CN116664568B - Retina layer segmentation method and system based on multi-visible light spectrum OCT (optical coherence tomography) image - Google Patents
Retina layer segmentation method and system based on multi-visible light spectrum OCT (optical coherence tomography) image Download PDFInfo
- Publication number
- CN116664568B CN116664568B CN202310942612.1A CN202310942612A CN116664568B CN 116664568 B CN116664568 B CN 116664568B CN 202310942612 A CN202310942612 A CN 202310942612A CN 116664568 B CN116664568 B CN 116664568B
- Authority
- CN
- China
- Prior art keywords
- visible light
- spectrum
- retina
- oct images
- oct
- 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
- 238000001228 spectrum Methods 0.000 title claims abstract description 119
- 210000001525 retina Anatomy 0.000 title claims abstract description 118
- 230000011218 segmentation Effects 0.000 title claims abstract description 97
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000012014 optical coherence tomography Methods 0.000 title description 83
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000001429 visible spectrum Methods 0.000 claims description 42
- 230000004927 fusion Effects 0.000 claims description 22
- 230000002207 retinal effect Effects 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 10
- 238000010586 diagram Methods 0.000 claims description 9
- 238000002372 labelling Methods 0.000 claims description 9
- 230000003595 spectral effect Effects 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000003709 image segmentation Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 2
- 238000007499 fusion processing Methods 0.000 claims description 2
- 239000010410 layer Substances 0.000 abstract description 97
- 239000011229 interlayer Substances 0.000 abstract description 13
- 238000003384 imaging method Methods 0.000 abstract description 9
- 230000008859 change Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000002834 transmittance Methods 0.000 description 1
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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10101—Optical tomography; Optical coherence tomography [OCT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
Abstract
The invention provides a retina layer segmentation method and a retina layer segmentation system based on multiple visible light spectrum OCT images, relates to the field of medical image processing, aims at the problem that the precision of a retina layer segmentation model based on single spectrum OCT images at present is poor, intercepts spectrum segments in different ranges in a relatively wide range of total visible light wave band for imaging, acquires retina layer images with different backscattering rates, acquires a plurality of retina layer OCT images with unique interlayer contrast information, and combines a corresponding retina layer segmentation model to improve the accuracy of retina segmentation.
Description
Technical Field
The invention relates to the field of medical image processing, in particular to a retina layer segmentation method and a retina layer segmentation system based on multiple visible spectrum OCT images.
Background
Optical coherence tomography (Vis-OCT) based on visible light is an emerging imaging mode, and compared with the traditional infrared OCT based imaging mode, the visible light based optical coherence tomography (Vis-OCT) based on infrared OCT has the advantages that due to the fact that the wavelength is short, the high axial resolution is achieved, the imaging contrast between tissues is improved, and extremely strong clinical application potential is displayed. After the retina is scanned by adopting the visible light OCT technology, the automatic segmentation of different tissue layers of the OCT image of the fovea area has important clinical significance, and on one hand, the automatic segmentation can improve the working efficiency and the segmentation precision of a professional ophthalmologist; on the other hand, the segmented result can be used to assist the ophthalmologist in judging the progress of some early diseases.
The main difficulty of segmentation is that the interlayer contrast of the retina layers in the OCT image is low, the existing retina layer segmentation research (such as deep learning and traditional threshold segmentation) is mostly based on the retina layer OCT image of a single spectrum section for processing, and the retina layer OCT image of the single spectrum section only contains one interlayer contrast information, so that the difficulty of retina layer segmentation is caused; particularly, the precision of retina layer segmentation based on a deep learning method and segmentation based on single interlayer contrast information is still difficult to meet the requirement.
Disclosure of Invention
Aiming at the defects existing in the prior art, the invention provides a retina layer segmentation method and a retina layer segmentation system based on multiple visible light spectrum OCT images, which are used for capturing different ranges of spectrum segments in a relatively wide range of total visible light wave band for imaging, obtaining retina layer images with different backscattering rates, obtaining multiple retina layer OCT images with unique interlayer contrast information, and improving the accuracy of retina segmentation by combining a corresponding retina layer segmentation model.
