CN103854284A - Cutting method for serous pigment epithelium layer disengagement retina based on three-dimensional diagram searching - Google Patents
Cutting method for serous pigment epithelium layer disengagement retina based on three-dimensional diagram searching Download PDFInfo
- Publication number
- CN103854284A CN103854284A CN201410066419.7A CN201410066419A CN103854284A CN 103854284 A CN103854284 A CN 103854284A CN 201410066419 A CN201410066419 A CN 201410066419A CN 103854284 A CN103854284 A CN 103854284A
- Authority
- CN
- China
- Prior art keywords
- interface
- retina
- layer
- image
- pigment epithelium
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Eye Examination Apparatus (AREA)
Abstract
The invention discloses a cutting method for a serous pigment epithelium layer disengagement retina based on three-dimensional diagram searching. The cutting method comprises (1) the method of rapid B scanning image alignment method based on retina upper interface cutting, (2) the method of multi-resolution image searching cutting method according to an interface obviousness degree sequence and with the cut interface as constraint conditions, (3) the method of obtaining a pigment epithelium layer lower interface with an upheaval area and a smooth retina bottom interface by utilizing the different restraint conditions to carry out an image searching algorithm, (4) the cutting method of serous pigment epithelium layer disengagement based on the position difference between the pigment epithelium layer lower interface and the retina bottom interface and by combining the size and brightness information of the area, and (5) the method of carrying out outer-layer retina hierarchical segmentation and correction after the image is smoothened. The cutting result of the cutting method for the serous pigment epithelium layer disengagement retina based on three-dimensional diagram searching has high accuracy, and the cutting method can replace manual cutting and has an important auxiliary function on diagnosis and treatment of clinical related ophthalmic diseases.
Description
Technical field
The invention belongs to retinal image segmentation method, especially to SD-OCT(domain optical coherence fault imaging) retinal images in organisational level and the dividing method of lesion region.
Background technology
Retina is the extension of brain neuroblastoma tissue, has complicated multi-level institutional framework.SD-OCT technology has become the strong instrument of one of nondestructive evaluation retinal disease, and it can provide fast, high-resolution, the 3-D view that shows retina interior laminate layer, for diagnosis and the treatment of Clinical Ophthalmology doctor to disease provides help.Cutting apart clinical practice of retina OCT image is significant: the cutting apart and the quantitative test of its shape, size, position is had to key effect to medical diagnosis on disease and treatment of lesion region; Retinal tissue level cut apart and to the quantitative test of various organization form, brightness for find early lesion, observation the course of disease and research pathology all play an important role.But the PVR that current most of oculist adopts manual mode to show OCT carries out quantitative test, subjectivity is strong, cannot guarantee accuracy and consistance, and is difficult to the mass data that multianalysis 3-D scanning brings.
There is following defect in current retina OCT Image Automatic Segmentation algorithm: (1) most of algorithm is all two-dimentional algorithm, at each sectioning image (x-z plane picture, be called B scan image) in independently cut apart, these class methods do not make full use of three-dimensional contextual information, be easier to be subject to the impact of picture noise or artifact, cause segmentation errors.(2) most of existing retinal tissue level partitioning algorithm is all for normal retina design, and in the time that retinal tissue produces larger deformation due to pathology, these algorithms will lose efficacy.
Serosity detachment of pigment epithelium may be caused by multiple choroid/retinal disease, as AMD, polypoidal choroidal vasculopathy in Chinese patients, central serous chorioretinopathy, uveitis etc.So far, also there is no the relevant report for the three-dimensional automatic division method of the system of all distinguishable organisational levels and lesion region in the retina OCT image of serosity detachment of pigment epithelium.
Summary of the invention
The invention provides a kind of solution of the above problems, a kind of three-dimensional automatic division method for the system of all distinguishable organisational levels and lesion region in the retina OCT image of serosity detachment of pigment epithelium with feasibility and validity is provided first.Wherein organisational level comprises 10 layers: nerve fibre layer (NFL), ganglion-cell layer (GCL), inner molecular layer (IPL), inner nuclear layer (INL), external plexiform layer (OPL), outer nuclear layer (ONL)+interior ganglionic layer (ISL), connection cilium (CL), outer ganglionic layer (OSL), verhoeff's membrane (VM), pigment epithelial layer (RPE), have 11 each interphases.Add due to pigment epithelial layer and the disengaging of retina bottom, an independent interface is formed on retina bottom, and therefore the present invention can detect 12 interphases altogether.
