CN114445421B - Identification and segmentation method, device and system for nasopharyngeal carcinoma lymph node region - Google Patents

Abstract

The invention discloses a method, a device and a system for identifying and segmenting nasopharyngeal carcinoma lymph node areas. The apparatus includes a data acquisition unit and an identification segmentation unit. The system includes an identification segmentation module and a data storage module. The method, the device and the system for identifying and segmenting the segmented magnetic resonance image to be identified and segmented through the three-dimensional depth supervision convolutional neural network three-dimensional model from end to end improve the accuracy of identifying and segmenting the nasopharyngeal carcinoma lymph node region; further, the method, the device and the system for identifying and dividing the nasopharyngeal carcinoma lymph node area further process the first training image data set through the preset double-examination data processing method to obtain the second training image data set, so that a reasonable model is designed according to the morphological characteristics of the lymph nodes, and accuracy of identifying and dividing the nasopharyngeal carcinoma lymph node area is improved.

Description

A method, device and system for identifying and segmenting lymph node areas in nasopharyngeal carcinoma

Technical field

The invention relates to the field of identification and segmentation of lymph node areas of nasopharyngeal cancer, and relates to a method, device and system for identification and segmentation of lymph node areas of nasopharyngeal cancer.

Background technique

Among patients with newly diagnosed nasopharyngeal carcinoma, metastatic lymph nodes are detected in approximately 70%-80% of cases at the time of first diagnosis. Accurate spatial modeling of metastatic lymph nodes is important for the success of treatment. In the past decade, artificial intelligence (AI) has developed rapidly and has shown good performance in the identification and automatic segmentation of normal anatomical structures or lesions in medical images. Whether in imaging studies during radiation therapy or in delineation studies of gross tumor volume (GTV), segmentation of regions of interest (ROI) or lesions is labor-intensive and therefore needs to be able to reduce physician workload. and auxiliary tools to improve diagnostic efficiency.

In the existing technology, deep learning is usually applied in directions such as nasopharyngeal cancer lymph node delineation modeling based on dicom images.

However, the existing technology still has the following shortcomings: these methods have cumbersome steps, are not end-to-end structured, and do not fully design reasonable models based on the morphological characteristics of lymph nodes.

Therefore, there is currently a need for a method, device and system for identifying and segmenting lymph node areas of nasopharyngeal carcinoma, so as to overcome the above-mentioned defects in the existing technology.

Contents of the invention

In view of the above-mentioned existing technical problems, the purpose of the present invention is to provide a method, device and system for identifying and segmenting the lymph node area of nasopharyngeal carcinoma, thereby improving the accuracy of identifying and segmenting the lymph node area of nasopharyngeal carcinoma.

The present invention provides a method for identifying and segmenting lymph node areas of nasopharyngeal carcinoma. The identification and segmentation method includes: obtaining a magnetic resonance image to be identified and segmented; and identifying and segmenting the magnetic resonance image through a preset lymphatic recognition and segmentation model. , thereby obtaining the segmented region image; the lymphatic recognition segmentation model is an end-to-end three-dimensional deep supervised convolutional neural network three-dimensional model from coarse to fine.

In one embodiment, identifying and segmenting the magnetic resonance image through a preset lymphatic recognition segmentation model to obtain a segmented region image specifically includes: performing data enhancement processing on the magnetic resonance image to obtain magnetic resonance enhancement. data group; performing filtering and convolution processing on the magnetic resonance enhanced data group to obtain a characteristic data group of the magnetic resonance enhanced data group; performing deconvolution processing on the characteristic data group to obtain noses with different segmentation granularities. Pharyngeal cancer metastasis lymph node region images and corresponding existence probabilities; filtering out the corresponding one or more first nasopharyngeal cancer metastasis lymph node region images whose existence probability is greater than the preset probability threshold; screening out the first nasopharyngeal cancer one or more second nasopharyngeal cancer metastasis lymph node region images whose segmentation granularity reaches a preset granularity threshold; obtain the third nasopharyngeal cancer metastasis lymph node region image with the highest segmentation granularity among the second nasopharyngeal cancer metastasis lymph node region images Pharyngeal cancer metastasis lymph node region image, and the third nasopharyngeal cancer metastasis lymph node region image is output as a segmented region image.

