CN109691980A - A kind of diabetic retina image lesion detection method - Google Patents
A kind of diabetic retina image lesion detection method Download PDFInfo
- Publication number
- CN109691980A CN109691980A CN201910012785.7A CN201910012785A CN109691980A CN 109691980 A CN109691980 A CN 109691980A CN 201910012785 A CN201910012785 A CN 201910012785A CN 109691980 A CN109691980 A CN 109691980A
- Authority
- CN
- China
- Prior art keywords
- image
- layer
- brightness
- detection method
- segmented
- 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
-
- 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Veterinary Medicine (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Biophysics (AREA)
- Public Health (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Ophthalmology & Optometry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Pathology (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of diabetic retina image lesion detection methods, comprising the following steps: A, image capture device acquisition retinal images are simultaneously pre-processed;B, pretreated image is split, is divided into multiple subgraphs;C, to each subgraph carry out feature extraction, and be input to handled in convolutional neural networks CNN model after export clear image;D, the clear image of output is sent in retinal defects database and carries out Auto-matching, the detection method that the present invention uses is easy to operate, and accuracy is high, effectively improves the accuracy and timeliness of detection.
Description
Technical field
The present invention relates to retinal images detection technique fields, specially a kind of diabetic retina image lesion detection side
Method.
Background technique
The normal abbreviation retina of pars optica retinae, the transparent film for one layer of softness, is tightly attached to choroid inner face, thoughts light
The effect of stimulation.Retinal thickness is different, generally 0.4mm, and optic disk edge is most thick, about 0.5mm, and central fovea is most thin, is
0.1mm, until ora serrata is 0.15mm.Retina is mainly thin by pigment epithelial cell, visual cell, Beale's ganglion cells, ganglion cell, level
The composition such as born of the same parents, amakrine, interplexiform cell and Muller cell.These cells and its protrusion are orderly aligned, can be accordingly by view
Nethike embrane is divided into 10 layers from the outside to the core.1. pigment epithelial layer: being made of single layer pigment epithelial cell;2. layer of rods and cones: by retinal rod
The evagination of cell and cone cell is constituted;3. external limiting membrane: being formed by connecting by the evagination end of Muller cell;4. outer nuclear layer: by
Rod cell and the cell body of cone cell composition;5. outer plexiform layer: by prominent and Beale's ganglion cells in rod cell and cone cell
Dendron constitute;6. inner nuclear layer: being made of the cell space of Beale's ganglion cells, horizontal cell, amakrine and Muller cell;In 7.
Stratum reticulare: it is made of the dendron of the aixs cylinder of Beale's ganglion cells and amakrine and ganglion cell;8. ganglion-cell layer: by the cell space of ganglion cell
Composition;9. nerve fibre layer: being made of the aixs cylinder of ganglion cell;10. internal limiting membrane: being formed by connecting for end of dashing forward in M ü ller cell.
Most important pathogenic eye illness in US and European population when diabetic retinopathy.It is pre- according to the World Health Organization
It surveys, to the year two thousand thirty, global patient with retinopathy will be added to 3.66 hundred million, and diabetes control will become an even more serious generation
Criticality problem.
Studies have shown that carrying out early diagnosis and therapy to patients with diabetic retinopathy can effectively prevent the loss of vision
And blindness, and the key prevented and treated is then by eye-ground photography inspection, regular follow-up finds the progress of the state of an illness, carries out laser in time
Therapeutic intervention.But be at present in the world more than that 50% patient does not receive any type of examination of eyes, based on eye fundus image
Diabetic retinopathy check that work visually observes progress substantially or by oculist.
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.
Summary of the invention
The purpose of the present invention is to provide a kind of diabetic retina image lesion detection methods, to solve above-mentioned background skill
The problem of being proposed in art.
To achieve the above object, the invention provides the following technical scheme: a kind of diabetic retina image lesion detection side
Method, comprising the following steps:
A, image capture device acquires retinal images and is pre-processed;
B, pretreated image is split, is divided into multiple subgraphs;
C, feature extraction is carried out to each subgraph, and be input to handled in convolutional neural networks CNN model after export it is clear
Clear image;
D, the clear image of output is sent in retinal defects database and carries out Auto-matching.
