CN111724401A - Image segmentation method and system based on boundary constraint cascade U-Net - Google Patents

Image segmentation method and system based on boundary constraint cascade U-Net Download PDF

Info

Publication number
CN111724401A
CN111724401A CN202010383001.4A CN202010383001A CN111724401A CN 111724401 A CN111724401 A CN 111724401A CN 202010383001 A CN202010383001 A CN 202010383001A CN 111724401 A CN111724401 A CN 111724401A
Authority
CN
China
Prior art keywords
boundary
network
segmentation
image
net
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
Application number
CN202010383001.4A
Other languages
Chinese (zh)
Inventor
李强
郭薇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huazhong University of Science and Technology
Original Assignee
Huazhong University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Huazhong University of Science and Technology filed Critical Huazhong University of Science and Technology
Priority to CN202010383001.4A priority Critical patent/CN111724401A/en
Publication of CN111724401A publication Critical patent/CN111724401A/en
Pending legal-status Critical Current

Links

Images

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
    • G06T7/12—Edge-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
    • 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
    • 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
    • G06T7/11—Region-based segmentation

Abstract

The invention discloses an image segmentation method and system based on boundary constraint cascade U-Net, and belongs to the field of medical image segmentation. The method comprises the following steps: constructing an image segmentation network with boundary constraint; the network comprises a cascaded U-Net boundary extraction sub-network and a U-Net region segmentation sub-network; the boundary extraction sub-network is used for extracting the boundary of the input image to obtain a probability map of the target boundary; the region segmentation sub-network is used for segmenting the input image under the constraint of the target boundary probability to obtain a probability map of the target region; taking a large number of images, corresponding manually-drawn regions and boundary labels as a training set, and carrying out iterative training on a boundary constraint cascade U-Net segmentation network to obtain a trained image segmentation model; and inputting the image to be segmented into the trained image segmentation model to obtain the segmentation result of the target area. The invention can fully utilize the detail information of the target boundary and obtain better segmentation performance.