The first object of the present invention is to provide a retina layer segmentation method based on multi-visible spectrum OCT images, which adopts the following scheme:
comprising the following steps:
acquiring OCT images of multiple visible light spectrums in different wave bands;
processing all OCT images of the visible light spectrum by using a retina layer segmentation model to obtain a segmented retina layer characteristic map;
the training process of the retina layer segmentation model comprises the following steps:
acquiring OCT images of wide visible light spectrums of human eye retina areas, segmenting the wide visible light spectrums to obtain multiple segments of visible light spectrums in different wave bands, and acquiring OCT images of each segment of visible light spectrums;
OCT images of a wide visible light spectrum and OCT images of all segments of a visible light spectrum are processed and input into a multi-channel semantic segmentation frame to obtain a retina layer segmentation model.
Further, the visible light spectrum of different wave bands corresponds to different wavelength ranges, and the spectrum width of each segment of visible light spectrum is equal.
Further, the different spectral bands correspond to different retinal layer contrasts on OCT images.
Further, in the process of obtaining the retina layer segmentation model, the obtained OCT image corresponding to the wide visible spectrum is subjected to truth labeling of retina layer segmentation, so that a truth labeling diagram is obtained and is used for training the retina layer segmentation model.
Further, after the OCT image of each section of visible light spectrum is obtained, data format conversion is carried out on the OCT images of all sections of visible light spectrum so as to meet the input requirement of a multi-channel semantic segmentation framework.
Further, the inputting the multi-channel semantic segmentation framework to obtain the retina layer segmentation model includes:
all the OCT images of the visible light spectrum are input into a multi-channel joint input module for feature extraction;
weighting the characteristics output by the multi-channel joint input module in the channel dimension, and outputting a characteristic diagram with multiple scales;
carrying out fusion processing on all the feature graphs, outputting a fusion feature graph, and carrying out similarity evaluation on the fusion feature graph and the true value labeling graph;
and after the similarity evaluation meets the requirement, storing the corresponding model and model internal parameters as retina layer segmentation models.
Further, the fusing processing of all the feature maps includes:
acquiring output four-scale feature graphs F1, F2, F3 and F4;
the feature map F1 is fused with the feature map F2 after up-sampling reduction to obtain a fused 1 feature map;
the feature map of the fusion 1 is fused with the feature map F3 after being sampled and restored to obtain a feature map of the fusion 2;
the feature map of the fusion 2 is fused with the feature map F4 after being sampled and restored to obtain a feature map of the fusion 3;
and (3) after the fusion 3 feature images are sampled and restored, obtaining and outputting the fusion feature images for similarity evaluation.
Further, after the similarity evaluation of the fusion feature map and the true value label map is judged, if the similarity does not meet the requirement, the result is fed back to each execution module of the multi-channel semantic segmentation frame, and the internal parameters of each execution module are adjusted to adjust the output result.
Further, in the step of weighting the characteristics output by the multi-channel joint input module in the channel dimension, the characteristics are weighted by multi-spectrum information based on the channel attention.
A second object of the present invention is to provide a retinal layer segmentation system based on multiple visible spectrum OCT images, comprising:
an image acquisition module configured to: acquiring OCT images of multiple visible light spectrums in different wave bands;
an image segmentation module configured to: processing all OCT images of the visible light spectrum by using a retina layer segmentation model to obtain a segmented retina layer characteristic map;
the training process of the retina layer segmentation model comprises the following steps:
acquiring OCT images of wide visible light spectrums of human eye retina areas, segmenting the wide visible light spectrums to obtain multiple segments of visible light spectrums in different wave bands, and acquiring OCT images of each segment of visible light spectrums;
OCT images of a wide visible light spectrum and OCT images of all segments of a visible light spectrum are processed and input into a multi-channel semantic segmentation frame to obtain a retina layer segmentation model.
Compared with the prior art, the invention has the advantages and positive effects that:
(1) Aiming at the problem that the precision of a retina layer segmentation model based on the retina layer OCT image of a single spectrum section is poor at present, spectrum sections of different ranges are intercepted in a total visible light wave band of a relatively wide range to image, retina layer images with different backscattering rates are obtained, a plurality of retina layer OCT images with unique interlayer contrast information are obtained, and the precision of retina segmentation is improved by combining the corresponding retina layer segmentation model.
