CN116310330A - Novel lung tumor segmentation method and device for federal semi-supervised learning - Google Patents
Novel lung tumor segmentation method and device for federal semi-supervised learning Download PDFInfo
- Publication number
- CN116310330A CN116310330A CN202310205430.6A CN202310205430A CN116310330A CN 116310330 A CN116310330 A CN 116310330A CN 202310205430 A CN202310205430 A CN 202310205430A CN 116310330 A CN116310330 A CN 116310330A
- Authority
- CN
- China
- Prior art keywords
- supervised
- data
- client
- unsupervised
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000011218 segmentation Effects 0.000 title claims abstract description 82
- 208000037841 lung tumor Diseases 0.000 title claims abstract description 61
- 208000020816 lung neoplasm Diseases 0.000 title claims abstract description 60
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000012549 training Methods 0.000 claims abstract description 75
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000012795 verification Methods 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 16
- 210000004072 lung Anatomy 0.000 claims description 14
- 206010028980 Neoplasm Diseases 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 11
- 230000002685 pulmonary effect Effects 0.000 claims description 10
- 230000002776 aggregation Effects 0.000 claims description 7
- 238000004220 aggregation Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 7
- 238000012952 Resampling Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 206010035664 Pneumonia Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000002599 functional magnetic resonance imaging Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000002595 magnetic resonance imaging Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000153 supplemental effect Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30061—Lung
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30096—Tumor; Lesion
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a novel lung tumor segmentation method and device for federal semi-supervised learning, wherein the method comprises the following steps: preprocessing lung tumor data to obtain preprocessed data, and dividing the preprocessed data into a supervised data set and an unsupervised data set; training a 2D U-Net segmentation model based on a federal semi-supervised learning framework by utilizing the preprocessed data to obtain a trained 2D U-Net segmentation model; the training comprises supervised client training and unsupervised client training; and inputting the lung tumor data to be segmented into the trained 2D U-Net segmentation model, and outputting a lung tumor segmentation result. The invention can relieve the problem of data heterogeneity of different clients through the federal semi-supervised learning framework so as to reduce the offset of the model.
Description
Technical Field
The invention belongs to the technical field of machine learning, and particularly relates to a novel lung tumor segmentation method and device for federal semi-supervised learning.
Background
Federal learning is an advanced distributed learning concept that utilizes datasets across multiple institutions without any explicit dataset or sharing. Although Federal Learning (FL) was originally designed for mobile edge devices, it has attracted increasing attention in the medical field due to its ability to protect patient information privacy. FL is independent of the type of input data. It is capable of analyzing various medical data patterns, from free text clinical reports to high-dimensional medical images. The FL is employed to train the predictive model and solve the support vector machine problem for analyzing Electronic Health Record (EHR) data. FL is applied to wearable medical treatment using personalized machine learning models. Some researchers have built a FL framework for multi-site fMRI classification to protect privacy. Recently, FL has been successfully applied to multi-institution brain MRI to perform tumor segmentation through deep neural networks and to improve privacy protection of patient information.
Semi-supervised learning utilizes available information of unlabeled data and supervision of the labeled data to improve the effectiveness and versatility of the machine learning model. In computer vision, semi-supervised learning has investigated various applications (e.g., image recognition) from different perspectives. To exploit unlabeled data, coherency constraints are studied to mitigate gaps in and between the labeled and unlabeled data domains. One trend is the framework of teacher-student models that can make good use of the consistency constraints between labeled and unlabeled data models. Another similar framework, "noisy students," when trained in conjunction with large amounts of unlabeled data, achieved the most advanced performance on ImageNet (the name of one dataset) classification. Meanwhile, consistency-based model regularization may be achieved by model prediction using unlabeled data, as well as data enhancement or preprocessing. Another trend is to design additional supervision tasks for unlabeled data, such as solving puzzle challenges, predicting rotation angles. Alternatively, co-training has been applied to semi-supervised image recognition, where different models are trained from different "views" in order to learn supplemental information from the data. With the disclosure of some medical datasets, semi-supervised learning techniques for medical image segmentation are also increasing.