Description

Image segmentation method and system based on boundary constraint cascade U-Net
Technical Field
The invention belongs to the field of medical image segmentation, and particularly relates to a medical image segmentation method and system based on boundary constraint cascade U-Net.
Background
The medical image segmentation is a complex and key step in the field of medical image understanding and analysis, provides reliable basis for clinical diagnosis and pathology research, and assists doctors to make more accurate diagnosis. In recent years, with the rapid development of the establishment of large-scale image databases and the rapid development of computing power, the deep learning method has achieved breakthrough research results in the fields of computer vision and image processing, and the strong feature learning capability of the deep learning method has attracted extensive attention. Therefore, the related method is applied to medical image segmentation, repeated and tedious work of doctors can be effectively reduced, the working efficiency is improved, and an auxiliary effect is provided for subsequent focus analysis.
The outstanding performance of a Convolutional Neural Network (CNN) in an ImageNet game shows that the CNN has great potential in the fields of image feature extraction, image segmentation and classification and the like. CNNs were initially applied to segment images, using equal-sized image patches around each pixel as input to a convolutional neural network to train and predict the network. But this would cause too much storage overhead due to the division of the image into small blocks of storage. Moreover, most convolutions are repeatedly calculated many times, which causes a disadvantage of low efficiency. To address the above problems, Long et al proposed a full Convolutional neural Network (FCN) in 2015. The network uses convolution layers to replace full connection layers in the common CNN, and deconvolution is adopted to enable the size of a network output image to be consistent with that of an input image, so that end-to-end segmentation is achieved. Meanwhile, the network also uses jump connection to fuse the extracted different features, so that the segmentation precision is improved. Subsequently, many researchers have improved the FCN structure for its widespread use in medical image segmentation.
After the advent of the FCN network, Ronneberger et al proposed a U-Net network in 2015 and achieved breakthrough performance in biomedical image segmentation. The U-shaped full convolution network is formed by a contraction descending path and a following symmetrical expansion ascending path, so that the U-Net is formed. As the U-Net directly transmits the characteristic diagram of the contraction path to the expansion path by using the jump connection, the U-Net is more accurate in pixel positioning, and the segmentation result is more accurate than that of FCN. Therefore, since the proposal of U-Net, most researchers use U-Net when performing medical image segmentation, and put forward improved models such as ResNet U-Net,3D U-Net,3D V-Net and Unet + +, on the basis of the U-Net.
Although the golden standard of medical image segmentation is determined by marking boundaries manually by doctors, most of the existing image segmentation networks realize segmentation by learning regional information of targets, neglecting detailed information of the target boundaries manually calibrated by doctors, and reducing the accuracy of segmentation results.
Disclosure of Invention
In view of the above defects or improvement needs of the prior art, the present invention provides a method and a system for segmenting a medical image based on boundary constraint cascade U-Net, which aims to introduce boundary constraint in the medical image segmentation, thereby improving the accuracy of the segmentation result.
To achieve the above object, according to an aspect of the present invention, there is provided an image segmentation method based on boundary constraint cascade U-Net, including:
s1, constructing an image segmentation network with boundary constraint;
the image segmentation network comprises a boundary extraction sub-network and a region segmentation sub-network which are cascaded; the boundary extraction sub-network and the region segmentation sub-network have the same structure and adopt U-Net networks;
the boundary extraction sub-network is used for extracting the boundary of the input image to obtain a probability map of the target boundary; the region segmentation sub-network is used for segmenting the input image under the constraint of the target boundary probability to obtain a probability map of the target region;
s2, using a large number of images and manually drawn boundary labels as a training set, and iteratively training a boundary extraction sub-network; then, a large number of images, corresponding extracted boundary probability graphs and manually drawn region labels are used as training sets, and a region segmentation sub-network is iteratively trained; finally, obtaining a trained image segmentation model with boundary constraint;
and S3, inputting the image to be segmented into the trained image segmentation model to obtain the segmentation result of the target area.
Furthermore, the loss functions adopted by the training region segmentation sub-network and the boundary extraction sub-network are all loss functions
Figure BDA0002482763330000031
Wherein p isiAnd giThe pixel values of points i in the output probability map P and the boundary label map G are shown, respectively, and N is the sum of pixels of the object in the image.
Further, the output of the boundary extraction sub-network is used as the input of the region segmentation sub-network together with the original image, and the segmentation of the region is constrained by extracting the boundary probability map.
Further, the U-Net network comprises a contraction path and an expansion path which are distributed in a symmetrical structure.
According to another aspect of the present invention, there is provided an image segmentation system based on boundary constraint cascade U-Net, comprising:
the image segmentation network construction module is used for constructing an image segmentation network with boundary constraint; the image segmentation network comprises a boundary extraction sub-network and a region segmentation sub-network which are cascaded; the boundary extraction sub-network and the region segmentation sub-network have the same structure and adopt U-Net networks; the boundary extraction sub-network is used for extracting the boundary of the input image to obtain a probability map of the target boundary; the region segmentation sub-network is used for segmenting the input image under the constraint of the target boundary probability to obtain a probability map of the target region;
the image segmentation network training module is used for iteratively training a boundary extraction sub-network by taking a large number of images and manually drawn boundary labels as a training set; then, a large number of images, corresponding extracted boundary probability graphs and manually drawn region labels are used as training sets, and a region segmentation sub-network is iteratively trained; finally, obtaining a trained image segmentation model with boundary constraint;
and the image segmentation model application module is used for inputting the image to be segmented into the trained image segmentation model to obtain the segmentation result of the target area.
Furthermore, the loss functions adopted by the training region segmentation sub-network and the boundary extraction sub-network are all loss functions
Figure BDA0002482763330000041
Wherein p isiAnd giThe pixel values of points i in the output probability map P and the label map G are shown, respectively, and N is the sum of pixels of the object in the image.
Further, the output of the boundary extraction sub-network is used as the input of the region segmentation sub-network together with the original image, and the segmentation of the region is constrained by extracting the boundary probability map.
Further, the U-Net network comprises a contraction path and an expansion path which are distributed in a symmetrical structure.
In general, the above technical solutions contemplated by the present invention can achieve the following advantageous effects compared to the prior art.
In the invention, in the process of training the segmentation network, the target boundary is added to constrain the training of the segmentation model, so that the model obtained by training focuses more on the shape of the boundary. Compared with the target internal region, the boundary plays a more important role in the segmentation process, so the cascaded U-Net network with the boundary constraint provided by the invention can more fully utilize the target boundary information which is more critical to the segmentation while learning the image region information, thereby being capable of obtaining better segmentation performance.
