CN116664953A - 2.5D pneumonia medical CT image classification device and equipment - Google Patents
2.5D pneumonia medical CT image classification device and equipment Download PDFInfo
- Publication number
- CN116664953A CN116664953A CN202310781182.XA CN202310781182A CN116664953A CN 116664953 A CN116664953 A CN 116664953A CN 202310781182 A CN202310781182 A CN 202310781182A CN 116664953 A CN116664953 A CN 116664953A
- Authority
- CN
- China
- Prior art keywords
- medical
- image
- pneumonia
- lung
- convnext
- 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
- 206010035664 Pneumonia Diseases 0.000 title claims abstract description 132
- 210000004072 lung Anatomy 0.000 claims abstract description 110
- 238000012549 training Methods 0.000 claims abstract description 68
- 238000005457 optimization Methods 0.000 claims abstract description 63
- 238000000034 method Methods 0.000 claims abstract description 36
- 230000008569 process Effects 0.000 claims abstract description 20
- 238000012545 processing Methods 0.000 claims description 24
- 238000012360 testing method Methods 0.000 claims description 22
- 238000007781 pre-processing Methods 0.000 claims description 18
- 238000003745 diagnosis Methods 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 14
- 238000010606 normalization Methods 0.000 claims description 14
- 238000012952 Resampling Methods 0.000 claims description 12
- 201000004813 Bronchopneumonia Diseases 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 208000029523 Interstitial Lung disease Diseases 0.000 claims description 4
- 238000000137 annealing Methods 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 208000024891 symptom Diseases 0.000 claims description 4
- 238000013508 migration Methods 0.000 claims description 3
- 230000005012 migration Effects 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 10
- 238000003860 storage Methods 0.000 description 10
- 238000004891 communication Methods 0.000 description 7
- 238000013135 deep learning Methods 0.000 description 6
- 238000013145 classification model Methods 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 230000003902 lesion Effects 0.000 description 5
- 238000013526 transfer learning Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000007792 addition Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 210000004204 blood vessel Anatomy 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000005520 cutting process Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 208000032923 Lobar pneumonia Diseases 0.000 description 1
- 206010035737 Pneumonia viral Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000003748 differential diagnosis Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 238000000227 grinding Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000008595 infiltration Effects 0.000 description 1
- 238000001764 infiltration Methods 0.000 description 1
- 238000011068 loading method Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 208000023504 respiratory system disease Diseases 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000026676 system process Effects 0.000 description 1
- 208000009421 viral pneumonia Diseases 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4007—Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
-
- 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
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06T2207/10—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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30061—Lung
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The application provides a 2.5D pneumonia medical CT image classification device and equipment, the device includes: the target image slice module is used for acquiring a plurality of continuous slices corresponding to the target 3D lung medical CT image data so as to form target 2.5D lung medical CT image data corresponding to the target 3D lung medical CT image data; the target image classification module is used for inputting target 2.5D lung medical CT image data into a ConvNeXt optimization model adopting a 2D convolution layer, so that the ConvNeXt optimization model correspondingly outputs a pneumonia focus type identification result corresponding to the target 2.5D lung medical CT image data. The application can realize automatic classification of the medical CT images of the pneumonia, can effectively improve the execution efficiency and convenience of the classification process of the CT images of the pneumonia and can improve the reliability and accuracy of the classification result of the CT images of the pneumonia on the basis of reducing the complexity and time consumption of model training.
Description
Technical Field
The application relates to the technical field of medical auxiliary equipment, in particular to a 2.5D pneumonia medical CT image classification device and equipment.
Background
In recent years, research for deep learning has made breakthrough progress in various sub-fields of computer vision. The method learns that the human visual system processes external information by simulating neurons of a human brain region, automatically extracts multi-level features of images and maps the images into a high-level abstract feature space to realize specified tasks. Because of its excellent image feature extraction capability, there is also a great deal of related research in the field of medical image segmentation. Essentially, medical image diagnosis is a problem of computer vision, pneumonia diagnosis is a type commonly found in clinical medical image diagnosis, and pneumonia is a clinically common respiratory disease, and its image appearance is diversified, and the CT image mainly shows single or multiple infiltration, glass grinding and real variegation. However, in clinic, the identification of similarly represented viral pneumonia is difficult, and high workload may also lead to missed diagnosis of early lesions.
The AI assists radiologists to improve the effectiveness of daily differential diagnosis of pneumonia to a certain extent. Deep learning is an important branch in the AI field, and has attracted increasing attention in the medical field as a tool for image classification.
At present, in the existing medical CT image classification method, a deep learning mode may be adopted, and a CT image of the patient 2D may be input into, for example, an AACN-UNet neural network model, so as to obtain a classification result corresponding to the CT image of the patient 2D, but due to the self-characteristics of the medical CT image of pneumonia, the existing medical CT image classification method cannot accurately determine whether a focus is included in the image. For example, in the case of lobular pneumonia, it is not possible to determine whether a white circular hole in an image is a lesion or a lumen of a blood vessel by only a single CT image. If a 3D CT image is input into the 3D image classification network, training of the model is more difficult than that of a 2D convolution network, especially for CT data with multiple slices, the memory occupation becomes large during the training of the network, a large amount of computer resources are required, the parameter number of the model becomes relatively large, and the model is easier to be overfitted for a small amount of medical data.
Therefore, there is a need to design a device that can reduce the complexity of model and training and can also guarantee the accuracy of the classification of the pneumonia CT images on the basis of realizing the automation of the classification of the pneumonia CT images.
Disclosure of Invention
In view of this, embodiments of the present application provide a 2.5D pneumonia medical CT image classification apparatus and device to obviate or ameliorate one or more of the disadvantages of the prior art.
One aspect of the present application provides a 2.5D pneumonia medical CT image classification apparatus, comprising:
the target image slicing module is used for acquiring a plurality of continuous slices corresponding to target 3D lung medical CT image data so as to form target 2.5D lung medical CT image data corresponding to the target 3D lung medical CT image data;
and the target image classification module is used for inputting the target 2.5D lung medical CT image data into a preset ConvNeXt optimization model adopting a 2D convolution layer so that the ConvNeXt optimization model correspondingly outputs a pneumonia focus type identification result corresponding to the target 2.5D lung medical CT image data.
In some embodiments of the application, further comprising: the model training module is connected with the target image classification module;
the model training module is used for training a ConvNeXt optimization model with a 2D convolution layer according to each 2.5D lung medical CT image sample provided with a label, so that the ConvNeXt optimization model is used for correspondingly outputting a pneumonia focus type identification result according to the input 2.5D lung medical CT image data;
Wherein the tag comprises: is used for respectively representing the symptoms of the pneumonia, the lobular pneumonia and the interstitial pneumonia.