Preferably, preprocess method is as follows in the step A:
A, the pixel of the retinal images of acquisition being divided into several figure layers according to brightness value, the brightness of each figure layer is different,
And each figure layer is pressed into brightness value, it is arranged from high to low, and the boundary of the image in each figure layer is all by closed curve
It constitutes;
B, the figure layer minimum for brightness and the maximum figure layer of brightness first individually carry out histogram equalization processing, then remove
Ambient noise finally carries out noise removal;
C, the figure layer among minimum brightness and maximum brightness is first removed into noise, then removes ambient noise, finally carried out straight
Square figure equalization processing;
D, the All Layers after finally will be processed merge into the enhanced image of piece image.
Preferably, image partition method is as follows in the step B:
A, corresponding first segmented image of image to be split is obtained, the first segmented image is after treating segmented image progress hyperfractionated
It is formed by, includes multiple regions in the first segmented image;
B, feature extraction is carried out to the two neighboring region of the first segmented image, and obtains current first segmentation figure according to preset algorithm
The edges of regions gradient difference value feature of every two adjacent area as in;
C, enhancing processing is carried out to the first segmented image according to edges of regions gradient difference value feature, obtains the second segmented image;
D, the second segmented image is input in image segmentation network, obtains segmentation result.
Preferably, in the step C, the pixel value of image, definition values, picture size size is extracted and is ranked up.
Compared with prior art, the beneficial effects of the present invention are:
(1) detection method that the present invention uses is easy to operate, and accuracy is high, effectively improves the accuracy and timeliness of detection
Property;
(2) image pre-processing method that the present invention uses reduces the global luminance difference of image, enhances picture contrast, has
Effect inhibits noise, further improves the clarity of image, further improves the accuracy in detection of lesion region;
(3) image partition method that the present invention uses reduces the situation for image segmentation mistake occur, improves the standard of image segmentation
Exactness.
Detailed description of the invention
Fig. 1 is flow diagram of the present invention.
Fig. 2 is image pre-processing method flow diagram of the present invention.
Fig. 3 is image partition method flow diagram of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Embodiment one:
The present invention provides a kind of technical solution referring to FIG. 1-2: a kind of diabetic retina image lesion detection method, including
Following steps:
A, image capture device acquires retinal images and is pre-processed;
B, pretreated image is split, is divided into multiple subgraphs;
C, feature extraction is carried out to each subgraph, extract the pixel value of image, definition values, picture size size and is arranged
Sequence, and be input to handled in convolutional neural networks CNN model after export clear image;
D, the clear image of output is sent in retinal defects database and carries out Auto-matching.
In the present invention, preprocess method is as follows in step A:
A, the pixel of the retinal images of acquisition being divided into several figure layers according to brightness value, the brightness of each figure layer is different,
And each figure layer is pressed into brightness value, it is arranged from high to low, and the boundary of the image in each figure layer is all by closed curve
It constitutes;
B, the figure layer minimum for brightness and the maximum figure layer of brightness first individually carry out histogram equalization processing, then remove
Ambient noise finally carries out noise removal;
C, the figure layer among minimum brightness and maximum brightness is first removed into noise, then removes ambient noise, finally carried out straight
Square figure equalization processing;
D, the All Layers after finally will be processed merge into the enhanced image of piece image.
The image pre-processing method that the present invention uses reduces the global luminance difference of image, enhances picture contrast,
Noise is effectively inhibited, the clarity of image is further improved, further improves the accuracy in detection of lesion region.
Embodiment two:
Fig. 1-3 is please referred to, the present invention provides a kind of technical solution: a kind of diabetic retina image lesion detection method, including
Following steps:
A, image capture device acquires retinal images and is pre-processed;
B, pretreated image is split, is divided into multiple subgraphs;
C, feature extraction is carried out to each subgraph, extract the pixel value of image, definition values, picture size size and is arranged
Sequence, and be input to handled in convolutional neural networks CNN model after export clear image;
D, the clear image of output is sent in retinal defects database and carries out Auto-matching.
In the present invention, preprocess method is as follows in step A:
A, the pixel of the retinal images of acquisition being divided into several figure layers according to brightness value, the brightness of each figure layer is different,
And each figure layer is pressed into brightness value, it is arranged from high to low, and the boundary of the image in each figure layer is all by closed curve
It constitutes;
B, the figure layer minimum for brightness and the maximum figure layer of brightness first individually carry out histogram equalization processing, then remove
Ambient noise finally carries out noise removal;
C, the figure layer among minimum brightness and maximum brightness is first removed into noise, then removes ambient noise, finally carried out straight
Square figure equalization processing;
D, the All Layers after finally will be processed merge into the enhanced image of piece image.
The image pre-processing method that the present invention uses reduces the global luminance difference of image, enhances picture contrast,
Noise is effectively inhibited, the clarity of image is further improved, further improves the accuracy in detection of lesion region.