Federal semi-supervised learning presents challenges and complexities to how to utilize unlabeled data in a distributed learning environment. Research literature has found that there is currently very limited research on federal semi-supervised learning of unlabeled medical imaging data. Document 1 proposes a framework of federal semi-supervised learning for segmentation of the pneumonia area. The method comprises the following basic steps: 1) Three-dimensional lung data is converted to the most common resolution of the chest CT dataset 0.8mm x 0.8mm x 5.0mm; 2) Clipping the CT intensity value to be in the range of-1000 to 0; 3) Normalizing the data to the [0,1] interval; 4) Clipping the image size of the lung data to an input size suitable for the network model; 5) The processed data is input into the federal semi-supervised learning framework for iterative training. The above is a general procedure for image segmentation for federal semi-supervised learning.
Disclosure of Invention
The invention aims to overcome the defects and shortcomings of the prior art, and provides a novel lung tumor segmentation method and device for federal semi-supervised learning, which can well relieve potential field gap between supervised and unsupervised clients through a federal semi-supervised learning framework.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
the invention provides a novel lung tumor segmentation method and device for federal semi-supervised learning, which comprises the following steps:
the invention provides a novel lung tumor segmentation method for federal semi-supervised learning, which comprises the following steps:
cutting, resampling and normalizing the lung tumor data to obtain preprocessed data, and dividing the preprocessed data into a supervised data set and an unsupervised data set;
training a 2D U-Net segmentation model based on a federal semi-supervised learning framework by utilizing the preprocessed data to obtain a trained 2D U-Net segmentation model; the training comprises supervised client training and unsupervised client training;
and inputting the lung tumor data to be segmented into the trained 2D U-Net segmentation model, and outputting a lung tumor segmentation result.
As a preferable technical scheme, the cutting, resampling and normalizing pretreatment are performed on the lung tumor data, specifically:
firstly, lung tumor data are segmented out of a lung through a segmentation model, and then the lung tumor data are cut to the size of a lung region on an original image;
secondly, because the resolution of each client data is different, the resolution of each client data needs to be resampled to the preset resolution;
finally, the normalization is carried out, the window width and window level suitable for the lung are considered, the intensity value of the lung data is firstly cut to be between [ -250,400], and then the normalization is carried out by using the Z-Score, so that the data is prevented from being compressed during the normalization.
As a preferable technical scheme, the supervised data set is equally proportioned into a supervised training set, a supervised test set and a supervised verification set; the unsupervised data set is divided into an unsupervised training set, an unsupervised test set and an unsupervised verification set.
As a preferable technical scheme, the specific process of the supervised client training is as follows:
first, a training set of the supervised client generates a predictive label Y through a first basic model H (x; θ) 1 The method comprises the steps of carrying out a first treatment on the surface of the Will predict tag Y 1 Performing loss calculation with the label Y;
then, performing verification on the supervised verification set, wherein each supervised client generates a verification result P;
secondly, the supervised client uploads gradient parameters to the serverThe unsupervised client uploads gradient parameters to the server side +.>
Server-side aggregation gradientAnd gradient->And then sending the updated gradient to each supervised client, wherein the aggregation policy is as follows:
assume that the result of the verification set of the supervised client at the t-th iteration training is P i (i∈C s ) At time t+1, the supervised client gradient update strategy is as follows:
wherein C is s Representing a supervised client, C u Representing an unsupervised client, w s And w u Initial weights for supervised and unsupervised clients respectively,representing dynamically updated weights, +.>And->Gradient parameters of supervised and unsupervised clients are shown, respectively, +.>The formula is:
finally, the gradient parameters of the best accuracy obtained on the supervised validation set are saved and tested on the supervised test set.
As a preferable technical solution, the process of the unsupervised client training is as follows:
first, the training set of the unsupervised client generates a pseudo tag through the second basic model H (u; θ)
Training set data of an unsupervised client side is subjected to overturn, rotation and Gaussian noise data enhancement and then passes through a second basic model H (u; theta) again, and a prediction label is generated
Pseudo tag generated by twice passing training set data of unsupervised client through second basic modelPredictive tagsThe consistency loss is made, so that the predicted value tends to be consistent;
assume that the gradient generated by the unsupervised client during t times of iterative training is set asWill->Uploading to a server, and after the server aggregates gradient parameters, returning a gradient global gradient of theta U ;
Defining an index moving average MODEL EMA_MODEL for each unsupervised client, wherein the index moving average MODEL EMA_MODEL is the same as the basic MODEL, but the gradient parameter update of the index moving average MODEL EMA_MODEL is different from that of the second basic MODEL, and the gradient parameter of the index moving average MODEL EMA_MODEL is the same as that of the basic MODEL when t+1 times of iterative trainingThe updated formula is:
gradient parameters of EMA_MODELIs updated with global gradient θ U And local gradient of unsupervised client +.>Wherein β represents a threshold of the weighted summation;
finally, the unsupervised client uses the exponential moving average MODEL ema_model to test on the unsupervised test set.