The invention provides a kind of retina dividing method based on three-dimensional plot search serosity detachment of pigment epithelium, the method mainly comprises 5 steps:
Step S01, image pre-service: mainly carry out the alignment between OCT denoising and B scan image;
Step S02, cutting apart of the each level of inner retina: adopt multi-resolution images search algorithm, cut apart successively according to interphase contrast order from high in the end, obtain nerve fibre layer (NFL), ganglion-cell layer (GCL), inner molecular layer (IPL), inner nuclear layer (INL), external plexiform layer (OPL), the interphase of outer nuclear layer (ONL)+interior ganglionic layer (ISL);
Step S03, pigment epithelial layer is cut apart and retina bottom is estimated: in outer retinal area, carry out figure search algorithm obtain interphase bottom the sliding retina of interface peace under the pigment epithelial layer of upwarped district by different constraint condition;
Step S04, detachment of pigment epithelium Region Segmentation: the region between the retina bottom interphase that under pigment epithelial layer, interface peace is slided is detachment of pigment epithelium region, and removes flase drop region according to area size and monochrome information;
Step S05, cutting apart of the each level of outer retina: will detect the each level of outer retina with figure search algorithm after image planarization according to interface under pigment epithelial layer, obtain connecting cilium (CL), interphase between outer ganglionic layer (OSL), verhoeff's membrane (VM), pigment epithelial layer (RPE).
Above-mentioned 5 steps specifically describe as follows,
(1) image pre-service
Image pre-service mainly comprises following two steps: denoising and the alignment of B scan image.
(a) OCT image denoising
The 3-D view that OCT ocular imaging instrument obtains contains more speckle noise.For guaranteeing the follow-up effect of cutting apart, must in effectively removing noise, retain as far as possible the marginal information in image.The present invention adopts a kind of two-sided filter fast to carry out denoising to each B scan image.Bilateral filtering result is:
Wherein
Here p is the pixel when pre-treatment, the pixel in the neighborhood S that q is p, I
pand I
qbe respectively the gray-scale value of p and q,
for the gray-scale value of filtering result,
for normalization coefficient,
with
that standard deviation is respectively σ
sand σ
rgaussian function, σ
sand σ
rthese two parameters are carried out value according to picture size size and contrast on border size respectively.
(b) B scan image alignment
In imaging process, the motion of eye can cause retina position in continuous B scan image to fluctuate up and down, and image is upper discontinuous at slice direction (y-direction).This can cause difficulty to three-dimensional segmentation.The present invention is based on the segmentation result at interface on retina is carried out to the alignment of B scan image, because this interface contrast in all levels interphase is the highest, even also can correctly cut apart on the image of dislocation.The process of cutting apart with multi-resolution images search algorithm is as follows: the 3-D view after denoising is carried out on vertical direction (z-direction) to down-sampling and make this direction pixel number become half, repeat once this process, obtain the image of three different resolutions, be expressed as from low to high yardstick 1,2,3 by resolution.Cut apart first and carry out on the yardstick 1 of lowest resolution, on the basis of acquired results, on yardstick 2, near zone, carry out further Accurate Segmentation, the like, the segmentation result on original image finally obtained.Cutting procedure is the process of finding the divisional plane of Least-cost, is completed by figure search algorithm.On nerve fibre layer, the cost function at interface calculates with Sobel Operator, by secretly less to bright marginal position cost function.Distinguish in order to be connected interface on cilium with ectonexine interphase, on yardstick 1, cost function has added another component, and this component is the brightness sum of the some pixels in each picture point top.Like this, because top, interface on nerve fibre layer is darker background area, the cost of its correspondence is just less than the cost that connects interface location on cilium, can correctly be detected.After interface segments on nerve fibre layer, on every B scan image, calculate its average height, i.e. average z value.The point in rejection image centre position in computation process, because these points are subject to the impact of central fovea or pathology to have larger displacement.According to moving or move down this image in the average height of interface on the nerve fibre layer obtaining in every B scanning, making in result interface average height on nerve fibre layer is a constant, has just played the effect of each image that aligns
(2) cutting apart of the each level of inner retina
Inner retina is subject to the impact of pathology less, therefore first cuts apart.Its dividing method be with step (1) in similar multi-resolution images search method.First detect and connect interface on cilium, the interphase of ectonexine is as constraint condition.Be segmented on nerve fibre layer and carry out in the subgraph below interface, because pathology causes hydrops in retina, epiretinal portion of tissue is not developed in OCT image, therefore the ectonexine interphase that detects is actual to be merged and forms for connecting interface and pigment epithelial layer lower surface on cilium, and it is defined as and connects cilium pigment epithelial layer and merge interface.Then, according to each interphase contrast on border order, take the interface that is partitioned into as constraint condition, cut apart ganglion-cell layer (GCL), inner molecular layer (IPL), inner nuclear layer (INL), external plexiform layer (OPL), the interphase of outer nuclear layer (ONL)+interior ganglionic layer (ISL).The interface that boundary contrast is lower may cannot effectively be cut apart on low yardstick, therefore need start to cut apart from higher resolution.For eliminating the impact of noise, the result obtaining is carried out in x direction to mean filter, to obtain the level and smooth interface of cutting apart.
(3) pigment epithelial layer is cut apart and the estimation of retina bottom
In the retina OCT of serosity detachment of pigment epithelium image, pigment epithelial layer is smooth protuberance in disengaging region, and the variation in the B scan image of front and back is larger.And below is darker hydrops region, its bottom interface may not developed.The present invention, in the time detecting this two interfaces based on figure search algorithm, adopts identical cost function and different interface smoothness constraint condition.In the time that smoothness constrained parameters are larger,, generally in the time that smoothness constrained parameters get 5~10, segmentation result is to have interface under the pigment epithelial layer of local eminence.When parameter value hour, smoothness constrained parameters get 1~4, time, segmentation result is level and smooth retina bottom interphase.
