CN116883420A - Choroidal neovascularization segmentation method and system in optical coherence tomography image - Google Patents
Choroidal neovascularization segmentation method and system in optical coherence tomography image Download PDFInfo
- Publication number
- CN116883420A CN116883420A CN202310655638.8A CN202310655638A CN116883420A CN 116883420 A CN116883420 A CN 116883420A CN 202310655638 A CN202310655638 A CN 202310655638A CN 116883420 A CN116883420 A CN 116883420A
- Authority
- CN
- China
- Prior art keywords
- segmentation
- features
- network
- optical coherence
- choroidal neovascularization
- 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
- 230000011218 segmentation Effects 0.000 title claims abstract description 109
- 238000012014 optical coherence tomography Methods 0.000 title claims abstract description 66
- 206010060823 Choroidal neovascularisation Diseases 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 51
- 208000005590 Choroidal Neovascularization Diseases 0.000 title claims abstract description 49
- 230000001419 dependent effect Effects 0.000 claims abstract description 9
- 238000011176 pooling Methods 0.000 claims description 44
- 230000004927 fusion Effects 0.000 claims description 35
- 230000002457 bidirectional effect Effects 0.000 claims description 29
- 230000003902 lesion Effects 0.000 claims description 19
- 239000011159 matrix material Substances 0.000 claims description 18
- 238000005070 sampling Methods 0.000 claims description 13
- 230000009977 dual effect Effects 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 7
- 230000002708 enhancing effect Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 abstract description 18
- 230000000694 effects Effects 0.000 abstract description 11
- 238000010586 diagram Methods 0.000 description 11
- 238000013527 convolutional neural network Methods 0.000 description 9
- 238000003709 image segmentation Methods 0.000 description 8
- 238000004590 computer program Methods 0.000 description 7
- 230000004044 response Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000002347 injection Methods 0.000 description 3
- 239000007924 injection Substances 0.000 description 3
- 239000004973 liquid crystal related substance Substances 0.000 description 3
- 239000000243 solution Substances 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 210000001525 retina Anatomy 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 108010041308 Endothelial Growth Factors Proteins 0.000 description 1
- 102000005789 Vascular Endothelial Growth Factors Human genes 0.000 description 1
- 108010019530 Vascular Endothelial Growth Factors Proteins 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000002137 anti-vascular effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 210000003161 choroid Anatomy 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000004438 eyesight Effects 0.000 description 1
- 230000004402 high myopia Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 238000007634 remodeling Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000006467 substitution reaction 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
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/0895—Weakly supervised learning, e.g. semi-supervised or self-supervised learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
-
- 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
- G06V10/267—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 by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/52—Scale-space analysis, e.g. wavelet analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/54—Extraction of image or video features relating to texture
-
- 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
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- 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
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a choroidal neovascularization segmentation method and a system in an optical coherence tomography image, which are characterized in that a segmentation network based on UNet is constructed, two paths are arranged in downsampling and used for respectively extracting multi-scale global features and multi-scale global features with local semantic information and long-distance dependent information, the extracted features on the two paths are fused, the fused features of each scale are output, and the feature information can be fully extracted; the fused features and the upsampled features with the same scale are enhanced and then fused during upsampling, so that the generation of focus features in the upsampling process can be focused, focus boundary information in the upsampling process is used as supervision information to enhance the expression of the boundary information, fine segmentation can be performed, and the segmentation effect is improved.
Description
Technical Field
The invention relates to the technical field of image segmentation, in particular to a choroidal neovascularization segmentation method and system in an optical coherence tomography image.
Background
Myopic choroidal neovascularization (myopic choroidal neovascularization, mCNV) is a common complication of high myopia, the lesions of which involve the macular area of the retina, with a severe impact on vision. Currently, the main treatment for mCNV is the injection of anti-Vascular Endothelial Growth Factor (VEGF) into the eye, which requires multiple injections to change the lesion from active to inactive.
The best tool to observe the change in lesions from active to inactive is optical coherence tomography (Optical Coherence Tomography, OCT). Because it is a non-invasive examination, it allows for the observation of the various levels of fine structures of the retina and choroid, and can be used to observe the size of lesions to help the physician judge the effectiveness of each injection treatment or the likelihood of recurrence.
With the development of deep learning, although an image segmentation method based on deep learning is used for segmenting an OCT image, in the OCT image, because a focal region of a choroidal neovascularization is small, a contrast is low, a boundary is blurred, a large amount of speckle noise exists, and the like, the problems cause insufficient extracted feature information, insufficient segmentation accuracy and poor segmentation effect when the choroidal neovascularization is segmented by the existing image segmentation method.
Accordingly, there is a need for improvement and advancement in the art.
Disclosure of Invention
The invention mainly aims to provide a choroidal neovascularization segmentation method, a system, an intelligent terminal and a storage medium in an optical coherence tomography image, which can solve the problems of insufficient feature information, insufficient segmentation accuracy and poor segmentation effect in the prior art of segmenting the choroidal neovascularization.