The invention also provides a novel lung tumor segmentation system for federal semi-supervised learning, which is applied to the novel lung tumor segmentation method for federal semi-supervised learning and comprises a preprocessing module, a model training module and a segmentation module;
the pretreatment module is used for carrying out pretreatment on lung tumor data to obtain pretreated data, and dividing the pretreated data into a supervised data set and an unsupervised data set;
the model training module is used for training a 2DU-Net segmentation model based on a federal semi-supervised learning framework by utilizing the preprocessed data to obtain a trained 2D U-Net segmentation model; the training comprises supervised client training and unsupervised client training;
the segmentation module is used for inputting the lung tumor data to be segmented into the trained 2D U-Net segmentation model and outputting a lung tumor segmentation result.
A further aspect of the present invention provides an electronic device, characterized in that the electronic device includes:
at least one processor; the method comprises the steps of,
a memory communicatively coupled to the at least one processor; wherein,,
the memory stores computer program instructions executable by the at least one processor to enable the at least one processor to perform the novel federal semi-supervised learning lung tumor segmentation method.
In yet another aspect, the present invention provides a computer readable storage medium storing a program which, when executed by a processor, implements the novel federal semi-supervised learning lung tumor segmentation method.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. the method can relieve the problem of data heterogeneity of different clients through the federal semi-supervised learning framework so as to reduce the offset of the model.
2. The invention is dynamic aggregation gradient parameter at the supervised client, and effectively utilizes the quality information of the data of each client.
3. According to the invention, an index moving average MODEL EMA_MODEL is introduced into the unsupervised client, and gradient parameters at the previous moment are considered during each iterative training, so that each unsupervised client can pay more attention to own data set.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a novel lung tumor segmentation method for Federal semi-supervised learning according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a supervised client training process according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an unsupervised client training process according to an embodiment of the present invention;
FIG. 4 is a block diagram of a novel Federal semi-supervised learning pulmonary tumor segmentation system according to an embodiment of the present invention;
fig. 5 is a block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to enable those skilled in the art to better understand the present application, the following description will make clear and complete descriptions of the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. It will be apparent that the described embodiments are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.
Reference in the specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.
Referring to fig. 1, in one embodiment of the present application, a novel pulmonary tumor segmentation method for federal semi-supervised learning is provided, comprising the steps of:
s1, cutting, resampling and normalizing pretreatment are carried out on lung tumor data to obtain pretreated data, and the pretreated data are divided into a supervised data set and an unsupervised data set.
The pretreatment process comprises the following steps:
s11, cutting; firstly, lung tumor data are segmented out of a lung through a segmentation model, and then the lung tumor data are cut to the size of a lung region on an original image;
s12, resampling; the resolution of each client data is different, and the resolution of each client data needs to be resampled to 1mm x 1mm;
s13, standardization; to consider a window width suitable for the lungs, the intensity values of the lung data are first truncated to between [ -250,400], and then normalized with the Z-Score to prevent compression during normalization of the data.
Further, the supervised data set is equally proportioned into a supervised training set, a supervised test set and a supervised verification set; the unsupervised data set is divided into an unsupervised training set, an unsupervised test set and an unsupervised verification set.
S2, training a 2D U-Net segmentation model based on a federal semi-supervised learning framework by utilizing the preprocessed data to obtain a trained 2D U-Net segmentation model; the training comprises supervised client training and unsupervised client training; the 2DU-Net is an end-to-end segmentation network, and can effectively segment out tumor areas.