(4) depart from Region Segmentation
It is detachment of pigment epithelium region that step (3) is cut apart the region between the interphase of interface and retina bottom under the pigment epithelial layer obtaining.But cut apart local error because the impact of noise causes interface, may occur the situation of flase drop.First all pixels between interface under pigment epithelial layer and retina bottom interphase are formed to several three-dimensional communication regions, calculate respectively volume and the mean flow rate in these regions.In the time that volume is less than a certain predetermined value or mean flow rate and is greater than a certain predetermined value, think that this region is that the disengaging region of flase drop is removed.
In obtaining three-dimensional disengaging region, also can obtain the two-dimensional distribution of territory, abscission zone in x-y direction.This is by correction to outer retina segmentation result for next step.
(5) cutting apart of the each level of outer retina
Outer retina is respectively organized in and in normal retina, is more smooth shape.In the time there is detachment of pigment epithelium, these are organized also and swell along with the protuberance of pigment epithelial layer, and may not develop above upwarped district, therefore in OCT image, these organize discontinuous, therefore must add the correctness of certain constraint condition to guarantee to cut apart.In the present invention, the result of cutting apart based on interface under pigment epithelial layer, by retinal images planarization, moves up and down by each row in image, makes interface under pigment epithelial layer become a plane.So just pigment epithelial layer bump is reverted to smooth shape, also just can be similar to and recover outer retina even shape under normal circumstances.On image after planarization, first merge interface based on cutting apart the connection cilium pigment epithelial layer obtaining in step (2), according to step (4) result, carry out interpolation correction to obtain connecting interface on cilium in detachment of pigment epithelium region by second order polynomial.Then cut apart the interface of outer ganglionic layer (OSL), verhoeff's membrane (VM), pigment epithelial layer (RPE) with figure search algorithm, externally the interphase of ganglionic layer (OSL), verhoeff's membrane (VM), also needs in detachment of pigment epithelium region to carry out interpolation correction by second order polynomial.The interphase of the connection cilium (CL) obtaining in this step, outer ganglionic layer (OSL), verhoeff's membrane (VM), pigment epithelial layer (RPE) is remapped back on original image and obtains final segmentation result.
The present invention has been merged bilateral filtering denoising, B scanning alignment, three-dimensional plot and has been cut the steps such as technology, connected region are cut apart, image planarization, segmentation result correction, segmentation result has higher accuracy, can substitute manually and cut apart, can play important booster action for the Clinics and Practices of clinical relevant ophthalmology disease.
Accompanying drawing explanation
Fig. 1 is structural representation of the present invention;
Fig. 2 is retinal tissue hierarchy chart picture, and Fig. 2 (a) is former B scan image, 10 layer tissues that Fig. 2 (b) is normal retina and 11 interphases;
Fig. 3 is two-sided filter denoising result, and Fig. 3 (a) is original image, and Fig. 3 (b) is image after denoising;
Fig. 4 is B scanning alignment result x-y planimetric map, Fig. 4 (a) original image, image after Fig. 4 (b) alignment;
Fig. 5 is for connecting cilium (CL), outer ganglionic layer (OSL), verhoeff's membrane (VM), the interface segmentation result of pigment epithelial layer (RPE), Fig. 5 (a) is interface and retina bottom interphase segmentation result under pigment epithelial layer, Fig. 5 (b) is according to the image after interface planarization under pigment epithelial layer, Fig. 5 (c) is the segmentation result to interface 7-10 in Fig. 5 (b), Fig. 5 (d) connects cilium (CL) for after shining upon back original image, outer ganglionic layer (OSL), verhoeff's membrane (VM), result (the wherein verhoeff's membrane (VM) at the interface of pigment epithelial layer (RPE), the interface of pigment epithelial layer (RPE) is substantially overlapping in this image),
Fig. 6 is serous pigmentary epithelial detachment OCT image segmentation result, the two dimension that Fig. 6 (a) is layering result shows, be 12 interphases from top to bottom, Fig. 6 (b) is interface on nerve fibre layer, connect the 3-D display of interface under interface on cilium, pigment epithelial layer, retina bottom interphase segmentation result, Fig. 6 (c) is for departing from the two dimension demonstration of Region Segmentation, and Fig. 6 (d) is for departing from the 3-D display of Region Segmentation.
In Fig. 2, Reference numeral is as follows, 1 nerve fibre layer (NFL), 2 ganglion-cell layers (GCL), 3 inner molecular layers (IPL), 4 inner nuclear layers (INL), 5 external plexiform layers (OPL), 6 outer nuclear layers (ONL)+interior ganglionic layer (ISL), 7 connect cilium (CL), 8 outer ganglionic layers (OSL), 9 verhoeff's membranes (VM), 10 pigment epithelial layers (RPE).
Embodiment
Below in conjunction with embodiment, further set forth the present invention.
