CN112102237A - Brain tumor recognition model training method and device based on semi-supervised learning - Google Patents
Brain tumor recognition model training method and device based on semi-supervised learning Download PDFInfo
- Publication number
- CN112102237A CN112102237A CN202010794964.3A CN202010794964A CN112102237A CN 112102237 A CN112102237 A CN 112102237A CN 202010794964 A CN202010794964 A CN 202010794964A CN 112102237 A CN112102237 A CN 112102237A
- Authority
- CN
- China
- Prior art keywords
- training
- brain
- trained
- network
- tumor
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 274
- 208000003174 Brain Neoplasms Diseases 0.000 title claims abstract description 164
- 238000000034 method Methods 0.000 title claims abstract description 71
- 238000001514 detection method Methods 0.000 claims abstract description 164
- 210000004556 brain Anatomy 0.000 claims description 123
- 206010028980 Neoplasm Diseases 0.000 claims description 112
- 239000013598 vector Substances 0.000 claims description 45
- 230000006870 function Effects 0.000 claims description 29
- 230000004927 fusion Effects 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000009827 uniform distribution Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 238000013527 convolutional neural network Methods 0.000 abstract description 3
- 238000002591 computed tomography Methods 0.000 description 18
- 230000008569 process Effects 0.000 description 11
- 239000000284 extract Substances 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 241000282326 Felis catus Species 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 238000002059 diagnostic imaging Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
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
- 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
- G06T2207/20—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
- G06T2207/30016—Brain
-
- 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
- G06T2207/30096—Tumor; Lesion
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides a brain tumor recognition model training method and device based on semi-supervised learning, wherein the brain tumor recognition model comprises a detection network and a classification network, and the method comprises the following steps: acquiring a first training sample set and a second training sample set; performing unsupervised learning on the detection network and the classification network respectively by using the first training sample set to generate a pre-training detection network and a pre-training classification network; training the pre-training detection network and the pre-training classification network by using the second training sample set to generate a trained detection network and a trained classification network; outputting a trained brain tumor recognition model, wherein the trained brain tumor recognition model comprises the trained detection network and the trained classification network. According to the invention, the fine-scale and coarse-scale data are fully utilized in a semi-supervised learning mode, so that a more robust deep convolutional neural network is obtained.
Description
Technical Field
The invention relates to the technical field of computer vision, in particular to a medical image classification and deep learning technology, and more particularly relates to a brain tumor recognition model training method and device based on semi-supervised learning.
Background
Medical images (such as CT images and MRI images) are important data in the medical field, and play a significant role in assisting doctors in diagnosis, pathological research and the like. The medical image is intelligently and automatically analyzed by using the artificial intelligence technology, so that the method has important significance in the aspects of improving medical efficiency, saving medical cost, reducing pain of patients and the like, and provides powerful guarantee for the informatization and intelligent construction of medical treatment in China and the improvement of medical treatment level in China. The classification of medical images is a most basic task in intelligent analysis based on medical images, and has important requirements in various specific scenes such as identification of disease types, judgment of lesion severity, quantification of recovery conditions of patients and the like. Therefore, there is an urgent need and great significance in developing accurate automatic classification method and system for medical images in practical scenes.
The related task of brain CT is one of the more popular topics in the field of medical imaging at present. However, since the medical image data is difficult to acquire and label and the characteristics of the position, size and the like of the tumor in the CT image are not fixed, the tumor identification is also difficult, and the accuracy of the brain tumor identification by using the model is low.
Disclosure of Invention
Aiming at the problems, the invention provides a method and a device for training a brain tumor recognition model based on semi-supervised learning, aiming at solving the problem that in the tasks such as the recognition of a brain CT sequence, available precise standard data of a medical image is rare so as to bring challenges to model training, and the precise standard data and coarse standard data are fully utilized in a semi-supervised learning mode so as to obtain a more robust deep convolutional neural network, thereby improving the accuracy of medical image classification and brain tumor recognition.
The embodiment of the first aspect of the present invention provides a method for training a brain tumor recognition model based on semi-supervised learning, where the brain tumor recognition model includes a detection network and a classification network, and the method includes:
acquiring a first training sample set and a second training sample set, wherein the first training sample set comprises a plurality of coarse standard data and/or non-standard data, the second training sample set comprises a plurality of fine standard data, and the data are brain medical images;
performing unsupervised learning on the detection network and the classification network respectively by using the first training sample set to generate a pre-training detection network and a pre-training classification network;
training the pre-training detection network and the pre-training classification network by using the second training sample set to generate a trained detection network and a trained classification network;
outputting a trained brain tumor recognition model, wherein the trained brain tumor recognition model comprises the trained detection network and the trained classification network.
In a second embodiment of the present invention, a brain tumor recognition method based on a brain tumor recognition model is provided, where the brain tumor recognition model is obtained by training the brain tumor recognition model based on semi-supervised learning as described in the first embodiment of the present invention, the brain tumor recognition model includes a detection network and a classification network, and the method includes:
acquiring a brain medical image to be identified;
inputting the brain medical image into the detection network of the brain tumor identification model to acquire a tumor potential region in the brain medical image;
inputting the tumor potential region and the brain medical image into the classification network of the brain tumor identification model to obtain a predicted probability that the brain medical image contains a tumor;
and when the prediction probability is larger than a preset threshold value, determining that the brain medical image contains the tumor.
The embodiment of the third aspect of the invention provides a training device for a brain tumor recognition model based on semi-supervised learning, wherein the brain tumor recognition model comprises a detection network and a classification network, and the training device comprises:
the system comprises an acquisition module, a comparison module and a display module, wherein the acquisition module is used for acquiring a first training sample set and a second training sample set, the first training sample set comprises a plurality of coarse standard data and/or non-standard data, the second training sample set comprises a plurality of fine standard data, and the data is brain medical images;
the first training module is used for performing unsupervised learning on the detection network and the classification network respectively by using the first training sample set to generate a pre-training detection network and a pre-training classification network;
the second training module is used for training the pre-training detection network and the pre-training classification network by using the second training sample set to generate a trained detection network and a trained classification network;
and the output module is used for outputting the trained brain tumor recognition model, and the trained brain tumor recognition model comprises the trained detection network and the trained classification network.