(2) The traditional mode does not consider spectrum, and the invention is based on the fact that the structure and the composition of each layer of retina are different, so that the spectrum performance is also different, not only is the information of each independent spectrum section comprehensively utilized, but also the change information of different spectrum sections can be utilized, and the accurate segmentation of retina layers is facilitated.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
Fig. 1 is a flow chart of a retina layer segmentation method based on multiple visible spectrum OCT images in embodiments 1 and 2 of the present invention.
FIG. 2 is a schematic diagram showing the pretreatment flow in examples 1 and 2 of the present invention.
Fig. 3 is a schematic diagram of a multi-channel semantic segmentation framework in embodiments 1 and 2 of the present invention.
Detailed Description
Example 1
In an exemplary embodiment of the present invention, as shown in fig. 1-3, a retinal layer segmentation method based on multiple visible spectrum OCT images is presented.
In the existing retina layer segmentation method, the retina layer OCT image based on a single spectrum is processed, but the retina layer OCT image of the single spectrum only contains one piece of interlayer contrast information, so that the retina layer segmentation is difficult, and in the retina layer segmentation method based on deep learning, the segmentation precision of a model trained by the interlayer contrast information on the retina layer is poor.
Based on this, the embodiment provides a retina layer segmentation method based on multiple visible spectrum OCT images, which adopts spectrum information of different visible spectrum segments, so that each retina layer OCT image of different spectrum segments has specific interlayer contrast information, and comprehensively considers the retina layer OCT images of multiple spectrum segments, thereby enriching the interlayer contrast information of the retina layers and improving the segmentation precision of the retina layers.
Next, a retinal layer segmentation method based on the multi-visible light spectrum OCT image will be described in detail with reference to the accompanying drawings.
The wavelength range of the visible light spectrum is 390-760nm, so that the width of the visible light spectrum is 370nm, while the visible light spectrum width adopted by the common visible light OCT is about 100nm (some lower 80nm and some higher 140 nm), in this embodiment, the OCT imaging system is based on the OCT imaging system to obtain the OCT image with the visible light spectrum width of 200nm, and the wavelength range of the visible light spectrum is 400-600 nm, so that the visible light spectrum interval of a larger range can be covered.
The retina of a human eye has a plurality of layers of tissues, and different retina layers have different transmittance and backscatter rate on different retina layers due to different compositions, the backscatter rate is a basic factor of OCT imaging, different retina layers have different backscatter rates, so that brightness and darkness of different retina layers on an image are different, and different retina layers have different backscatter rate changes on different spectrum sections.
The retinal layers cause significantly different spectral distributions of the retinal layer back-scatter ratios to be predominantly distributed over the visible spectrum, and thus images of the retinal layers with different back-scatter ratios are acquired by capturing spectral segments of different ranges in the visible band.
Alternatively, taking a spectrum width of 100nm as an example, a spectrum section of wavelengths ranging from 400nm to 500nm, from 430nm to 530nm, from 460nm to 560nm, from 490nm to 590nm and from 500nm to 600nm is selected, so as to obtain multiple sections of different visible spectrum OCT images with 100nm widths.
Aiming at the problem that the prior retina layer OCT images based on single spectrum are processed to cause difficult retina layer segmentation, a plurality of retina layer OCT images of visible spectrum are adopted to contain spectrum information of different visible spectrum, so that each retina layer OCT image of different spectrum has unique interlayer contrast information, the retina layer OCT images of different spectrum are integrated together, the interlayer contrast information of retina layers is greatly enriched, thereby improving the accuracy of training retina layer segmentation models and further improving the precision of retina layer segmentation.
It will be appreciated that the structure and composition of each layer of the retina is different, so is its spectral behavior. The traditional mode does not consider spectrum, but in the embodiment, not only the information of each independent spectrum segment is comprehensively utilized, but also the change information of different spectrum segments can be utilized, so that accurate segmentation of the retina layers is facilitated.