Shown in Figure 1, this method mainly comprises 5 steps: the cutting apart of image pre-service, the each level of inner retina, pigment epithelial layer are cut apart and retina bottom is estimated, departed from cutting apart of Region Segmentation, the each level of outer retina.
As shown in Figure 2, retinal tissue level comprises 10 layers: nerve fibre layer (NFL) 1, ganglion-cell layer (GCL) 2, inner molecular layer (IPL) 3, inner nuclear layer (INL) 4, external plexiform layer (OPL) 5, outer nuclear layer (ONL)+interior ganglionic layer (ISL) 6, connection cilium (CL) 7, outer ganglionic layer (OSL) 8, verhoeff's membrane (VM) 9, pigment epithelial layer (RPE) 10, have 11 interphases.
A kind of retina dividing method based on three-dimensional plot search serosity detachment of pigment epithelium of the present invention, specifically describe as follows,
(1) image pre-service
Image pre-service mainly comprises following two steps: denoising and the alignment of B scan image.
(a) OCT image denoising
The 3-D view that OCT ocular imaging instrument obtains contains more speckle noise.For guaranteeing the follow-up effect of cutting apart, must in effectively removing noise, retain as far as possible the marginal information in image.The present invention adopts a kind of two-sided filter fast to carry out denoising to each B scan image.Bilateral filtering result is:
Wherein
Here p is the pixel when pre-treatment, the pixel in the neighborhood S that q is p.I
pand I
qbe respectively the gray-scale value of p and q.
for the gray-scale value of filtering result,
for normalization coefficient.
with
that standard deviation is respectively σ
sand σ
rgaussian function.σ
sand σ
rthese two parameters are carried out value according to picture size size and contrast on border size respectively.Denoising result as shown in Figure 3.
(b) B scan image alignment
In imaging process, the motion of eye can cause retina position in continuous B scan image to fluctuate up and down, and image is upper discontinuous at slice direction (y-direction).This can cause difficulty to three-dimensional segmentation.The present invention is based on interface on nerve fibre layer 1, on retina, the segmentation result at interface carries out the alignment of B scan image, because this interface contrast in all levels interphase is the highest, even also can correctly cut apart on the image of dislocation.The process of cutting apart interface on nerve fibre layer 1 with multi-resolution images search algorithm is as follows: the 3-D view after denoising is carried out on vertical direction (z-direction) to down-sampling and make this direction pixel number become half, repeat once this process, obtain the image of three different resolutions, be expressed as from low to high yardstick 1, yardstick 2 and yardstick 3 by resolution.Cut apart first and carry out on the yardstick 1 of lowest resolution, on the basis of acquired results, on yardstick 2, near zone, carry out further Accurate Segmentation, the like, the segmentation result on original image finally obtained.
Cutting procedure is the process of finding the divisional plane of Least-cost, is completed by figure search algorithm.On nerve fibre layer 1, the cost function at interface calculates with Sobel Operator, by secretly less to bright marginal position cost function.Distinguish in order to be connected interface on cilium 7 with ectonexine interphase, on yardstick 1, cost function has added another component, and this component is the brightness sum of the some pixels in each picture point top.Like this, because top, interface on nerve fibre layer 1 is darker background area, the cost of its correspondence is just less than the cost that connects interface location on cilium 7, can correctly be detected.After interface segments on nerve fibre layer 1, on every B scan image, calculate its average height, i.e. average z value.The point in rejection image centre position in computation process, because these points are subject to the impact of central fovea or pathology to have larger displacement.According to moving or move down this image in the average height of interface on the nerve fibre layer 1 obtaining in every B scanning, making in result interface average height on nerve fibre layer 1 is a constant, has just played the effect of each image that aligns.Image effect after alignment can find out from x-y direction, as shown in Figure 4.
(2) cutting apart of the each level of inner retina
Inner retina is subject to the impact of pathology less, therefore first cuts apart.Its dividing method be with step (1) in similar multi-resolution images search method.
First detect and connect interface on cilium 7, the interphase of ectonexine is as constraint condition.Be segmented on nerve fibre layer 1 and carry out in the subgraph below interface, because pathology causes hydrops in retina, epiretinal portion of tissue is not developed in OCT image, therefore the ectonexine interphase that detects is actual to be merged and forms for connecting on cilium 7 interphase on interface and pigment epithelial layer 10, and it is defined as and connects cilium pigment epithelial layer and merge interface.
Then, according to each interphase contrast on border order, take the interface that is partitioned into as constraint condition, cut apart ganglion-cell layer (GCL) 2, inner molecular layer (IPL) 3, inner nuclear layer (INL) 4, external plexiform layer (OPL) 5, the interface of outer nuclear layer (ONL)+interior ganglionic layer (ISL) 6.The interface that boundary contrast is lower may cannot effectively be cut apart on low yardstick, therefore need start to cut apart from higher resolution.Specifically cut apart order, up and down the border variation of containment surfaces, correspondence and the initial yardstick of multi-resolution segmentation as shown in table 1.For eliminating the impact of noise, the result obtaining is carried out in x direction to mean filter, to obtain the level and smooth interface of cutting apart.
