CN111429421A - Model generation method, medical image segmentation method, device, equipment and medium - Google Patents
Model generation method, medical image segmentation method, device, equipment and medium Download PDFInfo
- Publication number
- CN111429421A CN111429421A CN202010197600.7A CN202010197600A CN111429421A CN 111429421 A CN111429421 A CN 111429421A CN 202010197600 A CN202010197600 A CN 202010197600A CN 111429421 A CN111429421 A CN 111429421A
- Authority
- CN
- China
- Prior art keywords
- image
- lung
- segmentation
- network
- knowledge
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- 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
- G06T2207/10081—Computed x-ray tomography [CT]
-
- 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
- G06T2207/10116—X-ray image
-
- 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
- G06T2207/30061—Lung
-
- 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
- G06T2207/30204—Marker
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Medical Informatics (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention discloses a model generation method, a medical image segmentation device, equipment and a medium. The model generation method comprises the following steps: acquiring a sample image of a sample chest, a lung mask image of a known lung in the sample image and shape prior knowledge of the lung mask image; taking the sample image, the lung mask image and the shape priori knowledge as a group of training samples, training an original segmentation model based on a plurality of groups of training samples, and generating a lung segmentation model; the original segmentation model comprises a feature extraction network, an image segmentation network and a priori knowledge regression network, wherein the image segmentation network and the priori knowledge regression network are respectively connected with the feature extraction network. The technical scheme of the embodiment of the invention solves the problem that the prior segmentation model does not effectively utilize the shape prior knowledge of the lung region, and the prior knowledge regression network set based on the shape prior knowledge is matched with the existing image segmentation network, so that the matching degree of the lung segmentation model and the lung region segmentation is improved.
Description
Technical Field
The embodiment of the invention relates to the technical field of medical image processing, in particular to a model generation method, a medical image segmentation device, equipment and a medium.
Background
X-ray imaging is the most common imaging modality used in medical examinations due to its low radiation dose and low cost. The X-ray chest radiograph obtained based on the X-ray imaging technology can be used for researching various structures in the chest cavity, for example, lung region information such as size, irregular shape and total lung volume extracted from the X-ray chest radiograph can be used as an important reference factor for medical staff to diagnose emphysema, lung cancer, tuberculosis, emphysema, pneumothorax, heart disease, pneumoconiosis and other clinical diseases. Therefore, the lung region is accurately segmented from the X-ray chest radiograph, and the method plays an important role in the subsequent medical image analysis process.
However, accurate segmentation of lung regions from chest X-ray images presents a number of difficulties and challenges, firstly, the shape and appearance of lung regions vary greatly due to differences in gender, age and health of individual patients; secondly, the existence of external objects such as a sternum line, an operation clamp, a pacemaker and the like can further increase the segmentation difficulty of the lung region; again, certain anatomical structures of the lungs may also lead to segmentation difficulties.
In recent years, with the rapid development of deep learning, in particular, Convolutional Neural Network (CNN) models, a large number of network models applicable to medical image segmentation have been proposed in succession, and these network models are also applicable to lung region segmentation in X-ray chest radiographs. However, these network models do not take into account the unique characteristics of the lung region at the time of setting, and they employ the same network structure for almost all segmentation tasks, which leaves the segmentation accuracy of the lung region to be improved.
Disclosure of Invention
The embodiment of the invention provides a model generation method, a medical image segmentation method, a device, equipment and a medium, which are used for generating a lung segmentation model matched with lung region segmentation.
In a first aspect, an embodiment of the present invention provides a model generation method, which may include:
acquiring a sample image of a sample chest, a lung mask image of a known lung in the sample image and shape prior knowledge of the lung mask image;
taking the sample image, the lung mask image and the shape priori knowledge as a group of training samples, training an original segmentation model based on a plurality of groups of training samples, and generating a lung segmentation model;
the original segmentation model comprises a feature extraction network, an image segmentation network and a priori knowledge regression network, wherein the image segmentation network and the priori knowledge regression network are respectively connected with the feature extraction network.
Optionally, training the original segmentation model based on multiple sets of training samples to generate a lung segmentation model, which may include:
inputting the sample image into an original segmentation model, and obtaining a lung predicted image and shape prediction knowledge according to an output result of the original segmentation model;
determining a segmentation loss function of an image segmentation network according to the lung predicted image and the lung mask image, and determining a regression loss function of a priori knowledge regression network according to the lung predicted image, the shape prediction knowledge and the shape priori knowledge;
and determining a loss function of the original segmentation model according to the segmentation loss function and the regression loss function, reversely inputting the loss function to the original segmentation model, adjusting network parameters of the original segmentation model, and generating the lung segmentation model.
Optionally, determining a regression loss function of the priori knowledge regression network according to the lung predicted image, the shape prediction knowledge and the shape priori knowledge may include:
determining a first regression loss function according to the shape prior knowledge and the shape prediction knowledge;
calculating shape calculation knowledge according to the lung predicted image, and determining a second regression loss function according to the shape priori knowledge and the shape calculation knowledge;
and determining the regression loss function of the prior knowledge regression network according to the first regression loss function and the second regression loss function.
Optionally, the image segmentation network and the prior knowledge regression network are respectively connected to a bottleneck layer of the feature extraction network, and/or the feature extraction network includes a residual connection layer.
Optionally, the shape prior knowledge includes at least one of the number, the area, the boundary perimeter, and the boundary length and width of the connected domain in the lung mask image; and/or the presence of a gas in the gas,
the priori knowledge regression network comprises an adaptive average pooling layer and a continuous full-connection layer; and/or the presence of a gas in the gas,
the sample image is acquired based on X-ray imaging techniques.