In order to achieve the above object, a first aspect of the present invention provides a choroidal neovascularization segmentation system in an optical coherence tomography image, comprising a segmentation network constructed based on UNet for segmenting choroidal neovascularization in an optical coherence tomography image, said segmentation network having an encoder for downsampling and a decoder for upsampling disposed therein;
a dual-path network is arranged in the encoder, a first path of the dual-path network is provided with a plurality of transducer blocks, and the first path is used for extracting first global features of a plurality of scales; a plurality of mixing units for extracting local semantic information and long-distance dependent information are arranged on a second path of the dual-path network, the second path is used for extracting second global features with multiple scales, and the scales of the first global features are in one-to-one correspondence with the scales of the second global features;
the encoder is also provided with a plurality of dual-path fusion units, and the dual-path fusion units are used for fusing the first global features and the second global features on each scale;
the decoder is provided with a plurality of bidirectional attention gating units and a cascading unit, the cascading unit is used for cascading the characteristics output by the dual-path fusion unit and the characteristics output by the bidirectional attention gating unit, the bidirectional attention gating unit is used for fusing up-sampling characteristics and the characteristics output by the dual-path fusion unit on adjacent layers, and the up-sampling characteristics are obtained by up-sampling the characteristics output by the cascading unit or the dual-path fusion unit;
and the output module is used for obtaining and outputting a segmentation result according to the characteristics output by the segmentation network.
Optionally, the mixing unit includes a connected residual convolution network and a Transformer network.
Optionally, a channel attention coding module is disposed on one path of the dual-path fusion unit, and is used for enhancing the interdependence relationship of the channel attention map extracted by the mixing unit, and a position attention coding module is disposed on the other path, and is used for enhancing the spatial information of the features extracted by the transducer block.
Optionally, the bidirectional attention gating unit includes two bidirectional attention gating modules, and a summing module, where the two bidirectional attention gating modules are respectively used to enhance the features output by the dual-path fusion unit and the upsampling features, and the summing module is used to add the features output by the two bidirectional attention gating modules in matrix.
Optionally, the bidirectional attention gate module is provided with a first pooling layer, a dot multiplication layer and a second pooling layer, the first pooling layer and the second pooling layer are both provided with a maximum pooling branch and an average pooling branch, the first pooling layer is used for performing pooling operation and full connection operation on the characteristics input into the bidirectional attention gate module, the dot multiplication layer is used for performing matrix dot multiplication on the characteristics output by the first pooling layer and the characteristics input into the bidirectional attention gate module, and the second pooling layer is used for performing pooling operation on the characteristics output by the dot multiplication layer.
Optionally, training the loss function employed by the segmentation network includes: lesion classification loss, boundary classification loss, and Dice loss.
A second aspect of the invention provides a method of choroidal neovascularization segmentation in an optical coherence tomography image, said method comprising:
pre-training a segmentation network in a choroidal neovascularization segmentation system in an optical coherence tomography image according to any one of the preceding claims;
inputting the optical coherence tomography image into a trained segmentation network;
acquiring the characteristics of the output of the trained segmentation network;
based on the features, a segmentation result is obtained and output.
Optionally, the loss function of the training segmentation network is:
wherein, the liquid crystal display device comprises a liquid crystal display device,classification of loss for lesions->For boundary classification loss, L Dice For the Dice loss, GT i And->Representing the ith pixel as the true mark of the focus and the focus boundary, PO i And PB i The predicted values of the ith pixel belonging to the focus and the focus boundary are respectively shown, and alpha, beta and gamma are respectively the weights of focus classification loss, boundary classification loss and price loss.
A third aspect of the present invention provides a smart terminal comprising a memory, a processor, and a choroidal neovascularization segmentation routine in an optical coherence tomography image stored on the memory and operable on the processor, the choroidal neovascularization segmentation routine in the optical coherence tomography image, when executed by the processor, performing the steps of any one of the method of choroidal neovascularization segmentation in the optical coherence tomography image.
A fourth aspect of the present invention provides a computer readable storage medium having stored thereon a choroidal neovascularization segmentation program in an optical coherence tomography image, which when executed by a processor, performs the steps of any one of the above-described choroidal neovascularization segmentation methods in an optical coherence tomography image.
Compared with the prior art, the method has the advantages that by constructing the split network based on UNet, two paths are arranged in the downsampling process and used for respectively extracting multi-scale global features and multi-scale global features with local semantic information and long-distance dependent information, the features extracted from the two paths are fused, the fused features of all scales are output, and the feature information can be fully extracted; the fused features and the upsampled features with the same scale are enhanced and then fused during upsampling, so that the generation of focus features in the upsampling process can be focused, focus boundary information in the upsampling process is used as supervision information to enhance the expression of the boundary information, fine segmentation can be performed, and the segmentation effect is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a network architecture diagram of a segmentation network in a choroidal neovascularization segmentation system in an optical coherence tomography image provided by an embodiment of the present invention;
FIG. 2 is a network architecture diagram of the hybrid unit of the embodiment of FIG. 1;
FIG. 3 is a network architecture diagram of the dual path fusion unit of the embodiment of FIG. 1;
FIG. 4 is a schematic diagram of a network architecture of the bi-directional attention gating unit of the embodiment of FIG. 1;
FIG. 5 is a graph of qualitative analysis effects of segmentation results for each network model;
FIG. 6 is a schematic illustration of a specific flow chart of a method for dividing choroidal neovascularization in an optical coherence tomography image according to an embodiment of the present invention;
fig. 7 is a schematic block diagram of an internal structure of an intelligent terminal according to an embodiment of the present invention.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth such as the particular system architecture, techniques, etc., in order to provide a thorough understanding of the embodiments of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.