Drawings
FIG. 1 is a flowchart of a boundary constraint cascade U-Net-based medical image segmentation method provided by an embodiment of the invention;
FIG. 2 is a structural diagram of a U-Net according to an embodiment of the present invention;
fig. 3 is a two-dimensional esophageal CT image, wherein the white curves in fig. 3(a) -3 (c) are the esophageal boundaries obtained by physician manual delineation, conventional U-Net segmentation, and cascaded U-Net segmentation with boundary constraints, respectively.
Fig. 4 is a two-dimensional cardiac CT image in which the white curves in fig. 4(a) -4 (c) are the cardiac boundaries obtained by physician manual delineation, conventional U-Net segmentation, and cascaded U-Net segmentation with boundary constraints, respectively.
Fig. 5 is a two-dimensional tracheal CT image in which the white curves in fig. 5(a) -5 (c) are the tracheal boundaries obtained by physician manual delineation, conventional U-Net segmentation, and cascaded U-Net segmentation with boundary constraints, respectively.
Fig. 6 is a two-dimensional aorta CT image in which the white curves in fig. 6(a) -6 (c) are the aorta boundaries obtained by physician manual delineation, conventional U-Net segmentation, and cascaded U-Net segmentation with boundary constraints, respectively.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
As shown in fig. 1, the present invention provides an image segmentation method based on boundary constraint cascade U-Net, including:
s1, constructing an image segmentation network with boundary constraint;
the image segmentation network comprises a boundary extraction sub-network and a region segmentation sub-network which are cascaded; the boundary extraction sub-network and the region segmentation sub-network have the same structure and adopt U-Net networks; the boundary extraction sub-network is used for extracting the boundary of the input image to obtain a probability map of the target boundary; the region segmentation sub-network is used for segmenting the input image under the constraint of the target boundary probability to obtain a probability map of a target region;
specifically, the input data of the boundary extraction sub-network is an original image, and is output as a probability graph of a target boundary; the input of the area segmentation sub-network is a two-channel image consisting of an original image and an extracted boundary probability map, and the output is a probability map of a target area; the network adopts a U-Net structure as shown in FIG. 2, which consists of a left contraction path and a right symmetrical expansion path. The systolic path uses 2 volumes of 3 x 3 and 1 maximum pooling operation of 2 x 2 multiple times, and the convolution operations are followed by the ReLU activation function. The number of eigen-channels is doubled after each downsampling operation is completed. The expansion path is an up-sampling process that is the reverse of the contraction path. The dilation path gradually restores the details and corresponding spatial dimensions of the target. Each step in the dilation path includes upsampling of the feature map, which is followed by a convolution operation with a convolution kernel size of 2 x 2 and halving the number of feature channels.
The fusion of the low-level characteristic diagram and the high-level characteristic diagram is realized through the copy channel, the capability of capturing different levels of characteristic information by a network is enhanced, and the segmentation accuracy is effectively improved. After replication, two convolution operations with kernel size 3 × 3 were performed, each convolution operation followed by a ReLU as activation function. And in the last layer, the number of the characteristic graph channels is changed from 32 to 1 by using convolution of 1 x 1, and the characteristic graph is changed into the prediction probability by using a sigmoid function.
S2, using a large number of images and manually drawn boundary labels as a training set, and iteratively training a boundary extraction sub-network; then, a large number of images, corresponding extracted boundary probability graphs and manually drawn region labels are used as training sets, and a region segmentation sub-network is iteratively trained; finally, obtaining a trained image segmentation model with boundary constraint;
specifically, compared with the target internal region, the boundary plays a more important role in the segmentation process, so that the target boundary is added to restrict the training of the region segmentation sub-network in the network training process, so that the model obtained by training focuses more on the shape of the boundary, and better segmentation performance is obtained.
The loss functions adopted by the training region segmentation sub-network and the boundary extraction sub-network are as follows:
Figure BDA0002482763330000061
wherein p isiAnd giThe pixel values of points i in the output probability map P and the label map G are shown, respectively, and N is the sum of pixels of the object in the image.
And S3, inputting the image to be segmented into the trained medical image segmentation model to obtain the segmentation result of the target area.
In particular, images that can be segmented by the method of the present invention include, but are not limited to, medical images.
In order to verify the effectiveness of the method, the esophagus, the heart, the trachea and the aorta in 40 groups of CT images are segmented, and the performance of the algorithm is verified by adopting a 5-fold method, namely 4 groups of data training and 1 group of testing are adopted each time. The experiment was repeated 5 times until all data were tested. In the embodiment of the invention, the average Dice coefficient (the intersection of the 2-fold segmentation result and the gold standard is divided by the area sum of the segmentation result and the gold standard) of all 40 sets of test data is used for measuring the segmentation performance of the cascaded U-Net network with boundary constraint. The Dice coefficient can measure the overall consistency between the segmentation result and the gold standard, and the value ranges from 0 to 1, wherein 0 means that the segmentation result and the gold standard are not overlapped, and 1 means that the segmentation result and the gold standard are completely overlapped.
The specific experimental process is as follows,
(1) verifying the organ segmentation performance of the boundary constraint cascade U-Net network;
esophagus, heart, trachea and aorta in CT images were segmented using conventional U-Net and cascaded U-Net with boundary constraints, respectively, with Dice coefficients for the segmentation as shown in table 1. When the boundary segmentation network is trained, the golden standard is an annular area with 5 pixel widths by taking a doctor to manually draw a boundary as a center.
TABLE 1
Organ Esophagus Heart and heart Trachea Aorta
Traditional U-Net 0.750 0.944 0.921 0.875
Cascade network 0.781 0.947 0.930 0.925
As can be seen from table 1, the segmentation performance was slightly improved after the boundary constraint was imposed on the heart and trachea, and the Dice coefficients were 0.947 and 0.930, respectively. For esophagus and aorta, after the boundary is added, the segmentation performance is obviously improved. The Dice coefficient of the esophagus segmentation is increased from 0.765 to 0.781, and the Dice coefficient of the aorta segmentation is increased from 0.875 to 0.925. The results show that organ segmentation performance with the boundary constrained cascaded U-Net network is better than that of the traditional U-Net.
(2) Organ segmentation result comparison with boundary constraint U-Net network
Fig. 3-6 are graphs of the results of segmentation of the esophagus, heart, trachea, and aorta, respectively. Wherein, the graphs (a), (b) and (c) respectively represent the boundary calibration manually by the doctor, the traditional U-Net segmentation result and the cascading U-Net segmentation result with the boundary constraint. As can be seen from the figure, compared with the cascading U-Net with boundary constraint provided by the application, the boundary segmentation of the traditional U-Net method is not accurate, and even the segmentation is mistaken or missed.
The embodiment of the invention also provides an image segmentation system based on boundary constraint cascade U-Net, and the functions and implementation modes of all the modules of the system are consistent with the method, and are not repeated herein.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims (8)