In a second aspect, an embodiment of the present invention further provides a medical image segmentation method, which may include:
acquiring a detected image of a detected chest and a trained lung segmentation model generated according to any one of the model generation methods;
and inputting the detected image into a lung segmentation model, and extracting a segmentation image of a lung region from the detected image according to an output result of the lung segmentation model.
In a third aspect, an embodiment of the present invention further provides a model generating apparatus, where the apparatus may include:
the first acquisition module is used for acquiring a sample image of a sample chest, a lung mask image of a known lung in the sample image and shape prior knowledge of the lung mask image;
and the model generation module is used for taking the sample image, the lung mask image and the shape priori knowledge as a group of training samples, training the original segmentation model based on a plurality of groups of training samples and generating the lung segmentation model, wherein the original segmentation model comprises a feature extraction network, and an image segmentation network and a priori knowledge regression network which are respectively connected with the feature extraction network.
In a fourth aspect, an embodiment of the present invention further provides a medical image segmentation apparatus, which may include:
a second obtaining module, configured to obtain a detected image of a detected chest and a trained lung segmentation model generated according to the model generation method provided in any embodiment of the present invention;
and the image segmentation module is used for inputting the detected image into the lung segmentation model and extracting a segmentation image of the lung region from the detected image according to an output result of the lung segmentation model.
In a fifth aspect, an embodiment of the present invention further provides an apparatus, where the apparatus may include:
one or more processors;
a memory for storing one or more programs;
when the one or more programs are executed by the one or more processors, the one or more processors implement the model generation method or the medical image segmentation method provided by any of the embodiments of the present invention.
In a sixth aspect, the embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the model generation method or the medical image segmentation method provided in any embodiment of the present invention.
According to the technical scheme of the embodiment of the invention, a group of training samples is formed by the acquired sample image of the chest of the sample, the lung mask image of the known lung in the sample image and the shape priori knowledge of the lung mask image, the original segmentation model is trained on the basis of a plurality of groups of training samples, the original segmentation model fully considers the shape priori knowledge of the known lung, and the original segmentation model is a multi-task learning framework capable of simultaneously learning an image segmentation task and a global image regression task, so that the generated trained lung segmentation model has strong generalization performance. According to the technical scheme, the problem that the prior segmentation model does not effectively utilize the shape priori knowledge of the lung region is solved, the prior knowledge regression network set based on the shape priori knowledge is matched with the prior image segmentation network, and the matching degree of the lung segmentation model and the lung region segmentation is improved.
Drawings
FIG. 1 is a flow chart of a model generation method according to a first embodiment of the present invention;
FIG. 2 is a diagram illustrating segmentation of lung regions in a model generation method according to a first embodiment of the present invention;
FIG. 3 is a schematic diagram of residual join in a model generation method according to a first embodiment of the present invention;
FIG. 4a is a schematic diagram of an existing segmentation model in a model generation method according to a first embodiment of the present invention;
FIG. 4b is a schematic diagram of an original segmentation model in a model generation method according to a first embodiment of the present invention;
fig. 4c is a schematic diagram of a local enlargement of a bottleneck layer and a priori knowledge regression network in an original segmentation model in a model generation method according to a first embodiment of the present invention;
FIG. 5 is a flowchart of a medical image segmentation method according to a second embodiment of the present invention;
fig. 6 is a block diagram of a model generation apparatus according to a third embodiment of the present invention;
FIG. 7 is a block diagram of a medical image segmentation apparatus according to a fourth embodiment of the present invention;
fig. 8 is a schematic structural diagram of an apparatus in the fifth embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.
Example one
Fig. 1 is a flowchart of a model generation method according to a first embodiment of the present invention. The present embodiment is applicable to the case of generating a lung segmentation model matching with lung region segmentation, and is particularly applicable to the case of generating a lung segmentation model matching with lung region segmentation based on a multitask learning framework. The method can be executed by the model generation device provided by the embodiment of the invention, the device can be realized by software and/or hardware, and the device can be integrated on various user terminals or servers.
Referring to fig. 1, the method of the embodiment of the present invention specifically includes the following steps:
s110, acquiring a sample image of the chest of the sample, a lung mask image of the known lung in the sample image and shape prior knowledge of the lung mask image.
The sample image may be obtained based on an X-ray imaging technology, which may be a Digital Radiography (DR) technology, a Computed Tomography (CR) technology, a Computed Tomography (CT) technology, or the like, and is not specifically limited herein, and the obtained sample image may be referred to as an X-ray chest film.
The lung mask image is usually a binary image of a known lung labeled manually, which can be used as a desired output in the model training process, and illustratively, as shown in fig. 2, the left side of fig. 2 is a schematic diagram of a sample image, and the right side of fig. 2 is a schematic diagram of a lung mask image corresponding to the left side of fig. 2.
The shape prior knowledge may be at least one of the number, the area, the boundary perimeter, and the boundary length and width of the connected regions in the lung mask image, and of course, it may also be the other shape attributes of the lung mask image, which are not specifically limited herein. For example, if the shape prior knowledge is the number of connected components, the shape prior knowledge is 2, because the lung is generally divided into a left lung and a right lung, and the mask image of the lung includes two connected components, i.e., the number of connected components is 2, which can also be understood as the number of segmentation targets a priori. The connected domain means that there are 8 adjacent pixels around each pixel in the sample image, and common adjacent relations include two types, namely 4 adjacent and 8 adjacent. If the pixel point A is adjacent to the pixel point B, the pixel point A is communicated with the pixel point B; on the basis, if the pixel point A is communicated with the pixel point B, and the pixel point B is communicated with the pixel point C, the pixel point A is communicated with the pixel point C. Visually, the pixels which are communicated with each other form the same area, and the pixels which are not communicated form different areas. Thus, a set of interconnected pixels is referred to as a connected domain.
