CN115578394B - Pneumonia image processing method based on asymmetric network - Google Patents
Pneumonia image processing method based on asymmetric network Download PDFInfo
- Publication number
- CN115578394B CN115578394B CN202211576953.3A CN202211576953A CN115578394B CN 115578394 B CN115578394 B CN 115578394B CN 202211576953 A CN202211576953 A CN 202211576953A CN 115578394 B CN115578394 B CN 115578394B
- Authority
- CN
- China
- Prior art keywords
- pneumonia
- identifier
- area
- network model
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30061—Lung
-
- 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 Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Databases & Information Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The application discloses a pneumonia image processing method based on an asymmetric network, which comprises the following steps: acquiring a pneumonia image to be processed; determining the pneumonia type corresponding to the pneumonia image to be processed according to a first network model trained in advance; the first network model is an asymmetric network; determining a lesion area in the pneumonia image to be processed according to a second network model trained in advance; the second network model is an asymmetric network; determining a classification identifier corresponding to the pneumonia image to be processed according to the pneumonia type corresponding to the pneumonia image to be processed and a lesion area in the pneumonia image to be processed; and performing annotation processing on the pneumonia image to be processed based on the classification identifier corresponding to the pneumonia image to be processed to obtain an annotated pneumonia image. The pneumonia image processing method can realize classification processing of pneumonia images based on an asymmetric network, and improves the applicability of the pneumonia images.
Description
Technical Field
The present application relates to the field of image processing technologies, and in particular, to a pneumonia image processing method based on an asymmetric network.
Background
With the development of image processing technology and medical imaging technology, in the medical field, image processing is usually performed based on medical images to obtain corresponding medical image processing results. The medical image processing results have great influence significance for doctors or patients.
In the related art, a large number of image processing methods are available for pneumonia images, the pneumonia images are processed and analyzed, and the corresponding processing and analysis results can be applied in a plurality of ways. However, in the related art, a classification processing scheme for these pneumonia images is lacking, and effective application of the pneumonia images is not facilitated.
Disclosure of Invention
The invention aims to provide a pneumonia image processing method based on an asymmetric network, which can realize classification processing of pneumonia images based on the asymmetric network and improve the applicability of the pneumonia images.
In order to achieve the above object, an embodiment of the present application provides an asymmetric network-based pneumonia image processing method, including: acquiring a pneumonia image to be processed; determining the pneumonia type corresponding to the pneumonia image to be processed according to a first network model trained in advance; the first network model is an asymmetric network; determining a lesion area in the pneumonia image to be processed according to a pre-trained second network model; the second network model is an asymmetric network; determining a classification identifier corresponding to the pneumonia image to be processed according to the pneumonia type corresponding to the pneumonia image to be processed and a lesion area in the pneumonia image to be processed; and performing labeling processing on the pneumonia image to be processed based on the classification identifier corresponding to the pneumonia image to be processed to obtain a labeled pneumonia image.
In one possible embodiment, the acquiring the pneumonia image to be processed includes: acquiring a plurality of original pneumonia images from a target database; the target database is used for storing medical image data; determining image similarity among a plurality of original pneumonia images; aiming at a plurality of original pneumonia images with image similarity larger than a first threshold value, determining a first number of original pneumonia images in the original pneumonia images as the pneumonia images to be processed; determining a second number of original pneumonia images in a plurality of original pneumonia images with image similarity smaller than or equal to the first threshold value as the pneumonia images to be processed; the second number is larger than the first number, and the sum of the second number and the first number is a preset value.
In one possible embodiment, the pneumonia image processing method further includes: acquiring a first training data set; the first training data set comprises a plurality of pneumonia images, and the pneumonia images are provided with pneumonia type labels; training an initial first network model based on the first training data set to obtain the trained first network model; when the initial first network model is trained, the initial first network model corresponds to preset training times, and when the training times of the initial first network model reach the preset training times, the training of the initial first network model is completed; alternatively, the pneumonia image processing method further includes: testing the trained first network model based on a first test data set; and carrying out optimization training on the trained first network model based on the test result.
In one possible embodiment, the pneumonia image processing method further includes: acquiring a second training data set; the second training data set comprises a plurality of pneumonia images, and lesion areas are marked in the pneumonia images; training an initial second network model based on the second training data set to obtain the trained second network model; when the initial second network model is trained, the initial second network model corresponds to preset training times, and when the training times of the initial second network model reach the preset training times, the training of the initial second network model is completed; alternatively, the pneumonia image processing method further includes: testing the trained second network model based on a second test data set; and carrying out optimization training on the trained second network model based on the test result.