In a fourth aspect, the present invention provides a brain tumor recognition apparatus based on a brain tumor recognition model, where the brain tumor recognition model is obtained by training through the training method of the brain tumor recognition model based on semi-supervised learning as described in the foregoing first aspect, the brain tumor recognition model includes a detection network and a classification network, and the apparatus includes:
the image acquisition module is used for acquiring a brain medical image to be identified;
a first input module for inputting the brain medical image into the detection network of the brain tumor identification model to obtain a tumor potential region in the brain medical image;
a second input module, configured to input the tumor potential region and the brain medical image into the classification network of the brain tumor identification model to obtain a predicted probability that the brain medical image contains a tumor;
a determining module, configured to determine that a tumor is included in the brain medical image when the prediction probability is greater than a preset threshold.
A fifth aspect of the present invention provides a computer device, comprising a processor, a memory and a computer program stored on the memory and executable on the processor, wherein the computer program, when executed by the processor, implements a method for training a brain tumor recognition model based on semi-supervised learning as described in the first aspect of the present invention, or implements a method for brain tumor recognition based on a brain tumor recognition model as described in the second aspect of the present invention.
A sixth aspect of the present invention provides a non-transitory computer-readable storage medium, having stored thereon a computer program, which, when being executed by a processor, implements a method for training a brain tumor recognition model based on semi-supervised learning as described in the first aspect, or implements a method for brain tumor recognition based on a brain tumor recognition model as described in the second aspect.
The technical scheme provided by the invention can at least bring the following beneficial effects:
the method divides the training of the brain tumor recognition model into two stages, firstly uses the rough mark data and/or the non-mark data to perform unsupervised learning on the whole model, so that the characteristics output by the model have certain clustering effect, and the generalization capability of the model can be enhanced; then, the detection network and the classification network are supervised and learned by utilizing the precise standard data, so that the discrimination capability of the model can be improved, the tumor automatic identification of the brain CT sequence in the medical image field is realized, and a foundation is laid for the clinical pre-detection and the secondary correction of the focus identified by a doctor. According to the method, a large amount of non-standard and coarse-standard data in the medical image are utilized through unsupervised learning, the brain tumor recognition model is obtained through semi-supervised learning training, and the fine-standard data and the coarse-standard data are fully considered in the training process, so that more effective feature extraction and more accurate classification are realized, the model with stronger discrimination capability and generalization capability is trained, and the trained model has the advantages of high precision, wide application range, strong robustness and the like, and has a stronger practical application prospect. The trained brain tumor recognition model is used for brain tumor recognition, so that the brain tumor can be automatically recognized, and the accuracy of a recognition result is improved.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The foregoing and/or additional aspects and advantages of the present application will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
fig. 1 is a schematic flowchart of a training method of a brain tumor recognition model based on semi-supervised learning according to an embodiment of the present invention;
fig. 2 is a schematic flowchart of a method for training a brain tumor recognition model based on semi-supervised learning according to another embodiment of the present invention;
fig. 3 is a schematic flowchart of a method for training a brain tumor recognition model based on semi-supervised learning according to another embodiment of the present invention;
fig. 4 is a schematic flow chart of a brain tumor identification method based on a brain tumor identification model according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a training apparatus for a brain tumor recognition model based on semi-supervised learning according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a brain tumor recognition apparatus based on a brain tumor recognition model according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are exemplary and intended to be used for explaining the present application and should not be construed as limiting the present application.
The following describes a method and an apparatus for training a brain tumor recognition model based on semi-supervised learning according to an embodiment of the present invention with reference to the drawings.
In recent years, due to the great success of deep learning in computer vision tasks, especially in the fields of image recognition, image segmentation, image detection, and the like, more and more scholars are continuously trying to solve corresponding problems in the field of medical images using a deep learning algorithm.
Electronic Computed Tomography (CT) is one of the most common tools in the medical field for diagnosing brain cases in humans, and thus much work in the medical imaging field is based on CT images, and the brain is the most important part of the human body, and tumors present at the location of the brain are more life-threatening than others. Therefore, tumor identification of brain CT images is clinically significant in practice, and the research applications thereof are wide, including but not limited to pre-detection or screening of actual CT images, secondary correction of doctor test results, and even replacing manual diagnosis procedures. For the above reasons, brain CT related tasks become one of the most popular topics in the field of medical images.
However, tumor identification for brain CT images is not a simple classification task, and in fact it faces many challenges.
On the one hand, in most medical image fields, the acquisition and labeling of medical image data is a difficult problem. Firstly, medical image data related to research is difficult to collect, the medical image data is usually provided by organizations such as hospitals with cooperative relations, and generally, in order to protect privacy of patients, the obtained medical image data needs to be strictly confidential, so that a few public medical image data sets are available, and the data volume is small; secondly, after the original data is collected, the marking of the data is very difficult, in general, the obtained data needs to be marked by a doctor with rich experience, and the accurate marking of the data is scarce for more complex tasks such as segmentation, detection and the like. The collected data is generally composed of a large portion of coarse-scale data and a small portion of fine-scale data. The rough-scale data is label data only containing rough information such as categories, and the fine-scale data is label data containing more specific information besides category information, such as the category of each pixel on the segmentation task, the specific frame position information of the object on the detection task, and the like.
On the other hand, identification of tumor cases in CT images is difficult. The tumor affects the CT image only in a local area, and the characteristics of the tumor such as position, size and the like are not fixed and have no obvious preference, so that the inter-class difference of data is small and the intra-class difference is large. In other words, the CT images with and without a tumor are generally slightly different from each other, and there is a certain difference between the CT images with a tumor. For the common classification problem in computer vision, the difference between classes is large, and the difference in classes is small, for example, the difference between cats and flowers far exceeds the difference between images of cats.
Because the marked data is divided into a small amount of fine-mark data and a large amount of coarse-mark data, if the fine-mark information is ignored, the supervision information is lost to a certain extent if all the data is directly regarded as coarse-mark data, and if only the fine-mark data is used, all the data cannot be completely utilized, and overfitting is easily caused. Therefore, from practical application, the method carries out supervised learning of detection and classification on the fine-scale data and unsupervised learning on the coarse-scale data, can fully utilize strong supervision information of the fine-scale data and weak supervision information of a large amount of coarse-scale data, finally carries out training on a brain CT sequence tumor recognition model based on semi-supervised learning, and trains a model with stronger discrimination capability and generalization capability.
Fig. 1 is a schematic flowchart of a training method of a brain tumor recognition model based on semi-supervised learning according to an embodiment of the present invention, where the method may be executed by the training apparatus of a brain tumor recognition model based on semi-supervised learning provided by the present invention, and the training apparatus of a brain tumor recognition model based on semi-supervised learning may be applied to a computer device provided by the present invention, and the computer device may be a server, or an electronic device such as a desktop computer or a notebook computer.