Acquiring OCT images of multiple visible light spectrums in different wave bands;
processing all OCT images of the visible light spectrum by using a retina layer segmentation model to obtain a segmented retina layer characteristic map;
the training process of the retina layer segmentation model comprises the following steps:
acquiring OCT images of wide visible light spectrums of human eye retina areas, segmenting the wide visible light spectrums to obtain multiple segments of visible light spectrums in different wave bands, and acquiring OCT images of each segment of visible light spectrums;
OCT images of a wide visible light spectrum and OCT images of all segments of a visible light spectrum are processed and input into a multi-channel semantic segmentation frame to obtain a retina layer segmentation model.
In this embodiment, the broad visible spectrum refers to a spectrum width of the visible spectrum being wider than a spectrum width of the sub-visible spectrum, for example, a spectrum width of the obtained broad visible spectrum is 200nm, and a wavelength range thereof is 400nm to 600nm. The spectrum width of the visible spectrum can be selected to be 100nm, and the wavelength range can be 400nm-500nm, 430nm-530nm, 460nm-560nm, 490nm-590nm or 500nm-600nm. Therefore, the broad visible spectrum in this embodiment is a relative description with respect to the sub-visible spectrum, and does not directly represent the spectrum width thereof, and the spectrum width of the broad visible spectrum is larger than that of the sub-visible spectrum, and at the same time, the wavelength range corresponding to the sub-visible spectrum is within the wavelength range of the broad visible spectrum.
Specifically, referring to fig. 1-3, in the retina layer segmentation method based on visible light spectrum OCT images, the establishment of a retina layer segmentation model mainly includes two parts of multi-visible light spectrum image preprocessing and multi-channel semantic segmentation methods.
Step 1: image preprocessing of multiple visible light spectrum segments, as shown in fig. 2, is used for obtaining image information for establishing a retina layer segmentation model;
step 1-1: collecting a wide visible spectrum OCT image of a retina area of a human eye, as shown in (a) of fig. 2;
step 1-2: dividing the wide visible spectrum of the acquired OCT image to obtain OCT images with several sections of visible spectrum, as shown in (b), (c) and (d) in fig. 2;
step 1-3: and (3) carrying out true value labeling on retina layer segmentation on the OCT image of the wide visible spectrum to obtain a true value labeling diagram shown in (e) of fig. 2, which is used for later model training. Because the scanned retinal position is unchanged, the divided spectral segments only change different spectral ranges, so that the OCT image positions of the scanning before and after the divided spectrum are consistent, and the truth images before and after the divided spectrum are the same.
In step 1-2, after dividing the wide visible spectrum of the collected OCT image, several sections of OCT images with relatively narrow visible spectrum can be obtained, where each section of OCT image with relatively narrow visible spectrum is the OCT image with the above-mentioned sub-visible spectrum.
For example, OCT images of a wide visible spectrum with a visible spectrum width of 200nm (the wavelength range is 400-600 nm) are acquired, and when the spectrum width is selected to be 100nm, segmentation is carried out, so that OCT images of 5-segment visible spectrums with the wavelength ranges of 400-500 nm, 430-530 nm, 460-560 nm, 490-590 nm and 500-600 nm are obtained.
Obvious changes of the contrast between retina layers in different spectral bands can be obviously observed, and the differences of the changes of brightness and darkness degree are shown on images.
Step 2: with reference to fig. 1 and 2, data format conversion;
specifically, the image format in fig. 2 (b), (c), (d) and (e) is converted into a tensor format which can be used by the semantic segmentation method;
step 3: performing an end-to-end semantic segmentation method using a multi-channel semantic segmentation framework, as shown in fig. 3;
step 3-1: inputting the images of multiple visible light spectrum segments into a multi-channel joint input module for extracting features;
step 3-2: weighting the characteristics output by the multi-channel joint input module in the channel dimension, improving the utilization of the network to the data in the channel dimension, and outputting characteristic diagrams with four scales, such as F1, F2, F3 and F4 in FIG. 3;
step 3-3: the feature map F1 is fused with the feature map F2 after up-sampling reduction to obtain a fused 1 feature map;
step 3-4: and processing the feature map F3 and the feature map F4 sequentially in the same way; the feature map of the fusion 1 is fused with the feature map F3 after being sampled and restored to obtain a feature map of the fusion 2; the feature map of the fusion 2 is fused with the feature map F4 after being sampled and restored to obtain a feature map of the fusion 3;
step 3-5: and 3, merging the results output by the 3 feature graphs through the last feature graph restoration module, and performing similarity evaluation with the truth value annotation graph shown in (e) of fig. 2. If the similarity cannot meet the requirement, feeding back the result to each execution module of the multi-channel semantic segmentation frame, and further optimizing the output result by adjusting internal parameters of each execution module until the similarity with the truth value feature diagram meets the requirement; in this embodiment, the similarity may be set to 98.2%.
