of the characteristic pattern of single scale is operated by linear " Core-Periphery is poor " to be calculated.It is raw
At brightness, color, direction character figure, as shown in Figure 7: where (a) is brightness Saliency maps, (b) is color Saliency maps, (c)
For direction Saliency maps.
Based on 3 characteristic patterns, 3 feature significance descriptions are establishedWithThe description of different characteristic conspicuousness is returned
One changes, and addition obtains Saliency maps S, and maximum of points is then the lime light of vision.Formula is as follows, as a result as shown in Figure 8.
Then, resulting Saliency maps are extracted using Local threshold segmentation algorithm, exports the maximum region of brightness value, as
The position of optic disk.
In addition, blood vessel segmentation module additionally uses a kind of better partitioning scheme in the present embodiment.Specifically, regarding
After disk positioning, in order to reduce computer capacity, the interference of other structures is reduced, the blood vessel segmentation module selection of the present embodiment includes
The a certain range of region of optic disk is as cut zone, as shown in Fig. 9 (a).
To reduce the interference that optic disk is divided at optic disk medium vessels edge, blood vessel segmentation module uses sort method to filter first
Device wipes optic disk blood vessel --- blood vessel of erasing, as shown in Fig. 9 (b).Sort method filter is a kind of nonlinear filter, it
It is the sequence based on pixel in image-region where filter, the value of center pixel is replaced by the value that ranking results determine.Such as Fig. 9
(b) as it can be seen that the optic disk edge of the filtered image of sort method is still very clear, it is easy to divide.
After the region of interest for obtaining optic disk segmentation, blood vessel segmentation module starts Fig. 9 (b) to carry out edge extracting.It notices
The close circle of the fringe region of optic disk, if the region of interest of image is switched to polar coordinate representation by rectangular co-ordinate, optic disk
As soon as edge will become the curve of a horizontal direction, this greatly facilitates the positioning and extraction at optic disk edge.Blood vessel segmentation mould
Block is using regional center as origin, using maximum radius in region as transition radius, carries out two-dimentional polar coordinates conversion.Result after conversion
For shown in Figure 10 (b).Conversion formula are as follows:
After converting through polar coordinates, optic disk edge becomes horizontal direction one clearly by approximate ellipsoidal closed curve before
Clear visible curve.Because the Accurate Segmentation of optic disk is particularly significant to subsequent medical diagnosis on disease, blood vessel segmentation module is to improve edge point
The precision cut is sought the gradient of the gray value at each pixel first, is utilized Canny using sub-pixel precision extracting method
Operator etc. acquires initial profile.Assuming that gray value gradient is parabolically distributed along edge normal direction.Based on such vacation
If the neighborhood gradient that can use initial profile point along normal direction fits the parabola, then the parabolical peak point pair
Answer profile point.The edge for noticing optic disk is to change most fast region on longitudinal direction, therefore, seeks the algorithm at optic disk edge just
As the position for seeking greatest gradient value on longitudinal direction, as a result as shown in Figure 11 (a).In order to accurately calculate, in embodiment, blood vessel
Dividing module image level cutting is 32 equal parts, greatest gradient is sought in each equal part, this method can effectively exclude one
A little interference of noise and the overlapping at edge.Finally the edge of the greatest gradient of each image slice is connected, as optic disk
Edge.As shown in Figure 11 (b).
It can more accurately be positioned using this optic disk positioning of the present embodiment and edge method of determination, optic disk locating module
The position of optic disk out, blood vessel segmentation module can more clearly from be partitioned into optic disk and other vasculature parts.
Although the principle of the present invention is described in detail above in conjunction with the preferred embodiment of the present invention, this field skill
Art personnel are it should be understood that above-described embodiment is only the explanation to exemplary implementation of the invention, not to present invention packet
Restriction containing range.Details in embodiment is simultaneously not meant to limit the scope of the invention, without departing substantially from spirit of the invention and
In the case where range, any equivalent transformation based on technical solution of the present invention, simple replacement etc. obviously change, and all fall within
Within the scope of the present invention.