In one embodiment, before acquiring the magnetic resonance image to be recognized and segmented, the recognition and segmentation method further includes: acquiring a preset first training image data group and a preset model to be trained; the model to be trained is an end An end-to-end, coarse-to-fine, three-dimensional deep supervised convolutional neural network three-dimensional model; the training image data group is processed through the preset double-review data processing method to obtain the second training image data group; the The second training image data set includes a plurality of second training images; the plurality of second training images are input into the model to be trained to calculate the corresponding predicted value, and based on each predicted value and the corresponding true value, update the Describe the model parameters of the model to be trained to obtain the lymphatic recognition segmentation model.

In one embodiment, after performing recognition and segmentation on the magnetic resonance image through a preset lymphatic recognition and segmentation model to obtain a segmented region image, the recognition and segmentation method further includes: sending the segmented region image to the user .

The present invention also provides a device for identifying and segmenting nasopharyngeal cancer lymph node areas. The identifying and segmenting device includes a data acquisition unit and an identification and segmentation unit, wherein the data acquisition unit is used to acquire magnetic resonance images to be identified and segmented; The identification and segmentation unit is used to identify and segment the nasopharyngeal carcinoma lymph nodes through a preset lymphatic identification and segmentation model, thereby obtaining a segmented region image; the lymphatic identification and segmentation model is end-to-end three-dimensional depth supervision from coarse to fine. Convolutional neural network 3D model.

In one embodiment, the recognition and segmentation device further includes a model training unit, which is used to: obtain a preset first training image data group and a preset model to be trained; the model to be trained is a terminal An end-to-end, coarse-to-fine, three-dimensional deep supervised convolutional neural network three-dimensional model; the training image data group is processed through the preset double-review data processing method to obtain the second training image data group; the The second training image data set includes a plurality of second training images; the plurality of second training images are input into the model to be trained to calculate the corresponding predicted value, and based on each predicted value and the corresponding true value, update the Describe the model parameters of the model to be trained to obtain the lymphatic recognition segmentation model.

In one embodiment, the recognition and segmentation device further includes an image sending unit, the image sending unit being configured to send the segmented area image to the user.

In one embodiment, the identification and segmentation unit is further configured to: perform data enhancement processing on the magnetic resonance image, thereby obtaining a magnetic resonance enhancement data set; perform filtering and convolution processing on the magnetic resonance enhancement data set, thereby obtaining The characteristic data group of the magnetic resonance enhanced data group; perform deconvolution processing on the characteristic data group to obtain images of nasopharyngeal carcinoma metastatic lymph node regions with different segmentation granularities and corresponding existence probabilities; filter out the existence probability greater than One or more first nasopharyngeal cancer metastasis lymph node region images corresponding to a preset probability threshold; filtering out one or more corresponding first nasopharyngeal cancer metastasis lymph node region images whose segmentation granularity reaches the preset granularity threshold. a second nasopharyngeal cancer metastasis lymph node region image; obtain the third nasopharyngeal cancer metastasis lymph node region image with the highest segmentation granularity among the second nasopharyngeal cancer metastasis lymph node region images, and divide the third nasopharyngeal cancer metastasis lymph node region image into The image output is a segmented area image.

The present invention also provides a system for identifying and segmenting lymph node areas of nasopharyngeal carcinoma. The identifying and segmenting system includes an identifying and segmenting module and a data storage module. The identifying and segmenting module is communicatively connected to the data storage module. The identifying and segmenting module It is used to perform the aforementioned identification and segmentation method of nasopharyngeal cancer lymph node region according to the data stored in the data storage module.