In other alternative embodiments, the similarity may be increased or decreased.
Step 3-6: and storing a model and model internal parameters corresponding to the optimal output result, and packaging the model and the model internal parameters into an executable program to serve as a retina layer segmentation model.
The retina layer images with different backscattering rates are obtained by intercepting spectrum bands with different ranges in a total visible light wave band with a relatively wide range for imaging, a plurality of retina layer OCT images with unique interlayer contrast information are obtained, and the accuracy of retina segmentation is improved by combining a corresponding retina layer segmentation model.
Example 2
In another exemplary embodiment of the present invention, as shown in fig. 1-3, a retinal layer segmentation system based on multiple visible light spectrum OCT images is presented.
A retinal layer segmentation system based on multiple visible light spectrum OCT images, comprising:
an image acquisition module configured to: acquiring OCT images of multiple visible light spectrums in different wave bands;
an image segmentation module configured to: processing all OCT images of the visible light spectrum by using a retina layer segmentation model to obtain a segmented retina layer characteristic map;
the training process of the retina layer segmentation model comprises the following steps:
acquiring OCT images of wide visible light spectrums of human eye retina areas, segmenting the wide visible light spectrums to obtain multiple segments of visible light spectrums in different wave bands, and acquiring OCT images of each segment of visible light spectrums;
OCT images of a wide visible light spectrum and OCT images of all segments of a visible light spectrum are processed and input into a multi-channel semantic segmentation frame to obtain a retina layer segmentation model.
It can be understood that the method for operating the above-mentioned retinal layer segmentation system based on multiple visible light spectrum OCT images is the same as that provided in embodiment 1, and reference may be made to the detailed description in embodiment 1, which is not repeated here.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (8)
1. The retina layer segmentation method based on the multi-visible light spectrum OCT image is characterized by comprising the following steps of:
acquiring OCT images of multiple visible light spectrums in different wave bands;
processing all OCT images of the visible light spectrum by using a retina layer segmentation model to obtain a segmented retina layer characteristic map;
the training process of the retina layer segmentation model comprises the following steps:
acquiring OCT images of wide visible light spectrums of human eye retina areas, segmenting the wide visible light spectrums to obtain multiple segments of visible light spectrums in different wave bands, and acquiring OCT images of each segment of visible light spectrums;
processing OCT images of a wide visible light spectrum and OCT images of all segments of a visible light spectrum, and inputting a multi-channel semantic segmentation frame to obtain a retina layer segmentation model;
the visible light spectrum of different wave bands corresponds to different wavelength ranges, and the spectrum width of each segment of visible light spectrum is equal; the contrast between retina layers on the OCT image is different corresponding to the different spectral bands.
2. The method for dividing retina layers based on multiple visible spectrum OCT images according to claim 1, wherein in the process of obtaining a retina layer dividing model, truth labeling of retina layer dividing is carried out on the OCT images corresponding to the obtained wide visible spectrum, so as to obtain a truth labeling diagram for training the retina layer dividing model.
3. The method for dividing the retina layers based on the multi-visible spectrum OCT image according to claim 2, wherein after the OCT image of each segment of the divided visible spectrum is acquired, the OCT images of all segments of the divided visible spectrum are subjected to data format conversion so as to meet the input requirement of a multi-channel semantic division framework.
4. A retinal layer segmentation method based on multi-visible spectrum OCT images according to claim 2 or 3, wherein the inputting the multi-channel semantic segmentation framework to obtain a retinal layer segmentation model includes:
all the OCT images of the visible light spectrum are input into a multi-channel joint input module for feature extraction;
weighting the characteristics output by the multi-channel joint input module in the channel dimension, and outputting a characteristic diagram with multiple scales;
carrying out fusion processing on all the feature graphs, outputting a fusion feature graph, and carrying out similarity evaluation on the fusion feature graph and the true value labeling graph;
and after the similarity evaluation meets the requirement, storing the corresponding model and model internal parameters as retina layer segmentation models.