S21, referring to FIG. 2, the training of the supervised client specifically comprises the following steps:
s211, firstly, generating a prediction label Y by a training set of the supervised client through a first basic model H (x; theta) 1 The method comprises the steps of carrying out a first treatment on the surface of the Will predict tag Y 1 Performing loss calculation with the label Y;
s212, performing verification on the supervised verification set, wherein each supervised client generates a verification result P;
s213, the supervised client uploads the gradient parameters to the serverThe unsupervised client uploads gradient parameters to the server side +.>
Server-side aggregation gradientAnd gradient->And then sending the updated gradient to each supervised client, wherein the aggregation policy is as follows:
s214, assuming that the result of the verification set of the supervised client is P in the t-th iterative training i (i∈C s ) At time t+ 1, the supervised client gradient update strategy is as follows:
wherein C is s Representing a supervised client, C u Representing an unsupervised client, w s And w u Initial weights for supervised and unsupervised clients respectively,representing dynamically updated weights, +.>And->Gradient parameters of supervised and unsupervised clients are shown, respectively, +.>The formula is:
and S215, finally, saving the gradient parameters with the best precision obtained on the supervised verification set, and testing on the supervised test set.
S22, referring to FIG. 3, the process of the unsupervised client training is as follows:
s221, firstly, generating a pseudo tag by the training set of the unsupervised client through a second basic model H (u; theta)
S222, training set data of the unsupervised client side is subjected to data enhancement such as overturn, rotation and Gaussian noise and then is subjected to a second basic model H (u; theta) again, and a prediction label is generated
S223, generating pseudo labels for training set data of the unsupervised client through the second basic model twiceAnd predictive tag->Make consistency loss, ensureEnsuring that predicted values tend to be consistent;
s224, assuming that the gradient generated by the unsupervised client during t times of iterative training is set asWill->Uploading to a server, and after the server aggregates gradient parameters, returning a gradient global gradient of theta U ;
S225, defining an index moving average MODEL EMA_MODEL by each unsupervised client, wherein the index moving average MODEL EMA_MODEL is the same as the basic MODEL, but the gradient parameter update of the EMA_MODEL is different from that of the second basic MODEL, and the gradient parameter of the index moving average MODEL EMA_MODEL is the same as that of the basic MODEL when t+1 times of iterative trainingThe updated formula is:
gradient parameters of EMA_MODELIs updated with global gradient θ U And local gradient of unsupervised client +.>Wherein β represents a threshold of the weighted summation;
and S226, finally, the unsupervised client uses the index moving average MODEL EMA_MODEL to test on the unsupervised test set.
S3, inputting the lung tumor data to be segmented into the trained 2D U-Net segmentation model, and outputting a lung tumor segmentation result.
In the invention, the supervised client side dynamically distributes weights, and potentially utilizes the quality of the data of each client side to aggregate gradient parameters; the unsupervised client updates the gradient parameters by reintroducing an exponential moving average MODEL ema_model.
It should be noted that, for the sake of simplicity of description, the foregoing method embodiments are all expressed as a series of combinations of actions, but it should be understood by those skilled in the art that the present invention is not limited by the order of actions described, as some steps may be performed in other order or simultaneously in accordance with the present invention.
Based on the same ideas the novel lung tumor segmentation method of the federal semi-supervised learning in the embodiment, the invention also provides a novel lung tumor segmentation system of the federal semi-supervised learning, which can be used for executing the novel lung tumor segmentation method of the federal semi-supervised learning. For ease of illustration, only those portions relevant to embodiments of the present invention are shown in the structural schematic diagram of one embodiment of a novel federal semi-supervised learning pulmonary tumor segmentation system, and it will be understood by those skilled in the art that the illustrated structure is not limiting of the apparatus, and may include more or fewer components than illustrated, or may combine certain components, or a different arrangement of components.
Referring to fig. 4, in another embodiment of the present application, a novel federal semi-supervised learning pulmonary tumor segmentation system 100 is provided, comprising a preprocessing module 101, a model training module 102, and a segmentation module 103;
the preprocessing module 101 is configured to preprocess lung tumor data to obtain preprocessed data, and divide the preprocessed data into a supervised data set and an unsupervised data set;
the model training module 102 is configured to train a 2DU-Net segmentation model based on a federal semi-supervised learning framework by using the preprocessed data to obtain a trained 2D U-Net segmentation model; the training comprises supervised client training and unsupervised client training;
the segmentation module 103 is configured to input the lung tumor data to be segmented into the trained 2D U-Net segmentation model, and output a lung tumor segmentation result.
It should be noted that, the novel federal semi-supervised learning lung tumor segmentation system and the novel federal semi-supervised learning lung tumor segmentation method according to the present invention are in one-to-one correspondence, and technical features and beneficial effects described in the embodiments of the novel federal semi-supervised learning lung tumor segmentation method are applicable to the embodiments of the novel federal semi-supervised learning lung tumor segmentation system, and specific content can be seen in the description of the embodiments of the novel federal semi-supervised learning lung tumor segmentation system, which is not repeated herein, thereby making a statement.