In some embodiments of the application, further comprising: the historical image slicing module is connected with the model training module;
the historical image slicing module is used for respectively acquiring a plurality of continuous slices corresponding to each historical 3D lung medical CT image sample in an image enhancement mode based on a preset slice quantity threshold value so as to obtain 2.5D lung medical CT image samples corresponding to each historical 3D lung medical CT image sample.
In some embodiments of the application, further comprising: the historical image preprocessing module is connected with the historical image slicing module;
the history image preprocessing module is used for preprocessing the acquired history 3D lung medical CT image data for the history diagnosis of the pneumonia focus to obtain corresponding history 3D lung medical CT image samples.
In some embodiments of the application, further comprising: the model structure optimization module is connected with the model training module;
the model structure optimization module is used for performing adjustment optimization for training on a 2.5D lung medical CT image sample on a network structure of the ConvNeXt model so as to generate a ConvNeXt optimization model adopting a 2D convolution layer.
In some embodiments of the application, further comprising: the model generalization test module is connected between the target image classification module and the model training module;
the model generalization test module is used for carrying out generalization capability test on the ConvNeXt optimization model obtained through training and selecting a ConvNeXt optimization model of which the corresponding test result meets the preset generalization performance requirement.
In some embodiments of the application, the model training module comprises:
the pre-training unit is used for pre-training the ConvNeXt optimization model by adopting the 2D convolution layer by adopting pre-acquired same-domain historical lung medical CT image data to obtain a pre-trained ConvNeXt optimization model;
the migration learning unit is used for iteratively training the pre-training ConvNeXt optimization model by adopting each 2.5D lung medical CT image sample with a label, and obtaining the ConvNeXt optimization model for correspondingly outputting a pneumonia focus type identification result according to the input 2.5D lung medical CT image data by adopting a cosine annealing learning rate mode in the training process.
In some embodiments of the present application, the history image preprocessing module includes:
the type conversion unit is used for converting the disclosed DICOM sequence data into NIFIT format to obtain each historical 3D medical CT image data for the historical diagnosis of the pneumonia focus;
The clipping unit is used for clipping out the lung areas in each historical 3D medical CT image data so as to obtain corresponding historical 3D lung medical CT image data;
the resampling unit is used for resampling and linear interpolation processing of each historical 3D lung medical CT image data;
the standardized processing unit is used for carrying out standardized processing on each historical 3D lung medical CT image data subjected to resampling and linear interpolation processing, and converting each historical 3D lung medical CT image data subjected to standardized processing into npy format respectively so as to obtain each corresponding historical 3D lung medical CT image sample of each historical 3D medical CT image data.
In some embodiments of the application, the ConvNeXt optimization model includes: the device comprises an input layer, a 2D convolution layer, a layer normalization unit, 3 ConvNeXt blocks, a first downsampling group, a second downsampling group, a third downsampling group, a global average pooling layer, a layer normalization unit, a linear layer and an output layer which are sequentially connected;
the first downsampling group and the third downsampling group comprise downsampling layers and 3 ConvNeXt blocks which are connected in sequence, and the second downsampling group comprises downsampling layers and 9 ConvNeXt blocks which are connected in sequence;
The ConvNeXt block includes: the depth separable 2D convolution layer, the layer normalization unit, the 2D convolution layer, the Gaussian error linear unit, the 2D convolution layer, the layer scaling unit and the path discarding unit are sequentially connected;
the downsampling layer includes: and the layer normalization unit and the 2D convolution layer are sequentially connected.
In a second aspect, the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, where the processor implements the functions of the 2.5D pneumonia medical CT image classification apparatus when executing the computer program.
A third aspect of the present application provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the functions of the 2.5D pneumonia medical CT image classification apparatus.
The application provides a 2.5D pneumonia medical CT image classification device, which is characterized by comprising a target image slice module, a target image processing module and a target image processing module, wherein the target image slice module is used for acquiring a plurality of continuous slices corresponding to target 3D lung medical CT image data so as to form target 2.5D lung medical CT image data corresponding to the target 3D lung medical CT image data; the target image classification module is used for inputting the target 2.5D lung medical CT image data into a preset ConvNeXt optimization model adopting a 2D convolution layer, so that the ConvNeXt optimization model correspondingly outputs a pneumonia focus type identification result corresponding to the target 2.5D lung medical CT image data, automatic classification of the pneumonia medical CT image can be realized, on the basis of reducing model training complexity and time consumption, the execution efficiency and convenience of a pneumonia CT image classification process can be effectively improved, and the reliability and accuracy of a pneumonia CT image classification result can be improved. The deep learning classification model is used for assisting a doctor in classifying pneumonia, so that the range of the pneumonia focus is provided for the doctor, the film reading pressure of the doctor can be effectively reduced, and the diagnosis efficiency and accuracy of the doctor are improved.
Additional advantages, objects, and features of the application will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the application. The objectives and other advantages of the application may be realized and attained by the structure particularly pointed out in the written description and drawings.
It will be appreciated by those skilled in the art that the objects and advantages that can be achieved with the present application are not limited to the above-described specific ones, and that the above and other objects that can be achieved with the present application will be more clearly understood from the following detailed description.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate and together with the description serve to explain the application. The components in the figures are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the application. Corresponding parts in the drawings may be exaggerated, i.e. made larger relative to other parts in an exemplary device actually manufactured according to the present application, for convenience in showing and describing some parts of the present application. In the drawings:
fig. 1 is a schematic diagram of a first structure of a 2.5D pneumonia medical CT image classification apparatus according to an embodiment of the application.
Fig. 2 is a schematic diagram illustrating a first connection relationship between a 2.5D pneumonia medical CT image classification apparatus and other devices according to an embodiment of the present application.
Fig. 3 is a schematic diagram illustrating a second structure of a 2.5D pneumonia medical CT image classification apparatus according to an embodiment of the application.
Fig. 4 is a schematic diagram of a third structure of a 2.5D pneumonia medical CT image classification apparatus according to an embodiment of the application.
Fig. 5 is a schematic diagram of a fourth structure of a 2.5D pneumonia medical CT image classification apparatus according to an embodiment of the application.
Fig. 6 is a schematic diagram illustrating a second connection relationship between the 2.5D pneumonia medical CT image classification apparatus and other devices according to an embodiment of the present application.
Fig. 7 is a schematic diagram illustrating an example of a total network architecture of a ConvNeXt optimization model according to an embodiment of the application.
Fig. 8 is a schematic diagram illustrating an example network architecture of a ConvNeXt block in a ConvNeXt optimization model according to an embodiment of the application.
Fig. 9 is a schematic diagram illustrating an example network architecture of a downsampling layer in a ConvNeXt optimization model according to an embodiment of the application.