In a possible implementation manner, the determining, according to the pneumonia type corresponding to the pneumonia image to be processed and the lesion area in the pneumonia image to be processed, a classification identifier corresponding to the pneumonia image to be processed includes: generating a lesion area identifier according to the area information of the lesion area in the pneumonia image to be processed; judging whether the pneumonia type corresponding to the pneumonia image to be processed is a target pneumonia type; if the pneumonia type corresponding to the pneumonia image to be processed is a target pneumonia type, acquiring a type identifier corresponding to the target pneumonia type; determining a classification identifier corresponding to the pneumonia image to be processed according to the lesion area identifier, the type identifier corresponding to the target pneumonia type, a preset target identifier and a first identifier generation algorithm; wherein the preset target identifier is: an identification generated based on an impact weight of the target pneumonia type on a pneumonia patient and a prevalence of the target pneumonia type; the impact weight is used to characterize the impact time and/or the impact range.
In one possible embodiment, the pneumonia image processing method further includes: if the pneumonia type corresponding to the pneumonia image to be processed is not the target pneumonia type, acquiring a type identifier corresponding to the pneumonia type corresponding to the pneumonia image to be processed; determining a classification identifier corresponding to the pneumonia image to be processed according to the lesion area identifier, a type identifier corresponding to the pneumonia type corresponding to the pneumonia image to be processed and a second identifier generation algorithm; the second identifier generation algorithm is different from the first identifier generation algorithm, and the complexity of the second identifier generation algorithm is lower than that of the first identifier generation algorithm.
In one possible embodiment, the region information includes: region area, region position, and region shape; generating a lesion area identifier according to the area information of the lesion area in the pneumonia image to be processed, including: determining a region area identifier according to the region area; determining an area position identifier according to the area position; determining a region shape identifier according to the region shape; and generating the lesion area identifier according to the area identifier, the area position identifier and the area shape identifier.
In a possible implementation, the determining a region area identifier according to the region area includes: if the area of the region is larger than a preset area, determining that the area identifier of the region is a first area identifier; if the area of the region is smaller than or equal to the preset area, determining that the area identifier of the region is a second area identifier; the determining of the area position identifier according to the area position comprises: if the area position belongs to a preset position range, determining the area position identification as a first area position identification; if the area position does not belong to the preset position range, determining the area position identification as a second area position identification; the determining the region shape identifier according to the region shape includes: if the region shape belongs to a preset shape set, determining that the region shape identifier is a first region shape identifier; and if the region shape does not belong to a preset shape set, determining that the region shape identifier is a second region shape identifier.
In one possible embodiment, the generating the lesion region identifier according to the region area identifier, the region position identifier and the region shape identifier includes: judging whether repeated marks exist in the area marks, the area position marks and the area shape marks or not; if repeated marks exist in the area mark, the area position mark and the area shape mark, generating a first lesion area mark based on the repeated marks and the repeated times of the repeated marks; generating a second lesion area identifier according to the area identifier, the area position identifier and other identifiers except the repeated identifier in the area shape identifier; generating the lesion area identifier according to the first lesion area identifier, the second lesion area identifier and a third identifier generation algorithm.
In one possible embodiment, the pneumonia image processing method further includes: generating a third training data set based on the labeled pneumonia image; training an initial third network model based on the third training data set to obtain a trained third network model; the third network model is an asymmetric network, and the trained third network model is used for labeling the pneumonia image; when the initial third network model is trained, the initial third network model corresponds to preset training times, and when the training times of the initial third network model reach the preset training times, the training of the initial third network model is completed; alternatively, the pneumonia image processing method further includes: testing the trained third network model based on a third test data set; and carrying out optimization training on the trained third network model based on the test result.
An embodiment of the present application provides a pneumonia image processing apparatus based on an asymmetric network, including: the functional modules are used for realizing the method and one or more corresponding embodiments of the pneumonia image processing method based on the asymmetric network.
An embodiment of the present application further provides an electronic device, including: a processor and a memory, the processor and the memory communicatively coupled; wherein the memory stores instructions executable by the processor to enable the processor to perform the method for processing an asymmetric network-based pneumonia image according to any one of the above embodiments.
An embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a computer, the method for processing an asymmetric network-based pneumonia image according to any one of the above-mentioned embodiments is executed.
Compared with the prior art, according to the pneumonia image processing method and device based on the asymmetric network, the electronic equipment and the computer readable storage medium, after the pneumonia image to be processed is obtained, the corresponding pneumonia type is determined according to the first network model; determining a lesion area according to the second network model; the first network model and the second network model are both asymmetric networks; and further, determining a classification identifier according to the determined pneumonia type and the lesion area, and labeling the pneumonia image by using the classification identifier to realize classification processing of the pneumonia image. Therefore, the pneumonia image processing scheme realizes the determination of the classification identification of the pneumonia image based on the prediction result of the asymmetric network, further realizes the classification processing of the pneumonia image based on the asymmetric network, and improves the applicability of the pneumonia image.