S120, taking the sample image, the lung mask image and the shape priori knowledge as a group of training samples, training an original segmentation model based on a plurality of groups of training samples, and generating the lung segmentation model, wherein the original segmentation model comprises a feature extraction network, an image segmentation network and a priori knowledge regression network which are respectively connected with the feature extraction network.
The existing segmentation model existing in the prior art generally includes only a feature extraction network and an image segmentation network, the feature extraction network can extract a feature map from an inputted medical image, and the image segmentation network can segment the medical image according to the extracted feature map. That is, for different segmentation tasks, the existing segmentation models all adopt the same network structure, and the unique characteristics of each segmentation task are not considered, so that the segmentation accuracy of each segmentation task needs to be improved.
In order to improve the segmentation accuracy of the lung region, the embodiment of the invention improves the existing segmentation model, and sets a priori knowledge regression network which is connected with the existing feature extraction network and fully considers the unique characteristics of the lung region, wherein the priori knowledge regression network can be used for learning the shape attribute of a global image, thereby forming the original segmentation model based on multi-task learning.
It should be noted that, firstly, the existing segmentation models may be those convolution neural network models that can be used to implement image segmentation, such as U-Net, Res-U-Net, SegNet, Mask R-CNN, FCN, etc., which are not specifically limited herein. The a priori knowledge regression network may be any network structure that converts a plurality of feature maps into at least one numerical value, such as 1 × 1 convolution layer, an adaptive average pooling layer plus continuous full-link layers, and the like, which are not specifically limited herein. Thirdly, the difference of the type of the shape priori knowledge mainly influences the network parameters of the priori knowledge regression network, and the difference usually does not influence the network structure of the priori knowledge regression network, for example, if the shape priori knowledge is the number of the connected domains, the expected output of the priori knowledge regression network is 1 numerical value; if the shape priori knowledge is the number of the connected domains and the area of the connected domains, the expected output of the priori knowledge regression network is 2 numerical values, and the difference of the expected output in the number and the numerical values can affect the network parameters.
In summary, the feature extraction network and the image segmentation network are matched with each other and can be used for executing an image segmentation task; the feature extraction network and the prior knowledge regression network are matched with each other and can be used for executing the global image regression task, namely, the feature extraction network is a network structure shared by the image segmentation task and the global image regression task. Therefore, in the model training process, the original segmentation model can simultaneously learn the image segmentation task and the global image regression task, and the multi-task learning scheme improves the generalization performance of the trained lung segmentation model and further improves the segmentation precision of the lung region.
According to the technical scheme of the embodiment of the invention, a group of training samples is formed by the acquired sample image of the chest of the sample, the lung mask image of the known lung in the sample image and the shape priori knowledge of the lung mask image, the original segmentation model is trained on the basis of a plurality of groups of training samples, the original segmentation model fully considers the shape priori knowledge of the known lung, and the original segmentation model is a multi-task learning framework capable of simultaneously learning an image segmentation task and a global image regression task, so that the generated trained lung segmentation model has strong generalization performance. According to the technical scheme, the problem that the prior segmentation model does not effectively utilize the shape priori knowledge of the lung region is solved, the prior knowledge regression network set based on the shape priori knowledge is matched with the prior image segmentation network, and the matching degree of the lung segmentation model and the lung region segmentation is improved.
On this basis, optionally, the image segmentation network and the prior knowledge regression network may be connected behind the same convolutional layer of the feature extraction network, which is mostly the last convolutional layer of the feature extraction network. At this time, after the feature extraction network extracts a plurality of feature maps from the inputted medical image, the plurality of feature maps can be simultaneously inputted to the image segmentation network and the prior knowledge regression network, thereby realizing the parallel execution of the image segmentation task and the global image regression task. Of course, the image segmentation network and the prior knowledge regression network may also be connected behind different convolution layers of the feature extraction network, and are not specifically limited herein.
Alternatively, the same convolutional layer may be a bottleneck layer (bottleneck layer), which may be considered as a convolutional layer outputting a feature map with a minimum resolution in the original segmentation model. It should be noted that the bottleneck layer may or may not be the last convolutional layer of the original segmentation model. For example, if the original segmentation model is obtained by setting a priori knowledge regression network on the existing U-Net network model, since the U-Net network model includes an encoder and a decoder, the bottleneck layer is usually the last convolutional layer of the encoder, but not the last convolutional layer of the original segmentation model; for another example, if the original segmentation model does not include the codec process, the bottleneck layer may be the last convolution layer of the original segmentation model, and is not specifically limited herein.
Optionally, the feature extraction network may include a Residual connection (Res) layer, where the reason for setting is that the convolutional neural network model may cause a problem of gradient disappearance or gradient explosion as the number of convolutional layer layers increases, and to solve this problem, a Residual connection layer may be set in the feature extraction network, and may connect an output feature map of a subsequent convolutional layer with an input feature map of a previous convolutional layer, thereby forming Residual connection. Illustratively, as shown in fig. 3, the output characteristic diagram f (X) of the subsequent volume of lamination layer is connected with the input characteristic diagram X of the previous volume of lamination layer to obtain f (X) + X.
An optional technical solution, training an original segmentation model based on a plurality of groups of training samples to generate a lung segmentation model, may specifically include: inputting a sample image into an original segmentation model, and obtaining a lung predicted image and shape prediction knowledge according to an output result of the original segmentation model, wherein the lung predicted image is obtained according to an output result of an image segmentation network, and the shape prediction knowledge is obtained according to an output result of a priori knowledge regression network; determining a segmentation loss function of an image segmentation network according to the lung predicted image and the lung mask image, and determining a regression loss function of a priori knowledge regression network according to the lung predicted image, the shape prediction knowledge and the shape priori knowledge; and determining a loss function of the original segmentation model according to the segmentation loss function and the regression loss function, reversely inputting the loss function to the original segmentation model, adjusting network parameters of the original segmentation model, and generating the lung segmentation model.