In the present embodiment, image partition method is as follows in step B:
A, corresponding first segmented image of image to be split is obtained, the first segmented image is after treating segmented image progress hyperfractionated
It is formed by, includes multiple regions in the first segmented image;
B, feature extraction is carried out to the two neighboring region of the first segmented image, and obtains current first segmentation figure according to preset algorithm
The edges of regions gradient difference value feature of every two adjacent area as in;
C, enhancing processing is carried out to the first segmented image according to edges of regions gradient difference value feature, obtains the second segmented image;
D, the second segmented image is input in image segmentation network, obtains segmentation result.
The image partition method that the present invention uses reduces the situation for image segmentation mistake occur, improves the standard of image segmentation
Exactness.
In conclusion the detection method that the present invention uses is easy to operate, accuracy is high, effectively improves the accurate of detection
Property and timeliness.
It although an embodiment of the present invention has been shown and described, for the ordinary skill in the art, can be with
A variety of variations, modification, replacement can be carried out to these embodiments without departing from the principles and spirit of the present invention by understanding
And modification, the scope of the present invention is defined by the appended.
Claims (4)
1. a kind of diabetic retina image lesion detection method, it is characterised in that: the following steps are included:
A, image capture device acquires retinal images and is pre-processed;
B, pretreated image is split, is divided into multiple subgraphs;
C, feature extraction is carried out to each subgraph, and be input to handled in convolutional neural networks CNN model after export it is clear
Clear image;
D, the clear image of output is sent in retinal defects database and carries out Auto-matching.
2. a kind of diabetic retina image lesion detection method according to claim 1, it is characterised in that: the step
Preprocess method is as follows in A:
A, the pixel of the retinal images of acquisition being divided into several figure layers according to brightness value, the brightness of each figure layer is different,
And each figure layer is pressed into brightness value, it is arranged from high to low, and the boundary of the image in each figure layer is all by closed curve
It constitutes;
B, the figure layer minimum for brightness and the maximum figure layer of brightness first individually carry out histogram equalization processing, then remove
Ambient noise finally carries out noise removal;
C, the figure layer among minimum brightness and maximum brightness is first removed into noise, then removes ambient noise, finally carried out straight
Square figure equalization processing;
D, the All Layers after finally will be processed merge into the enhanced image of piece image.
3. a kind of diabetic retina image lesion detection method according to claim 1, it is characterised in that: the step
Image partition method is as follows in B:
A, corresponding first segmented image of image to be split is obtained, the first segmented image is after treating segmented image progress hyperfractionated
It is formed by, includes multiple regions in the first segmented image;
B, feature extraction is carried out to the two neighboring region of the first segmented image, and obtains current first segmentation figure according to preset algorithm
The edges of regions gradient difference value feature of every two adjacent area as in;
C, enhancing processing is carried out to the first segmented image according to edges of regions gradient difference value feature, obtains the second segmented image;
D, the second segmented image is input in image segmentation network, obtains segmentation result.
4. a kind of diabetic retina image lesion detection method according to claim 1, it is characterised in that: the step
In C, extracts the pixel value of image, definition values, picture size size and be ranked up.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910012785.7A CN109691980A (en) | 2019-01-07 | 2019-01-07 | A kind of diabetic retina image lesion detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910012785.7A CN109691980A (en) | 2019-01-07 | 2019-01-07 | A kind of diabetic retina image lesion detection method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109691980A true CN109691980A (en) | 2019-04-30 |
Family
ID=66233166
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910012785.7A Pending CN109691980A (en) | 2019-01-07 | 2019-01-07 | A kind of diabetic retina image lesion detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109691980A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110267029A (en) * | 2019-07-22 | 2019-09-20 | 广州铭维软件有限公司 | A kind of long-range holographic personage's display technology based on AR glasses |