(3) pigment epithelial layer is cut apart and the estimation of retina bottom
In the retina OCT of serosity detachment of pigment epithelium image, pigment epithelial layer is smooth protuberance in disengaging region, and the variation in the B scan image of front and back is larger.And below is darker hydrops region, its bottom interface may not developed.The present invention, in the time detecting this two interfaces based on figure search algorithm, adopts identical cost function and different interface smoothness constraint condition.In the time that smoothness constrained parameters get 5~10, segmentation result is to have interface under the pigment epithelial layer of local eminence.When smoothness constrained parameters get 1~4 constantly, segmentation result is level and smooth retina bottom interphase.
(4) depart from Region Segmentation
The region that step (3) is cut apart between 10 times interfaces of pigment epithelial layer and the retina bottom interphase obtaining is detachment of pigment epithelium region.But cut apart local error because the impact of noise causes interface, may occur the situation of flase drop.First all pixels between 10 times interfaces of pigment epithelial layer and retina bottom interphase are formed to several three-dimensional communication regions, calculate respectively volume and the mean flow rate in these regions.In the time that volume is less than a certain predetermined value or mean flow rate and is greater than a certain predetermined value, think that this region is that the disengaging region of flase drop is removed.
In obtaining three-dimensional disengaging region, also can obtain the two-dimensional distribution of territory, abscission zone in x-y direction.This is by correction to outer retina segmentation result for next step.
(5) cutting apart of the each level of outer retina
Outer retina is respectively organized in and in normal retina, is more smooth shape.In the time there is detachment of pigment epithelium, these are organized also and swell along with the protuberance of pigment epithelial layer, and may not develop above upwarped district, therefore in OCT image, these organize discontinuous, therefore must add the correctness of certain constraint condition to guarantee to cut apart.
In the present invention, the result of cutting apart based on 10 times interfaces of pigment epithelial layer, by retinal images planarization, moves up and down by each row in image, makes 10 times interfaces of pigment epithelial layer become a plane.So just pigment epithelial layer bump is reverted to smooth shape, also just can be similar to and recover outer retina even shape under normal circumstances.On image after planarization, first merge interface based on cutting apart the connection cilium pigment epithelial layer obtaining in step (2),
According to step (4) result, carry out interpolation correction to obtain connecting interface on cilium 7 in detachment of pigment epithelium region by second order polynomial.Then cut apart the interphase of outer ganglionic layer (OSL) 8, verhoeff's membrane (VM) 9, pigment epithelial layer (RPE) 10 with figure search algorithm, cut apart order and constraint condition etc. equally in table 1.To cutting apart the interphase of outer ganglionic layer (OSL) 8, verhoeff's membrane (VM) 9, also need in detachment of pigment epithelium region to carry out interpolation correction by second order polynomial.Remapped back on original image and obtain final segmentation result in the interface of the connection cilium (CL) 7 obtaining in this step, outer ganglionic layer (OSL) 8, verhoeff's membrane (VM) 9, pigment epithelial layer (RPE) 10.
This step results as shown in Figure 5.
The each level interphase of table 1 retina dividing method
(6) experimental result
Part of test results as shown in Figure 6.
Level is cut apart, take the mean value of the independent manually segmentation results of two experts as goldstandard.Automatic segmentation result and goldstandard are cut apart the absolute value of difference of interface z value for cutting apart error.The absolute value of the difference of two expert's segmentation result z values represents difference between observer.Experimental result on 20 samples shows, it is 2.25 ± 0.96 pixels (7.87 ± 3.36 microns) that average error is cut apart in the present invention, compared with 2.23 ± 0.73 pixels of difference between observer (7.81 ± 2.56 microns), without obvious statistical discrepancy, can think basic identical.Therefore this method can substitute manual dividing method.
Pigment epithelial layer is departed to Region Segmentation, using a manual segmentation result of expert as goldstandard, adopts True Positive Rate TPVF and false positive rate FPVF as the objective indicator of appraisal procedure, be calculated as follows:
Wherein || represent volume, C
tPrepresent true positives point set, C
ePrepresent false positive point set, C
gTrepresent to depart from region point set in goldstandard, V represents all pixel set of whole retinal area.Experimental result shows, this method Average True positive rate is 87.9%, and average false positive rate is 0.36%.
So far, a kind of automatic division method of the retina SD-OCT image that is applicable to serosity detachment of pigment epithelium has been realized and has verified.The present invention has been merged bilateral filtering denoising, B scanning alignment, three-dimensional plot and has been cut the steps such as technology, connected region are cut apart, image planarization, segmentation result correction, segmentation result has higher accuracy, can substitute manually and cut apart, can play important booster action for the Clinics and Practices of clinical relevant ophthalmology disease.
Ultimate principle of the present invention, principal character and advantage have more than been described.The technician of the industry should understand; the present invention is not restricted to the described embodiments; that in above-described embodiment and instructions, describes just illustrates principle of the present invention; without departing from the spirit and scope of the present invention; the present invention also has various changes and modifications, and these changes and improvements all fall in the claimed scope of the invention.The claimed scope of the present invention is defined by appending claims and equivalent thereof.