Drawings
Fig. 1 is a flowchart of an asymmetric network-based pneumonia image processing method according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of an asymmetric network-based pneumonia image processing apparatus according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The following detailed description of embodiments of the present application is provided in conjunction with the accompanying drawings, but it should be understood that the scope of the present application is not limited to the specific embodiments.
Throughout the specification and claims, unless explicitly stated otherwise, the word "comprise", or variations such as "comprises" or "comprising", will be understood to imply the inclusion of a stated element or component but not the exclusion of any other element or component.
The technical scheme provided by the embodiment of the application can be applied to various application scenes needing pneumonia image processing. In some application scenarios, the pneumonia images need to be classified, and the technical scheme provided by the embodiment of the application can be applied to realize the classification of the pneumonia images. In other embodiments, the pneumonia images need to be put into storage, and at this time, the technical scheme provided by the embodiment of the present application may be used to add the classification identifiers of the pneumonia images, so that the pneumonia images are put into storage based on the classification identifiers.
In the related art, the classification or warehousing of pneumonia images generally depends on manual work; alternatively, the pneumonia image is not subjected to classification processing. However, with the increase of patients with pneumonia, classification processing of pneumonia images is of great significance to both pneumonia patients and physicians.
Based on this, the embodiment of the application provides a pneumonia image processing scheme based on an asymmetric network, and after a pneumonia image to be processed is obtained, a corresponding pneumonia type is determined according to a first network model; determining a lesion area according to the second network model; the first network model and the second network model are both asymmetric networks; and further, determining a classification identifier according to the determined pneumonia type and the lesion area, and labeling the pneumonia image by using the classification identifier to realize classification processing of the pneumonia image. Therefore, the pneumonia image processing scheme realizes the determination of the classification identification of the pneumonia image based on the prediction result of the asymmetric network, further realizes the classification processing of the pneumonia image based on the asymmetric network, and improves the applicability of the pneumonia image.
Therefore, the asymmetric network-based pneumonia image processing can be applied to a pneumonia image processing system, and the hardware form of the pneumonia image processing system can include but is not limited to: terminal device, client, server, browser, etc., without limitation thereto.
Referring to fig. 1, a flowchart of a pneumonia image processing method based on an asymmetric network according to an embodiment of the present application is shown, where the pneumonia image processing method includes:
In some embodiments, the number of pneumonia images to be processed is one, and in this case, the pneumonia image is processed. In other embodiments, the number of pneumonia images to be processed is multiple, and in this case, the multiple pneumonia images may be processed in the same processing manner.
As an alternative implementation, step 101 includes: acquiring a plurality of original pneumonia images from a target database; the target database is used for storing medical image data; determining image similarity among a plurality of original pneumonia images; aiming at a plurality of original pneumonia images with image similarity larger than a first threshold value, determining a first number of original pneumonia images in the original pneumonia images as pneumonia images to be processed; determining a second number of original pneumonia images in a plurality of original pneumonia images with image similarity smaller than or equal to a first threshold value as pneumonia images to be processed; the second number is larger than the first number, and the sum of the second number and the first number is a preset value.
In some embodiments, the target database is used to store various medical image data, such as: lung image data, brain image data, stomach image data, and the like. The target database may be a database of a medical institution or a database of a public medical network. Accordingly, medical image data can be acquired from the corresponding medical institution, and a doctor or a patient having authority.
However, the data acquired from the target database are all raw data, and are not subjected to corresponding image processing. Therefore, some image preprocessing or image screening processes are required in combination with the acquired original pneumonia images.
Further, image similarity between a plurality of original pneumonia images is calculated, and here, the image similarity can be understood as an average value of the image similarity between one original pneumonia image and other original pneumonia images. The image similarity can be used for representing the specificity of one original pneumonia image in a plurality of original pneumonia images.
Further, for those original pneumonia images whose image similarity is greater than the first threshold, a first number of the original pneumonia images may be determined as pneumonia images to be processed.
The first number can be set by combining different application scenes; for example: if there are 100 pneumonia images to be processed, the first number may be 40. And when the first number of original pneumonia images are selected, selecting according to the similarity of the images, and preferentially selecting the original pneumonia images with low image similarity as pneumonia images to be processed.
Correspondingly, the first threshold is also set in conjunction with different application scenarios, for example: the first threshold may be fifty percent.