5. The method for segmenting the retina layer based on the OCT image of multiple visible light spectrums according to claim 4, wherein the fusing all feature maps includes:
acquiring output four-scale feature graphs F1, F2, F3 and F4;
the feature map F1 is fused with the feature map F2 after up-sampling reduction to obtain a fused 1 feature map;
the feature map of the fusion 1 is fused with the feature map F3 after being sampled and restored to obtain a feature map of the fusion 2;
the feature map of the fusion 2 is fused with the feature map F4 after being sampled and restored to obtain a feature map of the fusion 3;
and (3) after the fusion 3 feature images are sampled and restored, obtaining and outputting the fusion feature images for similarity evaluation.
6. The method for dividing retina layers based on multiple visible spectrum OCT images according to claim 5, wherein after the similarity between the fusion feature map and the truth labeling map is evaluated, if the similarity does not meet the requirement, the result is fed back to each execution module of the multi-channel semantic division frame, and the internal parameters of each execution module are adjusted to adjust the output result.
7. The multi-visible spectrum OCT image-based retinal layer segmentation method of claim 5, wherein the characteristics output by the multi-channel joint input module are weighted in the channel dimension by the channel attention-based multispectral information.
8. A retinal layer segmentation system based on multiple visible light spectrum OCT images, comprising:
an image acquisition module configured to: acquiring OCT images of multiple visible light spectrums in different wave bands;
an image segmentation module configured to: processing all OCT images of the visible light spectrum by using a retina layer segmentation model to obtain a segmented retina layer characteristic map;
the training process of the retina layer segmentation model comprises the following steps:
acquiring OCT images of wide visible light spectrums of human eye retina areas, segmenting the wide visible light spectrums to obtain multiple segments of visible light spectrums in different wave bands, and acquiring OCT images of each segment of visible light spectrums;
processing OCT images of a wide visible light spectrum and OCT images of all segments of a visible light spectrum, and inputting a multi-channel semantic segmentation frame to obtain a retina layer segmentation model;
the visible light spectrum of different wave bands corresponds to different wavelength ranges, and the spectrum width of each segment of visible light spectrum is equal; the contrast between retina layers on the OCT image is different corresponding to the different spectral bands.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310942612.1A CN116664568B (en) | 2023-07-31 | 2023-07-31 | Retina layer segmentation method and system based on multi-visible light spectrum OCT (optical coherence tomography) image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310942612.1A CN116664568B (en) | 2023-07-31 | 2023-07-31 | Retina layer segmentation method and system based on multi-visible light spectrum OCT (optical coherence tomography) image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116664568A CN116664568A (en) | 2023-08-29 |
CN116664568B true CN116664568B (en) | 2023-10-13 |
Family
ID=87722761
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310942612.1A Active CN116664568B (en) | 2023-07-31 | 2023-07-31 | Retina layer segmentation method and system based on multi-visible light spectrum OCT (optical coherence tomography) image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116664568B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104778668A (en) * | 2015-04-16 | 2015-07-15 | 电子科技大学 | Optical remote sensing image thin cloud removal method on basis of visible band spectrum statistical characteristics |
CN115443480A (en) * | 2020-04-29 | 2022-12-06 | 卡尔蔡司医疗技术公司 | OCT orthopathology segmentation using channel-encoded slabs |
CN115482384A (en) * | 2022-09-23 | 2022-12-16 | 山东大学 | Visible light OCT image retina layer segmentation method and system |
CN116503593A (en) * | 2023-03-27 | 2023-07-28 | 上海理工大学 | Retina OCT image hydrops segmentation method based on deep learning |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20230196572A1 (en) * | 2021-12-21 | 2023-06-22 | Carl Zeiss Meditec, Inc. | Method and system for an end-to-end deep learning based optical coherence tomography (oct) multi retinal layer segmentation |