In the embodiment of the invention, the brain tumor recognition model comprises a detection network and a classification network, and the training of the brain tumor recognition model is actually the training of the detection network and the classification network. Wherein the detection network can use fast R-CNN, and the classification network can use ResNet-50.
As shown in fig. 1, the method for training a brain tumor recognition model based on semi-supervised learning may include the following steps:
In the embodiment of the invention, a large amount of coarse data and/or non-standard data can be obtained to form a first training sample set, and a small amount of fine standard data is obtained to form a second training sample set. The rough mark data are marked data only containing rough information such as categories and the like, the non-marked data are data which are not marked, the rough mark data and the non-marked data can be directly used, and the information such as related categories and the like does not need to be given; the precise marking data is marking data which contains more specific information besides category information.
It can be understood that, in the present invention, the coarse scale data, the unmarked data, and the fine scale data are all brain medical images. That is, in the present invention, brain medical data is collected to obtain a first training sample set and a second training sample set.
And 102, performing unsupervised learning on the detection network and the classification network respectively by using the first training sample set to generate a pre-training detection network and a pre-training classification network.
In the embodiment of the present invention, after the first training sample set is obtained, the first training sample set may be used to perform unsupervised learning on the detection network and the classification network in the brain tumor recognition model, respectively, so as to generate a pre-training detection network and a pre-training classification network.
The process of generating the pre-trained detection network by performing unsupervised learning on the detection network using the first training sample set is similar to the process of generating the pre-trained classification network by performing unsupervised learning on the classification network using the first training sample set, and the unsupervised learning process will be described below by taking the generation of the pre-trained detection network by training as an example.
When the pre-training detection network is generated through unsupervised learning, the coarse brain medical image and/or the non-standard brain medical image in the first training data set are/is input into the detection network to be calculated through operations such as convolution and the like, the detection network generates corresponding features of the brain medical image, loss function calculation is conducted on the basis of the features of the images, and parameters of the detection network are updated to minimize a loss function. When the detection network after updating the parameters converges, the unsupervised learning is finished to obtain a pre-training detection network; and when the detection network after updating the parameters is not converged, continuously inputting the brain medical image for iterative training, updating the parameters of the detection network again, judging whether the detection network after updating the parameters is converged, and repeating the process until the detection network is converged. And judging the convergence mode only by calculating the loss function values in two iterations before and after, and if the loss function values are not obviously changed, determining that the network convergence is detected and unsupervised learning is finished.
And 103, training the pre-training detection network and the pre-training classification network by using the second training sample set to generate a trained detection network and a trained classification network.
In the embodiment of the invention, after the pre-training detection network and the pre-training classification network are obtained through unsupervised learning, the pre-training detection network and the pre-training classification network can be further trained by using the precise standard data in the second training sample set, so that the trained detection network and the trained classification network are generated.
In specific implementation, the pre-training detection network can be supervised and trained by using the second training sample set to generate a trained detection network, and the trained detection network can predict the position information of the tumor in the brain medical image. In the training process, parameters of the pre-training detection network are adjusted based on the loss function so as to minimize the loss function, and when the pre-training detection network after the parameter adjustment is converged, supervised learning is completed to generate the trained detection network.
And then, performing supervised learning on the pre-training classification network by using a second training sample set. Firstly, inputting the brain medical images in the second training sample set into a trained detection network to obtain a tumor image region, then inputting the tumor image region and the brain medical images into a pre-training classification network for supervision training to generate a trained classification network, wherein the trained classification network can realize tumor identification of the brain medical images and output the probability of whether the images contain tumors.
And 104, outputting the trained brain tumor recognition model, wherein the trained brain tumor recognition model comprises a trained detection network and a trained classification network.
In the embodiment of the invention, after the trained detection network and the trained classification network are generated, the brain tumor recognition model is trained, and then the trained brain tumor recognition model can be output so as to automatically recognize the brain tumor by using the brain tumor recognition model.
In the training method of the brain tumor recognition model based on semi-supervised learning according to the embodiment, a first training sample set including a plurality of coarse data and/or non-standard data and a second training sample set including a plurality of fine data are obtained, unsupervised learning is performed on a detection network and a classification network respectively by using the first training sample set, a pre-training detection network and a pre-training classification network are generated, the pre-training detection network and the pre-training classification network are trained by using the second training sample set, the trained detection network and the trained classification network are generated, and thus the trained brain tumor recognition model is obtained and output. Therefore, the brain tumor recognition model is obtained through semi-supervised learning training, and the accurate standard data and the coarse standard data are fully considered in the training process, so that more effective feature extraction and more accurate classification are realized, and the model with stronger discrimination capability and generalization capability is trained.
In order to more clearly illustrate a specific implementation process of performing unsupervised learning on the detection network and the classification network to generate the pre-trained detection network and the pre-trained classification network respectively by using the first training sample set in the foregoing embodiment, the following description is made in detail with reference to fig. 2.
Fig. 2 is a schematic flowchart of a method for training a brain tumor recognition model based on semi-supervised learning according to another embodiment of the present invention, as shown in fig. 2, based on the embodiment shown in fig. 1, step 102 may include the following steps:
The basic flow of training is to continuously optimize the model by using an iterative mode. Due to the complexity of the training process and the limitation of computational resources, it is not possible to use all data for model update during each iteration. Generally, a small batch of training data is obtained by randomly sampling data and then training the small batch of training data in one iteration. Therefore, in the embodiment of the present invention, the first training sample set is randomly sampled on the basis of uniform distribution, and is divided into a plurality of sample subsets, each sample subset may include a preset number of brain medical images, for example, the preset number may be set to 64, and each sample subset includes 64 brain medical images.
In the embodiment of the invention, one sample subset can be selected from a plurality of sample subsets each time and respectively input into the detection network and the classification network, then the detection network and the classification network respectively extract the characteristics of the input brain medical image, unsupervised learning is carried out on the extracted characteristics based on the contrast loss function, the parameters of the detection network and the classification network are updated in an iterative manner, when the detection network after the parameters are updated is converged, a pre-trained detection network is generated, and when the classification network after the parameters are updated is converged, the pre-trained classification network is generated.
The comparison loss function is based on momentum, and the comparison loss function plays a role in unsupervised learning by drawing near homologous data features and drawing far non-homologous data features, and performs preliminary pre-training on a detection network and a classification network, so that features output by the model have a certain clustering effect, and the generalization capability of the model is enhanced.