And determining a second number of original pneumonia images as pneumonia images to be processed aiming at a plurality of original pneumonia images with image similarity smaller than or equal to a first threshold value.
The second quantity is larger than the first quantity, and the sum of the second quantity and the first quantity is a preset value. The preset value may be a value greater than or equal to the number of pneumonia images to be processed; and, the second number may be determined in conjunction with the number of pneumonia images to be processed, for example: the number of pneumonia images to be processed is 100, the second number may be 60.
Further, a plurality of pneumonia images to be processed can be determined by combining two different image similarity conditions.
In other embodiments, in addition to performing the screening process of the image to be processed, the image to be processed may be preprocessed, for example: image enhancement processing, image scaling processing, etc., to facilitate subsequent further processing.
And 102, determining the pneumonia type corresponding to the pneumonia image to be processed according to a pre-trained first network model. Wherein the first network model is an asymmetric network.
In some embodiments, the asymmetric network may be an asymmetric feedback neural network, an asymmetric concatenated neural network, an asymmetric dual-ring neural network, or the like, which is not limited herein.
Compared with the general neural network model, the asymmetric neural network model has higher recognizable image resolution and more accurate corresponding recognition result.
And the first network model is used for determining the pneumonia type corresponding to the pneumonia image to be processed. Types of pneumonia for example: new coronary pneumonia, common cold type pneumonia, nonspecific pneumonia, etc.
As an alternative embodiment, the training process of the first network model includes:
acquiring a first training data set; the first training data set comprises a plurality of pneumonia images, and the pneumonia images are provided with pneumonia type labels; training an initial first network model based on a first training data set to obtain a trained first network model; when the training times of the initial first network model reach the preset training times, the training of the initial first network model is completed; alternatively, the pneumonia image processing method further includes: testing the trained first network model based on the first test data set; and carrying out optimization training on the trained first network model based on the test result.
In some embodiments, the preset number of training times may be set in combination with the accuracy requirement, and the higher the accuracy requirement, the more the preset number of training times.
And, a first test data set, which may be part of a first training data set; or may be a data set provided together with the first training data set.
Therefore, the pneumonia image to be processed is input into the pre-trained first network model, and the pre-trained first network model can output the corresponding pneumonia type identification result.
And 103, determining a lesion area in the pneumonia image to be processed according to a second network model trained in advance. Wherein the second network model is an asymmetric network.
In some embodiments, the asymmetric network may be an asymmetric feedback neural network, an asymmetric concatenated neural network, an asymmetric dual-ring neural network, or the like, which is not limited herein. And the second network model and the first network model may adopt different asymmetric networks or the same asymmetric network.
And the second network model is used for determining a lesion area in the pneumonia image to be processed. For example: lesion in the center of the lung, lesion on both sides of the lung, etc.
As an alternative embodiment, the training process of the second network model includes: acquiring a second training data set; the second training data set comprises a plurality of pneumonia images, and lesion areas are marked in the pneumonia images; training the initial second network model based on a second training data set to obtain a trained second network model; when the training times of the initial second network model reach the preset training times, the training of the initial second network model is completed; alternatively, the pneumonia image processing method further includes: testing the trained second network model based on a second test data set; and carrying out optimization training on the trained second network model based on the test result.
In some embodiments, the preset number of training times may be set in combination with the accuracy requirement, and the higher the accuracy requirement, the more the preset number of training times.
And, a second set of test data, which may be part of a second set of training data; or may be a data set provided together with the second training data set.
Therefore, the pneumonia image to be processed is input into the pre-trained second network model, and the pre-trained second network model can output a corresponding pneumonia type identification result.
And 104, determining a classification identifier corresponding to the pneumonia image to be processed according to the pneumonia type corresponding to the pneumonia image to be processed and the lesion area in the pneumonia image to be processed.
As an alternative embodiment, step 104 includes: generating a lesion area identifier according to the area information of a lesion area in the pneumonia image to be processed; judging whether the pneumonia type corresponding to the pneumonia image to be processed is a target pneumonia type; if the pneumonia type corresponding to the pneumonia image to be processed is the target pneumonia type, obtaining a type identifier corresponding to the target pneumonia type; determining a classification identifier corresponding to the pneumonia image to be processed according to the lesion area identifier, the type identifier corresponding to the target pneumonia type, a preset target identifier and a first identifier generation algorithm; wherein the preset target identification is: an identification generated based on a weight of the impact of the target pneumonia type on the pneumonia patient and the prevalence of the target pneumonia type; the impact weight is used to characterize the impact time and/or impact range.
In this embodiment, the lesion area identifier is generated according to the area information of the lesion area in the pneumonia image to be processed.