It should be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It is also to be understood that the terminology used in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.
As used in this specification and the appended claims, the term "if" may be interpreted in context as "when …" or "upon" or "in response to a determination" or "in response to detection. Similarly, the phrase "if a condition or event described is determined" or "if a condition or event described is detected" may be interpreted in the context of meaning "upon determination" or "in response to determination" or "upon detection of a condition or event described" or "in response to detection of a condition or event described".
The following description of the embodiments of the present invention will be made more fully hereinafter with reference to the accompanying drawings, in which embodiments of the invention are shown, it being evident that the embodiments described are only some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.
Accurate segmentation of myopic choroidal neovascularization lesions in OCT is critical to assessing the progression of myopic choroidal neovascularization. Currently, the medical image segmentation mainly adopts a UNet network based on a convolutional neural network or various UNet deformation networks, such as Resunet, UNet++, nnUNet and the like, or adopts a visual transducer network. Although the networks are widely applied to medical image segmentation, because of small focal areas of myopic choroidal neovascularization, low contrast, blurred boundaries and the like in OCT images, the network model is required to have global and local feature information extraction and fusion capabilities so as to fully extract feature information, and is required to be sensitive to the boundary information of the focal areas so as to accurately segment the focal areas of myopic choroidal neovascularization, thereby obtaining better segmentation effect. At present, a single UNet network or UNet deformation network and a visual transducer network have limitations in extracting characteristic information, so that the characteristic information cannot be fully extracted, and the segmentation effect is poor. Although some hybrid networks such as UTNet network (a hybrid network of convolutional neural network and transducer network) have appeared at present, the identification of boundary information is still not sensitive enough and the segmentation result is not fine enough when segmentation is performed on the focal region of myopic choroidal neovascularization.
In order to improve the segmentation accuracy of focal areas of myopic choroidal neovascularization in OCT images, the invention provides a system and a method for segmenting the choroidal neovascularization in optical coherence tomography images, which are characterized in that a segmentation network based on UNet is constructed, two paths are arranged in downsampling and used for respectively extracting multi-scale global features and multi-scale global features with local semantic information and long-distance dependent information, the extracted features on the two paths are fused, and the fused features of each scale are output, so that the feature information can be fully extracted; the fused features and the upsampled features with the same scale are enhanced and then fused during upsampling, so that the generation of focus features in the upsampling process can be focused, focus boundary information in the upsampling process is used as supervision information to enhance the expression of the boundary information, the segmentation result can be refined, and the segmentation effect can be improved.
Exemplary System
The embodiment of the invention provides a choroidal neovascularization segmentation system in an optical coherence tomography image, which is operated on electronic equipment such as an intelligent terminal, a PC, a background server of a hospital and the like and is used for segmenting myopic choroidal neovascularization in the optical coherence tomography image. The segmentation of the choroidal neovascularization in the optical coherence tomography image is not limited to the segmentation of myopic choroidal neovascularization, and may be performed by segmentation of various choroidal neovascularization in the optical coherence tomography image. The segmentation system mainly comprises two parts: after the optical coherence tomography image is input into the segmentation network, the segmentation network outputs characteristic information such as classification of each pixel in the optical coherence tomography image belonging to a background area or a focus area, classification of pixels in the focus or focus boundaries and the like, then the output module determines a focus area according to the characteristic information, cuts the focus area in the optical coherence tomography image, segments out a focus area image of a myopic choroidal neovascularization and outputs the focus area image.
Fig. 1 shows the architecture of a partitioning network comprising an encoder for downsampling and a decoder for upsampling. The left part of fig. 1 is the downsampling stage and the middle part is the upsampling stage. The encoder is a dual path network, the top plurality of transform blocks in fig. 1 are first paths, each of which outputs a first global feature of one scale, for example: the first transducer block comprises 3 transducers, the output features have a scale ofThe second transducer block comprises 4 transducers with output features of scale +.>Etc. Four transducer blocks are arranged on the first path in this embodiment, so that the first global features of four scales can be downsampled. A plurality of mixing units (i.e. mixing modules in fig. 1) form a second path, the mixing units are used for extracting local semantic information and long-distance dependent information, the characteristics output by the mixing units are second global characteristics, and the mixing units are mixedThe scale of the second global feature output by the combining unit corresponds to the scale of the first global feature output by the transducer block one by one. In this embodiment, four mixing units are provided, and each mixing unit corresponds to one transducer block. That is, the second path can downsample the second global feature of four scales, and the sizes of the four scales of the second path correspond to the sizes of the four scales of the first path one to one.
The feature information of the optical coherence tomographic image can be sufficiently extracted from a plurality of angles of the local semantic information and the global texture information by the plurality of first global features extracted by the first path and the plurality of second global features extracted by the second path.