-
2023
- 2023-07-31 CN CN202310942612.1A patent/CN116664568B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104778668A (en) * | 2015-04-16 | 2015-07-15 | 电子科技大学 | Optical remote sensing image thin cloud removal method on basis of visible band spectrum statistical characteristics |
CN115443480A (en) * | 2020-04-29 | 2022-12-06 | 卡尔蔡司医疗技术公司 | OCT orthopathology segmentation using channel-encoded slabs |
CN115482384A (en) * | 2022-09-23 | 2022-12-16 | 山东大学 | Visible light OCT image retina layer segmentation method and system |
CN116503593A (en) * | 2023-03-27 | 2023-07-28 | 上海理工大学 | Retina OCT image hydrops segmentation method based on deep learning |
Non-Patent Citations (4)
Title |
---|
Exploiting multi-granularity visual features for retinal layer segmentation in human eyes;Xiang He et.al;《Frontiers in Bioengineering and Biotechnology》;第1-14页 * |
Nano sensitive study and fractal analysis of segmented retinal layers in Fourier domain OCT: promises for early disease detection;Nandan Das et.al;《PROCEEDINGS OF SPIE》;第1-10页 * |
王卓.基于OCT图像的视网膜椭圆体带损伤检测技术研究与应用.《中国优秀硕士学位论文全文数据库 信息科技辑》.2018,第I138-1234页. * |
视网膜OCT B-scan图像分割算法综述;袁坤;霍力;;半导体光电(05);第4-9页 * |
Also Published As
Publication number | Publication date |
---|---|
CN116664568A (en) | 2023-08-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhou et al. | Color retinal image enhancement based on luminosity and contrast adjustment | |
CN107527326A (en) | A kind of wheat scab diagnostic method based on high light spectrum image-forming | |
CN107203989A (en) | End-to-end chest CT image dividing method based on full convolutional neural networks | |
CN110852396A (en) | Sample data processing method for cervical image | |
Bian et al. | Optic disc and optic cup segmentation based on anatomy guided cascade network | |
CN108764342B (en) | Semantic segmentation method for optic discs and optic cups in fundus image | |
CN107977950B (en) | Rapid and effective video image fusion method based on multi-scale guide filtering | |
CN111488912B (en) | Laryngeal disease diagnosis system based on deep learning neural network | |
CN112927243B (en) | Micro-hemorrhage focus segmentation method based on convolutional neural network | |
CN106530294A (en) | Method for carrying out processing on meibomian gland image to obtain gland parameter information | |
CN114119950A (en) | Artificial intelligence-based oral cavity curved surface fault layer dental image segmentation method | |
CN116152500A (en) | Full-automatic tooth CBCT image segmentation method based on deep learning | |
CN111340773B (en) | Retinal image blood vessel segmentation method | |
Reddy et al. | Retinal fundus image enhancement using piecewise gamma corrected dominant orientation based histogram equalization | |
CN110969631A (en) | Fine photo hair dyeing method and system | |
CN115578406A (en) | CBCT jaw bone region segmentation method and system based on context fusion mechanism | |
CN116664568B (en) | Retina layer segmentation method and system based on multi-visible light spectrum OCT (optical coherence tomography) image | |
Dong et al. | Supervised learning-based retinal vascular segmentation by m-unet full convolutional neural network | |
KR102537470B1 (en) | Diagnosis method for Graves’orbitopathy using orbital computed tomography (CT) based on neural network-based algorithm | |
CN115482384A (en) | Visible light OCT image retina layer segmentation method and system | |
Javid et al. | Marking early lesions in labial colored dental images using a transfer learning approach | |
Grochowski et al. | Diagnosis of malignant melanoma by neural network ensemble-based system utilising hand-crafted skin lesion features | |
Vats et al. | Detecting oral cancer: the potential of artificial intelligence | |
Soliz et al. | Computer-aided methods for quantitative assessment of longitudinal changes in retinal images presenting with maculopathy | |
Meyer-Veit et al. | Hyperspectral endoscopy using deep learning for laryngeal cancer segmentation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20240313 Address after: Room 320, Building 14, Lishan Mingjun, No. 175 Lishan Road, Lixia District, Jinan City, Shandong Province, 250014 Patentee after: Shandong Tanwei Medical Technology Co.,Ltd. Country or region after: China Address before: 250061, No. ten, No. 17923, Lixia District, Ji'nan City, Shandong Province Patentee before: SHANDONG University Country or region before: China |
|
TR01 | Transfer of patent right |