As an optional implementation, the region information includes: region area, region position, and region shape; generating a lesion area identifier according to the area information of the lesion area in the pneumonia image to be processed, wherein the lesion area identifier comprises: determining a region area identifier according to the region area; determining a region position identifier according to the region position; determining a region shape identifier according to the region shape; and generating a lesion area identifier according to the area identifier, the area position identifier and the area shape identifier.
In this embodiment, first, according to different area information, an area information identifier corresponding to the area information is generated; and integrating the plurality of region information identifications to generate a lesion region identification.
As an optional implementation, determining the region area identifier according to the region area includes: if the area of the region is larger than the preset area, determining the area identifier of the region as a first area identifier; and if the area of the region is smaller than or equal to the preset area, determining the area identifier of the region as a second region area identifier.
In this embodiment, the preset area may be set in combination with different application scenarios, which may be set with different zone areas affecting the pneumonia patient.
And the first area identifier and the second area identifier can be set by combining different application scenes, and the two identifiers are different only by ensuring. For example: the first region area indicator may be a first character combination, the second region area indicator may be a second character combination, and so on.
As an optional implementation, determining the area location identifier according to the area location includes: if the area position belongs to the preset position range, determining the area position mark as a first area position mark; and if the area position does not belong to the preset position range, determining the area position mark as a second area position mark.
The preset position range may be set in combination with different application scenarios, for example: may be a fixed position range on both sides of the lung; which can be set with different location ranges for the degree of influence on the pneumonia patient.
And the first area position identification and the second area position identification can be set by combining different application scenes, and the two identifications are only required to be different. For example: the first region position identification may be a third character combination, the second region position identification may be a fourth character combination, and so on.
As an optional implementation, determining the region shape identifier according to the region shape includes: if the region shape belongs to the preset shape set, determining that the region shape identifier is a first region shape identifier; and if the region shape does not belong to the preset shape set, determining the region shape identifier as a second region shape identifier.
The preset shape set may be set in combination with different application scenarios, for example: may be a combination of shapes including a plurality of different shapes; which can be set with different zone shapes to the extent of the pneumonia patient's influence.
And the first area shape identifier and the second area shape identifier can be set by combining different application scenes, and the two identifiers are different only by ensuring. For example: the first region shape identification may be a fifth character combination, the second region shape identification may be a sixth character combination, and so on.
Further, as an optional implementation, generating a lesion region identifier according to the region area identifier, the region position identifier, and the region shape identifier includes: judging whether repeated marks exist in the area marks, the area position marks and the area shape marks or not; if repeated marks exist in the area mark, the area position mark and the area shape mark, generating a first lesion area mark based on the repeated marks and the repeated times of the repeated marks; generating a second lesion area identifier according to other identifiers except the repeated identifier in the area identifier, the area position identifier and the area shape identifier; and generating a lesion area identifier according to the first lesion area identifier, the second lesion area identifier and a third identifier generation algorithm.
In some embodiments, the first lesion area is identified as an identification after adding a character corresponding to the number of repetitions in the repetition identification.
In some embodiments, the region area identifier, the region position identifier, and the region shape identifier, except for the repeated identifier, are randomly combined to generate the second lesion region identifier.
In some embodiments, the third identity generation algorithm may be, for example: a hash algorithm, a standard character generation algorithm, etc., and are not limited herein.
Further, after the lesion area identification is generated, whether the pneumonia type corresponding to the pneumonia image to be processed is the target pneumonia type or not is judged. The target pneumonia type can be typical or pneumonia type with high incidence.
In some embodiments, if the pneumonia type corresponding to the pneumonia image to be processed is the target pneumonia type, obtaining a type identifier corresponding to the target pneumonia type; and determining a classification identifier corresponding to the pneumonia image to be processed according to the lesion area identifier, the type identifier corresponding to the target pneumonia type, a preset target identifier and a first identifier generation algorithm.
In some embodiments, the preset target identification is: an identification generated based on an impact weight of the target pneumonia type on the pneumonia patient and a prevalence rate of the target pneumonia type; the impact weight is used to characterize the impact time and/or impact range.
The influence weight may be a weight value determined by counting the influence time and/or the influence range of the target pneumonia type.
In some embodiments, the markers corresponding to the influence weight and the prevalence rate may be determined separately, and then the two markers are randomly combined or integrated to generate the target marker.
Further, the type identifier corresponding to the target pneumonia type, which is usually a label prescribed in the medical field, may be different in combination with different application scenarios.
In some embodiments, the implementation of the first identifier generation algorithm may refer to the implementation of the third identifier generation algorithm, and the first identifier generation algorithm and the third identifier generation algorithm may be the same or different.