Claims (8)
1. the retina dividing method based on three-dimensional plot search serosity detachment of pigment epithelium, is characterized in that, comprise the following steps,
Step S01, image pre-service: carry out the alignment between OCT image denoising and B scan image;
Step S02, cutting apart of the each level of inner retina: adopt multi-resolution images search algorithm, cut apart successively according to interphase contrast order from high in the end, obtain the interphase of nerve fibre layer, ganglion-cell layer, inner molecular layer, inner nuclear layer, external plexiform layer, outer nuclear layer+interior ganglionic layer;
Step S03, pigment epithelial layer is cut apart and retina bottom is estimated: in outer retinal area, carry out figure search algorithm obtain interphase bottom the sliding retina of interface peace under the pigment epithelial layer of upwarped district with different smoothness constrained parameters;
Step S04, detachment of pigment epithelium Region Segmentation: the region under pigment epithelial layer between interface and retina bottom interphase is detachment of pigment epithelium region, removes flase drop region according to area size and monochrome information;
Step S05, the cutting apart of the each level of outer retina: will detect the each level of outer retina with figure search algorithm after image planarization according to interface under pigment epithelial layer, and obtain connecting the interphase of cilium, outer ganglionic layer, verhoeff's membrane, pigment epithelial layer.
2. the retina dividing method based on three-dimensional plot search serosity detachment of pigment epithelium according to claim 1, other are characterised in that, in step S01, OCT image denoising specifically comprises the following steps, adopt a kind of quick two-sided filter to carry out obtaining the 3-D view after denoising after denoising to each B scan image, bilateral filtering result is suc as formula (1a):
Wherein
P is the pixel when pre-treatment, the pixel in the neighborhood S that q is p, I
pand I
qbe respectively the gray-scale value of p and q,
for the gray-scale value of filtering result,
for normalization coefficient,
with
that standard deviation is respectively σ
sand σ
rgaussian function, σ
sand σ
rthese two parameters are carried out value according to picture size size and contrast on border size respectively.
3. the retina dividing method based on three-dimensional plot search serosity detachment of pigment epithelium according to claim 1, other are characterised in that, being aligned to based on the segmentation result at interface on nerve fibre layer being carried out to the alignment of B scan image between described B scan image, specifically comprises the following steps
1-1) cut apart interface on nerve fibre layer with multi-resolution images search algorithm: the 3-D view in the vertical direction after denoising is carried out to down-sampling and make this direction pixel number become half, repeat once this process, obtain the image of three different resolutions, be expressed as from low to high yardstick 1, yardstick 2, yardstick 3 by resolution;
1-2) on the yardstick 1 of lowest resolution, cut apart, cut apart at yardstick 1 on the basis of acquired results, on yardstick 2, near zone, carry out further Accurate Segmentation, then on yardstick 3, near zone continues to cut apart, and finally obtains the segmentation result on original image;
1-3) after on nerve fibre layer, interface segments into, on every B scan image, calculate its average height, it is average z value, the point in rejection image centre position in computation process, according to moving or move down this image in the average height of interface on the nerve fibre layer obtaining in every B scanning, making in result interface average height on nerve fibre layer is constant, has just played the effect of each image that aligns.
4. the retina dividing method based on three-dimensional plot search serosity detachment of pigment epithelium according to claim 3, it is characterized in that: step 1-2) described in be divided into the process of divisional plane of finding Least-cost, completed by figure search algorithm, on nerve fibre layer, the cost function at interface calculates with Sobel Operator, by secretly less to bright marginal position cost function, distinguish in order to be connected interface on cilium with ectonexine interphase, on yardstick 1, cost function has added one-component, described component is the brightness sum of the some pixels in each picture point top.
5. the retina dividing method based on three-dimensional plot search serosity detachment of pigment epithelium according to claim 1, other are characterised in that, cutting apart specifically of the each level of described inner retina comprises the following steps,
First detect and connect interface on cilium, be that the interphase of ectonexine is as constraint condition, be segmented on nerve fibre layer and carry out in the subgraph below interface, because pathology causes hydrops in retina, epiretinal portion of tissue is not developed in OCT image, therefore the ectonexine interphase that detects is actual to be merged and forms for connecting on cilium interphase on interface and pigment epithelial layer, and it is defined as and connects cilium pigment epithelial layer and merge interface;
Then,, according to above-mentioned each interphase contrast on border order, take the interface that is partitioned into as constraint condition, cut apart ganglion-cell layer, inner molecular layer, inner nuclear layer, external plexiform layer, outer nuclear layer+internal segment bed interface.
6. the retina dividing method based on three-dimensional plot search serosity detachment of pigment epithelium according to claim 1, other are characterised in that, described step S03 pigment epithelial layer is cut apart and retina bottom is estimated specifically to comprise the following steps, during based on interface under figure search algorithm detection pigment epithelial layer and retina bottom interphase, adopt identical cost function and different interface smoothness constraint condition, in the time that smoothness constrained parameters get 5~10, segmentation result is to have interface under the pigment epithelial layer of local eminence, in the time that smoothness constrained parameters get 1~4, segmentation result is level and smooth retina bottom interphase.