In addition, in the implementation of the novel federal semi-supervised learning lung tumor segmentation system according to the above embodiment, the logic division of each program module is merely illustrative, and in practical application, the functional allocation may be performed by different program modules according to needs, for example, in view of configuration requirements of corresponding hardware or convenience of implementation of software, that is, the internal structure of the novel federal semi-supervised learning lung tumor segmentation system is divided into different program modules to perform all or part of the functions described above.
Referring to fig. 5, in one embodiment, an electronic device for implementing a method for lung tumor segmentation for novel federal semi-supervised learning is provided, where the electronic device 200 may include a first processor 201, a first memory 202, and a bus, and may further include a computer program stored in the first memory 202 and executable on the first processor 201, such as the lung tumor segmentation program 203 for novel federal semi-supervised learning.
The first memory 202 includes at least one type of readable storage medium, which includes flash memory, a mobile hard disk, a multimedia card, a card memory (e.g., SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. The first memory 202 may in some embodiments be an internal storage unit of the electronic device 200, such as a mobile hard disk of the electronic device 200. The first memory 202 may also be an external storage device of the electronic device 200 in other embodiments, such as a plug-in mobile hard disk, a Smart Media Card (SMC), a secure digital (SecureDigital, SD) Card, a Flash memory Card (Flash Card), etc. that are provided on the electronic device 200. Further, the first memory 202 may also include both an internal memory unit and an external memory device of the electronic device 200. The first memory 202 may be used to store not only application software installed in the electronic device 200 and various data, such as codes of a novel federal semi-supervised learning lung tumor segmentation program 203, but also temporarily store data that has been output or is to be output.
The first processor 201 may be formed by an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or may be formed by a plurality of integrated circuits packaged with the same function or different functions, including one or more central processing units (Central Processing unit, CPU), a microprocessor, a digital processing chip, a graphics processor, a combination of various control chips, and so on. The first processor 201 is a Control Unit (Control Unit) of the electronic device, connects various components of the entire electronic device using various interfaces and lines, and executes various functions of the electronic device 200 and processes data by running or executing programs or modules stored in the first memory 202 and calling data stored in the first memory 202.
Fig. 5 shows only an electronic device with components, and it will be understood by those skilled in the art that the structure shown in fig. 5 is not limiting of the electronic device 200 and may include fewer or more components than shown, or may combine certain components, or a different arrangement of components.
The novel federal semi-supervised learning lung tumor segmentation program 203 stored by the first memory 202 in the electronic device 200 is a combination of instructions that, when executed in the first processor 201, may implement:
preprocessing lung tumor data to obtain preprocessed data, and dividing the preprocessed data into a supervised data set and an unsupervised data set;
training a 2D U-Net segmentation model based on a federal semi-supervised learning framework by utilizing the preprocessed data to obtain a trained 2D U-Net segmentation model; the training comprises supervised client training and unsupervised client training;
and inputting the lung tumor data to be segmented into the trained 2D U-Net segmentation model, and outputting a lung tumor segmentation result.
Further, the modules/units integrated with the electronic device 200 may be stored in a non-volatile computer readable storage medium if implemented in the form of software functional units and sold or used as a stand-alone product. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM).
Those skilled in the art will appreciate that all or part of the processes in the methods of the above embodiments may be implemented by a computer program for instructing relevant hardware, where the program may be stored in a non-volatile computer readable storage medium, and where the program, when executed, may include processes in the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above examples, and any other changes, modifications, substitutions, combinations, and simplifications that do not depart from the spirit and principle of the present invention should be made in the equivalent manner, and the embodiments are included in the protection scope of the present invention.
Claims (8)
1. The novel lung tumor segmentation method for federal semi-supervised learning is characterized by comprising the following steps of:
cutting, resampling and normalizing the lung tumor data to obtain preprocessed data, and dividing the preprocessed data into a supervised data set and an unsupervised data set;
training a 2D U-Net segmentation model based on a federal semi-supervised learning framework by utilizing the preprocessed data to obtain a trained 2D U-Net segmentation model; the training comprises supervised client training and unsupervised client training;
and inputting the lung tumor data to be segmented into the trained 2D U-Net segmentation model, and outputting a lung tumor segmentation result.