In some embodiments, if the pneumonia type corresponding to the pneumonia image to be processed is not the target pneumonia type, obtaining a type identifier corresponding to the pneumonia type corresponding to the pneumonia image to be processed; determining a classification identifier corresponding to the pneumonia image to be processed according to the lesion area identifier, the type identifier corresponding to the pneumonia type corresponding to the pneumonia image to be processed and a second identifier generation algorithm; the second identifier generation algorithm is different from the first identifier generation algorithm, and the complexity of the second identifier generation algorithm is lower than that of the first identifier generation algorithm.
In this embodiment, the type identifier corresponding to the pneumonia type corresponding to the pneumonia image to be processed, as well as the type identifier corresponding to the target pneumonia type, may be determined based on the regulations in the medical field, and may be different in different application scenarios.
And 105, performing labeling processing on the pneumonia image to be processed based on the classification identifier corresponding to the pneumonia image to be processed to obtain a labeled pneumonia image.
In some embodiments, the classification identifier may be directly annotated in the pneumonia image to be processed.
In other embodiments, a corresponding image identifier is generated based on the classification identifier and then added to the pneumonia image to be processed.
Further, after the annotation-processed pneumonia image is obtained, it may be stored; when stored, images with the same identification may be stored in the same database.
In some embodiments, the pneumonia image processing method further includes: generating a third training data set based on the labeled pneumonia image; training the initial third network model based on a third training data set to obtain a trained third network model; the third network model is an asymmetric network, and the trained third network model is used for labeling the pneumonia image; when the initial third network model is trained, the initial third network model corresponds to preset training times, and when the training times of the initial third network model reach the preset training times, the training of the initial third network model is completed; alternatively, the pneumonia image processing method further includes: testing the trained third network model based on a third test data set; and performing optimization training on the trained third network model based on the test result.
In this embodiment, a training data set is generated based on the pneumonia image subjected to labeling processing, and the third network model is trained, so that the trained third network model can be directly used for the classification labeling processing of the pneumonia image.
As can be seen from the introduction of the foregoing embodiment, in the pneumonia image processing scheme provided in the embodiment of the present application, after obtaining a pneumonia image to be processed, a corresponding pneumonia type is determined according to a first network model; determining a lesion area according to the second network model; the first network model and the second network model are both asymmetric networks; and further, determining a classification identifier according to the determined pneumonia type and the lesion area, and labeling the pneumonia image by using the classification identifier to realize classification processing of the pneumonia image. Therefore, the pneumonia image processing scheme realizes the determination of the classification identification of the pneumonia image based on the prediction result of the asymmetric network, further realizes the classification processing of the pneumonia image based on the asymmetric network, and improves the applicability of the pneumonia image.
Referring to fig. 2, an asymmetric network-based pneumonia image processing apparatus according to an embodiment of the present application includes:
an obtaining module 201, configured to obtain a pneumonia image to be processed;
the processing module 202 is configured to determine a pneumonia type corresponding to the pneumonia image to be processed according to a pre-trained first network model; the first network model is an asymmetric network; determining a lesion area in the pneumonia image to be processed according to a pre-trained second network model; the second network model is an asymmetric network; determining a classification identifier corresponding to the pneumonia image to be processed according to the pneumonia type corresponding to the pneumonia image to be processed and a lesion area in the pneumonia image to be processed; and performing labeling processing on the pneumonia image to be processed based on the classification identifier corresponding to the pneumonia image to be processed to obtain a labeled pneumonia image.
In some embodiments, the obtaining module 201 is further configured to: acquiring a plurality of original pneumonia images from a target database; the target database is used for storing medical image data; determining image similarity among a plurality of original pneumonia images; aiming at a plurality of original pneumonia images with image similarity larger than a first threshold value, determining a first number of original pneumonia images in the original pneumonia images as the pneumonia images to be processed; determining a second number of original pneumonia images in a plurality of original pneumonia images with image similarity smaller than or equal to the first threshold value as the pneumonia images to be processed; the second number is larger than the first number, and the sum of the second number and the first number is a preset value.
In some embodiments, the obtaining module 201 is further configured to obtain a first training data set; the first training data set comprises a plurality of pneumonia images, and the pneumonia images are provided with pneumonia type labels; the processing module 202 is further configured to: training an initial first network model based on the first training data set to obtain the trained first network model; when the initial first network model is trained, the initial first network model corresponds to preset training times, and when the training times of the initial first network model reach the preset training times, the training of the initial first network model is completed; or testing the trained first network model based on a first test data set; and carrying out optimization training on the trained first network model based on the test result.