The network architecture of the hybrid unit is shown in fig. 2 and includes a connected residual convolution network and a transducer network. The inputs to the mixing unit are: features extracted from OCT images or features extracted from a mixing unit of a previous level, these input features first pass through a residual convolution network, then a transducer network, and output a second global feature. The residual convolution network comprises a convolution layer of 2 layers 3*3, local semantic information can be extracted, then the local semantic features output by the residual convolution network are diced, an embedded block sequence is obtained and then input into a transform network, and long-distance dependent information is extracted through multi-head attention and multi-layer perceptrons in the transform network.
In order to fuse the first global feature and the second global feature extracted by the Two paths respectively, a plurality of dual-path fusion units (TPFusion) are further arranged in the encoder, and the number of the dual-path fusion units is the same as the total number of feature scales. If the embodiment is correspondingly provided with four dual-path fusion units, one dual-path fusion unit is used for fusing the first global feature and the second global feature of the same scale, outputting the fused features and sending the fused features to decoders of the same level. Where a hierarchy refers to either an up-sampling phase of the decoder or a down-sampling phase of the encoder, each hierarchy outputs one scale of features. For example: the encoder and decoder in fig. 1 each include four levels, the first level features having the dimensions:the scale of the second level features is: />Each level of the encoder includes a dual path fusion unit and a mixing unit. The decoder includes a bi-directional attention gating unit (DAG in the figure) and a cascade unit (cascade C in the figure) from the first level to the third level.
The network architecture of the dual path fusion unit is shown in fig. 3, where the dual path fusion unit includes two parallel paths. Assume that the characteristics extracted by the transducer block areThe mixing unit extracts the characteristics ofSince the local features extracted by the convolutional neural network pay more attention to the semantic features of the image, a channel attention coding module is arranged on the left path in fig. 3, and the features extracted by the mixing unit are enhanced by the channel attention coding, namely the interdependence relationship of the channel attention map output by the mixing unit is enhanced. Since the features extracted by the transducer have long-distance contextual feature dependence and less concern about spatial information, a position attention encoding module is arranged on the right path in fig. 3, and the spatial information of the features extracted by the transducer block is enhanced by position attention encoding.
The post-fusion feature F obtained by the dual-path fusion unit according to the definition of the channel attention code and the position attention code TP Can be expressed as:
where Point_CNN (·) represents the Point convolution with kernel size of 1×1, sum (·) represents matrix addition, M (·) represents matrix multiplication, and Resh (·) represents feature dimension size remodeling (i.e., rearrangement).
The specific process is as follows: features F extracted by mixing units M Inputting a path provided with a channel attention coding module, and inputting a characteristic F M With transposed featuresMatrix multiplication, multiplication result and characteristic F M Matrix multiplying the rearranged results, rearranging the multiplied results again to obtain the result and the characteristic F M And adding to obtain the characteristics of the channel attention codes.
Transformer block extracted feature F T Inputting a path provided with a position attention coding module, and inputting a characteristic F T With transposed featuresMatrix multiplication, multiplication result and characteristic F T Matrix multiplying the rearranged results, rearranging the multiplied results again to obtain the result and the characteristic F T Adding to obtain the position attention coded characteristics;
then, the characteristics after channel attention coding and the characteristics after position attention coding are subjected to matrix addition, and then point convolution with the kernel size of 1 multiplied by 1 is carried out to obtain a fused characteristic F output by a double-path fusion unit TP 。
Through the double-path fusion unit, the features extracted by the four transducer blocks are supplemented in space information, the features extracted by the mixing unit are supplemented in the channel attention map, so that the features extracted by the two paths are effectively fused and are complemented in different feature dimensions, and the feature information extraction effect of the segmentation network is improved.
While the encoder is able to extract features with rich semantics and global texture, further fusion by the decoder is required to express features related to the foci of myopic choroidal neovascularization. Referring to fig. 1, a decoder is provided with a plurality of bidirectional attention gating units (DAGs) and a cascade unit, wherein the cascade unit is used for cascading the fused features output by the double-path fusion unit (TPFusion) and the features output by the bidirectional attention gating units (DAGs). The up-sampled characteristics obtained after up-sampling the characteristics of the cascade unit output of each level are the output characteristics of each level in the decoder. For example: the upsampled features obtained by upsampling the features of the cascade unit output of the first layer of the decoder as in fig. 1 are the output features of the first layer of the decoder. Because the fourth layer of the decoder is the bottommost layer, the upsampling feature of the layer is a feature obtained by directly upsampling the feature output by the dual path fusion unit. The bidirectional attention gating unit is used for fusing the upsampling characteristics and the characteristics output by the dual path fusion unit on adjacent levels, the bidirectional attention gating unit of the first level is used for fusing the characteristics output by the dual path fusion unit of the first level and the upsampling characteristics of the second level, and the bidirectional attention gating unit of the second level is used for fusing the characteristics output by the dual path fusion unit of the second level and the upsampling characteristics of the third level. The characteristics of the bi-directional attention gating unit output are used for inputting to the cascade units of the same hierarchy.
Specifically, in order to make the segmentation network pay more attention to the generation of lesion features in the upsampling process, the present invention designs a bidirectional attention gating unit using a channel attention mechanism. Referring to fig. 4, the bidirectional attention gating unit includes two bidirectional attention gating modules and a summing module, and one bidirectional attention gating module is used to enhance the fused characteristic F output by the dual path fusion unit TP A bi-directional attention gating module for enhancing the upsampling feature F up . The summing module is used for carrying out matrix addition on the characteristics output by the two bidirectional attention gating modules to obtain a double-attention gating characteristic diagram F DAG And output.