1. An image segmentation method based on boundary constraint cascade U-Net is characterized by comprising the following steps:
s1, constructing an image segmentation network with boundary constraint;
the image segmentation network comprises a boundary extraction sub-network and a region segmentation sub-network which are cascaded; the boundary extraction sub-network and the region segmentation sub-network have the same structure and adopt U-Net networks;
the boundary extraction sub-network is used for extracting the boundary of the input image to obtain a probability map of the target boundary; the region segmentation sub-network is used for segmenting the input image under the constraint of the target boundary probability to obtain a probability map of the target region;
s2, using a large number of images and manually-drawn area labels as a training set, and iteratively training a boundary extraction sub-network; then, a large number of images, corresponding extracted boundary probability graphs and manually drawn region labels are used as training sets, and a region segmentation sub-network is iteratively trained; finally, obtaining a trained image segmentation model with boundary constraint;
and S3, inputting the image to be segmented into the trained image segmentation model to obtain the segmentation result of the target area.
2. The image segmentation method based on boundary constraint cascade U-Net as claimed in claim 1, wherein the loss functions adopted by the training region segmentation sub-network and the boundary extraction sub-network are all loss functions
Figure FDA0002482763320000011
Wherein p isiAnd giAre respectively provided withThe pixel values of points i in the output probability map P and the label map G are shown, and N represents the sum of pixels of the object in the image.
3. A method for segmenting medical images based on boundary constrained cascade U-Net according to claim 1 or 2, characterized in that the output of the boundary extraction sub-network is taken as the input of the region segmentation sub-network together with the original image, and the region probability map is constrained by extracting the boundary probability map.
4. The boundary constraint cascade U-Net based medical image segmentation method according to any one of claims 1 to 3, wherein the U-Net network comprises a contraction path and an expansion path distributed in a symmetrical structure.
5. An image segmentation system based on boundary constraint cascade U-Net, which is characterized by comprising:
the image segmentation network construction module is used for constructing an image segmentation network with boundary constraint; the image segmentation network comprises a boundary extraction sub-network and a region segmentation sub-network which are cascaded; the boundary extraction sub-network and the region segmentation sub-network have the same structure and adopt U-Net networks; the boundary extraction sub-network is used for extracting the boundary of the input image to obtain a probability map of the target boundary; the region segmentation sub-network is used for segmenting the input image under the constraint of the target boundary probability to obtain a probability map of the target region;
the image segmentation network training module is used for iteratively training a boundary extraction sub-network by taking a large number of images and manually drawn boundary labels as a training set; then, a large number of images, corresponding extracted boundary probability graphs and manually drawn region labels are used as training sets, and a region segmentation sub-network is iteratively trained; finally, obtaining a trained image segmentation model with boundary constraint;
and the image segmentation model application module is used for inputting the image to be segmented into the trained image segmentation model to obtain the segmentation result of the target area.
6. The image segmentation system based on boundary constraint cascade U-Net as claimed in claim 5, wherein the loss functions adopted by the training region segmentation sub-network and the boundary extraction sub-network are all loss functions
Figure FDA0002482763320000021
Wherein p isiAnd giThe pixel values of points i in the output probability map P and the label map G are shown, respectively, and N is the sum of pixels of the object in the image.
7. A medical image segmentation system based on boundary constrained cascade U-Net as claimed in claim 5 or 6, characterized in that the output of the boundary extraction sub-network is used as the input of the region segmentation sub-network together with the original image, and the region segmentation is constrained by the extracted boundary probability map.
8. A medical image segmentation system based on boundary constrained cascade U-Net according to any one of claims 5 to 7, characterized in that the U-Net network comprises a contraction path and an expansion path distributed in a symmetrical structure.
CN202010383001.4A 2020-05-08 2020-05-08 Image segmentation method and system based on boundary constraint cascade U-Net Pending CN111724401A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010383001.4A CN111724401A (en) 2020-05-08 2020-05-08 Image segmentation method and system based on boundary constraint cascade U-Net