On this basis, optionally, determining a regression loss function of the priori knowledge regression network according to the lung predicted image, the shape prediction knowledge and the shape priori knowledge may specifically include: determining a first regression loss function according to the shape prior knowledge and the shape prediction knowledge; calculating shape calculation knowledge according to the lung predicted image, if the shape priori knowledge is the number of the connected domains, taking the number of the connected domains in the lung predicted image as the shape calculation knowledge, and further determining a second regression loss function according to the shape priori knowledge and the shape calculation knowledge; and determining the regression loss function of the prior knowledge regression network according to the first regression loss function and the second regression loss function.
In order to better understand the specific implementation process of the above steps, the model generation method of the present embodiment is exemplarily described below with reference to specific examples. As shown in fig. 4a-4c, for example, a U-Net network structure widely used in the field of medical image segmentation is taken as an existing segmentation model, and the existing segmentation model is composed of an encoder on the left side and a decoder on the right side. Specifically, in the encoder, feature maps of different resolutions may be extracted based on stacked 3 × 3 convolutional layers (Conv 3 × 3) and down-sampling (Max-firing) operations; in the decoder, the dimensions of the feature map may be restored based on the stacked 3 × 3 convolution layers and the Up-sampling (Up-Conv) operation, and the image segmentation result may be output based on the 1 × 1 convolution layer (Conv 1 × 1); meanwhile, the U-Net network structure further includes a skip-connection (i.e., Copy and Concat) layer, which performs feature fusion on the feature maps of the corresponding resolutions in the encoder and decoder.
On this basis, as shown in fig. 4b and 4c, the original segmentation model according to the embodiment of the present invention includes, with respect to the existing U-Net network structure, a feature extraction network (equivalent to an encoder), an image segmentation network (equivalent to a decoder) connected to a bottleneck layer of the feature extraction network, and an a priori knowledge regression network, and the feature extraction network includes a residual and Add (Copy and Add) layer.
Specifically, in the model training process, if the input image of the original segmentation model is an X-ray chest image I, the X-ray chest image I is usually a single-channel grayscale image, and I ∈ θ ═ Rw*hWhere w and h represent the width and height, respectively, of an X-ray chest film I, the output of the image segmentation task is a lung prediction image S, which is a binary image, S ∈ ═ 0,1}w*hAnd the output of the global image regression task is shape prediction knowledge C, which is a non-negative integer,thus, the image segmentation task may be defined as t1The global image regression task may be defined as θ →These two tasks are trained in the original segmentation model at the same time and share the encoder. When the feature map extracted by the encoder reaches the bottleneck layer, the feature map can be respectively input into the existing decoder of the U-Net network structure and the prior knowledge regression network arranged in the embodiment of the invention, wherein the decoder can be used for realizing pixel-by-pixel segmentation, and the prior knowledge regression network can be composed of a self-adaptive average pooling and continuous full connection layer and can be used for realizing global image regression.
Based on training samplesWhen the original segmentation model is trained, optionally, in an image segmentation task, a segmentation loss function can be calculated based on cross entropy or DICE coefficients. Wherein the definition of cross entropy isn is the total number of training samples, y is the true value of the artificial label in the lung mask image,predicting a predicted value in the image for the lung; the definition of the Dice coefficient isWhere | E ∩ Y | represents the intersection between the lung predicted image E and the lung mask image Y, | E | and | Y | represent the number of elements of E and Y, respectivelyn is the total number of training samples, y is the true value of the artificial label in the lung mask image,predict the predicted values in the image for the lungs. In particular, to improve the model training accuracy, the shape prior knowledge R can be usedtrueAnd shape prediction knowledge RpredCalculating a first regression loss function MSE (R) based on the mean square errortrue,Rpred) And calculating shape calculation knowledge R according to the predicted lung imagecalcFrom prior knowledge of the shape RtrueAnd shape computation knowledge RcalcCalculating a second regression loss function MSE (R) from the mean square errortrue,Rcalc) Further, the first regression loss function MSE (R)true,Rpred) And a second regression loss function MSE (R)true,Rcalc) After the addition, a regression loss function L is obtainedregThus, the loss function L of the original segmentation modeltotalCan be represented as Ltotal=αLseg+(1-α)LregWherein, LsegIs the segmentation loss function, α is in the range 0,1]And (4) the weight of each.
Example two
Fig. 5 is a flowchart of a medical image segmentation method provided in the second embodiment of the present invention. The embodiment can be applied to the condition of segmenting the lung region in the X-ray chest radiograph, in particular to the condition of segmenting the lung region in the X-ray chest radiograph based on the lung segmentation model, and the lung segmentation model effectively utilizes the shape prior knowledge of the lung region in the training process. The method can be executed by a medical image segmentation device provided by the embodiment of the invention, the device can be realized by software and/or hardware, and the device can be integrated on various user terminals or servers.
Referring to fig. 5, the method of the embodiment of the present invention specifically includes the following steps:
s210, acquiring an examined image of an examined chest, and generating a trained lung segmentation model according to the model generation method described in the embodiment I.
The detected image is obtained based on an X-ray imaging technology, the lung segmentation model comprises a feature extraction network, an image segmentation network and a priori knowledge regression network, wherein the image segmentation network and the priori knowledge regression network are respectively connected with the feature extraction network and are obtained by training a training sample formed based on a sample image, a lung mask image of the sample image and shape priori knowledge of the lung mask image.
S220, inputting the detected image into the lung segmentation model, and extracting a segmentation image of the lung region from the detected image according to an output result of the lung segmentation model.