The specific structure of the bidirectional attention gating module comprises a first pooling layer and a second pooling layer which are arranged up and down, and a dot multiplying layer is arranged between the first pooling layer and the second pooling layer. And the first pooling layer and the second pooling layer are respectively provided with a parallel maximum pooling branch and an average pooling branch. The first pooling layer is used for pooling operation and full connection operation on the features of the input bidirectional attention gating module, then matrix addition is carried out, then a dot multiplication layer is input, the dot multiplication layer is used for carrying out matrix dot multiplication on the features output by the first pooling layer and the features input into the bidirectional attention gating module, then the features are input into the second pooling layer, cascade connection is carried out after maximum pooling operation and average pooling operation are carried out, and the cascade connection result is subjected to matrix multiplication with the output features of the dot multiplication layer, so that the output features of the bidirectional attention gating module are obtained.
The specific process is as follows: first, the fusion feature F TP Respectively inputting the maximum pooling layer and the average pooling layer, respectively activating the obtained pooling result by a ReLu activation function, and changing the number of channels of the characteristics into the number of channels by two fully-connected layersWeighting and summing the features obtained by the maximum pooling branch and the average pooling branch, and fusing the features with the feature F TP Performing dot multiplication operation of the matrix to obtain attention force diagram F' TP . In order to maintain the integrity of the features, the obtained attention is subjected to a maximum pooling layer and an average pooling layer respectively, the features obtained by two pooling processes are subjected to cascading operation, and then are subjected to matrix multiplication with the obtained attention to obtain a fusion feature F TP Corresponding single attention gating feature F TP 。
For up-sampling feature F up The same operation as above is also performed to obtain an upsampling feature F up Corresponding single-attention gating features.
Finally, the feature F is fused TP Corresponding single attention gating feature F TP And upsampling feature F up Corresponding single attention gating feature F up Weighted summation is carried out to obtain a final dual-attention gating characteristic diagram F DAG . The specific expression is:
F′ TP ＝Sum(FC(ReLU(Pool Max (F TP ))),FC(ReLU(Pool Avg (F TP ))))，
F″ TP ＝M(F′ TP ,Concat(Pool Max (M(F TP ,F′ TP )),Pool Avg (M(F TP ,F′ TP ))))，
F′ up ＝Sum(FC(ReLU(Pool Max (F up ))),FC(ReLU(Pool Avg (F up ))))，
F″ up ＝M(F′ up ,Concat(Pool Max (M(F up ,F′ up )),Pool Avg (M(F up ,F′ up ))))，
F DAG ＝Sum(F″ TP ,F″ up )
wherein Pool Max (. Cndot.) and Pool Avg (. Cndot.) represents the max-pooling and average pooling operations, respectively. FC (·) represents fully connected layers, sum (·) represents matrix addition, and M (·) represents matrix multiplication. Concat (-) indicates a cascading operation.
Finally, after linear projection is carried out on the up-sampling characteristics of the first layer of the decoder, the classification information of each pixel belonging to a background area or a focus area and the classification information of each pixel belonging to a focus internal pixel or a focus boundary are output, a focus area is determined according to the classification information, clipping is carried out in an optical coherence tomography image, and a focus area image of a myopic choroidal neovascularization is segmented and output.
Through the bidirectional attention gating module, the up-sampling process of the decoder can pay more attention to the generation of the characteristic information of the focus area, and the prediction result of the classification information has a certain gain effect.
Since the segmentation task of choroidal neovascularization in optical coherence tomography images can be seen as a classification problem at one pixel level, i.e. only background and focus two categories. Thus, the binary cross entropy is chosen as the lesion classification loss. In order to use focus boundary information to supervise the segmentation task for fine segmentation, the boundary information of the focus is also used as a weak supervision information constraint segmentation network to generate a prediction segmentation mask map with rich boundary information, and the boundary information can also be regarded as a pixel-level classification problem, so that the weak supervision boundary loss function also selects a binary cross entropy function as boundary classification loss. Lesion classification lossAnd boundary classification loss->The specific expression of (2) is:
wherein GT i Andrepresenting the ith pixel as the true mark of the focus and the focus boundary, PO i And PB i Representing the predicted value that the ith pixel belongs to the lesion and the lesion boundary, respectively.
In order to prevent the network training process from over fitting and to segment the network so that the lesion area can be accurately predicted, the Dice loss is also used. The Dice loss has been demonstrated as a good performance and stability by multiple medical image segmentation tasks, dice loss L Dice The specific expression of (2) is:
wherein GT i Indicating that the ith pixel is the lesion, PO i Representing the predicted value that the ith pixel belongs to the lesion boundary.
Finally, the loss function of the split network includes: lesion classification loss, boundary classification loss, and Dice loss, specifically expressed as:
the α, β, γ are weights of lesion classification loss, boundary classification loss, and Dice loss, respectively, which are set to 0.4, 0.2, and 0.4 in this embodiment.
The framework of the split network 600 may be a UNet network based on the split network, or may be various modified UNet networks.