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010383001.4A CN111724401A (en) 2020-05-08 2020-05-08 Image segmentation method and system based on boundary constraint cascade U-Net

Publications (1)

Publication Number Publication Date
CN111724401A true CN111724401A (en) 2020-09-29

Family

ID=72564421

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010383001.4A Pending CN111724401A (en) 2020-05-08 2020-05-08 Image segmentation method and system based on boundary constraint cascade U-Net

Country Status (1)

Country Link
CN (1) CN111724401A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112489062A (en) * 2020-12-10 2021-03-12 中国科学院苏州生物医学工程技术研究所 Medical image segmentation method and system based on boundary and neighborhood guidance
CN112530554A (en) * 2020-12-17 2021-03-19 上海联影医疗科技股份有限公司 Scanning positioning method and device, storage medium and electronic equipment
CN112560857A (en) * 2021-02-20 2021-03-26 鹏城实验室 Character area boundary detection method, equipment, storage medium and device
CN112801968A (en) * 2021-01-22 2021-05-14 常州市第二人民医院 Double-layer depth network model, method and device for nuclear magnetic image segmentation
CN113077469A (en) * 2021-03-16 2021-07-06 中山大学 Sketch image semantic segmentation method and device, terminal device and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140270366A1 (en) * 2013-03-14 2014-09-18 Nec Laboratories America, Inc. Dimension-Wise Spatial Layout Importance Selection: An Alternative Way to Handle Object Deformation
CN110120033A (en) * 2019-04-12 2019-08-13 天津大学 Based on improved U-Net neural network three-dimensional brain tumor image partition method
US20190347557A1 (en) * 2018-05-14 2019-11-14 Tempus Labs, Inc. Generalizable and Interpretable Deep Learning Framework for Predicting MSI from Histopathology Slide Images
CN111028242A (en) * 2019-11-27 2020-04-17 中国科学院深圳先进技术研究院 Automatic tumor segmentation system and method and electronic equipment
CN111047606A (en) * 2019-12-05 2020-04-21 北京航空航天大学 Pathological full-section image segmentation algorithm based on cascade thought