Fig. 10 is a flow chart of a 2.5D pneumonia medical CT image classification method based on a ConvNeXt optimization model provided in an application example of the present application.
Fig. 11 is an exemplary logic diagram of the 2.5D pneumonia medical CT image classification method based on the ConvNeXt optimization model according to the present application.
Detailed Description
The present application will be described in further detail with reference to the following embodiments and the accompanying drawings, in order to make the objects, technical solutions and advantages of the present application more apparent. The exemplary embodiments of the present application and the descriptions thereof are used herein to explain the present application, but are not intended to limit the application.
It should be noted here that, in order to avoid obscuring the present application due to unnecessary details, only structures and/or processing steps closely related to the solution according to the present application are shown in the drawings, while other details not greatly related to the present application are omitted.
It should be emphasized that the term "comprises/comprising" when used herein is taken to specify the presence of stated features, elements, steps or components, but does not preclude the presence or addition of one or more other features, elements, steps or components.
It is also noted herein that the term "coupled" may refer to not only a direct connection, but also an indirect connection in which an intermediate is present, unless otherwise specified.
Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar components, or the same or similar steps.
The existing convolutional neural network classification model uses 2D convolution to process 2D data or uses 3D convolution to process 3D data, and the application provides a method using 2D convolution to realize classification of 2.5D pneumonia CT medical image data. In the training process, firstly, preprocessing is carried out on the 3D pneumonia CT image data to enable different 3D data to be in the same distribution, then a data enhancement method is adopted to balance data quantity among different categories, 2.5D data which can be processed by a model is taken out, and a network is trained in a transfer learning mode, so that a better result can be obtained under the condition of insufficient data quantity of the network.
Based on this, the application provides a 2.5D medical CT image classifying device based on the ConvNeXt optimizing model, and compared with the existing CNN-based classifying model, in order to introduce the context relation in the 3D data, a classifying network based on 3D convolution is adopted, the training of the model is more difficult than that of a 2D convolution network, especially for CT-based data with multiple slices, the occupation of the video memory becomes large when the network is trained, a large amount of computer resources are required, the parameter quantity of the model becomes relatively large, and the fitting is easier for precious and small amount of medical data. However, if a conventional 2D convolution classification network is adopted, the classification is generally performed on a single picture, so that context information, that is, information of adjacent slices, cannot be introduced. Thus, the present application processes the raw data from a 2.5D data perspective, views the data for a specified number of slices from a feature map perspective, and uses 2D convolution to achieve the introduction of context. Thereby achieving the aims of relatively small model, relatively stable training process and difficult overfitting.
Compared with the prior art, the method introduces a context relation, an image in an input network is not one slice, but a plurality of continuous slices, for a lobular pneumonia case, whether a white round hole in the image is a focus or a blood vessel cavity can not be determined by only a single slice, and the type of the lobular pneumonia case is determined by introducing information of adjacent upper and lower slices, but the prior art uses 2D convolution on the single slice to extract the information, so that the context relation can not be introduced.
The following examples are provided to illustrate the application in more detail.
In order to effectively improve the execution efficiency and convenience of the pneumonia CT image classification process on the basis of reducing the complexity and time consumption of model training, the embodiment of the application provides a 2.5D pneumonia medical CT image classification device which can be realized by a 2.5D pneumonia medical CT image classification device, referring to fig. 1, the 2.5D pneumonia medical CT image classification device 1 specifically comprises the following contents:
The target image slicing module 10 is configured to acquire a plurality of continuous slices corresponding to the target 3D lung medical CT image data, so as to form target 2.5D lung medical CT image data corresponding to the target 3D lung medical CT image data.
In one or more embodiments of the present application, the target 3D lung medical CT image data refers to 3D lung medical CT image data to be classified, and the target 2.5D lung medical CT image data refers to each successive image slice of the target 3D lung medical CT image data.
It is understood that 3D refers to three dimensions, 2D refers to two dimensions, 2.5D refers to pseudo 3D, and in the present application, 2.5D images refer to a plurality of consecutive 2D image slices in a 3D image.
In the target image slicing module 10, in order to further improve the effectiveness and accuracy of 2.5D pneumonia medical CT image classification, 3D medical CT image data may be received first, then the 3D medical CT image data is converted into a nifet format, lung regions in the 3D medical CT image data are cut out to obtain corresponding 3D lung medical CT image data, and resampling and linear interpolation processing are performed on the 3D lung medical CT image data to obtain target 3D lung medical CT image data corresponding to the 3D medical CT image data.
In a specific example, the total number of the plurality of serial slices may be the same as the total number of serial slices in each sample employed in training the ConvNeXt optimization model, and may be, for example, 30 serial slices. It is understood that a plurality of successive slices refers to respective image slices acquired at the same interval unit between the respective slices.
The target image classification module 20 is configured to input the target 2.5D lung medical CT image data into a preset ConvNeXt optimization model that adopts a 2D convolution layer, so that the ConvNeXt optimization model correspondingly outputs a pneumonia focus type identification result corresponding to the target 2.5D lung medical CT image data.
Since the conventional ConvNeXt model uses three channels of color images RGB or a single channel of gray images as input, the channels of the network frame of the ConvNeXt model need to be adjusted and optimized, so in the target image classification module 20, the ConvNeXt optimization model refers to adjusting and optimizing the network structure of the ConvNeXt model for training with 2.5D lung medical CT image samples, so that the ConvNeXt optimization model can adapt to the input of multiple slices.
The 2.5D pneumonia medical CT image classification device provided by the application can be used for classifying 2.5D pneumonia medical CT images in a server or in a client device. Specifically, the selection may be made according to the processing capability of the client device, and restrictions of the use scenario of the user. The application is not limited in this regard. If all operations are performed in the client device, the client device may further comprise a processor for specific processing of the 2.5D pneumonia medical CT image classification.
The client device may have a communication module (i.e. a communication unit) and may be connected to a remote server in a communication manner, so as to implement data transmission with the server. The server may include a server on the side of the task scheduling center, and in other implementations may include a server of an intermediate platform, such as a server of a third party server platform having a communication link with the task scheduling center server. The server may include a single computer device, a server cluster formed by a plurality of servers, or a server structure of a distributed device.
Any suitable network protocol may be used between the server and the client device, including those not yet developed on the filing date of the present application. The network protocols may include, for example, TCP/IP protocol, UDP/IP protocol, HTTP protocol, HTTPS protocol, etc. Of course, the network protocol may also include, for example, RPC protocol (Remote Procedure Call Protocol ), REST protocol (Representational State Transfer, representational state transfer protocol), etc. used above the above-described protocol.
On this basis, in order to further improve the automation degree and the intelligence degree of the 2.5D pneumonia medical CT image classification, referring to fig. 2, the 2.5D pneumonia medical CT image classification device 1 may be respectively in communication connection with the 3D lung medical CT image acquisition device 2 and the display device 3 (display or mobile terminal device, etc.), so as to receive the target 3D lung medical CT image data from the 3D lung medical CT image acquisition device 2 in real time, and send the pneumonia focus type identification result to the display device 3 for output display.
From the above description, it can be seen that the 2.5D pneumonia medical CT image classification device provided by the embodiments of the present application can implement automatic classification of the pneumonia medical CT image, and can effectively improve the execution efficiency and convenience of the pneumonia CT image classification process and improve the reliability and accuracy of the pneumonia CT image classification result on the basis of reducing the complexity and time consumption of model training. The deep learning classification model is used for assisting a doctor in classifying pneumonia, so that the range of the pneumonia focus is provided for the doctor, the film reading pressure of the doctor can be effectively reduced, and the diagnosis efficiency and accuracy of the doctor are improved.
In order to further improve the effectiveness and accuracy of 2.5D pneumonia medical CT image classification, in the 2.5D pneumonia medical CT image classification apparatus provided by the embodiment of the present application, referring to fig. 3, the 2.5D pneumonia medical CT image classification apparatus 1 further specifically includes the following contents:
A model training module 30 connected to the target image classification module 20;
the model training module 30 is configured to train a ConvNeXt optimization model using a 2D convolution layer according to each labeled 2.5D lung medical CT image sample, so that the ConvNeXt optimization model is configured to correspondingly output a pneumonia focus type identification result according to the input 2.5D lung medical CT image data;
wherein the tag comprises: is used for respectively representing the symptoms of the pneumonia, the lobular pneumonia and the interstitial pneumonia. For example: 0 represents normal, non-pneumonia symptoms, 1 represents lobar pneumonia, 2 represents lobular pneumonia, 3 represents interstitial pneumonia. The pneumonia focus type identification result refers to the probability value corresponding to each of the four identifications.
In order to further improve the effectiveness and accuracy of 2.5D pneumonia medical CT image classification, in the 2.5D pneumonia medical CT image classification apparatus provided by the embodiment of the present application, referring to fig. 3, the 2.5D pneumonia medical CT image classification apparatus 1 specifically includes the following contents:
a historical image slicing module 40 connected to the model training module 30;
the historical image slicing module 40 is configured to obtain, based on a preset slice number threshold, a plurality of continuous slices corresponding to each historical 3D lung medical CT image sample in an image enhancement manner, so as to obtain 2.5D lung medical CT image samples corresponding to each historical 3D lung medical CT image sample.
In order to further improve the effectiveness and accuracy of 2.5D pneumonia medical CT image classification, in the 2.5D pneumonia medical CT image classification apparatus provided by the embodiment of the present application, referring to fig. 3, the 2.5D pneumonia medical CT image classification apparatus specifically includes the following contents:
a history image preprocessing module 50 connected to the history image slicing module 40;
the history image preprocessing module 50 is configured to preprocess each acquired historical 3D lung medical CT image data for historic diagnosis of a pneumonia focus, so as to obtain a corresponding each historical 3D lung medical CT image sample.
In order to further improve the efficiency of 2.5D pneumonia medical CT image classification and reduce the computational complexity, in the 2.5D pneumonia medical CT image classification device provided by the embodiment of the present application, referring to fig. 3, the 2.5D pneumonia medical CT image classification device 1 further specifically includes the following contents:
a model structure optimization module 60 coupled to the model training module 30;
the model structure optimization module 60 is configured to perform adjustment optimization for training on the network structure of the ConvNeXt model with 2.5D lung medical CT image samples, so as to generate a ConvNeXt optimization model using a 2D convolution layer.
In order to further improve accuracy of 2.5D pneumonia medical CT image classification, in the 2.5D pneumonia medical CT image classification apparatus provided in the embodiment of the present application, referring to fig. 4, the 2.5D pneumonia medical CT image classification apparatus 1 further specifically includes the following:
a model generalization test module 70 connected between the target image classification module 20 and the model training module 30;
the model generalization testing module 70 is configured to perform a generalization capability test on the training-obtained convnex optimization model, and select a convnex optimization model whose corresponding test result meets a preset generalization performance requirement.
In order to further improve the efficiency of 2.5D pneumonia medical CT image classification and reduce the computational complexity, in the 2.5D pneumonia medical CT image classification device provided by the embodiment of the present application, referring to fig. 5, a model training module 30 in the 2.5D pneumonia medical CT image classification device 1 specifically includes the following:
the pre-training unit 31 is configured to pre-train the convnex optimization model using the 2D convolution layer by using the pre-acquired co-domain historical lung medical CT image data, so as to obtain a pre-trained convnex optimization model.
The migration learning unit 32 is configured to iteratively train the pre-trained convnex optimization model by using each labeled 2.5D lung medical CT image sample, and obtain a convnex optimization model for correspondingly outputting a pneumonia focus type identification result according to the input 2.5D lung medical CT image data by adopting a cosine annealing learning rate mode in the training process.
In order to further improve the efficiency of 2.5D pneumonia medical CT image classification and reduce the computational complexity, in the 2.5D pneumonia medical CT image classification device provided by the embodiment of the present application, referring to fig. 5, the history image preprocessing module 50 in the 2.5D pneumonia medical CT image classification device 1 specifically includes the following:
a type conversion unit 51 for converting the disclosed DICOM sequence data into a nifet format to obtain respective historical 3D medical CT image data for historic diagnosis of a pneumonia lesion;
a clipping unit 52, configured to clip out lung regions in each of the historical 3D medical CT image data, so as to obtain corresponding historical 3D lung medical CT image data;
a resampling unit 53, configured to resample and linearly interpolate each of the historical 3D lung medical CT image data;
the normalization processing unit 54 is configured to perform normalization processing on each of the resampled and linear interpolation processed historical 3D lung medical CT image data, and convert each of the normalized historical 3D lung medical CT image data into npy format, so as to obtain each of the historical 3D lung medical CT image samples corresponding to each of the historical 3D medical CT image data.
On this basis, in order to further improve the automation degree and the intelligence degree of the classification of the 2.5D pneumonia medical CT image, referring to fig. 6, the 2.5D pneumonia medical CT image classification apparatus 1 may be further connected to a 3D lung medical CT image database 4 in a communication manner, so as to collect the disclosed DICOM sequence data from the 3D lung medical CT image database 4.
In order to further reduce the complexity and time consumption of model training and improve the execution efficiency and convenience of the pneumonia CT image classification process, referring to fig. 7 to fig. 9, the ConvNeXt optimization model in the 2.5D pneumonia medical CT image classification device specifically includes the following contents:
the device comprises an input layer, a 2D convolution layer, a layer normalization unit, 3 ConvNeXt blocks, a first downsampling group, a second downsampling group, a third downsampling group, a global average pooling layer, a layer normalization unit, a linear layer and an output layer which are sequentially connected;
the first downsampling group and the third downsampling group comprise downsampling layers and 3 ConvNeXt blocks which are connected in sequence, and the second downsampling group comprises downsampling layers and 9 ConvNeXt blocks which are connected in sequence;
the ConvNeXt block includes: the depth separable 2D convolution layer, the layer normalization unit, the 2D convolution layer, the Gaussian error linear unit, the 2D convolution layer, the layer scaling unit and the path discarding unit are sequentially connected;
The downsampling layer includes: and the layer normalization unit and the 2D convolution layer are sequentially connected.
In order to further illustrate the 2.5D pneumonia medical CT image classification device in the foregoing embodiment, the present application further provides a specific application example of a 2.5D pneumonia medical CT image classification method implemented by using the 2.5D pneumonia medical CT image classification device, referring to fig. 10 and 11, the 2.5D pneumonia medical CT image classification method specifically includes the following contents:
1. step S1:3D data preprocessing: and acquiring a plurality of pieces of pneumonia CT medical image data, and preprocessing each piece of pneumonia CT medical image data to enable different pieces of 3D data to be in the same distribution. May be implemented by the history image preprocessing module 50.
The original DICOM data needs to be preprocessed, and the inconsistency of machine parameters for generating CT images can cause that all CT images are not in the same data space, and the actual space represented by each voxel in different CT images is inconsistent, so that the data needs to be resampled, and training, verification and testing are performed on the convnex model after the preprocessing is completed.
And acquiring original DICOM sequence data of the target medical image, reading the data, resampling the data according to Spacing parameters thereof, unifying the distribution of images, avoiding the reduction of model performance caused by inconsistent thickness of slices, carrying out standard orthomorphism on the images, and converting image information thereof into npy format files for storage for subsequent use.
DICOM sequence data collected from a hospital can be used as a 3D historical pneumonia CT image and converted into NIFIT format (suffix is nii), personal privacy information such as patient name, age and the like can be hidden in the process, and data only useful for the image itself is reserved. Preprocessing the obtained nii sequence data comprises the following steps:
s11: based on the HU (Henry Unit) value statistics of the lungs, lung regions in each 3D historical lung CT image data are cut out to form each 3D lung CT image data.
S12: and uniformly resampling the 3D lung CT image data according to the Spacing parameters corresponding to the 3D lung CT image data, so that the actual space represented by the voxels of all the 3D lung CT image data is kept consistent. The resampled 3D lung CT image data are then fixed in the form of (H, C, W) respectively by means of linear interpolation, wherein H, C and W represent the height, the number of channels and the width of the 3D lung CT image data respectively, and the values of the height H and the width W are equal. And further, uniform image distribution is realized, and the degradation of model performance caused by inconsistent thickness of slices is avoided.
S13: standard orthogonalization of each 3D lung CT image data after resampling is performed.
S14: the image information of the processed 3D lung CT image data is extracted, and the data is converted into a npy format file (namely, a NumPy special binary file), so that the subsequent reading of the image data is convenient. Wherein NumPy (Numerical Python) is an open source numerical extension of Python. Such tools can be used to store and process large matrices.
That is, the 3D lung CT image data is subjected to operations such as type conversion, clipping and resampling through steps S11 to S14 to regenerate a data set belonging to the same distribution space, and the lung region obtained after clipping is also subjected to enhancement processing.
2. Step S2: data set construction: and acquiring 2.5D data of different categories by adopting a data enhancement mode, and keeping the data of different categories consistent in quantity so as to construct a data set of the training network. May be implemented by the historical image slicing module 40.
According to the characteristics of the data set, a proper number C of continuous slices is selected, then the 3D data after preprocessing is subjected to data enhancement to cut out C continuous slices, so that the data quantity of each category is kept balanced, and in the process of cutting out the slices, random and uniform cutting out is carried out according to the slice range of a focus, thereby completing the construction of a training set, a verification set and a test set.
The number of slices C can be selected according to different characteristics of the data set, such as the size of different types of pneumonia areas, the degree of variation among different slices and the like, and the number of slices C can jointly determine the sample composition when the data set is sent to the network training.
The specific operation of sample composition is as follows:
s21: c successive slices are taken along the axis of the 3D lung CT image data to form a sample.
In a special case, when c=1, the sample is the sample required by the 2D classification network, namely, the sample Zhang Qiepian, but when C is very large, irrelevant information input into the network is increased, interference is introduced, performance of the model is reduced, and the reasoning speed of the model is also reduced.
Based on this, it is important to select a proper number of slices, and through experiments, it is finally determined that 30 consecutive slices can be selected as inputs to the network, which results in a better effect.
In the selection process, a plurality of continuous 30 slices are uniformly distributed in the 3D lung CT image data to serve as one sample, and finally a plurality of samples are obtained through data enhancement, but the number of samples in each category needs to be kept in an equilibrium state, so that the decline of model performance caused by sample unbalance is avoided. In addition, due to the relatively small size of the data set used in the application, when slicing and sampling the data, a slight overlap phenomenon possibly occurs between different data.
3. Step S3: pre-training the model in the same domain data to obtain pre-training weights, and then performing transfer learning by using the pneumonia 25D data, and performing iterative updating to obtain a plurality of groups of weights. May be implemented sequentially by model structure optimization module 60 and model training module 30.
Pre-training an improved ConvNeXt model and performing transfer learning: firstly, pretraining a model by using data in the same domain, loading weights obtained by pretraining into the model, further, training the model by using the training set, carrying out inverse gradient propagation on the model by using a cross entropy function, updating the weights, taking the accuracy of category judgment as an index for judging whether the model performance is good or not, and storing a plurality of groups of model weights with good performance on a verification set according to the index.
S31: the ConvNeXt network is adopted as a basic model structure of the model, and the conventional ConvNeXt model adopts three channels of a color picture RGB as input or a single channel of a gray picture as input, so that the channels of a network frame of the ConvNeXt model are required to be adjusted and optimized, and the corresponding ConvNeXt optimization model can adapt to the input of 30 slices.
Further, the model may be pre-trained using data in the same domain, and the obtained pre-training weights are used for transfer learning, that is, the obtained training set data is used to train the network, so that the network performs multiple iterative training based on the pre-training weights, and an Adam optimizer is selected, and the loss function of training is cross EntropyLoss, where the formula is as follows:
Wherein N represents the number of samples of a batch; i represents the sample number in the lot; y is i A true category label representing the sample;class labels representing model predictions.
The cosine annealing learning rate mode is adopted in the training process so as to ensure that the model can stably converge, and the learning rate is continuously reduced and some super parameters are continuously improved through multiple training.
And the accuracy rate of category judgment is used as an index for measuring the quality of the model, the network is trained by using the strategy, and a plurality of groups of weights with good index effects on the verification set are stored.
4. Step S4: and testing the generalization performance of the model on the test set, and selecting a group of weights with the best generalization performance to obtain a final target classification model. May be implemented by a model generalization test module 70.
Test of model generalization ability: the generalization performance of the plurality of groups of weights is tested on the test set, the accuracy of class judgment is also used as an index, the weight with the best performance on the test set is the required target, so far, the 2.5D pneumonia medical CT image classification method based on the ConvNeXt optimization model is completely finished, and the model can be used for assisting doctors in judging the pneumonia class.
And the accuracy of category judgment is adopted as an index for measuring the generalization capability of the model, and the model with the best generalization performance is tested on a test set and used as a final model. In addition, in the present application, a many-to-one approach is adopted, and the criterion is that as long as a lesion is included in the slice sequence, the output result is the type of the lesion. Therefore, the obtained model is used for auxiliary diagnosis of pneumonia, and can assist doctors to quickly locate the range of the pneumonia focus.
Wherein the correspondence between the number of samples trained using the ConvNeXt optimization model and the corresponding number of test samples is shown in Table 1, wherein each sample is 30 consecutive slices.
And then, a plurality of continuous slices corresponding to the target 3D lung medical CT image data are acquired by adopting a target image slicing module 10 to form target 2.5D lung medical CT image data corresponding to the target 3D lung medical CT image data, and then the target 2.5D lung medical CT image data are input into a preset ConvNeXt optimization model adopting a 2D convolution layer by adopting a target image classification module 20, so that the ConvNeXt optimization model correspondingly outputs a pneumonia focus type identification result corresponding to the target 2.5D lung medical CT image data.
TABLE 1
In summary, the application example provided by the application can realize automatic classification of the pneumonia medical CT images, can effectively improve the execution efficiency and convenience of the pneumonia CT image classification process and can improve the reliability and accuracy of the pneumonia CT image classification result on the basis of reducing the complexity and time consumption of model training. The deep learning classification model is used for assisting a doctor in classifying pneumonia, so that the range of the pneumonia focus is provided for the doctor, the film reading pressure of the doctor can be effectively reduced, and the diagnosis efficiency and accuracy of the doctor are improved.
The embodiment of the application also provides an electronic device, which may include a processor, a memory, a receiver and a transmitter, where the processor is configured to perform the functions of the 2.5D pneumonia medical CT image classification apparatus mentioned in the foregoing embodiment, and the processor and the memory may be connected by a bus or other manners, for example, by a bus connection. The receiver may be connected to the processor, memory, by wire or wirelessly.
The processor may be a central processing unit (Central Processing Unit, CPU). The processor may also be any other general purpose processor, digital signal processor (Digital Signal Processor, DSP), application specific integrated circuit (Application Specific Integrated Circuit, ASIC), field programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof.
The memory is used as a non-transitory computer readable storage medium for storing a non-transitory software program, a non-transitory computer executable program and a module, such as program instructions/modules corresponding to the 2.5D pneumonia medical CT image classification device in the embodiment of the present application. The processor executes various functional applications and data processing of the processor by running non-transitory software programs, instructions and modules stored in the memory, i.e. the functions of the 2.5D pneumonia medical CT image classification device in the above-described method embodiments are implemented.
The memory may include a memory program area and a memory data area, wherein the memory program area may store an operating system, at least one application program required for a function; the storage data area may store data created by the processor, etc. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory may optionally include memory located remotely from the processor, the remote memory being connectable to the processor through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The one or more modules are stored in the memory that, when executed by the processor, perform the functions of the 2.5D pneumonia medical CT image classification apparatus in embodiments.
In some embodiments of the present application, a user equipment may include a processor, a memory, and a transceiver unit, which may include a receiver and a transmitter, the processor, the memory, the receiver, and the transmitter may be connected by a bus system, the memory being configured to store computer instructions, the processor being configured to execute the computer instructions stored in the memory to control the transceiver unit to transmit and receive signals.
As an implementation manner, the functions of the receiver and the transmitter in the present application may be considered to be implemented by a transceiver circuit or a dedicated chip for transceiver, and the processor may be considered to be implemented by a dedicated processing chip, a processing circuit or a general-purpose chip.
As another implementation manner, a manner of using a general-purpose computer may be considered to implement the server provided by the embodiment of the present application. I.e. program code for implementing the functions of the processor, the receiver and the transmitter are stored in the memory, and the general purpose processor implements the functions of the processor, the receiver and the transmitter by executing the code in the memory.
The embodiment of the application also provides a computer readable storage medium, on which a computer program is stored, which when being executed by a processor, is used for realizing the function of the 2.5D pneumonia medical CT image classification device. The computer readable storage medium may be a tangible storage medium such as Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disk, a removable memory disk, a CD-ROM, or any other form of storage medium known in the art.
Those of ordinary skill in the art will appreciate that the various illustrative components, systems, and methods described in connection with the embodiments disclosed herein can be implemented as hardware, software, or a combination of both. The particular implementation is hardware or software dependent on the specific application of the solution and the design constraints. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application. When implemented in hardware, it may be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), suitable firmware, a plug-in, a function card, or the like. When implemented in software, the elements of the application are the programs or code segments used to perform the required tasks. The program or code segments may be stored in a machine readable medium or transmitted over transmission media or communication links by a data signal carried in a carrier wave.
It should be understood that the application is not limited to the particular arrangements and instrumentality described above and shown in the drawings. For the sake of brevity, a detailed description of known methods is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method processes of the present application are not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between steps, after appreciating the spirit of the present application.
In the present application, features that are described and/or illustrated with respect to one embodiment may be used in the same way or in a similar way in one or more other embodiments and/or in combination with or instead of the features of the other embodiments
The above description is only of the preferred embodiments of the present application and is not intended to limit the present application, and various modifications and variations can be made to the embodiments of the present application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (10)
1. A 2.5D pneumonia medical CT image classification device, comprising:
the target image slicing module is used for acquiring a plurality of continuous slices corresponding to target 3D lung medical CT image data so as to form target 2.5D lung medical CT image data corresponding to the target 3D lung medical CT image data;
and the target image classification module is used for inputting the target 2.5D lung medical CT image data into a preset ConvNeXt optimization model adopting a 2D convolution layer so that the ConvNeXt optimization model correspondingly outputs a pneumonia focus type identification result corresponding to the target 2.5D lung medical CT image data.
2. The 2.5D pneumonia medical CT image classification apparatus according to claim 1, further comprising: the model training module is connected with the target image classification module;
the model training module is used for training a ConvNeXt optimization model with a 2D convolution layer according to each 2.5D lung medical CT image sample provided with a label, so that the ConvNeXt optimization model is used for correspondingly outputting a pneumonia focus type identification result according to the input 2.5D lung medical CT image data;
wherein the tag comprises: is used for respectively representing the symptoms of the pneumonia, the lobular pneumonia and the interstitial pneumonia.
3. The 2.5D pneumonia medical CT image classification apparatus according to claim 2, further comprising: the historical image slicing module is connected with the model training module;
the historical image slicing module is used for respectively acquiring a plurality of continuous slices corresponding to each historical 3D lung medical CT image sample in an image enhancement mode based on a preset slice quantity threshold value so as to obtain 2.5D lung medical CT image samples corresponding to each historical 3D lung medical CT image sample.
4. The 2.5D pneumonia medical CT image classification apparatus according to claim 3, further comprising: the historical image preprocessing module is connected with the historical image slicing module;
the history image preprocessing module is used for preprocessing the acquired history 3D lung medical CT image data for the history diagnosis of the pneumonia focus to obtain corresponding history 3D lung medical CT image samples.
5. The 2.5D pneumonia medical CT image classification apparatus according to claim 2, further comprising: the model structure optimization module is connected with the model training module;
the model structure optimization module is used for performing adjustment optimization for training on a 2.5D lung medical CT image sample on a network structure of the ConvNeXt model so as to generate a ConvNeXt optimization model adopting a 2D convolution layer.
6. The 2.5D pneumonia medical CT image classification apparatus according to claim 2, further comprising: the model generalization test module is connected between the target image classification module and the model training module;
the model generalization test module is used for carrying out generalization capability test on the ConvNeXt optimization model obtained through training and selecting a ConvNeXt optimization model of which the corresponding test result meets the preset generalization performance requirement.
7. The 2.5D pneumonia medical CT image classification apparatus according to claim 2, wherein said model training module comprises:
the pre-training unit is used for pre-training the ConvNeXt optimization model by adopting the 2D convolution layer by adopting pre-acquired same-domain historical lung medical CT image data to obtain a pre-trained ConvNeXt optimization model;
the migration learning unit is used for iteratively training the pre-training ConvNeXt optimization model by adopting each 2.5D lung medical CT image sample with a label, and obtaining the ConvNeXt optimization model for correspondingly outputting a pneumonia focus type identification result according to the input 2.5D lung medical CT image data by adopting a cosine annealing learning rate mode in the training process.
8. The 2.5D pneumonia medical CT image classification apparatus according to claim 4, wherein said history image preprocessing module comprises:
The type conversion unit is used for converting the disclosed DICOM sequence data into NIFIT format to obtain each historical 3D medical CT image data for the historical diagnosis of the pneumonia focus;
the clipping unit is used for clipping out the lung areas in each historical 3D medical CT image data so as to obtain corresponding historical 3D lung medical CT image data;
the resampling unit is used for resampling and linear interpolation processing of each historical 3D lung medical CT image data;
the standardized processing unit is used for carrying out standardized processing on each historical 3D lung medical CT image data subjected to resampling and linear interpolation processing, and converting each historical 3D lung medical CT image data subjected to standardized processing into npy format respectively so as to obtain each corresponding historical 3D lung medical CT image sample of each historical 3D medical CT image data.
9. The 2.5D pneumonia medical CT image classification apparatus according to any one of claims 1-8, wherein said ConvNeXt optimization model comprises: the device comprises an input layer, a 2D convolution layer, a layer normalization unit, 3 ConvNeXt blocks, a first downsampling group, a second downsampling group, a third downsampling group, a global average pooling layer, a layer normalization unit, a linear layer and an output layer which are sequentially connected;
The first downsampling group and the third downsampling group comprise downsampling layers and 3 ConvNeXt blocks which are connected in sequence, and the second downsampling group comprises downsampling layers and 9 ConvNeXt blocks which are connected in sequence;
the ConvNeXt block includes: the depth separable 2D convolution layer, the layer normalization unit, the 2D convolution layer, the Gaussian error linear unit, the 2D convolution layer, the layer scaling unit and the path discarding unit are sequentially connected;
the downsampling layer includes: and the layer normalization unit and the 2D convolution layer are sequentially connected.
10. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor performs the functions of the 2.5D pneumonia medical CT image classification apparatus according to any one of claims 1-8 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310781182.XA CN116664953A (en) | 2023-06-28 | 2023-06-28 | 2.5D pneumonia medical CT image classification device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310781182.XA CN116664953A (en) | 2023-06-28 | 2023-06-28 | 2.5D pneumonia medical CT image classification device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116664953A true CN116664953A (en) | 2023-08-29 |
Family
ID=87709834
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310781182.XA Pending CN116664953A (en) | 2023-06-28 | 2023-06-28 | 2.5D pneumonia medical CT image classification device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116664953A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117315380A (en) * | 2023-11-30 | 2023-12-29 | 深圳市健怡康医疗器械科技有限公司 | Deep learning-based pneumonia CT image classification method and system |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106780460A (en) * | 2016-12-13 | 2017-05-31 | 杭州健培科技有限公司 | A kind of Lung neoplasm automatic checkout system for chest CT image |
| CN109523521A (en) * | 2018-10-26 | 2019-03-26 | 复旦大学 | Lung neoplasm classification and lesion localization method and system based on more slice CT images |
| CN110110617A (en) * | 2019-04-22 | 2019-08-09 | 腾讯科技（深圳）有限公司 | Medical image dividing method, device, electronic equipment and storage medium |
| US20190287242A1 (en) * | 2018-03-16 | 2019-09-19 | Infervision | Computed tomography pulmonary nodule detection method based on deep learning |
| CN110766051A (en) * | 2019-09-20 | 2020-02-07 | 四川大学华西医院 | Lung nodule morphological classification method based on neural network |
| CN110993064A (en) * | 2019-11-05 | 2020-04-10 | 北京邮电大学 | Deep learning-oriented medical image labeling method and device |
| CN112365962A (en) * | 2020-11-17 | 2021-02-12 | 中南大学湘雅二医院 | Medical resource allocation method and device based on lung CT (computed tomography) image and clinical information |
| CN113269230A (en) * | 2021-04-23 | 2021-08-17 | 复旦大学 | Multi-pneumonia CT classification method and device based on time sequence high-dimensional feature extraction |
| CN113920109A (en) * | 2021-10-29 | 2022-01-11 | 沈阳东软智能医疗科技研究院有限公司 | Medical image recognition model training method, recognition method, device and equipment |
| WO2022063200A1 (en) * | 2020-09-24 | 2022-03-31 | 上海健康医学院 | Non-small cell lung cancer prognosis survival prediction method, medium and electronic device |
| CN115908449A (en) * | 2022-10-21 | 2023-04-04 | 北京邮电大学 | 2.5D medical CT image segmentation method and device based on improved UNet model |
| CN116205967A (en) * | 2023-04-27 | 2023-06-02 | 中国科学院长春光学精密机械与物理研究所 | Medical image semantic segmentation method, device, equipment and medium |
| CN116245892A (en) * | 2022-12-05 | 2023-06-09 | 北京建筑大学 | Image processing model generation method, image processing method and device |
-
2023
- 2023-06-28 CN CN202310781182.XA patent/CN116664953A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106780460A (en) * | 2016-12-13 | 2017-05-31 | 杭州健培科技有限公司 | A kind of Lung neoplasm automatic checkout system for chest CT image |
| US20190287242A1 (en) * | 2018-03-16 | 2019-09-19 | Infervision | Computed tomography pulmonary nodule detection method based on deep learning |
| CN109523521A (en) * | 2018-10-26 | 2019-03-26 | 复旦大学 | Lung neoplasm classification and lesion localization method and system based on more slice CT images |
| CN110110617A (en) * | 2019-04-22 | 2019-08-09 | 腾讯科技（深圳）有限公司 | Medical image dividing method, device, electronic equipment and storage medium |
| CN110766051A (en) * | 2019-09-20 | 2020-02-07 | 四川大学华西医院 | Lung nodule morphological classification method based on neural network |
| CN110993064A (en) * | 2019-11-05 | 2020-04-10 | 北京邮电大学 | Deep learning-oriented medical image labeling method and device |
| WO2022063200A1 (en) * | 2020-09-24 | 2022-03-31 | 上海健康医学院 | Non-small cell lung cancer prognosis survival prediction method, medium and electronic device |
| CN112365962A (en) * | 2020-11-17 | 2021-02-12 | 中南大学湘雅二医院 | Medical resource allocation method and device based on lung CT (computed tomography) image and clinical information |
| CN113269230A (en) * | 2021-04-23 | 2021-08-17 | 复旦大学 | Multi-pneumonia CT classification method and device based on time sequence high-dimensional feature extraction |
| CN113920109A (en) * | 2021-10-29 | 2022-01-11 | 沈阳东软智能医疗科技研究院有限公司 | Medical image recognition model training method, recognition method, device and equipment |
| CN115908449A (en) * | 2022-10-21 | 2023-04-04 | 北京邮电大学 | 2.5D medical CT image segmentation method and device based on improved UNet model |
| CN116245892A (en) * | 2022-12-05 | 2023-06-09 | 北京建筑大学 | Image processing model generation method, image processing method and device |
| CN116205967A (en) * | 2023-04-27 | 2023-06-02 | 中国科学院长春光学精密机械与物理研究所 | Medical image semantic segmentation method, device, equipment and medium |
Non-Patent Citations (2)
| Title |
|---|
| QT-SMILE: "ConvNeXt网络模型", pages 1, Retrieved from the Internet <URL:https://blog.csdn.net/qq_34848334/article/details/122668728?ops_request_misc=&request_id=&biz_id=102&utm_term=convnext%20pytorch&utm_medium=distribute.pc_search_result.none-task-blog-2~all~sobaiduweb~default-1-122668728.142^v100^control&spm=1018.2226.3001.4187> * |
| 唐迁;杜博;恽爽;高莉;吴爽;张超;兰猛;陈紫业;李亮;查云飞;张良培;李平湘;: "COVID-19 CT影像智能诊断系统", 武汉大学学报(信息科学版), no. 06, 5 June 2020 (2020-06-05) * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117315380A (en) * | 2023-11-30 | 2023-12-29 | 深圳市健怡康医疗器械科技有限公司 | Deep learning-based pneumonia CT image classification method and system |
| CN117315380B (en) * | 2023-11-30 | 2024-02-02 | 深圳市健怡康医疗器械科技有限公司 | Deep learning-based pneumonia CT image classification method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yu et al. | Deep-learning-empowered breast cancer auxiliary diagnosis for 5GB remote E-health | |
| CN109671094B (en) | Fundus image blood vessel segmentation method based on frequency domain classification | |
| CN113781397B (en) | Medical image focus detection modeling method, device and system based on federal learning | |
| JP2016168046A (en) | Plant disease diagnostic system, plant disease diagnostic method, and program | |
| CN111275714A (en) | Prostate MR image segmentation method based on attention mechanism 3D convolutional neural network | |
| CN108764342B (en) | Semantic segmentation method for optic discs and optic cups in fundus image | |
| CN108830149B (en) | Target bacterium detection method and terminal equipment | |
| US11967181B2 (en) | Method and device for retinal image recognition, electronic equipment, and storage medium | |
| CN113450328A (en) | Medical image key point detection method and system based on improved neural network | |
| CN108231190A (en) | Handle the method for image and nerve network system, equipment, medium, program | |
| CN116664953A (en) | 2.5D pneumonia medical CT image classification device and equipment | |
| CN114140478B (en) | Federal learning method, system, device and medium for medical image segmentation | |
| CN110738660A (en) | Spine CT image segmentation method and device based on improved U-net | |
| JP2021514228A (en) | Image processing methods and devices, and training methods for neural network models | |
| CN112102259A (en) | Image segmentation algorithm based on boundary guide depth learning | |
| CN116503668A (en) | Medical image classification method based on small sample element learning | |
| CN114881105A (en) | Sleep staging method and system based on transformer model and contrast learning | |
| CN112489062B (en) | Medical image segmentation method and system based on boundary and neighborhood guidance | |
| Alaguselvi et al. | Quantitative analysis of fundus image enhancement in the detection of diabetic retinopathy using deep convolutional neural network | |
| CN117253071B (en) | Semi-supervised target detection method and system based on multistage pseudo tag enhancement | |
| CN117765252A (en) | breast cancer identification system and method based on Swin transducer and contrast learning | |
| CN113421228A (en) | Thyroid nodule identification model training method and system based on parameter migration | |
| CN115937129B (en) | Method and device for processing left and right half brain relations based on multi-mode magnetic resonance image | |
| US20230297646A1 (en) | System and methods for classifying magnetic resonance imaging (mri) image characteristics | |
| CN116958154A (en) | Image segmentation method and device, storage medium and electronic equipment |
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 |