In order to demonstrate the effectiveness of the segmentation network of the present invention, several networks that have been shown to be optimal in medical image segmentation in recent years were chosen for comparison experiments, including U-Net, atten U-Net, swin-UNet, UTNet, TMUNet, res _UTNet. For the OCT lesion segmentation task of mCNV of the present invention, the comparative quantization results for each network are shown in table 1:
table 1 quantitative comparison results (%)
As can be seen from table 1, the comparison networks selected in the present invention are all network structures designed by using the U-Net network as the base network model, wherein, swin-UNet, UTNet, TMUNet and res_utnet are both network models combining the CNN structure and the transducer structure, and the 4 networks are also classical networks in which the CNN and the transducer are deeply fused. Compared with a comparison network, the segmentation network of the invention obtains the optimal result on the segmentation indexes participating in comparison. In particular, the most common and objective index and the index IoU of the segmentation are evaluated, and compared with a baseline U-Net network, the index of the segmentation network is improved by approximately 3 percentage points, and the IoU score is improved by more than 4.5 percentage points. By comparing the two algorithms of U-Net and UTNet, the segmentation performance of the network is greatly improved after the tranformer framework is introduced, wherein the score of the Dice coefficient and the score of IoU are improved by more than 1 percent, which indicates that the introduction of the tranformer framework can enhance the extraction of the network to the global features, thereby improving the expression of the network to the focus features. Compared with the second-ranking model in table 1, the split network of the invention is improved by 1.14% and 2.65% in the Dice coefficient and IoU index respectively. The result shows that the segmentation network can fully integrate the local semantic features extracted by the CNN network and the global features depended on the long-distance features extracted by the transducer network, and plays a role in a certain gain on the overall segmentation performance of the network. Further, the results of the division of the respective network models were also qualitatively analyzed, and the results are shown in fig. 5. As can be seen from fig. 5, the two models U-Net and nten U-Net have poor segmentation results, and the segmentation of the focal region in the image is not complete enough, and some of the focal region cannot be detected, which indicates that the network focuses more on semantic feature information in the image and has limited learning ability on spatial and global feature information under the condition of using only the CNN model for image segmentation. In addition, the OCT image of mCNV has more noise and more fuzzy boundary information, which easily causes that the CNN network learns a lot of interference information when learning local semantic features, so that the network learns some wrong information, and the focus features, such as the segmentation result of the second row in fig. 5, cannot be completely expressed, wherein the two methods of the attribute U-Net and the Swin-UNet hardly detect focus areas, and the segmentation result is only a background area. By observing the first and third lines, when the network introduces a transducer framework (such as UTNet, TMUNet, res _UTNet and the like), the global features which are dependent on long distances can be well learned, so that the lesion feature information can be completely expressed. Compared with the network models, the segmentation network provided by the invention can learn local semantic features and long-distance dependent global features better, and learn focus boundary information better, so that the defect of fuzzy focus boundary information of OCT images is overcome.
In summary, since the encoding stage extracts and fuses the multi-scale features on two paths, and the decoding stage fuses the features obtained in the encoding stage by using the bidirectional attention gating unit, the model focuses more on the generation of focus features in the up-sampling process, so that the boundary information of the focus is used as supervision information to enhance the expression of the boundary information, and the fine segmentation effect is improved.
Exemplary method
The embodiment of the invention also provides a choroidal neovascularization segmentation method in the optical coherence tomography image. The above segmentation network is applied to the segmentation of choroidal neovascularization and the segmentation network is trained in advance. Specifically, as shown in fig. 6, the present embodiment includes the following steps:
step S100: the optical coherence tomography image is input into the trained segmentation network.
Step S200: and acquiring the characteristics of the output of the trained segmentation network.
Step S300: based on the above features, a segmentation result is obtained and output.
Specifically, after the optical coherence tomography image is input into the trained segmentation network, the segmentation network outputs classification information of which each pixel belongs to a background area or a focus area and classification information of which each pixel belongs to a focus inner pixel or a focus boundary, the focus area can be determined according to the classification information, then the focus area image of the myopic choroidal neovascularization is segmented and output after cutting in the optical coherence tomography image.
In particular, in this embodiment, the specific functions of each step of the method for dividing a choroidal neovascularization in the optical coherence tomography image may refer to the corresponding descriptions in the system for dividing a choroidal neovascularization in the optical coherence tomography image, which are not described herein.
Based on the above embodiment, the present invention further provides an intelligent terminal, and a functional block diagram thereof may be shown in fig. 7. The intelligent terminal comprises a processor, a memory, a network interface and a display screen which are connected through a system bus. The processor of the intelligent terminal is used for providing computing and control capabilities. The memory of the intelligent terminal comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a choroidal neovascularization segmentation program in optical coherence tomography images. The internal memory provides an environment for an operating system in a non-volatile storage medium and for the operation of a choroidal neovascularization segmentation program in optical coherence tomography images. The network interface of the intelligent terminal is used for communicating with an external terminal through network connection. The method comprises the steps of performing a choroidal neovascularization segmentation method in any one of the above-described optical coherence tomography images when the choroidal neovascularization segmentation program in the optical coherence tomography image is executed by a processor. The display screen of the intelligent terminal can be a liquid crystal display screen or an electronic ink display screen.