Claims (4)
1. a kind of system for diabetic retinopathy screening, which is characterized in that the system comprises: eye fundus image obtains
Equipment, image procossing and screening equipment and report output equipment are taken,
The eye fundus image obtains the eye fundus image that equipment is used to acquire or receive tested personnel；
Described image processing and screening equipment are for handling the eye fundus image and being detected wherein with the presence or absence of lesion, so
After will test result and be conveyed to the report output equipment；
The report output equipment is based on the testing result output phase and answers examining report,
Wherein, described image processing and screening equipment include: image pre-processing module, image correction module, blood vessel segmentation module,
Optic disk locating module, red lesion detection module, brightness lesion detection module and categorization module, the categorization module are based on institute
The output of red lesion detection module, brightness lesion detection module, blood vessel segmentation module and optic disk locating module is stated to determine quilt
Whether patient corresponding to detection picture has diabetic retinopathy,
Wherein, the optic disk locating module is used to generate the characteristic pattern of multiple and different scales, extracts image in multiple and different scales
Gaussian pyramid, after establishing gaussian pyramid, extract the characteristic pattern of single scale, to different characteristic conspicuousness description return
One changes, and addition obtains Saliency maps S, and maximum of points is then the lime light of vision；
The optic disk locating module is also used to extract resulting Saliency maps using Local threshold segmentation algorithm, and output brightness value is most
Big region, as the position of optic disk,
The blood vessel segmentation module is chosen comprising a certain range of region of optic disk as cut zone, is filtered using sort method
Device wipes optic disk blood vessel, and the sort method filter is a kind of nonlinear filter, based on image district where filter
The sequence of pixel in domain, the value of center pixel is replaced by the value that ranking results determine, and then obtains the region of interest of optic disk segmentation
Domain, and the blood vessel segmentation module carries out edge extracting, the blood vessel to using the image after sort method filter process
Divide module using regional center as origin, using maximum radius in region as transition radius, carries out two-dimentional polar coordinates conversion, and institute
The gradient for stating the gray value that blood vessel segmentation module is sought at each pixel acquires initial profile using Canny operator, it is assumed that
Gray value gradient is parabolically distributed along edge normal direction, based on such it is assumed that using initial profile point along normal side
To neighborhood gradient fit the parabola, then the parabolical peak point corresponding contour point.
2. the system according to claim 1 for diabetic retinopathy screening, which is characterized in that the eyeground
Image acquisition equipment is colored fundus camera, alternatively, it is eye fundus image receiving device that the eye fundus image, which obtains equipment,.
3. the system according to claim 1 for diabetic retinopathy screening, which is characterized in that described image
Correction module is used to carry out image the normalized of color, brightness and exposure distribution, and normalized detailed process includes:
According to the exposure distribution of image, a quadratic surface is fitted, is then carried out according to Luminance Distribution of the curved surface of fitting to image
Counter-bonification；Then, a preferable image of width quality is selected as reference picture, and the gray scale calculated in its color and luminance channel is straight
The grey level histogram of the tone of image processed, saturation degree and luminance information is all normalized to reference picture by square figure peak value
In reference value.