7. the retina dividing method based on three-dimensional plot search serosity detachment of pigment epithelium according to claim 1, other are characterised in that, described step S04, detachment of pigment epithelium Region Segmentation specifically comprises,
All pixels between interface under pigment epithelial layer and retina bottom interphase are formed to several three-dimensional communication regions, calculate respectively volume and the mean flow rate in described three-dimensional communication region, in the time that the volume in region is less than a certain predetermined value or mean flow rate and is greater than a certain predetermined value, the disengaging region that described region is flase drop.
8. the retina dividing method based on three-dimensional plot search serosity detachment of pigment epithelium according to claim 1, other are characterised in that, cutting apart specifically of the each level of described step S05 outer retina comprises the following steps,
The result of cutting apart based on interface under pigment epithelial layer, by retinal images planarization, move up and down by each row in image, make interface under pigment epithelial layer become a plane, so just pigment epithelial layer bump is reverted to smooth shape, recover outer retina even shape under normal circumstances;
On retinal images after planarization, merge interface based on cutting apart the connection cilium pigment epithelial layer obtaining in step S02, according to the result of step S04, carry out interpolation correction to obtain connecting interface on cilium in detachment of pigment epithelium region by second order polynomial, cut apart outer ganglionic layer with figure search algorithm, verhoeff's membrane, pigment epithelium bed interface, externally ganglionic layer, verhoeff's membrane interphase, carry out interpolation correction in detachment of pigment epithelium region by second order polynomial, by connection cilium obtained above, outer ganglionic layer, verhoeff's membrane, remap back on original image and obtain final segmentation result in pigment epithelium bed interface.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410066419.7A CN103854284B (en) | 2014-02-26 | 2014-02-26 | Based on graphics search serous pigmentary epithelial pull-up from retina dividing method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410066419.7A CN103854284B (en) | 2014-02-26 | 2014-02-26 | Based on graphics search serous pigmentary epithelial pull-up from retina dividing method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103854284A true CN103854284A (en) | 2014-06-11 |
CN103854284B CN103854284B (en) | 2016-08-31 |
Family
ID=50861896
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410066419.7A Active CN103854284B (en) | 2014-02-26 | 2014-02-26 | Based on graphics search serous pigmentary epithelial pull-up from retina dividing method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103854284B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106997594A (en) * | 2016-01-26 | 2017-08-01 | 上海联影医疗科技有限公司 | The localization method and device of a kind of part tissue of eye |
WO2017171643A1 (en) * | 2016-03-28 | 2017-10-05 | Agency For Science, Technology And Research | Three-dimensional representation of skin structure |
WO2018175654A1 (en) * | 2017-03-23 | 2018-09-27 | Doheny Eye Institute | Systems, methods, and devices for optical coherence tomgraphy multiple enface angiography averaging |
CN109886969A (en) * | 2019-01-15 | 2019-06-14 | 南方医科大学 | A kind of three-dimensional automatic division method of respiratory tract endoscopic optical coherent faultage image |
CN110010219A (en) * | 2019-03-13 | 2019-07-12 | 杭州电子科技大学 | Optical coherence tomography image retinopathy intelligent checking system and detection method |
CN110163838A (en) * | 2019-04-01 | 2019-08-23 | 江西比格威医疗科技有限公司 | A kind of retina OCT image hierarchical algorithm |
CN117893742A (en) * | 2024-03-15 | 2024-04-16 | 深圳大学总医院 | Uveitis image feature recognition method |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120020539A1 (en) * | 2010-07-21 | 2012-01-26 | Peter Derr | Method and System for Analyzing Optical Coherence Tomography (OCT) Results using Color Reflectivity Discretization Analysis |
CN103514605A (en) * | 2013-10-11 | 2014-01-15 | 南京理工大学 | Choroid layer automatic partitioning method based on HD-OCT retina image |
-
2014
- 2014-02-26 CN CN201410066419.7A patent/CN103854284B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120020539A1 (en) * | 2010-07-21 | 2012-01-26 | Peter Derr | Method and System for Analyzing Optical Coherence Tomography (OCT) Results using Color Reflectivity Discretization Analysis |
CN103514605A (en) * | 2013-10-11 | 2014-01-15 | 南京理工大学 | Choroid layer automatic partitioning method based on HD-OCT retina image |
Non-Patent Citations (3)
Title |
---|
MONA K. GARVIN ET AL.: "Intraretinal Layer Segmentation of Macular Optical Coherence Tomography Images Using Optimal 3-D Graph Search", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》, vol. 27, no. 10, 10 October 2008 (2008-10-10), XP011226538, DOI: doi:10.1109/TMI.2008.923966 * |