Compared with the prior art, embodiments of the present invention have the following beneficial effects:

The present invention provides a method, device and system for identifying and segmenting nasopharyngeal cancer lymph node areas. The magnetic resonance image to be identified and segmented is identified and segmented through an end-to-end three-dimensional deep supervised convolutional neural network three-dimensional model from coarse to fine, thereby A segmented area including nasopharyngeal cancer lymph nodes is obtained. The identification and segmentation method, device and system improve the accuracy of identification and segmentation of nasopharyngeal cancer lymph node areas.

Furthermore, the method, device and system for identifying and segmenting nasopharyngeal cancer lymph node areas provided by the present invention also first process the first training image data set through a preset double-review data processing method to obtain the second training image data set. , and then input the second training image data set into the model to be trained for training, thereby fully designing a reasonable model based on the morphological characteristics of the lymph nodes, thereby improving the accuracy of identifying and segmenting the lymph node region of nasopharyngeal carcinoma.

Description of the drawings

The present invention will be further described below in conjunction with the accompanying drawings, wherein:

Figure 1 shows a flow chart of one embodiment of a method for identifying and segmenting nasopharyngeal cancer lymph node regions according to the present invention;

Figure 2 shows a flow chart of another embodiment of a method for identifying and segmenting nasopharyngeal cancer lymph node regions according to the present invention;

Figure 3 shows a structural diagram of an embodiment of a device for identifying and segmenting nasopharyngeal cancer lymph node areas according to the present invention;

Figure 4 shows a structural diagram of an embodiment of a system for identifying and segmenting nasopharyngeal cancer lymph node regions according to the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.

Specific embodiment one

The embodiment of the present invention first describes a method for identifying and segmenting lymph node areas of nasopharyngeal carcinoma. FIG. 1 shows a flow chart of one embodiment of a method for identifying and segmenting lymph node regions of nasopharyngeal carcinoma according to the present invention.

As shown in Figure 1, the method includes the following steps:

S1: Obtain the magnetic resonance image to be identified and segmented.

S2: Identify and segment the magnetic resonance image through a preset lymphatic recognition and segmentation model to obtain a segmented region image.

The lymphatic recognition segmentation model is an end-to-end, coarse-to-fine, three-dimensional deep supervised convolutional neural network three-dimensional model. In order to improve the accuracy of identification and segmentation of lymph node areas in nasopharyngeal carcinoma, embodiments of the present invention design an end-to-end, coarse-to-fine, three-dimensional deep supervised convolutional neural network three-dimensional model (3D CF-CNN) for magnetic resonance imaging. Image recognition and segmentation.

The model starts from the coarse scale, gradually increases the recognition segmentation size according to the recognition results to further refine the positioning, and draws the fine-scale attention in the area that has been detected at the coarse scale, so that it can quickly and accurately segment metastatic lymph nodes.

In one embodiment, identifying and segmenting the magnetic resonance image through a preset lymphatic recognition segmentation model to obtain a segmented region image specifically includes: performing data enhancement processing on the magnetic resonance image to obtain magnetic resonance enhancement. data group; performing filtering and convolution processing on the magnetic resonance enhanced data group to obtain a characteristic data group of the magnetic resonance enhanced data group; performing deconvolution processing on the characteristic data group to obtain noses with different segmentation granularities. Pharyngeal cancer metastasis lymph node region images and corresponding existence probabilities; filtering out the corresponding one or more first nasopharyngeal cancer metastasis lymph node region images whose existence probability is greater than the preset probability threshold; screening out the first nasopharyngeal cancer one or more second nasopharyngeal cancer metastasis lymph node region images whose segmentation granularity reaches a preset granularity threshold; obtain the third nasopharyngeal cancer metastasis lymph node region image with the highest segmentation granularity among the second nasopharyngeal cancer metastasis lymph node region images Pharyngeal cancer metastasis lymph node region image, and the third nasopharyngeal cancer metastasis lymph node region image is output as a segmented region image. In one embodiment, the preset probability threshold is 0.5. Among them, the nasopharyngeal cancer metastatic lymph node region images with different segmentation granularities are characteristic images, and the data representation form of the nasopharyngeal cancer metastasis lymph node region image and the corresponding existence probability includes a probability map.