4. the system according to claim 3 for diabetic retinopathy screening, which is characterized in that the blood vessel
Segmentation module is for dividing blood vessel and optic disk.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510920921.4A CN105513077B (en) | 2015-12-11 | 2015-12-11 | A kind of system for diabetic retinopathy screening |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510920921.4A CN105513077B (en) | 2015-12-11 | 2015-12-11 | A kind of system for diabetic retinopathy screening |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105513077A CN105513077A (en) | 2016-04-20 |
| CN105513077B true CN105513077B (en) | 2019-01-04 |
Family
ID=55721034
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510920921.4A Active CN105513077B (en) | 2015-12-11 | 2015-12-11 | A kind of system for diabetic retinopathy screening |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105513077B (en) |
Families Citing this family (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106530295A (en) * | 2016-11-07 | 2017-03-22 | 首都医科大学 | Fundus image classification method and device of retinopathy |
| CN106725295A (en) * | 2016-11-29 | 2017-05-31 | 瑞达昇科技(大连)有限公司 | A kind of miniature check-up equipment, device and its application method |
| CN106846301B (en) * | 2016-12-29 | 2020-06-23 | 北京理工大学 | Retina image classification method and device |
| CN106934798B (en) * | 2017-02-20 | 2020-08-21 | 苏州体素信息科技有限公司 | Diabetic retinopathy classification and classification method based on deep learning |
| CN108771530B (en) * | 2017-05-04 | 2021-03-30 | 深圳硅基智能科技有限公司 | Fundus lesion screening system based on deep neural network |
| CN107423571B (en) * | 2017-05-04 | 2018-07-06 | 深圳硅基仿生科技有限公司 | Diabetic retinopathy identifying system based on eye fundus image |
| CN107038704B (en) * | 2017-05-04 | 2020-11-06 | 季鑫 | Retina image exudation area segmentation method and device and computing equipment |
| CN107256410B (en) * | 2017-05-26 | 2021-05-14 | 上海鹰瞳医疗科技有限公司 | Fundus image classification method and device |
| WO2018222136A1 (en) * | 2017-05-30 | 2018-12-06 | 正凯人工智能私人有限公司 | Image processing method and system |
| CN110348428B (en) | 2017-11-01 | 2023-03-24 | 腾讯科技（深圳）有限公司 | Fundus image classification method and device and computer-readable storage medium |
| CN108272434B (en) * | 2017-12-07 | 2020-06-19 | 柯鑫 | Method and device for processing fundus images |
| CN108416344B (en) * | 2017-12-28 | 2021-09-21 | 中山大学中山眼科中心 | Method for locating and identifying eyeground color optic disk and yellow spot |
| CN108416371A (en) * | 2018-02-11 | 2018-08-17 | 艾视医疗科技成都有限公司 | A kind of diabetic retinopathy automatic testing method |
| CN108615051B (en) * | 2018-04-13 | 2020-09-15 | 博众精工科技股份有限公司 | Diabetic retina image classification method and system based on deep learning |
| CN108596895B (en) * | 2018-04-26 | 2020-07-28 | 上海鹰瞳医疗科技有限公司 | Fundus image detection method, device and system based on machine learning |
| CN109166117B (en) * | 2018-08-31 | 2022-04-12 | 福州依影健康科技有限公司 | Automatic eye fundus image analysis and comparison method and storage device |
| CN111292286B (en) * | 2018-11-21 | 2023-07-11 | 福州依影健康科技有限公司 | Analysis method and system for change of characteristic data of sugar mesh bottom and storage device |
| CN111222361B (en) * | 2018-11-23 | 2023-12-19 | 福州依影健康科技有限公司 | Method and system for analyzing characteristic data of change of blood vessel of retina in hypertension |
| CN109816637B (en) * | 2019-01-02 | 2023-03-07 | 电子科技大学 | Method for detecting hard exudation area in fundus image |
| CN109691980A (en) * | 2019-01-07 | 2019-04-30 | 哈尔滨理工大学 | A kind of diabetic retina image lesion detection method |
| CN109887582A (en) * | 2019-03-01 | 2019-06-14 | 中山大学中山眼科中心 | A kind of intelligence ophthalmology image management and analysis system and image processing method |
| CN110175983B (en) * | 2019-04-17 | 2024-05-07 | 平安科技（深圳）有限公司 | Fundus focus screening method, fundus focus screening device, fundus focus screening computer equipment and fundus focus storage medium |
| CN110084809B (en) * | 2019-05-06 | 2021-03-16 | 成都医云科技有限公司 | Diabetic retinopathy data processing method and device and electronic equipment |
| CN110335253B (en) * | 2019-06-05 | 2023-12-15 | 平安科技（深圳）有限公司 | Eye image processing method, device, equipment and computer readable storage medium |
| CN110838358A (en) * | 2019-10-09 | 2020-02-25 | 南通大学 | Personalized intelligent diabetes information management system combining fundus images |