2. The novel federal semi-supervised learning pulmonary tumor segmentation method according to claim 1, wherein the clipping, resampling and normalizing pretreatment of the pulmonary tumor data comprises:
firstly, lung tumor data are segmented out of a lung through a segmentation model, and then the lung tumor data are cut to the size of a lung region on an original image;
secondly, because the resolution of each client data is different, the resolution of each client data needs to be resampled to the preset resolution;
finally, the normalization is carried out, the window width and window level suitable for the lung are considered, the intensity value of the lung data is firstly cut to be between [ -250,400], and then the normalization is carried out by using the Z-Score, so that the data is prevented from being compressed during the normalization.
3. The method for segmenting lung tumors for novel federal semi-supervised learning as set forth in claim 1, wherein said supervised data set is equally proportioned into a supervised training set, a supervised testing set and a supervised validation set; the unsupervised data set is divided into an unsupervised training set, an unsupervised test set and an unsupervised verification set.
4. The novel pulmonary tumor segmentation method for federal semi-supervised learning as set forth in claim 1, wherein the specific process of the supervised client training is:
first, a training set of the supervised client generates a predictive label Y through a first basic model H (x; θ) 1 The method comprises the steps of carrying out a first treatment on the surface of the Will predict tag Y 1 Performing loss calculation with the label Y;
then, performing verification on the supervised verification set, wherein each supervised client generates a verification result P;
secondly, the supervised client uploads gradient parameters to the serverUnsupervised client uploads gradient parameters to serverServer-side aggregation gradient->And gradient->And then sending the updated gradient to each supervised client, wherein the aggregation policy is as follows:
assume that the result of the verification set of the supervised client at the t-th iteration training is P i (i∈C s ) At time t+1, the supervised client gradient update strategy is as follows:
wherein C is s Representing a supervised client, C u Representing an unsupervised client, w s And w u Initial weights for supervised and unsupervised clients respectively,representing dynamically updated weights, +.>And->Gradient parameters of supervised and unsupervised clients are shown, respectively, +.>The formula is:
finally, the gradient parameters of the best accuracy obtained on the supervised validation set are saved and tested on the supervised test set.
5. The novel federal semi-supervised learning pulmonary tumor segmentation method of claim 1, wherein the unsupervised client training process is:
first, the training set of the unsupervised client generates a pseudo tag through the second basic model H (u; θ)
Training set data of an unsupervised client side is subjected to overturn, rotation and Gaussian noise data enhancement and then passes through a second basic model H (u; theta) again, and a prediction label is generated
Pseudo tag generated by twice passing training set data of unsupervised client through second basic modelAnd predictive tag->The consistency loss is made, so that the predicted value tends to be consistent;
assume that the gradient generated by the unsupervised client during t times of iterative training is set asWill->Uploading to a server, and after the server aggregates gradient parameters, returning a gradient global gradient of theta U ;
Defining an index moving average MODEL EMA_MODEL for each unsupervised client, wherein the index moving average MODEL EMA_MODEL is the same as the basic MODEL, but the gradient parameter update of the index moving average MODEL EMA_MODEL is different from that of the second basic MODEL, and the gradient parameter of the index moving average MODEL EMA_MODEL is the same as that of the basic MODEL when t+1 times of iterative trainingThe updated formula is:
gradient parameters of EMA_MODELIs updated with global gradient θ U And local gradient of unsupervised client +.>Wherein β represents a threshold of the weighted summation;
finally, the unsupervised client uses the exponential moving average MODEL ema_model to test on the unsupervised test set.
6. A novel pulmonary tumor segmentation system for federal semi-supervised learning, which is characterized by being applied to the novel pulmonary tumor segmentation method for federal semi-supervised learning as set forth in any one of claims 1-5, and comprising a preprocessing module, a model training module and a segmentation module;
the pretreatment module is used for carrying out pretreatment on lung tumor data to obtain pretreated data, and dividing the pretreated data into a supervised data set and an unsupervised data set;
the model training module is used for training a 2DU-Net segmentation model based on a federal semi-supervised learning framework by utilizing the preprocessed data to obtain a trained 2D U-Net segmentation model; the training comprises supervised client training and unsupervised client training;
the segmentation module is used for inputting the lung tumor data to be segmented into the trained 2D U-Net segmentation model and outputting a lung tumor segmentation result.