STEPHANIE J. CHIU ET AL.: "Validated Automatic Segmentation of AMD Pathology Including Drusen and Geographic Atrophy in SD-OCT Images", 《CLINICAL TRIALS》, vol. 53, no. 1, 31 January 2012 (2012-01-31) * |
李俊峰: "双边滤波算法的快速实现及其在图像处理的应用", 《中国优秀硕士学位论文全文数据库-信息科技辑》, 31 December 2013 (2013-12-31) * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106997594A (en) * | 2016-01-26 | 2017-08-01 | 上海联影医疗科技有限公司 | The localization method and device of a kind of part tissue of eye |
CN106997594B (en) * | 2016-01-26 | 2020-12-01 | 上海联影医疗科技股份有限公司 | Method and device for positioning eye tissue |
US10856799B2 (en) | 2016-03-28 | 2020-12-08 | Agency For Science, Technology And Research | Three-dimensional representation of skin structure |
WO2017171643A1 (en) * | 2016-03-28 | 2017-10-05 | Agency For Science, Technology And Research | Three-dimensional representation of skin structure |
WO2018175654A1 (en) * | 2017-03-23 | 2018-09-27 | Doheny Eye Institute | Systems, methods, and devices for optical coherence tomgraphy multiple enface angiography averaging |
US11452443B2 (en) | 2017-03-23 | 2022-09-27 | Doheny Eye Institute | Systems, methods, and devices for optical coherence tomography multiple enface angiography averaging |
US10918274B2 (en) | 2017-03-23 | 2021-02-16 | Doheny Eye Institute | Systems, methods, and devices for optical coherence tomography multiple enface angiography averaging |
US10456031B2 (en) | 2017-03-23 | 2019-10-29 | Doheny Eye Institute | Systems, methods, and devices for optical coherence tomography multiple enface angiography averaging |
CN109886969B (en) * | 2019-01-15 | 2023-03-17 | 南方医科大学 | Three-dimensional automatic segmentation method for respiratory tract endoscopic optical coherence tomography image |
CN109886969A (en) * | 2019-01-15 | 2019-06-14 | 南方医科大学 | A kind of three-dimensional automatic division method of respiratory tract endoscopic optical coherent faultage image |
CN110010219B (en) * | 2019-03-13 | 2021-12-10 | 杭州电子科技大学 | Intelligent detection system and detection method for retinopathy by optical coherence tomography |
CN110010219A (en) * | 2019-03-13 | 2019-07-12 | 杭州电子科技大学 | Optical coherence tomography image retinopathy intelligent checking system and detection method |
CN110163838A (en) * | 2019-04-01 | 2019-08-23 | 江西比格威医疗科技有限公司 | A kind of retina OCT image hierarchical algorithm |
CN110163838B (en) * | 2019-04-01 | 2023-04-18 | 江西比格威医疗科技有限公司 | Retina OCT image layering algorithm |
CN117893742A (en) * | 2024-03-15 | 2024-04-16 | 深圳大学总医院 | Uveitis image feature recognition method |
CN117893742B (en) * | 2024-03-15 | 2024-05-28 | 深圳大学总医院 | Uveitis image feature recognition method |
Also Published As
Publication number | Publication date |
---|---|
CN103854284B (en) | 2016-08-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103854284A (en) | Cutting method for serous pigment epithelium layer disengagement retina based on three-dimensional diagram searching | |
JP6025311B2 (en) | Ophthalmic diagnosis support apparatus and method | |
EP1976424B1 (en) | A method of eye examination by optical coherence tomography | |
EP3568059B1 (en) | Method and system for three-dimensional thickness mapping of corneal micro-layers and corneal diagnoses | |
US9098742B2 (en) | Image processing apparatus and image processing method | |
US8699774B2 (en) | Image processing apparatus, control method thereof, and program | |
US9514513B2 (en) | Establishing compatibility between two- and three-dimensional optical coherence tomography scans | |
KR20130027506A (en) | Tomogram observation apparatus, processing method, and non-transitory computer-readable storage medium | |
US8870377B2 (en) | Image processing apparatus, image processing apparatus control method, ophthalmologic apparatus, ophthalmologic apparatus control method, ophthalmologic system, and storage medium | |
US20090268159A1 (en) | Automated assessment of optic nerve head with spectral domain optical coherence tomography | |
JP6526145B2 (en) | Image processing system, processing method and program | |
JP2017127397A (en) | Image processing device, estimation method, system and program | |
EP3129955B1 (en) | Method for the analysis of image data representing a three-dimensional volume of biological tissue | |
de Sisternes et al. | A machine learning approach for device-independent automated segmentation of retinal cysts in spectral domain optical coherence tomography images | |
WO2019157113A1 (en) | Segmentation-based corneal mapping | |
EP3129956B1 (en) | Method for the analysis of image data representing a three-dimensional volume of biological tissue | |
Jo et al. | Comparison of evaluation parameters in the retinal layer between diabetic cystoid macular edema and postoperative cystoid macular edema after cataract surgery based on a hierarchical approach | |
JP2017119226A (en) | Image processing apparatus, image processing method, and image processing program | |
JP2019115827A (en) | Image processing system, processing method and program |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20170810 Address after: 215011 No. 209 Chuk Yuen Road, Suzhou hi tech Development Zone, Jiangsu, Suzhou Patentee after: Suzhou were Medical Technology Co. Ltd. Address before: 215006, 1, ten Zi street, Suzhou District, Jiangsu, Suzhou Patentee before: Soochow University |