In the training method of the brain tumor recognition model based on semi-supervised learning of the embodiment, the first training sample set is randomly sampled based on uniform distribution, the first training sample set is divided into a plurality of sample subsets, each sample subset comprises a preset number of the brain medical images, one sample subset is selected from the plurality of sample subsets and is respectively input into the detection network and the classification network, so that the detection network and the classification network respectively extract the features of the brain medical images, and unsupervised learning is performed on the extracted features based on the contrast loss function to generate the pre-trained detection network and the pre-trained classification network, thereby a large amount of non-standard data and coarse-standard data in the brain medical images are fully utilized through unsupervised learning, so that the features output by the model have a certain clustering effect, and the generalization capability of the model can be enhanced, and the method is favorable for realizing more effective feature extraction and more accurate classification.
In order to more clearly illustrate a specific implementation process of training the pre-trained detection network and the pre-trained classification network to generate a trained detection network and a trained classification network by using the second training sample set in the foregoing embodiment, the following description is made in detail with reference to fig. 3.
Fig. 3 is a flowchart illustrating a method for training a brain tumor recognition model based on semi-supervised learning according to another embodiment of the present invention, as shown in fig. 3, based on the embodiment shown in fig. 1, step 103 may include the following steps:
and 301, training the pre-training detection network by using a second training sample set to generate a trained detection network.
In the embodiment of the invention, the pre-training detection network is trained by using the second training sample set, the parameters of the pre-training detection network are updated iteratively based on the loss function of the pre-training detection network so as to minimize the loss function of the pre-training detection network, and when the pre-training detection network after updating the parameters is converged, the training is completed, and the trained detection network is generated. The trained detection network can predict the tumor position in the brain medical image.
In the embodiment of the invention, when the trained classification network is generated by training, the training can be performed by combining the output result of the trained detection network. Firstly, inputting a plurality of fine-scale data in a second training sample set into a trained detection network to obtain a prediction result of the trained detection network on a tumor region in the plurality of fine-scale data. And the prediction result is the tumor position predicted by the trained detection network.
Then, a plurality of local tumor images are extracted from the plurality of precise target data according to the tumor position in the prediction result. That is to say, for each prediction result, according to the prediction result, a local image where the tumor indicated by the tumor position is located is cut out from the brain medical image corresponding to the prediction result, so as to obtain a local tumor image.
In the embodiment of the present invention, after obtaining the plurality of local tumor images, the size of the plurality of local tumor images may be adjusted, for example, for each local tumor image, the size of the local tumor image may be adjusted to be the same as that of the corresponding original brain medical image, so as to obtain the corresponding target local tumor image.
And 305, respectively inputting each target local tumor image and the corresponding brain medical image into a pre-training classification network for feature extraction, and obtaining a local feature vector of the target local tumor image and a global feature vector of the brain medical image.
And step 306, splicing the local feature vector and the global feature vector to generate a fusion feature vector.
In the embodiment of the invention, after the target local tumor images are obtained, each target local tumor image and the corresponding brain medical image can be respectively input into the pre-training classification network for feature extraction, and the local feature vector corresponding to the target local tumor images and the global feature vector corresponding to the brain medical images are obtained. And then, connecting the local feature vector of the target local tumor image with the global feature vector of the corresponding brain medical image on the channel dimension to obtain a fusion feature vector.
And 307, inputting the fusion feature vector into a pre-training classification network for training, and updating parameters of the pre-training classification network based on a cross entropy loss function and a back propagation algorithm so as to minimize the cross entropy loss function.
And 308, when the pre-training classification network after the parameters are updated is converged, determining that the pre-training classification network is trained, and generating the trained classification network.
In the embodiment of the invention, after the fusion feature vector is obtained, the fusion feature vector can be input into the pre-training classification network for training, and the parameters of the pre-training classification network are updated based on the cross entropy loss function and the back propagation algorithm so as to minimize the cross entropy loss function. And then judging whether the pre-training classification network after updating the parameters is converged, if so, determining that the pre-training classification network is trained, and generating the trained classification network.
The training method of the brain tumor recognition model based on semi-supervised learning according to the embodiment trains a pre-training detection network to generate a trained detection network by using a second training sample set, inputs a plurality of fine-scale data into the trained detection network to obtain the prediction results of the trained detection network on tumor regions in the plurality of fine-scale data, extracts a plurality of local tumor images from the plurality of fine-scale data according to the prediction results, further inputs each target local tumor image and the corresponding brain medical image into a pre-training classification network to perform feature extraction to obtain local feature vectors of the target local tumor images and global feature vectors of the brain medical images, splices the local feature vectors and the global feature vectors to generate fused feature vectors, and further inputs the fused feature vectors into the pre-training classification network to perform training, and updating parameters of the pre-training classification network based on a cross entropy loss function and a back propagation algorithm so as to minimize the cross entropy loss function, determining that the pre-training classification network is trained when the pre-training classification network after the parameters are updated is converged, and generating the trained classification network. Therefore, the detection network is obtained through training, the classification network is trained by utilizing the output of the trained detection network, the characteristic difference between data can be strengthened, the problems of large intra-class difference and small inter-class difference of the data are relieved to a certain extent, in addition, the local characteristic vector and the global characteristic vector are fused to generate a fusion characteristic vector to be used for training the classification network, the influence of the characteristics of the tumor region on classification is strengthened through a characteristic fusion mode, the tumor region can be concerned by the model, the tumor recognition is further realized, and the judgment capability of the final model can be obviously enhanced.
In order to implement the above embodiments, the present invention further provides a brain tumor identification method based on the brain tumor identification model.
Fig. 4 is a schematic flow chart of a brain tumor recognition method based on a brain tumor recognition model according to an embodiment of the present invention, where the method may be executed by the brain tumor recognition apparatus based on the brain tumor recognition model provided by the present invention, and the brain tumor recognition apparatus based on the brain tumor recognition model may be applied to a computer device provided by the present invention, and the computer device may be a server, or an electronic device such as a desktop computer or a notebook computer.
As shown in fig. 4, the brain tumor identification method based on the brain tumor identification model may include the following steps:
The medical brain image may be a CT brain image.
The brain tumor recognition model of the present embodiment is obtained by training the brain tumor recognition model based on semi-supervised learning according to the training method of the brain tumor recognition model of the foregoing embodiment, and includes a detection network and a classification network, both of which are trained networks. The brain tumor identification model has the advantages of high precision, wide application range, strong robustness and the like.
In this embodiment, when the brain medical image needs to be detected to identify whether the brain medical image contains a tumor region, the acquired brain medical image may be first input into a detection network of the brain tumor identification model, and the detection network may predict position information of the tumor in the input brain medical image, so that the potential tumor region in the brain medical image may be acquired from the detection network. For example, the detection network may mark a tumor in the brain medical image in the form of a rectangular frame, where the region where the rectangular frame is located is a potential tumor region.