7. An electronic device, the electronic device comprising:
at least one processor; the method comprises the steps of,
a memory communicatively coupled to the at least one processor; wherein,,
the memory stores computer program instructions executable by the at least one processor to enable the at least one processor to perform a novel federal semi-supervised learning lung tumor segmentation method as set forth in any one of claims 1-5.
8. A computer readable storage medium storing a program which, when executed by a processor, implements a novel federal semi-supervised learning lung tumor segmentation method according to any of claims 1-5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310205430.6A CN116310330A (en) | 2023-03-06 | 2023-03-06 | Novel lung tumor segmentation method and device for federal semi-supervised learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310205430.6A CN116310330A (en) | 2023-03-06 | 2023-03-06 | Novel lung tumor segmentation method and device for federal semi-supervised learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116310330A true CN116310330A (en) | 2023-06-23 |
Family
ID=86828177
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310205430.6A Pending CN116310330A (en) | 2023-03-06 | 2023-03-06 | Novel lung tumor segmentation method and device for federal semi-supervised learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116310330A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117011563A (en) * | 2023-08-04 | 2023-11-07 | 山东建筑大学 | Road damage inspection cross-domain detection method and system based on semi-supervised federal learning |
-
2023
- 2023-03-06 CN CN202310205430.6A patent/CN116310330A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117011563A (en) * | 2023-08-04 | 2023-11-07 | 山东建筑大学 | Road damage inspection cross-domain detection method and system based on semi-supervised federal learning |
CN117011563B (en) * | 2023-08-04 | 2024-03-01 | 山东建筑大学 | Road damage inspection cross-domain detection method and system based on semi-supervised federal learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Payette et al. | An automatic multi-tissue human fetal brain segmentation benchmark using the fetal tissue annotation dataset | |
Musallam et al. | A new convolutional neural network architecture for automatic detection of brain tumors in magnetic resonance imaging images | |
Nemoto et al. | Efficacy evaluation of 2D, 3D U-Net semantic segmentation and atlas-based segmentation of normal lungs excluding the trachea and main bronchi | |
Lee et al. | Applying artificial intelligence to longitudinal imaging analysis of vestibular schwannoma following radiosurgery | |
CN116310330A (en) | Novel lung tumor segmentation method and device for federal semi-supervised learning | |
EP4095758A1 (en) | Training large-scale vision transformer neural networks | |
WO2021179702A1 (en) | Method and apparatus for segmenting bone fragments from three-dimensional image, computer device and storage medium | |
Karthiga et al. | Machine learning based diagnosis of Alzheimer’s disease | |
Mabu et al. | Semi-supervised CycleGAN for domain transformation of chest CT images and its application to opacity classification of diffuse lung diseases | |
CN117058307A (en) | Method, system, equipment and storage medium for generating heart three-dimensional nuclear magnetic resonance image | |
Maier et al. | Predicting stroke lesion and clinical outcome with random forests | |
Zaridis et al. | Region-adaptive magnetic resonance image enhancement for improving CNN-based segmentation of the prostate and prostatic zones | |
Abdikerimova et al. | Detection of chest pathologies using autocorrelation functions | |
Seshadri Ramana et al. | Deep convolution neural networks learned image classification for early cancer detection using lightweight | |
Cai et al. | 3D Medical Image Segmentation with Sparse Annotation via Cross-Teaching Between 3D and 2D Networks | |
Jiang et al. | PET-diffusion: Unsupervised PET enhancement based on the latent diffusion model | |
Hao et al. | Iris segmentation using feature channel optimization for noisy environments | |
Xia et al. | HRNet: A hierarchical recurrent convolution neural network for retinal vessel segmentation | |
DE102022120117A1 (en) | On-device detection of digital objects and generation of object masks | |
Hsieh et al. | Based on DICOM RT structure and multiple loss function deep learning algorithm in organ segmentation of head and neck image | |
Wilms et al. | A kernelized multi-level localization method for flexible shape modeling with few training data | |
Han et al. | Video primal sketch: A unified middle-level representation for video | |
Mejia et al. | Template independent component analysis with spatial priors for accurate subject-level brain network estimation and inference | |
Zhou et al. | Image segmentation of rectal tumor based on improved u-net model with deep learning | |
Gohil et al. | Kidney and kidney tumor segmentation using spatial and channel attention enhanced U-Net |
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 |