CN110335263A (en) * | 2019-06-28 | 2019-10-15 | 珠海博明软件有限公司 | It is a kind of to identify the measurement scheme for improving 3D difference in height computational accuracy by brightness |
CN112168347A (en) * | 2020-11-10 | 2021-01-05 | 哈尔滨理工大学 | Computer-aided design method for fracture reduction |
CN112529902A (en) * | 2021-01-26 | 2021-03-19 | 江苏卓玉智能科技有限公司 | Hole checking method of PCB (printed circuit board) |
CN114494063A (en) * | 2022-01-25 | 2022-05-13 | 电子科技大学 | Night traffic image enhancement method based on biological vision mechanism |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH08150119A (en) * | 1994-11-30 | 1996-06-11 | Canon Inc | Image processor for ophthalmology |
CN104318542A (en) * | 2014-11-20 | 2015-01-28 | 上海华力创通半导体有限公司 | Image enhancement processing algorithm |
CN105513077A (en) * | 2015-12-11 | 2016-04-20 | 北京大恒图像视觉有限公司 | System for screening diabetic retinopathy |
CN107871321A (en) * | 2016-09-23 | 2018-04-03 | 南开大学 | Image partition method and device |
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 |
-
2019
- 2019-01-07 CN CN201910012785.7A patent/CN109691980A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH08150119A (en) * | 1994-11-30 | 1996-06-11 | Canon Inc | Image processor for ophthalmology |
CN104318542A (en) * | 2014-11-20 | 2015-01-28 | 上海华力创通半导体有限公司 | Image enhancement processing algorithm |
CN105513077A (en) * | 2015-12-11 | 2016-04-20 | 北京大恒图像视觉有限公司 | System for screening diabetic retinopathy |
CN107871321A (en) * | 2016-09-23 | 2018-04-03 | 南开大学 | Image partition method and device |
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 |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110335263A (en) * | 2019-06-28 | 2019-10-15 | 珠海博明软件有限公司 | It is a kind of to identify the measurement scheme for improving 3D difference in height computational accuracy by brightness |
CN110267029A (en) * | 2019-07-22 | 2019-09-20 | 广州铭维软件有限公司 | A kind of long-range holographic personage's display technology based on AR glasses |
CN112168347A (en) * | 2020-11-10 | 2021-01-05 | 哈尔滨理工大学 | Computer-aided design method for fracture reduction |
CN112529902A (en) * | 2021-01-26 | 2021-03-19 | 江苏卓玉智能科技有限公司 | Hole checking method of PCB (printed circuit board) |
CN114494063A (en) * | 2022-01-25 | 2022-05-13 | 电子科技大学 | Night traffic image enhancement method based on biological vision mechanism |
CN114494063B (en) * | 2022-01-25 | 2023-04-07 | 电子科技大学 | Night traffic image enhancement method based on biological vision mechanism |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109691980A (en) | A kind of diabetic retina image lesion detection method | |
Gao et al. | Automatic feature learning to grade nuclear cataracts based on deep learning | |
EP3659067B1 (en) | Method of modifying a retina fundus image for a deep learning model | |
CN109528155B (en) | Intelligent screening system suitable for high myopia complicated with open angle glaucoma and establishment method thereof | |
Bader Alazzam et al. | Identification of diabetic retinopathy through machine learning | |
Tsui et al. | Pediatric retinal conditions imaged by ultra wide field fluorescein angiography | |
CN109464120A (en) | A kind of screening for diabetic retinopathy method, apparatus and storage medium | |
US20230022921A1 (en) | System and method for analyzing corneal lesion using anterior ocular segment image, and computer-readable recording medium | |
CN109671049A (en) | A kind of medical image processing method, system, equipment, storage medium | |
Johari et al. | Early detection of diabetic retinopathy by using deep learning neural network | |
CN109730633A (en) | Choroidal artery angiographic method and equipment based on optical coherence tomography swept-volume | |
Alluwimi et al. | A basis for customising perimetric locations within the macula in glaucoma | |
CN105447874A (en) | Human eye bleeding image detection apparatus | |
Zhang et al. | Artificial intelligence to detect meibomian gland dysfunction from in-vivo laser confocal microscopy | |
An et al. | Association between the types of posterior staphyloma and their risk factors in pathological myopia | |
Saravanan et al. | Automated red lesion detection in diabetic retinopathy | |
Iqbal et al. | Automatic diagnosis of diabetic retinopathy using fundus images | |
Simon | Retinal image enhancement and eye disease identification | |
Kiresur et al. | Automatic detection of diabetic retinopathy in fundus image: a survey | |
Suwandi et al. | A Systematic Literature Review: Diabetic Retinopathy Detection Using Deep Learning | |
Kamothi et al. | Detection of diabetic retinopathy using transfer learning approach | |
Sathya et al. | Hybrid Enhancement Technique of Retinal Fundus Image Using Reconstruction Filter Segmentation with Various Classification and Optimization Methods | |
Li et al. | Computational Model-Based Estimation of Mouse Eyeball Structure From Two-Dimensional Flatmount Microscopy Images | |
Gayathri et al. | Extraction of Textural Features from Retinal Fundus Images | |
Alazzam et al. | Research Article Identification of Diabetic Retinopathy through Machine Learning |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190430 |