According to the technical scheme of the embodiment of the invention, the priori knowledge regression network in the lung segmentation model is set based on the shape priori knowledge of the lung region, so that the generated lung segmentation model is a multi-task learning framework capable of simultaneously learning an image segmentation task and a global image regression task, and the generalization performance of the trained lung segmentation model is improved. By the technical scheme, the problem that the segmentation precision of the lung region is low due to the fact that the prior knowledge of the shape of the lung region is not effectively utilized in the existing segmentation model is solved, and the effect of accurate segmentation of the lung region is achieved.
EXAMPLE III
Fig. 6 is a block diagram of a model generation apparatus according to a third embodiment of the present invention, which is configured to execute the model generation method according to any of the embodiments. The device and the model generating method of each embodiment belong to the same inventive concept, and details which are not described in detail in the embodiment of the model generating device can refer to the embodiment of the model generating method. Referring to fig. 6, the apparatus may specifically include: a first acquisition module 310 and a model generation module 320.
The first obtaining module 310 is configured to obtain a sample image of a chest of the sample, a lung mask image of a known lung in the sample image, and shape prior knowledge of the lung mask image;
the model generating module 320 is configured to use the sample image, the lung mask image, and the shape prior knowledge as a set of training samples, train an original segmentation model based on a plurality of sets of training samples, and generate a lung segmentation model, where the original segmentation model includes a feature extraction network, and an image segmentation network and a prior knowledge regression network that are respectively connected to the feature extraction network.
Optionally, the model generating module 320 may specifically include:
the actual output obtaining unit is used for inputting the sample image into the original segmentation model and obtaining a lung predicted image and shape prediction knowledge according to the output result of the original segmentation model;
the loss function determining unit is used for determining a segmentation loss function of the image segmentation network according to the lung predicted image and the lung mask image and determining a regression loss function of the priori knowledge regression network according to the lung predicted image, the shape prediction knowledge and the shape priori knowledge;
and the model generation unit is used for determining a loss function of the original segmentation model according to the segmentation loss function and the regression loss function, reversely inputting the loss function into the original segmentation model, adjusting network parameters of the original segmentation model and generating the lung segmentation model.
Optionally, the loss function determining unit may be specifically configured to:
determining a first regression loss function according to the shape prior knowledge and the shape prediction knowledge;
calculating shape calculation knowledge according to the lung predicted image, and determining a second regression loss function according to the shape priori knowledge and the shape calculation knowledge;
and determining the regression loss function of the prior knowledge regression network according to the first regression loss function and the second regression loss function.
Optionally, the image segmentation network and the prior knowledge regression network are respectively connected to a bottleneck layer of the feature extraction network, and/or the feature extraction network includes a residual connection layer.
Optionally, the shape prior knowledge includes at least one of the number, the area, the boundary perimeter, and the boundary length and width of the connected domain in the lung mask image; and/or the presence of a gas in the gas,
the priori knowledge regression network comprises an adaptive average pooling layer and a continuous full-connection layer; and/or the presence of a gas in the gas,
the sample image is acquired based on X-ray imaging techniques.
The model generation device provided by the third embodiment of the invention uses the acquired sample image of the chest of the sample, the known lung mask image in the sample image and the shape priori knowledge of the lung mask image as a group of training samples through the mutual cooperation of the first acquisition module and the model generation module, and trains the original segmentation model based on a plurality of groups of training samples, wherein the original segmentation model fully considers the known lung shape priori knowledge and is a multi-task learning framework capable of simultaneously learning an image segmentation task and a global image regression task, so that the generated trained lung segmentation model has strong generalization performance. The device solves the problem that the prior segmentation model does not effectively utilize the shape prior knowledge of the lung region, and the prior knowledge regression network set based on the shape prior knowledge is matched with the prior image segmentation network, so that the matching degree of the lung segmentation model and the lung region segmentation is improved.
The model generation device provided by the embodiment of the invention can execute the model generation method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method.
It should be noted that, in the embodiment of the model generating apparatus, the included units and modules are merely divided according to functional logic, but are not limited to the above division as long as the corresponding functions can be implemented; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.
Example four
Fig. 7 is a block diagram of a medical image segmentation apparatus according to a fourth embodiment of the present invention, which is configured to execute the medical image segmentation method according to any of the embodiments described above. The device and the medical image segmentation method of each embodiment belong to the same inventive concept, and details which are not described in detail in the embodiment of the medical image segmentation device can refer to the embodiment of the medical image segmentation method. Referring to fig. 7, the apparatus may specifically include: a second acquisition module 410 and an image segmentation module 420.
A second obtaining module 410, configured to obtain a detected image of a detected chest, and a trained lung segmentation model generated according to the model generation method according to any embodiment of the present invention;
and the image segmentation module 420 is configured to input the detected image into the lung segmentation model, and extract a segmented image of the lung region from the detected image according to an output result of the lung segmentation model.
In the medical image segmentation device provided by the fourth embodiment of the present invention, the second acquisition module and the image segmentation module are mutually matched, and the priori knowledge regression network in the lung segmentation model is set based on the shape priori knowledge of the lung region, so that the generated lung segmentation model is a multi-task learning framework capable of simultaneously learning an image segmentation task and a global image regression task, which improves the generalization performance of the trained lung segmentation model, and thus, after the acquired image to be examined is input to the lung segmentation model, an accurate segmentation image of the lung region can be extracted from the image to be examined. The device solves the problem that the segmentation precision of the lung region is low due to the fact that the prior shape knowledge of the lung region is not effectively utilized in the existing segmentation model, and achieves the effect of accurate segmentation of the lung region.
The medical image segmentation device provided by the embodiment of the invention can execute the medical image segmentation method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method.
It should be noted that, in the embodiment of the medical image segmentation apparatus, the included units and modules are only divided according to the functional logic, but are not limited to the above division as long as the corresponding functions can be realized; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.