In practical applications, since images can be represented at different scales, the model can also be supervised for training at different scales. Among them, s∈{1,…,S} indicates that the model outputs different coarse and fine granularity sets, where s=1 indicates the finest granularity. The resolution at this level is exactly the same as the resolution of the model input image.

The optimization process of the model is as follows:

1. The target pyramid is constructed through dilation and max pooling.

target[s=1] is the real label of the lymph node, then the coarse-scale target is iteratively derived as follows:

target’[s]=dilate(target[s]);

target[s+1]=maxpool(target’[s]);

2. The S-scale loss function is weighted by pixels by the true label target[s] of the lymph node at that scale and the coarse-scale prediction probability map, as shown below:

prediction’[s+1]=upsample(prediction[s+1]);

weightmap[s]=max(target’[s],prediction’[s+1]);

3. The decoding module of each scale s generates a correction map correction[s]. The prediction of a scale is the accumulation of its direct coarse-scale prediction logit*[s+1] and the correction map correction[s] of its own scale, as follows :

logit’[s+1]=upsample(logit[s+1]);

logits=merge(logit’[s+1],correction[s]);

On the 3D CF-CNN model side, the decoding module of each scale outputs a lymph node segmentation probability map. On the target side, a multi-scale pyramid is constructed based on manually outlined labels, i.e., lymph node regions, to match the size of the output probability map. The output at each scale is then compared to the corresponding output in the target pyramid. The losses at all scales are added together as the total loss.

In deep attention supervision of 3D CF-CNN, coarse layers are supervised to segment a superset of the foreground. Its predicted probability map, after upsampling to fine resolution, is used as a weight map to segment tighter supersets of the foreground. Different levels of weight map re-supervise foreground segmentation at different segmentation granularities until the finest resolution is achieved. Such a process mimics how humans gradually zoom in to locate an organ and eventually identify it at the pixel or voxel level. When you need to find an organ, you start with a rough localization; once you've localized it, you can start perusing the pixel or voxel level; in the process, once you've localized the organ at a high scale, the rest of the image becomes similar to that at a lower level. Search has nothing to do with it. Also in coarse-to-fine depth attention supervision, the predicted probability map at coarse scale is used as a weight map to exclude other irrelevant parts and focus on fine scale only within the foreground superset that has been identified at coarse scale. In this way, the model can identify lymph nodes of different shapes and can quickly and accurately segment metastatic lymph nodes.

The embodiment of the present invention describes a method for identifying and segmenting lymph node areas of nasopharyngeal carcinoma. The magnetic resonance image to be identified and segmented is identified and segmented through an end-to-end three-dimensional deep supervised convolutional neural network three-dimensional model from coarse to fine, thereby obtaining the following information: Segmentation area of nasopharyngeal cancer lymph nodes. This identification and segmentation method improves the accuracy of identification and segmentation of nasopharyngeal cancer lymph node areas.

Specific embodiment two

Furthermore, the embodiment of the present invention also describes a method for identifying and segmenting lymph node regions of nasopharyngeal carcinoma. Figure 2 shows a flow chart of another embodiment of a method for identifying and segmenting nasopharyngeal cancer lymph node regions according to the present invention.

As shown in Figure 2, the method includes the following steps:

A1: Obtain the preset first training image data group and the preset model to be trained.

The model to be trained is an end-to-end, coarse-to-fine, three-dimensional deep supervised convolutional neural network three-dimensional model. Wherein, the first training image data group includes nasopharyngeal cancer magnetic resonance imaging data collected from a hospital or medical center.