In this embodiment, after the tumor potential region in the brain medical image is acquired, the tumor potential region and the brain medical image may be input into the classification network of the brain tumor identification model together. The classification network calculates the brain medical image and the tumor potential area, extracts the global feature vector corresponding to the brain medical image and the local feature vector of the local image corresponding to the tumor potential area, fuses the local feature vectors and the global feature vectors, inputs the fusion feature vectors generated by fusion into the classifier, and calculates to obtain the prediction probability of the brain medical image containing the tumor. Therefore, in this embodiment, the prediction probability of the brain medical image including the tumor can be obtained from the classification network of the brain tumor identification model.
And step 404, when the prediction probability is larger than a preset threshold value, determining that the brain medical image contains the tumor.
In this embodiment, after the prediction probability that the brain medical image includes the tumor is obtained, whether the brain medical image includes the tumor may be determined according to the prediction probability. When the prediction probability is greater than a preset threshold, it can be determined that the tumor is included in the brain medical image.
The preset threshold may be preset, and the corresponding classification threshold may be selected according to an actual situation, for example, the preset threshold is set to 0.5.
Generally, in the medical field, negative and positive are used to characterize whether a certain test result is normal. Thus, as an example, when the probability of positive prediction output by the classification network is greater than a preset threshold (e.g., 0.5), the brain medical image may be determined to be positive, i.e., the brain medical image contains a tumor.
According to the brain tumor identification method based on the brain tumor identification model, the brain medical image to be identified is acquired, the brain medical image is input into the detection network of the brain tumor identification model to acquire the tumor potential area in the brain medical image, the tumor potential area and the brain medical image are further input into the classification network of the brain tumor identification model to acquire the prediction probability of the brain medical image containing the tumor, and when the prediction probability is larger than a preset threshold value, the brain medical image containing the tumor is determined, so that the brain tumor can be automatically identified, the accuracy of an identification result is improved, and the brain tumor can be accurately and robustly automatically identified.
In order to realize the embodiment, the invention further provides a training device of the brain tumor recognition model based on semi-supervised learning.
Fig. 5 is a schematic structural diagram of a training apparatus for a brain tumor recognition model based on semi-supervised learning according to an embodiment of the present invention, and as shown in fig. 5, the training apparatus 50 for a brain tumor recognition model based on semi-supervised learning includes: an acquisition module 510, a first training module 520, a second training module 530, and an output module 540.
The obtaining module 510 is configured to obtain a first training sample set and a second training sample set, where the first training sample set includes a plurality of coarse-scale data and/or non-scale data, the second training sample set includes a plurality of fine-scale data, and the data is a brain medical image.
A first training module 520, configured to perform unsupervised learning on the detection network and the classification network respectively by using the first training sample set, so as to generate a pre-training detection network and a pre-training classification network.
In a possible implementation manner of the embodiment of the present invention, the first training module 520 is specifically configured to:
randomly sampling the first training sample set based on uniform distribution, dividing the first training sample set into a plurality of sample subsets, wherein each sample subset contains a preset number of brain medical images;
selecting one sample subset from the plurality of sample subsets to be respectively input into the detection network and the classification network, so that the detection network and the classification network respectively perform feature extraction on the brain medical image, and performing unsupervised learning on the extracted features based on a contrast loss function to generate the pre-trained detection network and the pre-trained classification network.
A second training module 530, configured to train the pre-trained detection network and the pre-trained classification network by using the second training sample set, and generate a trained detection network and a trained classification network.
In a possible implementation manner of the embodiment of the present invention, the second training module 530 is specifically configured to:
training the pre-training detection network by using the second training sample set to generate a trained detection network;
inputting the plurality of fine-scale data into the trained detection network to obtain the prediction results of the trained detection network on the tumor regions in the plurality of fine-scale data;
extracting a plurality of local tumor images from the plurality of precise marking data according to the prediction result;
adjusting the sizes of the local tumor images to generate a plurality of target local tumor images;
inputting each target local tumor image and the corresponding brain medical image into the pre-training classification network respectively for feature extraction, and acquiring a local feature vector of the target local tumor image and a global feature vector of the brain medical image;
splicing the local feature vector and the global feature vector to generate a fusion feature vector;
inputting the fusion feature vector into the pre-training classification network for training, and updating parameters of the pre-training classification network based on a cross entropy loss function and a back propagation algorithm so as to minimize the cross entropy loss function;
and when the pre-training classification network after updating the parameters converges, determining that the pre-training classification network is trained completely, and generating the trained classification network.
An output module 540, configured to output the trained brain tumor recognition model, where the trained brain tumor recognition model includes the trained detection network and the trained classification network.
It should be noted that the foregoing explanation of the embodiment of the brain tumor recognition model training method based on semi-supervised learning is also applicable to the brain tumor recognition model training apparatus based on semi-supervised learning in this embodiment, and the implementation principle is similar, and is not described herein again.
According to the training device of the brain tumor recognition model based on semi-supervised learning, provided by the embodiment of the invention, a first training sample set comprising a plurality of coarse data and/or non-standard data and a second training sample set comprising a plurality of fine data are obtained, the first training sample set is used for carrying out unsupervised learning on a detection network and a classification network respectively to generate a pre-training detection network and a pre-training classification network, then the second training sample set is used for training the pre-training detection network and the pre-training classification network to generate the trained detection network and the trained classification network, so that the trained brain tumor recognition model is obtained and output. Therefore, the brain tumor recognition model is obtained through semi-supervised learning training, and the accurate standard data and the coarse standard data are fully considered in the training process, so that more effective feature extraction and more accurate classification are realized, and the model with stronger discrimination capability and generalization capability is trained.
In order to implement the above embodiments, the present invention further provides a brain tumor recognition apparatus based on the brain tumor recognition model.
Fig. 6 is a schematic structural diagram of a brain tumor recognition apparatus based on a brain tumor recognition model according to an embodiment of the present invention, where the brain tumor recognition model of this embodiment is obtained by training the brain tumor recognition model based on semi-supervised learning according to the training method of the brain tumor recognition model of the foregoing embodiment, and the brain tumor recognition model includes a detection network and a classification network. As shown in fig. 6, the brain tumor recognition apparatus 60 based on the brain tumor recognition model includes: an image acquisition module 610, a first input module 620, a second input module 630, and a determination module 640.
The image acquiring module 610 is configured to acquire a medical brain image to be identified.
A first input module 620, configured to input the brain medical image into the detection network of the brain tumor identification model to obtain a tumor potential region in the brain medical image.