EXAMPLE five
Fig. 8 is a schematic structural diagram of an apparatus according to a fifth embodiment of the present invention, as shown in fig. 8, the apparatus includes a memory 510, a processor 520, an input device 530, and an output device 540. The number of processors 520 in the device may be one or more, and one processor 520 is taken as an example in fig. 8; the memory 510, processor 520, input device 530, and output device 540 in the apparatus may be connected by a bus or other means, such as by bus 550 in fig. 8.
The memory 510 is used as a computer readable storage medium for storing software programs, computer executable programs, and modules, such as program instructions/modules corresponding to the model generation method in the embodiment of the present invention (for example, the first obtaining module 310 and the model generation module 320 in the model generation apparatus), or program instructions/modules corresponding to the medical image segmentation method in the embodiment of the present invention (for example, the second obtaining module 410 and the image segmentation module 420 in the medical image segmentation apparatus). The processor 520 executes various functional applications of the apparatus and data processing, i.e., the above-described model generation method or medical image segmentation method, by executing software programs, instructions, and modules stored in the memory 510.
The memory 510 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to use of the device, and the like. Further, the memory 510 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some examples, memory 510 may further include memory located remotely from processor 520, which may be connected to devices through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input device 530 may be used to receive input numeric or character information and generate key signal inputs related to user settings and function control of the device. The output device 540 may include a display device such as a display screen.
EXAMPLE six
An embodiment of the present invention provides a storage medium containing computer-executable instructions, which when executed by a computer processor, perform a method for model generation, the method comprising:
acquiring a sample image of a sample chest, a lung mask image of a known lung in the sample image and shape prior knowledge of the lung mask image;
taking the sample image, the lung mask image and the shape priori knowledge as a group of training samples, training an original segmentation model based on a plurality of groups of training samples, and generating a lung segmentation model;
the original segmentation model comprises a feature extraction network, an image segmentation network and a priori knowledge regression network, wherein the image segmentation network and the priori knowledge regression network are respectively connected with the feature extraction network.
Of course, the storage medium provided by the embodiment of the present invention contains computer-executable instructions, and the computer-executable instructions are not limited to the method operations described above, and may also perform related operations in the model generation method provided by any embodiment of the present invention.
EXAMPLE seven
A seventh embodiment of the present invention provides a storage medium containing computer-executable instructions, which when executed by a computer processor, are configured to perform a medical image segmentation method, including:
acquiring a detected image of a detected chest and a trained lung segmentation model generated according to the model generation method of any embodiment;
and inputting the detected image into a lung segmentation model, and extracting a segmentation image of a lung region from the detected image according to an output result of the lung segmentation model.
From the above description of the embodiments, it will be apparent to those skilled in the art that the present invention may be implemented by software and necessary general-purpose hardware, and certainly may be implemented by hardware, but in many cases, the foregoing is a better embodiment of the present invention, and according to this understanding, the technical solution of the present invention or portions contributing to the prior art may be embodied in the form of a software product, which may be stored in a computer-readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a flash Memory (F L ASH), a hard disk or an optical disk, etc., and includes instructions for enabling a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the embodiments of the present invention.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (10)
1. A method of model generation, comprising:
acquiring a sample image of a sample chest, a lung mask image of a known lung in the sample image, and shape prior knowledge of the lung mask image;
taking the sample image, the lung mask image and the shape priori knowledge as a group of training samples, and training an original segmentation model based on a plurality of groups of training samples to generate a lung segmentation model;
the original segmentation model comprises a feature extraction network, an image segmentation network and a priori knowledge regression network, wherein the image segmentation network and the priori knowledge regression network are respectively connected with the feature extraction network.
2. The method of claim 1, wherein training an original segmentation model based on the plurality of sets of training samples to generate a lung segmentation model comprises:
inputting the sample image into an original segmentation model, and obtaining a lung predicted image and shape prediction knowledge according to an output result of the original segmentation model;
determining a segmentation loss function of the image segmentation network according to the lung predicted image and the lung mask image, and determining a regression loss function of the priori knowledge regression network according to the lung predicted image, the shape prediction knowledge and the shape priori knowledge;
and determining a loss function of the original segmentation model according to the segmentation loss function and the regression loss function, reversely inputting the loss function to the original segmentation model, and adjusting network parameters of the original segmentation model to generate a lung segmentation model.
3. The method of claim 2, wherein said determining a regression loss function of said a priori knowledge regression network based on said lung predicted image, said shape prediction knowledge and said shape a priori knowledge comprises:
determining a first regression loss function according to the shape prior knowledge and the shape prediction knowledge;
calculating shape calculation knowledge according to the lung prediction image, and determining a second regression loss function according to the shape priori knowledge and the shape calculation knowledge;
and determining the regression loss function of the priori knowledge regression network according to the first regression loss function and the second regression loss function.
4. The method according to claim 1, wherein the image segmentation network and the a priori knowledge regression network are connected behind a bottleneck layer of the feature extraction network, respectively, and/or wherein the feature extraction network comprises a residual connection layer.
5. The method of claim 1, wherein the shape prior knowledge includes at least one of a number, an area, a boundary perimeter, and a boundary length and width of connected components in the lung mask image;
and/or the prior knowledge regression network comprises an adaptive average pooling layer and a continuous full-connected layer;
and/or the sample image is acquired based on X-ray imaging techniques.
6. A medical image segmentation method, comprising:
acquiring an examined image of an examined chest and a trained lung segmentation model generated according to the method of any one of claims 1-5;
and inputting the detected image into the lung segmentation model, and extracting a segmentation image of a lung region from the detected image according to an output result of the lung segmentation model.