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140270366A1 (en) * 2013-03-14 2014-09-18 Nec Laboratories America, Inc. Dimension-Wise Spatial Layout Importance Selection: An Alternative Way to Handle Object Deformation
US20190347557A1 (en) * 2018-05-14 2019-11-14 Tempus Labs, Inc. Generalizable and Interpretable Deep Learning Framework for Predicting MSI from Histopathology Slide Images
CN110120033A (en) * 2019-04-12 2019-08-13 天津大学 Based on improved U-Net neural network three-dimensional brain tumor image partition method
CN111028242A (en) * 2019-11-27 2020-04-17 中国科学院深圳先进技术研究院 Automatic tumor segmentation system and method and electronic equipment
CN111047606A (en) * 2019-12-05 2020-04-21 北京航空航天大学 Pathological full-section image segmentation algorithm based on cascade thought

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112489062A (en) * 2020-12-10 2021-03-12 中国科学院苏州生物医学工程技术研究所 Medical image segmentation method and system based on boundary and neighborhood guidance
CN112489062B (en) * 2020-12-10 2024-01-30 中国科学院苏州生物医学工程技术研究所 Medical image segmentation method and system based on boundary and neighborhood guidance
CN112530554A (en) * 2020-12-17 2021-03-19 上海联影医疗科技股份有限公司 Scanning positioning method and device, storage medium and electronic equipment
CN112530554B (en) * 2020-12-17 2023-08-18 上海联影医疗科技股份有限公司 Scanning positioning method and device, storage medium and electronic equipment
CN112801968A (en) * 2021-01-22 2021-05-14 常州市第二人民医院 Double-layer depth network model, method and device for nuclear magnetic image segmentation
CN112801968B (en) * 2021-01-22 2022-03-22 常州市第二人民医院 Double-layer depth network model, method and device for nuclear magnetic image segmentation
CN112560857A (en) * 2021-02-20 2021-03-26 鹏城实验室 Character area boundary detection method, equipment, storage medium and device
CN113077469A (en) * 2021-03-16 2021-07-06 中山大学 Sketch image semantic segmentation method and device, terminal device and storage medium
CN113077469B (en) * 2021-03-16 2023-01-24 中山大学 Sketch image semantic segmentation method and device, terminal device and storage medium

Similar Documents

Publication Publication Date Title
CN111724401A (en) Image segmentation method and system based on boundary constraint cascade U-Net
US11373305B2 (en) Image processing method and device, computer apparatus, and storage medium
CN108734290B (en) Convolutional neural network construction method based on attention mechanism and application
CN108230339B (en) Stomach cancer pathological section labeling completion method based on pseudo label iterative labeling
CN111445478B (en) Automatic intracranial aneurysm region detection system and detection method for CTA image
CN110298383B (en) Multi-mode deep learning-based pathology classification method and system
CN108335303B (en) Multi-scale palm skeleton segmentation method applied to palm X-ray film
CN111242288B (en) Multi-scale parallel deep neural network model construction method for lesion image segmentation
CN112862824A (en) Novel coronavirus pneumonia focus detection method, system, device and storage medium
CN113223005B (en) Thyroid nodule automatic segmentation and grading intelligent system
CN112085741B (en) Gastric cancer pathological section segmentation algorithm based on deep learning
CN116228792A (en) Medical image segmentation method, system and electronic device
CN116630971B (en) Wheat scab spore segmentation method based on CRF_Resunate++ network
CN112927237A (en) Honeycomb lung focus segmentation method based on improved SCB-Unet network
Mamdouh et al. A New Model for Image Segmentation Based on Deep Learning.
CN112634308B (en) Nasopharyngeal carcinoma target area and organ-at-risk delineating method based on different receptive fields
CN112686932B (en) Image registration method for medical image, image processing method and medium
CN116778346B (en) Pipeline identification method and system based on improved self-attention mechanism
CN116778164A (en) Semantic segmentation method for improving deep V < 3+ > network based on multi-scale structure
CN113192076B (en) MRI brain tumor image segmentation method combining classification prediction and multi-scale feature extraction
CN114627123A (en) Leucocyte detection method integrating double-current weighting network and spatial attention mechanism
CN114120359A (en) Method for measuring body size of group-fed pigs based on stacked hourglass network
CN112053311A (en) Image segmentation method based on expansion convolution
CN111144422A (en) Positioning identification method and system for aircraft component
CN110992309A (en) Fundus image segmentation method based on deep information transfer network

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20200929