A second input module 630, configured to input the tumor potential region and the brain medical image into the classification network of the brain tumor identification model to obtain a predicted probability that the brain medical image includes a tumor.
A determining module 640, configured to determine that a tumor is included in the brain medical image when the prediction probability is greater than a preset threshold.
It should be noted that the foregoing explanation of the embodiment of the brain tumor identification method based on the brain tumor identification model is also applicable to the brain tumor identification apparatus based on the brain tumor identification model of this embodiment, and the implementation principle is similar, and is not repeated here.
The brain tumor identification device based on the brain tumor identification model of the embodiment inputs the brain medical image into the detection network of the brain tumor identification model by acquiring the brain medical image to be identified so as to acquire the tumor potential region in the brain medical image, and further inputs the tumor potential region and the brain medical image into the classification network of the brain tumor identification model so as to acquire the prediction probability of the tumor contained in the brain medical image, and when the prediction probability is greater than a preset threshold value, the tumor contained in the brain medical image is determined, so that the brain tumor automatic identification can be realized, the accuracy of the identification result is improved, and the brain tumor automatic identification with accuracy and robustness is realized.
In order to implement the foregoing embodiments, the present invention further provides a computer device, which includes a processor, a memory and a computer program stored on the memory and executable on the processor, and when the computer program is executed by the processor, the computer device implements the brain tumor recognition model training method based on semi-supervised learning as described in the foregoing embodiments, or implements the brain tumor recognition model training method based on the brain tumor recognition model as described in the foregoing embodiments.
In order to achieve the above embodiments, the present invention further proposes a non-transitory computer readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the method for training a brain tumor recognition model based on semi-supervised learning as described in the foregoing embodiments, or implements the method for brain tumor recognition based on a brain tumor recognition model as described in the foregoing embodiments.
In the description herein, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present application, "plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
In addition, functional units in the embodiments of the present application may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium.
The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc. Although embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application, and that variations, modifications, substitutions and alterations may be made to the above embodiments by those of ordinary skill in the art within the scope of the present application.
Claims (10)
1. A method for training a brain tumor recognition model based on semi-supervised learning, wherein the brain tumor recognition model comprises a detection network and a classification network, and the method comprises the following steps:
acquiring a first training sample set and a second training sample set, wherein the first training sample set comprises a plurality of coarse standard data and/or non-standard data, the second training sample set comprises a plurality of fine standard data, and the data are brain medical images;
performing unsupervised learning on the detection network and the classification network respectively by using the first training sample set to generate a pre-training detection network and a pre-training classification network;
training the pre-training detection network and the pre-training classification network by using the second training sample set to generate a trained detection network and a trained classification network;
outputting a trained brain tumor recognition model, wherein the trained brain tumor recognition model comprises the trained detection network and the trained classification network.
2. The method of claim 1, wherein the unsupervised learning of the detection network and the classification network using the first training sample set to generate a pre-trained detection network and a pre-trained classification network comprises:
randomly sampling the first training sample set based on uniform distribution, dividing the first training sample set into a plurality of sample subsets, wherein each sample subset contains a preset number of brain medical images;
selecting one sample subset from the plurality of sample subsets to be respectively input into the detection network and the classification network, so that the detection network and the classification network respectively perform feature extraction on the brain medical image, and performing unsupervised learning on the extracted features based on a contrast loss function to generate the pre-trained detection network and the pre-trained classification network.
3. The method of claim 1, wherein training the pre-trained detection network and the pre-trained classification network using the second training sample set to generate a trained detection network and a trained classification network comprises:
training the pre-training detection network by using the second training sample set to generate a trained detection network;
inputting the plurality of fine-scale data into the trained detection network to obtain the prediction results of the trained detection network on the tumor regions in the plurality of fine-scale data;
extracting a plurality of local tumor images from the plurality of precise marking data according to the prediction result;
adjusting the sizes of the local tumor images to generate a plurality of target local tumor images;
inputting each target local tumor image and the corresponding brain medical image into the pre-training classification network respectively for feature extraction, and acquiring a local feature vector of the target local tumor image and a global feature vector of the brain medical image;
splicing the local feature vector and the global feature vector to generate a fusion feature vector;
inputting the fusion feature vector into the pre-training classification network for training, and updating parameters of the pre-training classification network based on a cross entropy loss function and a back propagation algorithm so as to minimize the cross entropy loss function;
and when the pre-training classification network after updating the parameters converges, determining that the pre-training classification network is trained completely, and generating the trained classification network.
4. A brain tumor recognition method based on a brain tumor recognition model, wherein the brain tumor recognition model is trained by the training method of the brain tumor recognition model based on semi-supervised learning according to any one of claims 1 to 3, the brain tumor recognition model comprises a detection network and a classification network, and the method comprises:
acquiring a brain medical image to be identified;
inputting the brain medical image into the detection network of the brain tumor identification model to acquire a tumor potential region in the brain medical image;
inputting the tumor potential region and the brain medical image into the classification network of the brain tumor identification model to obtain a predicted probability that the brain medical image contains a tumor;
and when the prediction probability is larger than a preset threshold value, determining that the brain medical image contains the tumor.
5. A training apparatus for a brain tumor recognition model based on semi-supervised learning, wherein the brain tumor recognition model comprises a detection network and a classification network, the apparatus comprising:
the system comprises an acquisition module, a comparison module and a display module, wherein the acquisition module is used for acquiring a first training sample set and a second training sample set, the first training sample set comprises a plurality of coarse standard data and/or non-standard data, the second training sample set comprises a plurality of fine standard data, and the data is brain medical images;
the first training module is used for performing unsupervised learning on the detection network and the classification network respectively by using the first training sample set to generate a pre-training detection network and a pre-training classification network;
the second training module is used for training the pre-training detection network and the pre-training classification network by using the second training sample set to generate a trained detection network and a trained classification network;
and the output module is used for outputting the trained brain tumor recognition model, and the trained brain tumor recognition model comprises the trained detection network and the trained classification network.
6. The apparatus of claim 5, wherein the first training module is specifically configured to:
randomly sampling the first training sample set based on uniform distribution, dividing the first training sample set into a plurality of sample subsets, wherein each sample subset contains a preset number of brain medical images;
selecting one sample subset from the plurality of sample subsets to be respectively input into the detection network and the classification network, so that the detection network and the classification network respectively perform feature extraction on the brain medical image, and performing unsupervised learning on the extracted features based on a contrast loss function to generate the pre-trained detection network and the pre-trained classification network.