It will be appreciated by those skilled in the art that the schematic block diagram shown in fig. 7 is merely a block diagram of a portion of the structure associated with the present inventive arrangements and is not limiting of the smart terminal to which the present inventive arrangements are applied, and that a particular smart terminal may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.
The embodiment of the invention also provides a computer readable storage medium, wherein the computer readable storage medium stores a choroidal neovascularization segmentation program in an optical coherence tomography image, and the choroidal neovascularization segmentation program in the optical coherence tomography image realizes the steps of any one of the choroidal neovascularization segmentation methods in the optical coherence tomography image provided by the embodiment of the invention when being executed by a processor.
It should be understood that the sequence number of each step in the above embodiment does not mean the sequence of execution, and the execution sequence of each process should be determined by its function and internal logic, and should not be construed as limiting the implementation process of the embodiment of the present invention.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions. The functional units and modules in the embodiment may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit, where the integrated units may be implemented in a form of hardware or a form of a software functional unit. In addition, the specific names of the functional units and modules are only for distinguishing from each other, and are not used for limiting the protection scope of the present invention. The specific working process of the units and modules in the above system may refer to the corresponding process in the foregoing method embodiment, which is not described herein again.
In the foregoing embodiments, the descriptions of the embodiments are emphasized, and in part, not described or illustrated in any particular embodiment, reference is made to the related descriptions of other embodiments.
Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus/terminal device and method may be implemented in other manners. For example, the apparatus/terminal device embodiments described above are merely illustrative, e.g., the division of the modules or units described above is merely a logical function division, and may be implemented in other manners, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed.
The integrated modules/units described above, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer-readable storage medium. Based on such understanding, the present invention may implement all or part of the flow of the method of the above embodiment, or may be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the steps of each method embodiment may be implemented. The computer program comprises computer program code, and the computer program code can be in a source code form, an object code form, an executable file or some intermediate form and the like. The computer readable medium may include: any entity or device capable of carrying the computer program code described above, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth. The content of the computer readable storage medium can be appropriately increased or decreased according to the requirements of the legislation and the patent practice in the jurisdiction.
The above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will understand that; the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions are not intended to depart from the spirit and scope of the various embodiments of the invention, which are also within the spirit and scope of the invention.
Claims (10)
1. A choroidal neovascularization segmentation system in an optical coherence tomography image, comprising:
a segmentation network constructed based on UNet for segmenting choroidal neovascularization in an optical coherence tomography image, the segmentation network being provided with an encoder for downsampling and a decoder for upsampling;
a dual-path network is arranged in the encoder, a first path of the dual-path network is provided with a plurality of transducer blocks, and the first path is used for extracting first global features of a plurality of scales; a plurality of mixing units for extracting local semantic information and long-distance dependent information are arranged on a second path of the dual-path network, the second path is used for extracting second global features with multiple scales, and the scales of the first global features are in one-to-one correspondence with the scales of the second global features;
the encoder is also provided with a plurality of dual-path fusion units, and the dual-path fusion units are used for fusing the first global features and the second global features on each scale;
the decoder is provided with a plurality of bidirectional attention gating units and a cascading unit, the cascading unit is used for cascading the characteristics output by the dual-path fusion unit and the characteristics output by the bidirectional attention gating unit, the bidirectional attention gating unit is used for fusing up-sampling characteristics and the characteristics output by the dual-path fusion unit on adjacent layers, and the up-sampling characteristics are obtained by up-sampling the characteristics output by the cascading unit or the dual-path fusion unit;
and the output module is used for obtaining and outputting a segmentation result according to the characteristics output by the segmentation network.
2. The choroidal neovascularization segmentation system in an optical coherence tomography image of claim 1, wherein said mixing unit comprises a connected residual convolution network and a transducer network.
3. The choroidal neovascularization segmentation system of claim 1, wherein one of said dual path fusion units has a path attention encoding module for enhancing the interdependence of the path attention map extracted by said mixing unit, and the other path has a position attention encoding module for enhancing the spatial information of the features extracted by said transducer block.
4. The choroidal neovascularization segmentation system in an optical coherence tomography image of claim 1, wherein said bi-directional attention gating unit comprises two bi-directional attention gating modules for enhancing features output by said dual path fusion unit and said upsampling features, respectively, and a summing module for matrix summing features output by said two bi-directional attention gating modules.
5. The system of claim 4, wherein the bi-directional attention gating module is provided with a first pooling layer, a dot-product layer, and a second pooling layer, the first pooling layer and the second pooling layer each being provided with a maximum pooling branch and an average pooling branch, the first pooling layer being configured to pool and fully connect features input to the bi-directional attention gating module, the dot-product layer being configured to dot-product features output from the first pooling layer with features input to the bi-directional attention gating module, and the second pooling layer being configured to pool features output from the dot-product layer.
6. The choroidal neovascularization segmentation system in an optical coherence tomography image of claim 1, wherein training the segmentation network employs a loss function comprising: lesion classification loss, boundary classification loss, and Dice loss.