In one embodiment, the model to be trained includes an input module, an encoding module, a decoding module, and an output module; wherein the input module performs data enhancement processing on the input data (including data standardization, x- and y-axis deformation, and rotation along the Z-axis, Horizontal flip); the encoding module is used to extract image features related to metastatic lymph nodes. The encoding module contains several different levels of residual convolution blocks. Each residual convolution block uses an "activation layer (used for preliminary filtering of data features) in front ", the convolutional layer comes after" mode (as the encoding module continues to deepen, the image features extracted by the convolution block become more abstract, and the more abstract features are more representative of the input data); the decoding module includes several residual convolution blocks, The several residual convolution blocks are used for deconvolution to output nasopharyngeal carcinoma metastatic lymph node regions with different segmentation granularities (each segmentation granularity represents a segmentation accuracy and a specific resolution of the output image) and the corresponding probability (the probability is greater than 0.5 The area is considered to contain lymph nodes), and the top layer is the probability map of the metastatic lymph node area with the same resolution as the input image.

A2: Process the first training image data group through the preset double-review data processing method to obtain the second training image data group.

The second training image data set includes a plurality of second training images. After obtaining the first training image data set, in order to ensure the accuracy and effectiveness of the data used for model training, it is necessary to first review the first training image data through a "double review mechanism" (that is, experienced doctors are responsible for delineation and experts are responsible for review) The team conducts audit screening and preprocesses the data obtained from the audit screening. In one embodiment, the preprocessing process includes operations such as image data cleaning, data normalization, plain scanning and enhanced image registration.

A3: Input the plurality of second training images into the model to be trained to calculate the corresponding predicted value, and update the model parameters of the model to be trained according to each predicted value and the corresponding true value, thereby obtaining lymphatic recognition Segmentation model.

In order to train the model to be trained, the second training image obtained above is divided into a training set, a verification set and a test set according to a preset division ratio. The training set is used to train the model, the verification set is used to evaluate and select the model, and the verification set is used to evaluate and select the model. Test set to test the robustness and generalization ability of the model.

A4: Obtain the magnetic resonance image to be identified and segmented.

A5: Use the preset lymphatic recognition and segmentation model to identify and segment the magnetic resonance image to obtain a segmented region image.

The lymphatic recognition segmentation model is an end-to-end, coarse-to-fine, three-dimensional deep supervised convolutional neural network three-dimensional model. In order to improve the accuracy of identification and segmentation of lymph node areas in nasopharyngeal carcinoma, embodiments of the present invention design an end-to-end, coarse-to-fine, three-dimensional deep supervised convolutional neural network three-dimensional model (3D CF-CNN) for magnetic resonance imaging. Image recognition and segmentation.

The model starts from the coarse scale, gradually increases the recognition segmentation size according to the recognition results to further refine the positioning, and draws the fine-scale attention in the area that has been detected at the coarse scale, so that it can quickly and accurately segment metastatic lymph nodes.

In one embodiment, identifying and segmenting the magnetic resonance image through a preset lymphatic recognition segmentation model to obtain a segmented region image specifically includes: performing data enhancement processing on the magnetic resonance image to obtain magnetic resonance enhancement. data group; performing filtering and convolution processing on the magnetic resonance enhanced data group to obtain a characteristic data group of the magnetic resonance enhanced data group; performing deconvolution processing on the characteristic data group to obtain noses with different segmentation granularities. Pharyngeal cancer metastasis lymph node region images and corresponding existence probabilities; filtering out the corresponding one or more first nasopharyngeal cancer metastasis lymph node region images whose existence probability is greater than the preset probability threshold; screening out the first nasopharyngeal cancer one or more second nasopharyngeal cancer metastasis lymph node region images whose segmentation granularity reaches a preset granularity threshold; obtain the third nasopharyngeal cancer metastasis lymph node region image with the highest segmentation granularity among the second nasopharyngeal cancer metastasis lymph node region images Pharyngeal cancer metastasis lymph node region image, and the third nasopharyngeal cancer metastasis lymph node region image is output as a segmented region image.

A6: Send the segmented area image to the user.

In one embodiment, the form of sending to the user includes displaying the divided area image to the user on the display screen, and sending the divided area image to the user's user terminal through the communication module.