In some embodiments, the obtaining module 201 is further configured to obtain a second training data set; the second training data set comprises a plurality of pneumonia images, and lesion areas are marked in the pneumonia images; the processing module 202 is further configured to: training an initial second network model based on the second training data set to obtain the trained second network model; when the initial second network model is trained, the initial second network model corresponds to preset training times, and when the training times of the initial second network model reach the preset training times, the training of the initial second network model is completed; or testing the trained second network model based on a second test data set; and carrying out optimization training on the trained second network model based on the test result.
In some embodiments, the processing module 202 is further configured to: generating a lesion area identifier according to the area information of the lesion area in the pneumonia image to be processed; judging whether the pneumonia type corresponding to the pneumonia image to be processed is a target pneumonia type; if the pneumonia type corresponding to the pneumonia image to be processed is a target pneumonia type, acquiring a type identifier corresponding to the target pneumonia type; determining a classification identifier corresponding to the pneumonia image to be processed according to the lesion area identifier, the type identifier corresponding to the target pneumonia type, a preset target identifier and a first identifier generation algorithm; wherein the preset target identifier is: an identification generated based on an impact weight of the target pneumonia type on a pneumonia patient and a prevalence of the target pneumonia type; the impact weight is used to characterize the impact time and/or the impact range.
In some embodiments, the processing module 202 is further configured to: if the pneumonia type corresponding to the pneumonia image to be processed is not the target pneumonia type, acquiring a type identifier corresponding to the pneumonia type corresponding to the pneumonia image to be processed; determining a classification identifier corresponding to the pneumonia image to be processed according to the lesion area identifier, a type identifier corresponding to the pneumonia type corresponding to the pneumonia image to be processed and a second identifier generation algorithm; the second identifier generation algorithm is different from the first identifier generation algorithm, and the complexity of the second identifier generation algorithm is lower than that of the first identifier generation algorithm.
In some embodiments, the region information comprises: region area, region position, and region shape; the processing module 202 is further configured to: determining a region area identifier according to the region area; determining a region position identifier according to the region position; determining an area shape identifier according to the area shape; and generating the lesion area identifier according to the area identifier, the area position identifier and the area shape identifier.
In some embodiments, the processing module 202 is further configured to: if the area of the region is larger than a preset area, determining that the area identifier of the region is a first area identifier; if the area of the region is smaller than or equal to the preset area, determining that the area identifier of the region is a second area identifier; if the area position belongs to a preset position range, determining the area position identification as a first area position identification; if the area position does not belong to the preset position range, determining the area position identifier as a second area position identifier; if the region shape belongs to a preset shape set, determining that the region shape identifier is a first region shape identifier; and if the region shape does not belong to the preset shape set, determining that the region shape identifier is a second region shape identifier.
In some embodiments, the processing module 202 is further configured to: judging whether repeated marks exist in the area marks, the area position marks and the area shape marks or not; if repeated marks exist in the area mark, the area position mark and the area shape mark, generating a first lesion area mark based on the repeated marks and the repeated times of the repeated marks; generating a second lesion area identifier according to the area identifier, the area position identifier and other identifiers except the repeated identifier in the area shape identifier; generating the lesion area identifier according to the first lesion area identifier, the second lesion area identifier and a third identifier generation algorithm.
In some embodiments, the processing module 202 is further configured to: generating a third training data set based on the labeled pneumonia images; training an initial third network model based on the third training data set to obtain a trained third network model; the third network model is an asymmetric network, and the trained third network model is used for labeling the pneumonia image; when the initial third network model is trained, the initial third network model corresponds to a preset training time, and when the training time of the initial third network model reaches the preset training time, the training of the initial third network model is completed; or testing the trained third network model based on a third test data set; and carrying out optimization training on the trained third network model based on the test result.
It can be understood that the pneumonia image processing device based on the asymmetric network corresponds to the pneumonia image processing method based on the asymmetric network, and therefore, the implementation of each functional module is referred to the foregoing embodiment, and will not be described repeatedly herein.
Referring to fig. 3, an embodiment of the present application further provides an electronic device, including: a processor 301 and a memory 302, the processor 301 and the memory 302 being communicatively coupled. The electronic equipment can be used as an execution main body of the pneumonia image processing method based on the asymmetric network.
The memory 302 stores instructions executable by the processor 301, and the instructions are executed by the processor 301, so that the processor 301 can execute the asymmetric network-based pneumonia image processing method described in the foregoing embodiment.
In some embodiments, the processor 301 and the memory 302 are communicatively coupled via a communication bus.