7. The apparatus of claim 5, wherein the second training module is specifically configured to:
training the pre-training detection network by using the second training sample set to generate a trained detection network;
inputting the plurality of fine-scale data into the trained detection network to obtain the prediction results of the trained detection network on the tumor regions in the plurality of fine-scale data;
extracting a plurality of local tumor images from the plurality of precise marking data according to the prediction result;
adjusting the sizes of the local tumor images to generate a plurality of target local tumor images;
inputting each target local tumor image and the corresponding brain medical image into the pre-training classification network respectively for feature extraction, and acquiring a local feature vector of the target local tumor image and a global feature vector of the brain medical image;
splicing the local feature vector and the global feature vector to generate a fusion feature vector;
inputting the fusion feature vector into the pre-training classification network for training, and updating parameters of the pre-training classification network based on a cross entropy loss function and a back propagation algorithm so as to minimize the cross entropy loss function;
and when the pre-training classification network after updating the parameters converges, determining that the pre-training classification network is trained completely, and generating the trained classification network.
8. A brain tumor recognition apparatus based on a brain tumor recognition model, wherein the brain tumor recognition model is trained by the training method of the brain tumor recognition model based on semi-supervised learning according to any one of claims 1 to 3, the brain tumor recognition model comprises a detection network and a classification network, the apparatus comprises:
the image acquisition module is used for acquiring a brain medical image to be identified;
a first input module for inputting the brain medical image into the detection network of the brain tumor identification model to obtain a tumor potential region in the brain medical image;
a second input module, configured to input the tumor potential region and the brain medical image into the classification network of the brain tumor identification model to obtain a predicted probability that the brain medical image contains a tumor;
a determining module, configured to determine that a tumor is included in the brain medical image when the prediction probability is greater than a preset threshold.
9. A computer arrangement comprising a processor, a memory and a computer program stored on the memory and being executable on the processor, the computer program, when being executed by the processor, implementing a method for brain tumor recognition based on a semi-supervised learning brain tumor recognition model according to any one of claims 1-3, or implementing a method for brain tumor recognition based on a brain tumor recognition model according to claim 4.
10. A non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the method for training a brain tumor recognition model based on semi-supervised learning according to any one of claims 1 to 3, or implements the method for brain tumor recognition based on a brain tumor recognition model according to claim 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010794964.3A CN112102237A (en) | 2020-08-10 | 2020-08-10 | Brain tumor recognition model training method and device based on semi-supervised learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010794964.3A CN112102237A (en) | 2020-08-10 | 2020-08-10 | Brain tumor recognition model training method and device based on semi-supervised learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112102237A true CN112102237A (en) | 2020-12-18 |
Family
ID=73753056
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010794964.3A Pending CN112102237A (en) | 2020-08-10 | 2020-08-10 | Brain tumor recognition model training method and device based on semi-supervised learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112102237A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112668492A (en) * | 2020-12-30 | 2021-04-16 | 中山大学 | Behavior identification method for self-supervised learning and skeletal information |
| CN113160243A (en) * | 2021-03-24 | 2021-07-23 | 联想(北京)有限公司 | Image segmentation method and electronic equipment |
| CN113298065A (en) * | 2021-05-13 | 2021-08-24 | 杭州电子科技大学 | Eye melanoma recognition method based on self-supervision learning |
| CN113470828A (en) * | 2021-06-30 | 2021-10-01 | 上海商汤智能科技有限公司 | Classification method and device, electronic equipment and storage medium |
| CN113657143A (en) * | 2021-06-25 | 2021-11-16 | 中国计量大学 | Garbage classification method based on classification and detection joint judgment |
| CN113793347A (en) * | 2021-09-18 | 2021-12-14 | 福建师范大学 | Brain tumor MR image segmentation method based on local-global adaptive information learning |
| CN114005073A (en) * | 2021-12-24 | 2022-02-01 | 东莞理工学院 | Upper limb mirror image rehabilitation training and recognition method and device |
| CN114299304A (en) * | 2021-12-15 | 2022-04-08 | 腾讯科技（深圳）有限公司 | Image processing method and related equipment |
| CN116342859A (en) * | 2023-05-30 | 2023-06-27 | 安徽医科大学第一附属医院 | Method and system for identifying lung tumor area based on imaging features |
| CN116385809A (en) * | 2023-06-05 | 2023-07-04 | 山东第一医科大学附属省立医院(山东省立医院) | MRI brain tumor classification method and system based on semi-supervised learning |
| CN116523914A (en) * | 2023-07-03 | 2023-08-01 | 智慧眼科技股份有限公司 | Aneurysm classification recognition device, method, equipment and storage medium |
| CN116524339A (en) * | 2023-07-05 | 2023-08-01 | 宁德时代新能源科技股份有限公司 | Object detection method, apparatus, computer device, storage medium, and program product |
| CN116740714A (en) * | 2023-06-12 | 2023-09-12 | 北京长木谷医疗科技股份有限公司 | Intelligent self-labeling method and device for hip joint diseases based on unsupervised learning |
| CN117152138A (en) * | 2023-10-30 | 2023-12-01 | 陕西惠宾电子科技有限公司 | Medical image tumor target detection method based on unsupervised learning |
| CN117352120A (en) * | 2023-06-05 | 2024-01-05 | 北京长木谷医疗科技股份有限公司 | GPT-based intelligent self-generation method, device and equipment for knee joint lesion diagnosis |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017207138A1 (en) * | 2016-05-31 | 2017-12-07 | Siemens Healthcare Gmbh | Method of training a deep neural network |
| CN111080596A (en) * | 2019-12-11 | 2020-04-28 | 浙江工业大学 | Auxiliary screening method and system for pneumoconiosis fusing local shadows and global features |
-
2020
- 2020-08-10 CN CN202010794964.3A patent/CN112102237A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017207138A1 (en) * | 2016-05-31 | 2017-12-07 | Siemens Healthcare Gmbh | Method of training a deep neural network |