The embodiment of the present invention describes a method for identifying and segmenting lymph node areas of nasopharyngeal carcinoma. The magnetic resonance image to be identified and segmented is identified and segmented through an end-to-end three-dimensional deep supervised convolutional neural network three-dimensional model from coarse to fine, thereby obtaining the following information: Segmentation area of nasopharyngeal cancer lymph node. This identification and segmentation method improves the accuracy of identification and segmentation of nasopharyngeal cancer lymph node area; further, the identification and segmentation method of nasopharyngeal cancer lymph node area described in the embodiment of the present invention also uses preset The dual-review data processing method first processes the first training image data set to obtain the second training image data set, and then inputs the second training image data set into the model to be trained for training, thereby fully designing the design based on the morphological characteristics of the lymph nodes. A reasonable model improves the accuracy of identification and segmentation of lymph node areas in nasopharyngeal carcinoma.

Specific embodiment three

In addition to the above methods, embodiments of the present invention also describe a device for identifying and segmenting lymph node areas in nasopharyngeal cancer. Figure 3 shows a structural diagram of an embodiment of a device for identifying and segmenting nasopharyngeal cancer lymph node regions according to the present invention.

As shown in FIG. 3 , the recognition and segmentation device includes a data acquisition unit 11 and a recognition and segmentation unit 12 .

Among them, the data acquisition unit 11 is used to acquire the magnetic resonance image to be identified and segmented.

The identification and segmentation unit 12 is used to identify and segment the nasopharyngeal carcinoma lymph nodes through a preset lymphatic identification and segmentation model, thereby obtaining a segmented region image; the lymphatic identification and segmentation model is end-to-end three-dimensional depth supervision from coarse to fine. Convolutional neural network 3D model. The model starts from the coarse scale, gradually increases the recognition segmentation size according to the recognition results to further refine the positioning, and draws the fine-scale attention in the area that has been detected at the coarse scale, so that it can quickly and accurately segment metastatic lymph nodes.

In one embodiment, the identification and segmentation unit 12 is further configured to: perform data enhancement processing on the magnetic resonance image, thereby obtaining a magnetic resonance enhancement data set; perform filtering and convolution processing on the magnetic resonance enhancement data set, so as to obtain the magnetic resonance enhancement data set. The characteristic data group of the magnetic resonance enhanced data group; perform deconvolution processing on the characteristic data group to obtain images of nasopharyngeal carcinoma metastatic lymph node regions with different segmentation granularities and corresponding existence probabilities; filter out the existence probability greater than the predetermined one or more first nasopharyngeal cancer metastasis lymph node region images corresponding to the set probability threshold; filter out one or more corresponding first nasopharyngeal cancer metastasis lymph node region images whose segmentation granularity reaches the preset granularity threshold The second nasopharyngeal cancer metastasis lymph node region image; obtain the third nasopharyngeal cancer metastasis lymph node region image with the highest segmentation granularity among the second nasopharyngeal cancer metastasis lymph node region images, and combine the third nasopharyngeal cancer metastasis lymph node region image with The output is a segmented region image.

In one embodiment, the recognition and segmentation device further includes a model training unit, which is used to: obtain a preset first training image data group and a preset model to be trained; the model to be trained is a terminal An end-to-end, coarse-to-fine, three-dimensional deep supervised convolutional neural network three-dimensional model; the training image data group is processed through the preset double-review data processing method to obtain the second training image data group; the The second training image data set includes a plurality of second training images; the plurality of second training images are input into the model to be trained to calculate the corresponding predicted value, and based on each predicted value and the corresponding true value, update the Describe the model parameters of the model to be trained to obtain the lymphatic recognition segmentation model.

In one embodiment, the recognition and segmentation device further includes an image sending unit, the image sending unit being configured to send the segmented area image to the user.

The embodiment of the present invention describes a device for identifying and segmenting nasopharyngeal cancer lymph node areas. It uses an end-to-end three-dimensional deep supervised convolutional neural network three-dimensional model from coarse to fine to identify and segment the magnetic resonance image to be identified and segmented, thereby obtaining the following information: The identification and segmentation device of the nasopharyngeal cancer lymph node region improves the accuracy of the identification and segmentation of the nasopharyngeal cancer lymph node region; further, the identification and segmentation device of the nasopharyngeal cancer lymph node region described in the embodiment of the present invention also uses preset The dual-review data processing method first processes the first training image data set to obtain the second training image data set, and then inputs the second training image data set into the model to be trained for training, thereby fully designing the design based on the morphological characteristics of the lymph nodes. A reasonable model improves the accuracy of identification and segmentation of lymph node areas in nasopharyngeal carcinoma.