7. A method of dividing a choroidal neovascularization in an optical coherence tomography image, characterized by pre-training a dividing network in the choroidal neovascularization dividing system in the optical coherence tomography image as defined in any one of claims 1 to 6;
inputting the optical coherence tomography image into a trained segmentation network;
acquiring the characteristics of the output of the trained segmentation network;
based on the features, a segmentation result is obtained and output.
8. The method of choroidal neovascularization segmentation in an optical coherence tomography image of claim 7, wherein the loss function of the training segmentation network is:
wherein the method comprises the steps ofClassification of loss for lesions->For boundary classification loss, L Dice For the Dice loss, GT i Andrepresenting the ith pixel as the true mark of the focus and the focus boundary, PO i And PB i The predicted values of the ith pixel belonging to the focus and the focus boundary are respectively shown, and alpha, beta and gamma are respectively the weights of focus classification loss, boundary classification loss and price loss.
9. A smart terminal comprising a memory, a processor, and a choroidal neovascularization segmentation program in an optical coherence tomography image stored on the memory and executable on the processor, the choroidal neovascularization segmentation program in the optical coherence tomography image, when executed by the processor, implementing the steps of the choroidal neovascularization segmentation method in an optical coherence tomography image as recited in claim 7.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a choroidal neovascularization segmentation procedure in an optical coherence tomography image, which when executed by a processor, implements the steps of the method of choroidal neovascularization segmentation in an optical coherence tomography image according to claim 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310655638.8A CN116883420A (en) | 2023-06-02 | 2023-06-02 | Choroidal neovascularization segmentation method and system in optical coherence tomography image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310655638.8A CN116883420A (en) | 2023-06-02 | 2023-06-02 | Choroidal neovascularization segmentation method and system in optical coherence tomography image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116883420A true CN116883420A (en) | 2023-10-13 |
Family
ID=88266913
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310655638.8A Pending CN116883420A (en) | 2023-06-02 | 2023-06-02 | Choroidal neovascularization segmentation method and system in optical coherence tomography image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116883420A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117726642A (en) * | 2024-02-07 | 2024-03-19 | 中国科学院宁波材料技术与工程研究所 | High reflection focus segmentation method and device for optical coherence tomography image |
| CN117726642B (en) * | 2024-02-07 | 2024-05-31 | 中国科学院宁波材料技术与工程研究所 | High reflection focus segmentation method and device for optical coherence tomography image |
-
2023
- 2023-06-02 CN CN202310655638.8A patent/CN116883420A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117726642A (en) * | 2024-02-07 | 2024-03-19 | 中国科学院宁波材料技术与工程研究所 | High reflection focus segmentation method and device for optical coherence tomography image |
| CN117726642B (en) * | 2024-02-07 | 2024-05-31 | 中国科学院宁波材料技术与工程研究所 | High reflection focus segmentation method and device for optical coherence tomography image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110111313B (en) | Medical image detection method based on deep learning and related equipment | |
| CN110807788B (en) | Medical image processing method, medical image processing device, electronic equipment and computer storage medium | |
| Li et al. | Connection sensitive attention U-NET for accurate retinal vessel segmentation | |
| CN110298844B (en) | X-ray radiography image blood vessel segmentation and identification method and device | |
| Chen et al. | PCAT-UNet: UNet-like network fused convolution and transformer for retinal vessel segmentation | |
| Hu et al. | AS-Net: Attention Synergy Network for skin lesion segmentation | |
| Guo et al. | Retinal vessel segmentation combined with generative adversarial networks and dense U-Net | |
| KR20190105180A (en) | Apparatus for Lesion Diagnosis Based on Convolutional Neural Network and Method thereof | |
| CN113205524B (en) | Blood vessel image segmentation method, device and equipment based on U-Net | |
| CN116309648A (en) | Medical image segmentation model construction method based on multi-attention fusion | |
| CN113012163A (en) | Retina blood vessel segmentation method, equipment and storage medium based on multi-scale attention network | |
| CN115546570A (en) | Blood vessel image segmentation method and system based on three-dimensional depth network | |
| Agrawal et al. | Deep dive in retinal fundus image segmentation using deep learning for retinopathy of prematurity | |
| CN115809998A (en) | Based on E 2 Glioma MRI data segmentation method based on C-Transformer network | |
| CN112884788A (en) | Cup optic disk segmentation method and imaging method based on rich context network | |
| Jiang et al. | MFI-Net: A multi-resolution fusion input network for retinal vessel segmentation | |
| CN112633416A (en) | Brain CT image classification method fusing multi-scale superpixels | |
| CN116363081A (en) | Placenta implantation MRI sign detection classification method and device based on deep neural network | |
| CN114418987A (en) | Retinal vessel segmentation method and system based on multi-stage feature fusion | |
| CN113610842A (en) | OCT image retina detachment and splitting automatic segmentation method based on CAS-Net | |
| CN110009641A (en) | Crystalline lens dividing method, device and storage medium | |
| Liu et al. | Feature pyramid U‐Net for retinal vessel segmentation | |
| CN111507950B (en) | Image segmentation method and device, electronic equipment and computer-readable storage medium | |
| Huang et al. | DBFU-Net: Double branch fusion U-Net with hard example weighting train strategy to segment retinal vessel | |
| Desiani et al. | A Novelty Patching of Circular Random and Ordered Techniques on Retinal Image to Improve CNN U-Net Performance. |
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 |