7. A model generation apparatus, comprising:
a first obtaining module, configured to obtain a sample image of a chest of a sample, a lung mask image of a known lung in the sample image, and shape prior knowledge of the lung mask image;
and the model generation module is used for taking the sample image, the lung mask image and the shape priori knowledge as a group of training samples, training an original segmentation model based on a plurality of groups of training samples, and generating a lung segmentation model, wherein the original segmentation model comprises a feature extraction network, and an image segmentation network and a priori knowledge regression network which are respectively connected with the feature extraction network.
8. A medical image segmentation apparatus, characterized by comprising:
a second acquisition module for acquiring an examined image of an examined chest and a trained lung segmentation model generated according to the method of any one of claims 1-5;
and the image segmentation module is used for inputting the detected image into the lung segmentation model and extracting a segmentation image of a lung region from the detected image according to an output result of the lung segmentation model.
9. An apparatus, characterized in that the apparatus comprises:
one or more processors;
a memory for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement a model generation method as claimed in any one of claims 1-5, or a medical image segmentation method as claimed in claim 6.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the model generation method as set forth in any one of claims 1 to 5 or the medical image segmentation method as set forth in claim 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010197600.7A CN111429421B (en) | 2020-03-19 | 2020-03-19 | Model generation method, medical image segmentation method, device, equipment and medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010197600.7A CN111429421B (en) | 2020-03-19 | 2020-03-19 | Model generation method, medical image segmentation method, device, equipment and medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111429421A true CN111429421A (en) | 2020-07-17 |
CN111429421B CN111429421B (en) | 2021-08-27 |
Family
ID=71548096
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010197600.7A Active CN111429421B (en) | 2020-03-19 | 2020-03-19 | Model generation method, medical image segmentation method, device, equipment and medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111429421B (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111950637A (en) * | 2020-08-14 | 2020-11-17 | 厦门美图之家科技有限公司 | Purple matter detection method, purple matter detection device, skin detector and readable storage medium |
CN111968134A (en) * | 2020-08-11 | 2020-11-20 | 影石创新科技股份有限公司 | Object segmentation method and device, computer readable storage medium and computer equipment |
CN112001856A (en) * | 2020-07-29 | 2020-11-27 | 东软医疗系统股份有限公司 | Training method of denoising model, image noise removing method and related device |
CN112613517A (en) * | 2020-12-17 | 2021-04-06 | 深圳大学 | Endoscopic instrument segmentation method, endoscopic instrument segmentation apparatus, computer device, and storage medium |
CN112669273A (en) * | 2020-12-22 | 2021-04-16 | 吉林大学 | Method and device for automatically segmenting drusen in fundus image and readable storage medium |
CN112750124A (en) * | 2021-01-22 | 2021-05-04 | 推想医疗科技股份有限公司 | Model generation method, image segmentation method, model generation device, image segmentation device, electronic equipment and storage medium |
CN112802040A (en) * | 2021-01-28 | 2021-05-14 | 上海藤核智能科技有限公司 | X-ray pneumothorax segmentation and evaluation method based on edge perception |
CN113449781A (en) * | 2021-06-17 | 2021-09-28 | 上海深至信息科技有限公司 | Generation method and system of thyroid nodule classification model |
CN113538530A (en) * | 2021-07-09 | 2021-10-22 | 深圳市深光粟科技有限公司 | Ear medical image segmentation method and device, electronic equipment and storage medium |
CN113610785A (en) * | 2021-07-26 | 2021-11-05 | 安徽理工大学 | Pneumoconiosis early warning method and device based on intelligent image and storage medium |
WO2023108968A1 (en) * | 2021-12-14 | 2023-06-22 | 北京邮电大学 | Image classification method and system based on knowledge-driven deep learning |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107248162A (en) * | 2017-05-18 | 2017-10-13 | 杭州全景医学影像诊断有限公司 | The method of preparation method and acute cerebral ischemia the image segmentation of acute cerebral ischemia Image Segmentation Model |
CN109064443A (en) * | 2018-06-22 | 2018-12-21 | 哈尔滨工业大学 | A kind of multi-model organ segmentation method and system based on abdominal ultrasound images |
US20190343477A1 (en) * | 2016-06-30 | 2019-11-14 | Shanghai United Imaging Healthcare Co., Ltd. | Methods and systems for extracting blood vessel |
CN110728675A (en) * | 2019-10-22 | 2020-01-24 | 慧影医疗科技(北京)有限公司 | Pulmonary nodule analysis device, model training method, device and analysis equipment |
CN110751187A (en) * | 2019-09-26 | 2020-02-04 | 上海联影智能医疗科技有限公司 | Training method of abnormal area image generation network and related product |
-
2020
- 2020-03-19 CN CN202010197600.7A patent/CN111429421B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190343477A1 (en) * | 2016-06-30 | 2019-11-14 | Shanghai United Imaging Healthcare Co., Ltd. | Methods and systems for extracting blood vessel |
CN107248162A (en) * | 2017-05-18 | 2017-10-13 | 杭州全景医学影像诊断有限公司 | The method of preparation method and acute cerebral ischemia the image segmentation of acute cerebral ischemia Image Segmentation Model |
CN109064443A (en) * | 2018-06-22 | 2018-12-21 | 哈尔滨工业大学 | A kind of multi-model organ segmentation method and system based on abdominal ultrasound images |
CN110751187A (en) * | 2019-09-26 | 2020-02-04 | 上海联影智能医疗科技有限公司 | Training method of abnormal area image generation network and related product |
CN110728675A (en) * | 2019-10-22 | 2020-01-24 | 慧影医疗科技(北京)有限公司 | Pulmonary nodule analysis device, model training method, device and analysis equipment |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112001856A (en) * | 2020-07-29 | 2020-11-27 | 东软医疗系统股份有限公司 | Training method of denoising model, image noise removing method and related device |
CN111968134A (en) * | 2020-08-11 | 2020-11-20 | 影石创新科技股份有限公司 | Object segmentation method and device, computer readable storage medium and computer equipment |
CN111968134B (en) * | 2020-08-11 | 2023-11-28 | 影石创新科技股份有限公司 | Target segmentation method, device, computer readable storage medium and computer equipment |
CN111950637B (en) * | 2020-08-14 | 2024-05-03 | 厦门美图宜肤科技有限公司 | Ultraviolet detection method, device, skin detector and readable storage medium |
CN111950637A (en) * | 2020-08-14 | 2020-11-17 | 厦门美图之家科技有限公司 | Purple matter detection method, purple matter detection device, skin detector and readable storage medium |
CN112613517A (en) * | 2020-12-17 | 2021-04-06 | 深圳大学 | Endoscopic instrument segmentation method, endoscopic instrument segmentation apparatus, computer device, and storage medium |
CN112613517B (en) * | 2020-12-17 | 2022-02-18 | 深圳大学 | Endoscopic instrument segmentation method, endoscopic instrument segmentation apparatus, computer device, and storage medium |
CN112669273A (en) * | 2020-12-22 | 2021-04-16 | 吉林大学 | Method and device for automatically segmenting drusen in fundus image and readable storage medium |
CN112750124B (en) * | 2021-01-22 | 2021-11-09 | 推想医疗科技股份有限公司 | Model generation method, image segmentation method, model generation device, image segmentation device, electronic equipment and storage medium |
CN112750124A (en) * | 2021-01-22 | 2021-05-04 | 推想医疗科技股份有限公司 | Model generation method, image segmentation method, model generation device, image segmentation device, electronic equipment and storage medium |
CN112802040A (en) * | 2021-01-28 | 2021-05-14 | 上海藤核智能科技有限公司 | X-ray pneumothorax segmentation and evaluation method based on edge perception |
CN112802040B (en) * | 2021-01-28 | 2024-05-31 | 上海藤核智能科技有限公司 | X-ray pneumothorax segmentation and assessment method based on edge perception |
CN113449781A (en) * | 2021-06-17 | 2021-09-28 | 上海深至信息科技有限公司 | Generation method and system of thyroid nodule classification model |
CN113538530A (en) * | 2021-07-09 | 2021-10-22 | 深圳市深光粟科技有限公司 | Ear medical image segmentation method and device, electronic equipment and storage medium |
CN113538530B (en) * | 2021-07-09 | 2024-03-01 | 深圳市深光粟科技有限公司 | Ear medical image segmentation method and device, electronic equipment and storage medium |
CN113610785A (en) * | 2021-07-26 | 2021-11-05 | 安徽理工大学 | Pneumoconiosis early warning method and device based on intelligent image and storage medium |
WO2023108968A1 (en) * | 2021-12-14 | 2023-06-22 | 北京邮电大学 | Image classification method and system based on knowledge-driven deep learning |
Also Published As
Publication number | Publication date |
---|---|
CN111429421B (en) | 2021-08-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111429421B (en) | Model generation method, medical image segmentation method, device, equipment and medium | |
US11861829B2 (en) | Deep learning based medical image detection method and related device | |
US20210365717A1 (en) | Method and apparatus for segmenting a medical image, and storage medium | |
CN110599528B (en) | Unsupervised three-dimensional medical image registration method and system based on neural network | |
CN108921851B (en) | Medical CT image segmentation method based on 3D countermeasure network | |
EP3611699A1 (en) | Image segmentation using deep learning techniques | |
US11810301B2 (en) | System and method for image segmentation using a joint deep learning model | |
WO2021128825A1 (en) | Three-dimensional target detection method, method and device for training three-dimensional target detection model, apparatus, and storage medium | |
US10929643B2 (en) | 3D image detection method and apparatus, electronic device, and computer readable medium | |
WO2022032824A1 (en) | Image segmentation method and apparatus, device, and storage medium | |
US10878564B2 (en) | Systems and methods for processing 3D anatomical volumes based on localization of 2D slices thereof | |
WO2023044605A1 (en) | Three-dimensional reconstruction method and apparatus for brain structure in extreme environments, and readable storage medium | |
Shu et al. | LVC-Net: Medical image segmentation with noisy label based on local visual cues | |
CN110570394A (en) | medical image segmentation method, device, equipment and storage medium | |
WO2023207743A1 (en) | Image detection method and apparatus, and computer device, storage medium and program product | |
CN114387317A (en) | CT image and MRI three-dimensional image registration method and device | |
US20230386067A1 (en) | Systems and methods for segmenting 3d images | |
CN116245832A (en) | Image processing method, device, equipment and storage medium | |
CN114972211A (en) | Training method, segmentation method, device, equipment and medium of image segmentation model | |
CN113822323A (en) | Brain scanning image identification processing method, device, equipment and storage medium | |
CN111209946B (en) | Three-dimensional image processing method, image processing model training method and medium | |
CN111507950B (en) | Image segmentation method and device, electronic equipment and computer-readable storage medium | |
Patel et al. | PTXNet: An extended UNet model based segmentation of pneumothorax from chest radiography images | |
CN117911432A (en) | Image segmentation method, device and storage medium | |
CN114586065A (en) | Method and system for segmenting images |
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 | ||
CB02 | Change of applicant information | ||
CB02 | Change of applicant information |
Address after: Room B401, floor 4, building 1, No. 12, Shangdi Information Road, Haidian District, Beijing 100085 Applicant after: Tuxiang Medical Technology Co., Ltd Address before: Room B401, floor 4, building 1, No. 12, Shangdi Information Road, Haidian District, Beijing 100085 Applicant before: Beijing Tuoxiang Technology Co.,Ltd. |
|
GR01 | Patent grant | ||
GR01 | Patent grant |