Specific embodiment four

In addition to the above methods and devices, the present invention also describes a system for identifying and segmenting lymph node areas of nasopharyngeal carcinoma. Figure 4 shows a structural diagram of an embodiment of a system for identifying and segmenting nasopharyngeal cancer lymph node regions according to the present invention.

As shown in Figure 4, the identification and segmentation system includes an identification and segmentation module 1 and a data storage module 2. The identification and segmentation module 1 is communicatively connected with the data storage module 2. The data storage module 2 is used to store all data, so The identification and segmentation module 1 is used to perform the aforementioned identification and segmentation method of nasopharyngeal cancer lymph node regions based on the data stored in the data storage module 2 .

In one embodiment, the recognition and segmentation system also includes a user interaction module, which is used to interact with the recognition and segmentation module 1 and the data storage module 2 according to instructions input by the user, and is used to send a final message to the user. Obtained segmented area images and other information.

In one embodiment, the user interaction module includes a touch display screen/non-touch display screen, an input keyboard, a virtual keyboard, an indicator light, a microphone, a speaker, and a combination of one or more of the foregoing.

The embodiment of the present invention describes a system for identifying and segmenting nasopharyngeal cancer lymph node areas. It uses an end-to-end three-dimensional deep supervised convolutional neural network three-dimensional model from coarse to fine to identify and segment the magnetic resonance image to be identified and segmented, thereby obtaining the following information: Segmentation area of nasopharyngeal cancer lymph node, this identification and segmentation system improves the accuracy of identification and segmentation of nasopharyngeal cancer lymph node area; further, the identification and segmentation system of nasopharyngeal cancer lymph node area described in the embodiment of the present invention also uses preset The dual-review data processing method first processes the first training image data set to obtain the second training image data set, and then inputs the second training image data set into the model to be trained for training, thereby fully designing the design based on the morphological characteristics of the lymph nodes. A reasonable model improves the accuracy of identification and segmentation of lymph node areas in nasopharyngeal carcinoma.

The above-mentioned specific embodiments further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above-mentioned are only specific embodiments of the present invention and are not intended to limit the scope of the present invention. . It is particularly pointed out that for those skilled in the art, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims (7)

1. A method for identifying and dividing a lymph node area of nasopharyngeal carcinoma, said method comprising:

acquiring a segmented magnetic resonance image to be identified;

identifying and segmenting the magnetic resonance image through a preset lymph identification segmentation model, so as to obtain a segmented region image; the lymph identification segmentation model is an end-to-end three-dimensional model of a three-dimensional depth supervision convolutional neural network from thick to thin;

the magnetic resonance image is identified and segmented through a preset lymph identification segmentation model, so that a segmented region image is obtained, and the method specifically comprises the following steps: performing data enhancement processing on the magnetic resonance image so as to obtain a magnetic resonance enhancement data set;

performing filtering convolution processing on the magnetic resonance enhancement data set so as to obtain a characteristic data set of the magnetic resonance enhancement data set;

deconvolution processing is carried out on the characteristic data set, so that nasopharyngeal carcinoma metastasis lymph node area images with different segmentation granularities and corresponding existence probabilities are obtained;

screening out one or more corresponding first nasopharyngeal carcinoma metastasis lymph node area images with the existence probability larger than a preset probability threshold value;

screening out one or more corresponding second nasopharyngeal carcinoma metastasis lymph node area images of which the segmentation granularity reaches a preset granularity threshold value;

acquiring a third nasopharyngeal carcinoma metastasis lymph node area image with highest segmentation granularity in the second nasopharyngeal carcinoma metastasis lymph node area image, and outputting the third nasopharyngeal carcinoma metastasis lymph node area image as a segmentation area image;

wherein in coarse-to-fine deep attention supervision, a coarse-scale predictive probability map is used as a weight map to exclude other irrelevant parts and focus fine-scale only within the superset of the foreground that the coarse-scale has identified.

2. The identification and segmentation method of a nasopharyngeal carcinoma lymph node area according to claim 1, further comprising, prior to acquiring a magnetic resonance image of a segmentation to be identified:

acquiring a preset first training image data set and a preset model to be trained; the model to be trained is an end-to-end three-dimensional model of a three-dimensional depth supervision convolutional neural network from thick to thin;

processing the training image data set by a preset double-examination data processing method so as to obtain a second training image data set; the second training image data set includes a plurality of second training images;

and inputting the plurality of second training images into the model to be trained to calculate corresponding predicted values, and updating model parameters of the model to be trained according to each predicted value and corresponding true values, so as to obtain a lymph identification segmentation model.

3. The identification and segmentation method of a nasopharyngeal carcinoma lymph node area according to claim 2, wherein after said magnetic resonance image is subjected to identification and segmentation by a preset lymph identification and segmentation model, thereby obtaining a segmented area image, said identification and segmentation method further comprises:

and sending the segmented region image to a user.

4. A recognition and segmentation device for a nasopharyngeal carcinoma lymph node area is characterized by comprising a data acquisition unit and a recognition and segmentation unit, wherein,

the data acquisition unit is used for acquiring a segmented magnetic resonance image to be identified;

the identification and segmentation unit is used for carrying out identification and segmentation on the nasopharyngeal carcinoma lymph nodes through a preset lymph identification and segmentation model so as to obtain segmented region images; the lymph identification segmentation model is an end-to-end three-dimensional depth supervision convolutional neural network three-dimensional model from thick to thin;

the identification and segmentation unit is further used for:

performing data enhancement processing on the magnetic resonance image so as to obtain a magnetic resonance enhancement data set;

performing filtering convolution processing on the magnetic resonance enhancement data set so as to obtain a characteristic data set of the magnetic resonance enhancement data set;

deconvolution processing is carried out on the characteristic data set, so that nasopharyngeal carcinoma metastasis lymph node area images with different segmentation granularities and corresponding existence probabilities are obtained;

screening out one or more corresponding first nasopharyngeal carcinoma metastasis lymph node area images with the existence probability larger than a preset probability threshold value;

screening out one or more corresponding second nasopharyngeal carcinoma metastasis lymph node area images of which the segmentation granularity reaches a preset granularity threshold value;

acquiring a third nasopharyngeal carcinoma metastasis lymph node area image with highest segmentation granularity in the second nasopharyngeal carcinoma metastasis lymph node area image, and outputting the third nasopharyngeal carcinoma metastasis lymph node area image as a segmentation area image;

wherein in coarse-to-fine deep attention supervision, a coarse-scale predictive probability map is used as a weight map to exclude other irrelevant parts and focus fine-scale only within the superset of the foreground that the coarse-scale has identified.

5. The apparatus according to claim 4, further comprising a model training unit for:

acquiring a preset first training image data set and a preset model to be trained; the model to be trained is an end-to-end three-dimensional model of a three-dimensional depth supervision convolutional neural network from thick to thin;

processing the training image data set by a preset double-examination data processing method so as to obtain a second training image data set; the second training image data set includes a plurality of second training images;

and inputting the plurality of second training images into the model to be trained to calculate corresponding predicted values, and updating model parameters of the model to be trained according to each predicted value and corresponding true values, so as to obtain a lymph identification segmentation model.

6. The apparatus according to claim 5, further comprising an image transmission unit configured to: and sending the segmented region image to a user.

7. A system for identifying and dividing a lymph node area of nasopharyngeal carcinoma, comprising an identification and division module and a data storage module, wherein the identification and division module is in communication connection with the data storage module, and the identification and division module is configured to perform the method for identifying and dividing a lymph node area of nasopharyngeal carcinoma according to the data stored in the data storage module.