| CN113283270A (en) * | 2020-02-20 | 2021-08-20 | 京东方科技集团股份有限公司 | Image processing method and apparatus, screening system, computer-readable storage medium |
| CN111340778A (en) * | 2020-02-25 | 2020-06-26 | 中国科学院深圳先进技术研究院 | Glaucoma image processing method and equipment |
| CN112890764B (en) * | 2021-01-18 | 2022-12-13 | 哈尔滨工业大学 | Unmanned low-cost portable eye ground disease detection system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101018346A (en) * | 2007-03-07 | 2007-08-15 | 北京中星微电子有限公司 | Method, device and imaging device for eliminating the gradual halation phenomena |
| CN103091330A (en) * | 2013-01-11 | 2013-05-08 | 河海大学常州校区 | Copper strip surface defect identifying device for simulating human visual perception mechanism and method thereof |
| CN104392463A (en) * | 2014-12-16 | 2015-03-04 | 西安电子科技大学 | Image salient region detection method based on joint sparse multi-scale fusion |
-
2015
- 2015-12-11 CN CN201510920921.4A patent/CN105513077B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101018346A (en) * | 2007-03-07 | 2007-08-15 | 北京中星微电子有限公司 | Method, device and imaging device for eliminating the gradual halation phenomena |
| CN103091330A (en) * | 2013-01-11 | 2013-05-08 | 河海大学常州校区 | Copper strip surface defect identifying device for simulating human visual perception mechanism and method thereof |
| CN104392463A (en) * | 2014-12-16 | 2015-03-04 | 西安电子科技大学 | Image salient region detection method based on joint sparse multi-scale fusion |
Non-Patent Citations (2)
| Title |
|---|
| 眼底图像处理与分析中的关键技术研究;王玉亮;《中国博士学位论文全文数据库 信息科技辑》;20141115;正文第9-11、15-16、22-23、126-137、141-144页 |
| 糖尿病视网膜病变筛查方法分析;邴寒 等;《黑龙江医学》;20111231;第35卷(第12期);第905-907页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105513077A (en) | 2016-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105513077B (en) | A kind of system for diabetic retinopathy screening | |
| Gegundez-Arias et al. | Locating the fovea center position in digital fundus images using thresholding and feature extraction techniques | |
| Köse et al. | Simple methods for segmentation and measurement of diabetic retinopathy lesions in retinal fundus images | |
| Antal et al. | An ensemble-based system for microaneurysm detection and diabetic retinopathy grading | |
| CN100530204C (en) | Assessment of lesions in an image | |
| Xiong et al. | An approach to evaluate blurriness in retinal images with vitreous opacity for cataract diagnosis | |
| Siddalingaswamy et al. | Automatic grading of diabetic maculopathy severity levels | |
| Jitpakdee et al. | A survey on hemorrhage detection in diabetic retinopathy retinal images | |
| US20120249959A1 (en) | Apparatus and method for non-invasive diabetic retinopathy detection and monitoring | |
| Siddalingaswamy et al. | Automatic detection of multiple oriented blood vessels in retinal images | |
| Chin et al. | Automatic fovea location in retinal images using anatomical priors and vessel density | |
| CN106530294A (en) | Method for carrying out processing on meibomian gland image to obtain gland parameter information | |
| CN104318565B (en) | Interactive method for retinal vessel segmentation based on bidirectional region growing of constant-gradient distance | |
| Suero et al. | Locating the Optic Disc in Retinal Images Using Morphological Techniques. | |
| CN104299242A (en) | Fluorescence angiography fundus image extraction method based on NGC-ACM | |
| CN102567734A (en) | Specific value based retina thin blood vessel segmentation method | |
| Wang et al. | Accurate disease detection quantification of iris based retinal images using random implication image classifier technique | |
| Prageeth et al. | Early detection of retinal nerve fiber layer defects using fundus image processing | |
| Wisaeng et al. | Automatic detection of exudates in retinal images based on threshold moving average models | |
| Siddalingaswamy et al. | Automated detection of anatomical structures in retinal images | |
| Aruchamy et al. | Automated glaucoma screening in retinal fundus images | |
| Giancardo et al. | Quality assessment of retinal fundus images using elliptical local vessel density | |
| Zhou et al. | Computer aided diagnosis for diabetic retinopathy based on fundus image | |
| CN112991289B (en) | Processing method and device for standard section of image | |
| CN109447948A (en) | A kind of optic disk dividing method based on lesion colour retinal fundus images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |