CN115761342A - Lung CT image pneumonia classification method, device and equipment - Google Patents
Lung CT image pneumonia classification method, device and equipment Download PDFInfo
- Publication number
- CN115761342A CN115761342A CN202211457766.3A CN202211457766A CN115761342A CN 115761342 A CN115761342 A CN 115761342A CN 202211457766 A CN202211457766 A CN 202211457766A CN 115761342 A CN115761342 A CN 115761342A
- Authority
- CN
- China
- Prior art keywords
- image data
- pneumonia
- classification
- lung
- image
- 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
Images
Classifications
-
- 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
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a lung CT image pneumonia classification method, a lung CT image pneumonia classification device and lung CT image pneumonia classification equipment, belongs to the technical field of image processing, and is used for solving the problems of low detection efficiency and low detection precision of manual data labeling in the prior art. The method comprises the following steps: acquiring lung CT image data to be identified of a target object; inputting lung CT image data to be recognized into a trained target classification model to obtain a classification result; the target classification model is obtained by adopting unlabeled pneumonia CT image data to carry out self-supervision training; the classification results include at least normal lung, viral pneumonia, and bacterial pneumonia. The self-supervision training is introduced based on the CT image data of the unlabelled pneumonia, so that the dependence on the labeled data can be reduced, the data labeling cost is further reduced, the detection efficiency is improved, and the detection precision can be improved in the process of better transferring generalization to a classification and identification task from the internal characterization of a large amount of unlabelled pneumonia image data.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a lung CT image pneumonia classification method, a lung CT image pneumonia classification device and lung CT image pneumonia classification equipment.
Background
In recent years, the development of deep learning gradually leads to the clinical application of intelligent image diagnosis, wherein the application of the lung CT image segmentation technology enables medical personnel to quickly and accurately position a lung lesion region and simultaneously assist in judging other lung diseases, such as lung cancer node identification and the like. However, there are relatively few reports from studies relating to the diagnosis of pneumonia types by pulmonary CT imaging. Pneumonia type diagnosis based on lung CT images is an image classification problem.
As an upstream task in the field of computer vision, image classification is a popular research hotspot, and the performance of the image classification indirectly affects the performance of visual downstream tasks, such as target detection, segmentation, tracking and the like. The traditional image classification mainly depends on manual feature setting for distinguishing, time is consumed, and the optimal distinguishing features selected manually cannot be guaranteed, so that the precision and the efficiency are not high. In recent years, the image classification technology based on deep learning shows good performance, optimal distinguishing characteristics can be automatically learned from image data, weights are updated through a back propagation algorithm to fit data, and the performance and the efficiency of a classification model are greatly improved.
Therefore, it is desirable to provide a more reliable lung CT image pneumonia classification scheme.
Disclosure of Invention
The invention aims to provide a lung CT image pneumonia classification method, a lung CT image pneumonia classification device and lung CT image pneumonia classification equipment, which are used for solving the problems of low detection efficiency and low detection precision caused by manual data labeling in the prior art.
In order to achieve the above purpose, the invention provides the following technical scheme:
in a first aspect, the present invention provides a lung CT image pneumonia classification method, including:
acquiring lung CT image data to be identified of a target object;
inputting the lung CT image data to be recognized into a trained target classification model to obtain a classification result; the target classification model is obtained by performing self-supervision learning training by adopting non-label pneumonia CT image data; the classification includes at least normal lung, viral pneumonia, and bacterial pneumonia.
Optionally, before inputting the lung CT image data to be recognized into the trained target classification model and obtaining a classification result, the method further includes:
acquiring a label-free pneumonia CT image data set;
carrying out random masking operation on image data in the non-tag pneumonia CT image data set to obtain masked image data;
converting the masked image data into label data, inputting the label data into a Transformer model, and outputting target image data;
calculating MSE loss between the target image data and the original image data;
and adjusting the weight of the Transformer model based on the MSE loss, performing fine tuning training of a multi-classification task, and taking the model with the highest accuracy on a verification set as the target classification model.
Optionally, the transform model includes a multi-head self-attention block and a multi-layer perceptron; the multilayer perceptron is composed of a hidden layer and an output layer.
Optionally, the acquiring a tag-free pneumonia CT image data set specifically includes:
determining a classification task scene corresponding to pneumonia;
and acquiring historical non-label pneumonia CT image data corresponding to the classification task scene to obtain a non-label pneumonia CT image data set.
Optionally, the randomly masking the image data in the unlabeled pneumonia CT image data set to obtain masked image data specifically includes:
sending image data in the non-tag pneumonia CT image data set into a random mask module;
and dividing the image corresponding to the image data into uniform squares to obtain the masked image data.
Optionally, converting the masked image data into tag data, inputting the tag data into a transform model, and outputting target image data, which specifically includes:
dividing the masked image data into blocks with fixed sizes, and performing convolution operation on each block corresponding to the image data by utilizing two-dimensional convolution;
carrying out one-dimensional processing on the data after the convolution operation to generate tokens;
and inputting the generated tokens into a Transformer model, and outputting target image data.
Optionally, a first part of data in the unlabeled pneumonia CT image data set is used for self-supervision pre-training of a classification model; and marking a second part of data in the unlabeled pneumonia CT image data set for self-supervision fine adjustment of a classification model.
In a second aspect, the present invention provides a lung pneumonia classification device by CT imaging, the device comprising:
the lung CT image data acquisition module to be identified is used for acquiring lung CT image data to be identified of the target object;
the target classification model identification module is used for inputting the lung CT image data to be identified into a trained target classification model to obtain a classification result; the target classification model is obtained by performing self-supervision learning training by adopting non-label pneumonia CT image data; the classification includes at least normal lung, viral pneumonia, and bacterial pneumonia.
In a third aspect, the present invention provides a lung pneumonia classification apparatus by CT imaging, the apparatus comprising:
the communication unit/communication interface is used for acquiring lung CT image data to be identified of the target object;
the processing unit/processor is used for inputting the lung CT image data to be recognized into a trained target classification model to obtain a classification result; the target classification model is obtained by performing self-supervision learning training on the unlabeled pneumonia CT image data; the classification results include at least normal lung, viral pneumonia, and bacterial pneumonia.
In a fourth aspect, the present invention provides a computer storage medium, which stores instructions that, when executed, implement the above method for classifying pneumonia in lung CT images.
Compared with the prior art, the invention provides a lung CT image pneumonia classification method, a lung CT image pneumonia classification device and lung CT image pneumonia classification equipment. The method comprises the following steps: acquiring lung CT image data to be identified of a target object; inputting the lung CT image data to be identified into a trained target classification model to obtain a classification result; the target classification model is obtained by performing self-supervision training on the unlabeled pneumonia CT image data; the classification results include at least normal lung, viral pneumonia, and bacterial pneumonia. The self-supervision training is introduced based on the CT image data of the unlabelled pneumonia, so that the dependence on the labeled data can be reduced, the data labeling cost is further reduced, the detection efficiency is improved, and the detection precision can be improved in the process of better transferring generalization to a classification and identification task from the internal characterization of a large amount of unlabelled pneumonia image data.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention and not to limit the invention. In the drawings:
FIG. 1 is a schematic flow chart of a lung CT image pneumonia classification method according to the present invention;
FIG. 2 is a mask diagram of the lung CT image pneumonia classification method according to the present invention;
FIG. 3 is a schematic structural diagram of a lung CT image pneumonia classification device provided by the present invention;
fig. 4 is a schematic structural view of the lung CT image pneumonia classification apparatus provided by the present invention.
Detailed Description
In order to facilitate clear description of technical solutions of the embodiments of the present invention, in the embodiments of the present invention, terms such as "first" and "second" are used to distinguish the same items or similar items having substantially the same functions and actions. For example, the first threshold and the second threshold are only used for distinguishing different thresholds, and the order of the thresholds is not limited. Those skilled in the art will appreciate that the terms "first," "second," etc. do not denote any order or quantity, nor do the terms "first," "second," etc. denote any order or importance.
It is to be understood that the terms "exemplary" or "such as" are used herein to mean serving as an example, instance, or illustration. Any embodiment or design described herein as "exemplary" or "e.g.," is not necessarily to be construed as preferred or advantageous over other embodiments or designs. Rather, use of the word "exemplary" or "such as" is intended to present concepts related in a concrete fashion.
In the present invention, "at least one" means one or more, "a plurality" means two or more. "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone, wherein A and B can be singular or plural. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. "at least one of the following" or similar expressions refer to any combination of these items, including any combination of the singular or plural items. For example, at least one (one) of a, b, or c, may represent: a, b, c, a and b combination, a and c combination, b and c combination, or a, b and c combination, wherein a, b and c can be single or multiple.
In the prior art, the deep learning technology is based on data driving in nature, and although the deep learning technology achieves good performance in image classification, the deep learning technology relies on a large amount of manual data, and high labor cost is inevitably brought about. Compared with label data, the label-free data is almost inexhaustible on the Internet, and the label-free data has the characteristics of low price and easy availability. How to train the model by using label-free data and improve the performance of the model is a key technical problem at present.
Chinese patent CN111415743B proposes a pneumonia classification method, a classification apparatus, a computer-readable storage medium, and an electronic device, which enable a neural network model to automatically identify the type of pneumonia by inputting acquired pneumonia CT images and labeling information into the neural network model and training. Although the combination of clinical information of a detected person further improves the identification accuracy, the data annotation cost is very high, and the wide application of the technology is further limited.
Therefore, the invention provides a more reliable lung CT image pneumonia classification scheme.
Next, a scheme provided by an embodiment of the present specification is explained with reference to the accompanying drawings:
fig. 1 is a schematic flow chart of a lung CT image pneumonia classification method provided by the present invention, and as shown in fig. 1, the flow chart may include the following steps:
step 110: and acquiring CT image data of the lung to be identified of the target object.
The target object may be a patient or other object requiring identification of lung CT images.
Step 120: and inputting the lung CT image data to be recognized into a trained target classification model to obtain a classification result.
The target classification model can be a model obtained by performing self-supervision learning training by adopting unlabeled pneumonia CT image data; the classification results include at least normal lung, viral pneumonia, and bacterial pneumonia.
The data of the target classification model obtained by training is unlabeled pneumonia CT image data, and manual labeling data are not needed to be adopted as training data.
Self-supervised learning (Self-supervised learning) aims to mine the characteristic characteristics of data as supervision information by designing auxiliary tasks (Proxy tasks) for label-free data, and improve the feature extraction capability of a model.
The method in fig. 1, by acquiring CT image data of a lung to be identified of a target object; inputting the lung CT image data to be identified into a trained target classification model to obtain a classification result; the target classification model is obtained by adopting unlabeled pneumonia CT image data to carry out self-supervision training; the classification results include at least normal lung, viral pneumonia, and bacterial pneumonia. The self-supervision training is introduced based on the CT image data of the unlabelled pneumonia, so that the dependence on the labeled data can be reduced, the data labeling cost is further reduced, the detection efficiency is improved, and the detection precision can be improved in the process of better transferring generalization to a classification and identification task from the internal characterization of a large amount of unlabelled pneumonia image data.
Based on the method of fig. 1, the embodiments of the present specification also provide some specific implementations of the method, which are described below.
Optionally, before inputting the lung CT image data to be recognized into the trained target classification model to obtain a classification result, the method may further include:
acquiring a CT image data set of the unlabeled pneumonia;
carrying out random masking operation on the image data in the non-tag pneumonia CT image data set to obtain masked image data;
converting the masked image data into label data, inputting the label data into a Transformer model, and outputting target image data;
calculating MSE loss between the target image data and the original image data;
and adjusting the weight of the Transformer model based on the MSE loss, performing fine tuning training of a multi-classification task, and taking the model with the highest accuracy on a verification set as the target classification model.
The acquiring of the unlabeled pneumonia CT image data set may specifically include:
determining a classification task scene corresponding to pneumonia;
and acquiring historical non-label pneumonia CT image data corresponding to the classification task scene to obtain a non-label pneumonia CT image data set.
Before training a classification model, a large number of label-free data sets can be collected, and when the data sets are obtained, a large number of historical label-free pneumonia CT image data can be obtained according to related classification task scenes. The method comprises the following steps that a first part of data in a label-free pneumonia CT image data set is used for self-supervision pre-training of a classification model; the second part of data in the unlabeled pneumonia CT image data set is labeled and then used for the self-supervised fine tuning of the classification model, for example, 70% of data may be selected for the self-supervised pre-training, and 30% of data may be manually labeled for the self-supervised fine tuning.
The Transformer model may include a multi-headed self-attention block multi-head self-attention (MSA) and a multi-layer perceptron Multi Layer Persistence (MLP). For each layer, the input from attention is a triplet (query, key, value). The multilayer perceptron is composed of a hidden layer and an output layer:
MLP(x)＝Linear(GELU(Linear(x)))：
MSA is a multiple attention head operation, with inputs:
x∈R B*N+1*C (ii) a Expanded by tensor into query space Q, key space K, value space V, as in equation (1):
wherein d is k Is the scaling factor.
The whole Transformer network is formed by stacking LayerNorm, MSA and MLP, namely:
Out1(x)＝MSA(LN(x))+x
Out2(x)＝MLP(LN(x))+x
Block(x)＝Out 2 (Out 1 (x))
Transformer(x)Block(Block(Block(…Blcok(x))))×D；
wherein D represents the network depth and represents that the Transformer network consists of D Block modules.
Optionally, the randomly masking the image data in the unlabeled pneumonia CT image data set to obtain masked image data specifically may include:
sending image data in the non-label pneumonia CT image data set into a random mask module;
and dividing the image corresponding to the image data into uniform squares to obtain the masked image data.
In the specific implementation process, the label-free CT image data is sent to a random mask module, and the input is set as follows: x belongs to R B *3*224*224 Where B is BatchSize, and 3, 224 are the number of channels, height, and width, respectively, of the input image. As shown in fig. 2, the left image is the original image, and the right image is the masked image in the two images in fig. 2. The original image is divided into uniform squares, and assuming that the sides of the squares are 28, the original image can be divided into
Blocks, 75% of which are randomly masked.
Optionally, converting the masked image data into tag data, inputting the tag data into a transform model, and outputting target image data, which may specifically include:
dividing the image data after the mask into blocks with fixed sizes, and performing convolution operation on each block corresponding to the image data by utilizing two-dimensional convolution;
carrying out one-dimensional processing on the data after the convolution operation to generate tokens;
and inputting the generated tokens into a Transformer model, and outputting target image data.
In specific implementation, the PatchEmbed operation is performed on the masked image, specifically, the original image is divided into uniform blocks, each block of the image is subjected to convolution operation by Conv2D convolution, and then tokens is generated by Flatten, and the inputs are set as follows: x belongs to R B*3*224*224 。
tokens＝PatchEmbed(x)+cls＝Flatten(Conv2D(x))+cls (2)
In formula (2), wherein:
tokens∈R B*N+1*C ；
C＝output channels Conv2d。
each Batch contains N tokens, and the corresponding artwork is divided into N uniform squares with side lengths equal to:
kernel size Conv2D each token dimension is C, where cls ∈ R B*1*C Is a token with position information, which is used to extract the information of the global token for classification.
And (3) carrying out inverse PatchEmbed operation on tokens output by the transducer model to obtain an image: x belongs to R B *3*224*224 Let the original image be: y is equal to R B*3*224*224 Calculating the MSE loss between x and y, as in equation (3):
where n is the number of pixels, x i Is a certain pixel value, y, in the restored image i Is the corresponding x in the original image i The pixel value of (2).
And setting the number of training rounds to iteratively update the transform weight until the MSE loss is minimum, wherein the MSE loss can indicate that the network has the capability of restoring the image, and the side indicates that the network effectively learns the inherent characteristics of the data.
And selecting 30% of data from the unlabeled data for manual labeling for self-supervision fine adjustment. Extracting cls epsilon R in output tokens B*1*C And obtaining score output of each class through MLP transformation, calculating cross entropy loss with the label, updating the weight, and iterating until the error is minimum to obtain a fine-tuned model with good classification precision.
According to the technical scheme provided by the invention, random masking operation is carried out on a certain amount of non-label pneumonia CT image data, the masked image is converted into token and is input into a transform model, and the MSE loss between the output and the original image is calculated, so that the transform model has certain resilience on the data. And then, marking related labels on part of unlabeled pneumonia CT image data, extracting transform model weight to perform fine tuning training of a multi-classification task, and taking a model with the highest accuracy on a verification set as a target classification model. The non-label data are usually cheap and available, self-supervision learning is carried out on a large number of non-label CT images, inherent abundant characteristics of the data are better learned, the characteristics have good generalization, migration training is carried out on the characteristics under a small number of labeled data sets, the identification capacity of the model on the pneumonia types is improved, the data labeling cost is effectively reduced, and the CT image pneumonia identification scheme is more efficient and economical.
Based on the same idea, the present invention further provides a lung pneumonia classification device by CT image, as shown in fig. 3, the device may include:
the to-be-identified lung CT image data obtaining module 310 is configured to obtain to-be-identified lung CT image data of the target object;
the target classification model identification module 320 is used for inputting the lung CT image data to be identified into a trained target classification model to obtain a classification result; the target classification model is obtained by performing self-supervision learning training by adopting non-label pneumonia CT image data; the classification includes at least normal lung, viral pneumonia, and bacterial pneumonia.
Based on the apparatus in fig. 3, some specific implementation units may also be included:
optionally, the apparatus may further include:
the data set acquisition module is used for acquiring a label-free pneumonia CT image data set;
the mask operation module is used for carrying out random mask operation on the image data in the non-tag pneumonia CT image data set to obtain masked image data;
the transform model primary identification module is used for converting the masked image data into label data, inputting the label data into a transform model and outputting target image data;
the MSE loss calculation module is used for calculating MSE loss between the target image data and the original image data;
and the weight adjusting module is used for adjusting the weight of the Transformer model based on the MSE loss, performing fine tuning training of a multi-classification task, and taking the model with the highest accuracy on the verification set as the target classification model.
Optionally, the transform model may include a multi-headed self-attention block and a multi-layered perceptron; the multi-layer perceptron may be composed of a hidden layer and an output layer.
Optionally, the data set obtaining module may specifically include:
the scene determining unit is used for determining a classification task scene corresponding to pneumonia;
and the data set acquisition unit is used for acquiring historical unlabeled pneumonia CT image data corresponding to the classification task scene to obtain an unlabeled pneumonia CT image data set.
Optionally, the mask operation module may be specifically configured to:
sending image data in the non-label pneumonia CT image data set into a random mask module;
and dividing the image corresponding to the image data into uniform squares to obtain the masked image data.
Optionally, the object classification model identification module 320 may be specifically configured to:
dividing the image data after the mask into blocks with fixed sizes, and performing convolution operation on each block corresponding to the image data by utilizing two-dimensional convolution;
carrying out one-dimensional processing on the data after the convolution operation to generate tokens;
and inputting the generated tokens into a Transformer model, and outputting target image data.
Optionally, a first part of data in the unlabeled pneumonia CT image data set may be used for an unsupervised pre-training of a classification model; the second part of the data in the unlabeled pneumonia CT image data set can be used for the self-supervision fine adjustment of the classification model after being labeled.
Based on the same idea, the embodiment of the present specification further provides a lung CT image pneumonia classification device.
As shown in fig. 4. The lung CT image pneumonia classification device can comprise:
the communication unit/communication interface is used for acquiring lung CT image data to be identified of the target object;
the processing unit/processor is used for inputting the lung CT image data to be recognized into a trained target classification model to obtain a classification result; the target classification model is obtained by performing self-supervision learning training by adopting non-label pneumonia CT image data; the classification includes at least normal lung, viral pneumonia, and bacterial pneumonia.
As shown in fig. 4, the terminal device may further include a communication line. The communication link may include a path for transmitting information between the aforementioned components.
Optionally, as shown in fig. 4, the terminal device may further include a memory. The memory is used for storing computer-executable instructions for implementing the inventive arrangements and is controlled by the processor for execution. The processor is used for executing the computer execution instructions stored in the memory, thereby realizing the method provided by the embodiment of the invention.
As shown in fig. 4, the memory may be a read-only memory (ROM) or other type of static storage device that may store static information and instructions, a Random Access Memory (RAM) or other type of dynamic storage device that may store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disk storage, optical disk storage (including compact disc, laser disc, optical disc, digital versatile disc, blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited to. The memory may be self-contained and coupled to the processor via a communication link. The memory may also be integral to the processor.
Optionally, the computer-executable instructions in the embodiment of the present invention may also be referred to as application program codes, which is not specifically limited in this embodiment of the present invention.
In one implementation, as shown in FIG. 4, a processor may include one or more CPUs, such as CPU0 and CPU1 of FIG. 4, for example.
In one embodiment, as shown in fig. 4, the terminal device may include a plurality of processors, such as the processor in fig. 4. Each of these processors may be a single core processor or a multi-core processor.
Based on the same idea, embodiments of the present specification further provide a computer storage medium corresponding to the foregoing embodiments, where the computer storage medium stores instructions, and when the instructions are executed, the method in the foregoing embodiments is implemented.
The above mainly introduces the scheme provided by the embodiment of the present invention from the perspective of interaction among the modules. It is understood that each module contains hardware structure and/or software unit for executing each function in order to realize the above functions. Those of skill in the art will readily appreciate that the present invention can be implemented in hardware or a combination of hardware and computer software, with the exemplary elements and algorithm steps described in connection with the embodiments disclosed herein. Whether a function is performed as hardware or computer software drives hardware depends upon the particular application and design constraints imposed on the solution. 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 invention.
The functional modules may be divided according to the above method examples, for example, the functional modules may be divided corresponding to the functions, or two or more functions may be integrated into one processing module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. It should be noted that, the division of the modules in the embodiment of the present invention is schematic, and is only one logic function division, and another division manner may be available in actual implementation.
The processor in this specification may also have the function of a memory. The memory is used for storing computer-executable instructions for implementing the inventive arrangements and is controlled for execution by the processor. The processor is used for executing the computer execution instructions stored in the memory, thereby realizing the method provided by the embodiment of the invention.
The memory may be, but is not limited to, a read-only memory (ROM) or other type of static storage device that may store static information and instructions, a Random Access Memory (RAM) or other type of dynamic storage device that may store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disk read-only memory (CD-ROM) or other optical disk storage, optical disk storage (including compact disk, laser disk, optical disk, digital versatile disk, blu-ray disk, etc.), magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. The memory may be separate and coupled to the processor via a communication link. The memory may also be integrated with the processor.
Optionally, the computer-executable instructions in the embodiment of the present invention may also be referred to as application program codes, which is not specifically limited in this embodiment of the present invention.
The method disclosed by the embodiment of the invention can be applied to a processor or realized by the processor. The processor may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in a processor or instructions in the form of software. The processor may be a general purpose processor, a Digital Signal Processor (DSP), an ASIC, an FPGA (field-programmable gate array) or other programmable logic device, discrete gate or transistor logic device, or discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in a memory, and a processor reads information in the memory and combines hardware thereof to complete the steps of the method.
In one possible implementation, a computer-readable storage medium is provided, in which instructions are stored, and when executed, are used to implement the method in the foregoing embodiments.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the procedures or functions described in the embodiments of the present invention are performed in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, a terminal, a user device, or other programmable apparatus. The computer program or instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer program or instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire or wirelessly. The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that integrates one or more available media. The usable medium may be a magnetic medium, such as a floppy disk, a hard disk, a magnetic tape; or optical media such as Digital Video Disks (DVDs); it may also be a semiconductor medium, such as a Solid State Drive (SSD).
While the invention has been described in connection with various embodiments, other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a review of the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the word "a" or "an" does not exclude a plurality. A single processor or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.
Although the present invention has been described in connection with the specific features and embodiments thereof, it is apparent that various modifications and combinations can be made thereto without departing from the spirit and scope of the invention. Accordingly, the specification and figures are merely exemplary of the invention as defined in the appended claims and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (10)
1. A lung CT image pneumonia classification method is characterized by comprising the following steps:
acquiring lung CT image data to be identified of a target object;
inputting the lung CT image data to be recognized into a trained target classification model to obtain a classification result; the target classification model is obtained by performing self-supervision learning training by adopting non-label pneumonia CT image data; the classification results include at least normal lung, viral pneumonia, and bacterial pneumonia.
2. The method as claimed in claim 1, wherein before inputting the lung CT image data to be identified into the trained target classification model and obtaining the classification result, the method further comprises:
acquiring a label-free pneumonia CT image data set;
carrying out random masking operation on the image data in the non-tag pneumonia CT image data set to obtain masked image data;
converting the masked image data into label data, inputting the label data into a Transformer model, and outputting target image data;
calculating MSE loss between the target image data and the original image data;
and adjusting the weight of the Transformer model based on the MSE loss, performing fine tuning training of a multi-classification task, and taking the model with the highest accuracy on a verification set as the target classification model.
3. The method of claim 2, wherein the Transformer model comprises a multi-headed self-attention block and a multi-layered perceptron; the multilayer perceptron is composed of a hidden layer and an output layer.
4. The method according to claim 2, wherein the acquiring the unlabeled pneumonia CT image data set specifically comprises:
determining a classification task scene corresponding to pneumonia;
and acquiring historical non-label pneumonia CT image data corresponding to the classification task scene to obtain a non-label pneumonia CT image data set.
5. The method according to claim 2, wherein the randomly masking the image data in the unlabeled pneumonia CT image data set to obtain masked image data specifically comprises:
sending image data in the non-tag pneumonia CT image data set into a random mask module;
and (3) dividing an image corresponding to the image data in the non-label pneumonia CT image data set into uniform squares to obtain masked image data.
6. The method of claim 2, wherein converting the masked image data into tag data, inputting the tag data into a transform model, and outputting target image data, specifically comprises:
dividing the masked image data into blocks with fixed sizes, and performing convolution operation on each block corresponding to the image data by utilizing two-dimensional convolution;
carrying out one-dimensional processing on the data after the convolution operation to generate tokens;
and inputting the generated tokens into a Transformer model, and outputting target image data.
7. The method of claim 4, wherein a first portion of data in the unlabeled pneumonia CT image data set is used for an unsupervised pre-training of a classification model; and marking a second part of data in the unlabeled pneumonia CT image data set for self-supervision fine adjustment of a classification model.
8. A lung CT image pneumonia classification device is characterized by comprising:
the lung CT image data acquisition module to be identified is used for acquiring lung CT image data to be identified of the target object;
the target classification model identification module is used for inputting the lung CT image data to be identified into a trained target classification model to obtain a classification result; the target classification model is obtained by performing self-supervision learning training by adopting non-label pneumonia CT image data; the classification results include at least normal lung, viral pneumonia, and bacterial pneumonia.
9. A lung CT image pneumonia classification device is characterized in that the device comprises:
the communication unit/communication interface is used for acquiring lung CT image data to be identified of the target object;
the processing unit/processor is used for inputting the lung CT image data to be recognized into a trained target classification model to obtain a classification result; the target classification model is obtained by performing self-supervision learning training by adopting non-label pneumonia CT image data; the classification results include at least normal lung, viral pneumonia, and bacterial pneumonia.
10. A computer storage medium having stored thereon instructions that, when executed, implement the lung CT image pneumonia classification method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211457766.3A CN115761342A (en) | 2022-11-21 | 2022-11-21 | Lung CT image pneumonia classification method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211457766.3A CN115761342A (en) | 2022-11-21 | 2022-11-21 | Lung CT image pneumonia classification method, device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115761342A true CN115761342A (en) | 2023-03-07 |
Family
ID=85333826
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211457766.3A Pending CN115761342A (en) | 2022-11-21 | 2022-11-21 | Lung CT image pneumonia classification method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115761342A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117496276A (en) * | 2023-12-29 | 2024-02-02 | 广州锟元方青医疗科技有限公司 | Lung cancer cell morphology analysis and identification method and computer readable storage medium |
-
2022
- 2022-11-21 CN CN202211457766.3A patent/CN115761342A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117496276A (en) * | 2023-12-29 | 2024-02-02 | 广州锟元方青医疗科技有限公司 | Lung cancer cell morphology analysis and identification method and computer readable storage medium |
| CN117496276B (en) * | 2023-12-29 | 2024-04-19 | 广州锟元方青医疗科技有限公司 | Lung cancer cell morphology analysis and identification method and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Detect globally, refine locally: A novel approach to saliency detection | |
| Min et al. | Hyperpixel flow: Semantic correspondence with multi-layer neural features | |
| US10410353B2 (en) | Multi-label semantic boundary detection system | |
| US10002313B2 (en) | Deeply learned convolutional neural networks (CNNS) for object localization and classification | |
| Shi et al. | Multiscale multitask deep NetVLAD for crowd counting | |
| US9767381B2 (en) | Similarity-based detection of prominent objects using deep CNN pooling layers as features | |
| US20170032222A1 (en) | Cross-trained convolutional neural networks using multimodal images | |
| CN108197532A (en) | The method, apparatus and computer installation of recognition of face | |
| CN110276406B (en) | Expression classification method, apparatus, computer device and storage medium | |
| US11640714B2 (en) | Video panoptic segmentation | |
| Wang et al. | Semi-supervised generative adversarial hashing for image retrieval | |
| CN109063719B (en) | Image classification method combining structure similarity and class information | |
| WO2022121485A1 (en) | Image multi-tag classification method and apparatus, computer device, and storage medium | |
| Huang et al. | Joint cross-modal and unimodal features for RGB-D salient object detection | |
| CN111182364B (en) | Short video copyright detection method and system | |
| CN112200041B (en) | Video motion recognition method and device, storage medium and electronic equipment | |
| Li et al. | Cross-level parallel network for crowd counting | |
| Maity et al. | A novel approach for human action recognition from silhouette images | |
| US20230104262A1 (en) | Panoptic segmentation refinement network | |
| Mao et al. | 3dg-stfm: 3d geometric guided student-teacher feature matching | |
| CN115761342A (en) | Lung CT image pneumonia classification method, device and equipment | |
| Maity et al. | Robust human action recognition using AREI features and trajectory analysis from silhouette image sequence | |
| Kapadia et al. | Improved CBIR system using Multilayer CNN | |
| CN112257628A (en) | Method, device and equipment for identifying identities of outdoor competition athletes | |
| Choudhury et al. | Scale aware deep pedestrian detection |
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 |