| CN111080596A (en) * | 2019-12-11 | 2020-04-28 | 浙江工业大学 | Auxiliary screening method and system for pneumoconiosis fusing local shadows and global features |
Non-Patent Citations (3)
| Title |
|---|
| 杨健 等: "基于集成DBN的肺部肿瘤计算机辅助诊断模型研究", 《现代电子技术》 * |
| 贾文娟: "基于深度学习的MRI病脑图像多分类方法研究", 《中国优秀硕士学位论文全文数据库》 * |
| 韩光辉 等: "肺部CT图像病变区域检测方法", 《自动化学报》 * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112668492B (en) * | 2020-12-30 | 2023-06-20 | 中山大学 | Behavior recognition method for self-supervision learning and skeleton information |
| CN112668492A (en) * | 2020-12-30 | 2021-04-16 | 中山大学 | Behavior identification method for self-supervised learning and skeletal information |
| CN113160243A (en) * | 2021-03-24 | 2021-07-23 | 联想(北京)有限公司 | Image segmentation method and electronic equipment |
| CN113298065A (en) * | 2021-05-13 | 2021-08-24 | 杭州电子科技大学 | Eye melanoma recognition method based on self-supervision learning |
| CN113298065B (en) * | 2021-05-13 | 2024-06-11 | 杭州电子科技大学 | Eye melanin tumor identification method based on self-supervision learning |
| CN113657143A (en) * | 2021-06-25 | 2021-11-16 | 中国计量大学 | Garbage classification method based on classification and detection joint judgment |
| CN113657143B (en) * | 2021-06-25 | 2023-06-23 | 中国计量大学 | Garbage classification method based on classification and detection combined judgment |
| CN113470828A (en) * | 2021-06-30 | 2021-10-01 | 上海商汤智能科技有限公司 | Classification method and device, electronic equipment and storage medium |
| CN113793347B (en) * | 2021-09-18 | 2023-05-09 | 福建师范大学 | Brain tumor MR image segmentation method based on local-global self-adaptive information learning |
| CN113793347A (en) * | 2021-09-18 | 2021-12-14 | 福建师范大学 | Brain tumor MR image segmentation method based on local-global adaptive information learning |
| CN114299304A (en) * | 2021-12-15 | 2022-04-08 | 腾讯科技（深圳）有限公司 | Image processing method and related equipment |
| CN114299304B (en) * | 2021-12-15 | 2024-04-12 | 腾讯科技（深圳）有限公司 | Image processing method and related equipment |
| CN114005073A (en) * | 2021-12-24 | 2022-02-01 | 东莞理工学院 | Upper limb mirror image rehabilitation training and recognition method and device |
| CN116342859A (en) * | 2023-05-30 | 2023-06-27 | 安徽医科大学第一附属医院 | Method and system for identifying lung tumor area based on imaging features |
| CN116342859B (en) * | 2023-05-30 | 2023-08-18 | 安徽医科大学第一附属医院 | Method and system for identifying lung tumor area based on imaging features |
| CN116385809A (en) * | 2023-06-05 | 2023-07-04 | 山东第一医科大学附属省立医院(山东省立医院) | MRI brain tumor classification method and system based on semi-supervised learning |
| CN117352120A (en) * | 2023-06-05 | 2024-01-05 | 北京长木谷医疗科技股份有限公司 | GPT-based intelligent self-generation method, device and equipment for knee joint lesion diagnosis |
| CN117352120B (en) * | 2023-06-05 | 2024-06-11 | 北京长木谷医疗科技股份有限公司 | GPT-based intelligent self-generation method, device and equipment for knee joint lesion diagnosis |
| CN116740714A (en) * | 2023-06-12 | 2023-09-12 | 北京长木谷医疗科技股份有限公司 | Intelligent self-labeling method and device for hip joint diseases based on unsupervised learning |
| CN116740714B (en) * | 2023-06-12 | 2024-02-09 | 北京长木谷医疗科技股份有限公司 | Intelligent self-labeling method and device for hip joint diseases based on unsupervised learning |
| CN116523914B (en) * | 2023-07-03 | 2023-09-19 | 智慧眼科技股份有限公司 | Aneurysm classification recognition device, method, equipment and storage medium |
| CN116523914A (en) * | 2023-07-03 | 2023-08-01 | 智慧眼科技股份有限公司 | Aneurysm classification recognition device, method, equipment and storage medium |
| CN116524339B (en) * | 2023-07-05 | 2023-10-13 | 宁德时代新能源科技股份有限公司 | Object detection method, apparatus, computer device, storage medium, and program product |
| CN116524339A (en) * | 2023-07-05 | 2023-08-01 | 宁德时代新能源科技股份有限公司 | Object detection method, apparatus, computer device, storage medium, and program product |
| CN117152138A (en) * | 2023-10-30 | 2023-12-01 | 陕西惠宾电子科技有限公司 | Medical image tumor target detection method based on unsupervised learning |
| CN117152138B (en) * | 2023-10-30 | 2024-01-16 | 陕西惠宾电子科技有限公司 | Medical image tumor target detection method based on unsupervised learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112102237A (en) | Brain tumor recognition model training method and device based on semi-supervised learning | |
| CN110689038B (en) | Training method and device for neural network model and medical image processing system | |
| CN110111313B (en) | Medical image detection method based on deep learning and related equipment | |
| US11610313B2 (en) | Systems and methods for generating normative imaging data for medical image processing using deep learning | |
| CN109741346B (en) | Region-of-interest extraction method, device, equipment and storage medium | |
| US11593943B2 (en) | RECIST assessment of tumour progression | |
| US10810735B2 (en) | Method and apparatus for analyzing medical image | |
| CN112102266B (en) | Attention mechanism-based cerebral infarction medical image classification model training method | |
| Zhou et al. | Salient region detection via integrating diffusion-based compactness and local contrast | |
| CN107480677B (en) | Method and device for identifying interest region in three-dimensional CT image | |
| CN110796199B (en) | Image processing method and device and electronic medical equipment | |
| Viji et al. | RETRACTED ARTICLE: An improved approach for automatic spine canal segmentation using probabilistic boosting tree (PBT) with fuzzy support vector machine | |
| CN110796659B (en) | Target detection result identification method, device, equipment and storage medium | |
| WO2019128460A1 (en) | Image significance detection method and device | |
| CN111192660B (en) | Image report analysis method, device and computer storage medium | |
| CN113743463B (en) | Tumor benign and malignant recognition method and system based on image data and deep learning | |
| Ma et al. | Combining population and patient-specific characteristics for prostate segmentation on 3D CT images | |
| CN114332132A (en) | Image segmentation method and device and computer equipment | |
| CN113160199B (en) | Image recognition method and device, computer equipment and storage medium | |
| EP4168931A1 (en) | Domain aware medical image classifier interpretation by counterfactual impact analysis | |
| CN116091427A (en) | Dermatosis classification device, classification method and storage medium | |
| CN111598144B (en) | Training method and device for image recognition model | |
| CN113222989A (en) | Image grading method and device, storage medium and electronic equipment | |
| CN113962957A (en) | Medical image processing method, bone image processing method, device and equipment | |
| Judice et al. | A novel assessment of various bio-imaging methods for lung tumor detection and treatment by using 4-D and 2-D CT 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201218 |
|
| RJ01 | Rejection of invention patent application after publication |