It is understood that the electronic device may further include more required general modules, which are not described in the embodiments of the present application.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing descriptions of specific exemplary embodiments of the present application have been presented for purposes of illustration and description. It is not intended to limit the application to the precise form disclosed, and obviously many modifications and variations are possible in light of the above teaching. The exemplary embodiments were chosen and described in order to explain certain principles of the present application and its practical application to enable one skilled in the art to make and use various exemplary embodiments of the present application and various alternatives and modifications thereof. It is intended that the scope of the application be defined by the claims and their equivalents.
Claims (9)
1. A pneumonia image processing method based on an asymmetric network is characterized by comprising the following steps:
acquiring a pneumonia image to be processed;
determining the pneumonia type corresponding to the pneumonia image to be processed according to a first network model trained in advance; the first network model is an asymmetric network;
determining a lesion area in the pneumonia image to be processed according to a second network model trained in advance; the second network model is an asymmetric network;
determining a classification identifier corresponding to the pneumonia image to be processed according to the pneumonia type corresponding to the pneumonia image to be processed and a lesion area in the pneumonia image to be processed;
performing labeling processing on the pneumonia image to be processed based on the classification identifier corresponding to the pneumonia image to be processed to obtain a labeled pneumonia image;
the acquiring of the pneumonia image to be processed comprises the following steps:
acquiring a plurality of original pneumonia images from a target database; the target database is used for storing medical image data;
determining image similarity among a plurality of original pneumonia images;
determining a first number of original pneumonia images in a plurality of original pneumonia images with image similarity larger than a first threshold value as the pneumonia images to be processed;
determining a second number of original pneumonia images in a plurality of original pneumonia images with image similarity smaller than or equal to the first threshold value as the pneumonia images to be processed; the second number is larger than the first number, and the sum of the second number and the first number is a preset value.
2. The pneumonia image processing method according to claim 1, characterized in that the pneumonia image processing method further includes:
acquiring a first training data set; the first training data set comprises a plurality of pneumonia images, and the pneumonia images are provided with pneumonia type labels;
training an initial first network model based on the first training data set to obtain the trained first network model;
when the initial first network model is trained, the initial first network model corresponds to preset training times, and when the training times of the initial first network model reach the preset training times, the training of the initial first network model is completed;
alternatively, the pneumonia image processing method further includes: testing the trained first network model based on a first test data set; and carrying out optimization training on the trained first network model based on the test result.
3. The pneumonia image processing method according to claim 1, characterized in that the pneumonia image processing method further includes:
acquiring a second training data set; the second training data set comprises a plurality of pneumonia images, and lesion areas are marked in the pneumonia images;
training an initial second network model based on the second training data set to obtain the trained second network model;
when the initial second network model is trained, the initial second network model corresponds to preset training times, and when the training times of the initial second network model reach the preset training times, the training of the initial second network model is completed;
or, the pneumonia image processing method further comprises: testing the trained second network model based on a second test data set; and carrying out optimization training on the trained second network model based on the test result.
4. The pneumonia image processing method according to claim 1, wherein the determining the classification identifier corresponding to the pneumonia image to be processed according to the pneumonia type corresponding to the pneumonia image to be processed and the lesion area in the pneumonia image to be processed comprises:
generating a lesion area identifier according to the area information of the lesion area in the pneumonia image to be processed;
judging whether the pneumonia type corresponding to the pneumonia image to be processed is a target pneumonia type;
if the pneumonia type corresponding to the pneumonia image to be processed is a target pneumonia type, acquiring a type identifier corresponding to the target pneumonia type;
determining a classification identifier corresponding to the pneumonia image to be processed according to the lesion area identifier, the type identifier corresponding to the target pneumonia type, a preset target identifier and a first identifier generation algorithm; wherein the preset target identifier is: an identification generated based on an impact weight of the target pneumonia type on a pneumonia patient and a prevalence of the target pneumonia type; the impact weight is used to characterize the impact time and/or the impact range.
5. The pneumonia image processing method according to claim 4, characterized in that the pneumonia image processing method further includes:
if the pneumonia type corresponding to the pneumonia image to be processed is not the target pneumonia type, acquiring a type identifier corresponding to the pneumonia type corresponding to the pneumonia image to be processed;
determining a classification identifier corresponding to the pneumonia image to be processed according to the lesion area identifier, a type identifier corresponding to the pneumonia type corresponding to the pneumonia image to be processed and a second identifier generation algorithm; the second identifier generation algorithm is different from the first identifier generation algorithm, and the complexity of the second identifier generation algorithm is lower than that of the first identifier generation algorithm.
6. The pneumonia image processing method according to claim 4, characterized in that the area information includes: region area, region position, and region shape; generating a lesion area identifier according to the area information of the lesion area in the pneumonia image to be processed, including:
determining a region area identifier according to the region area;
determining an area position identifier according to the area position;
determining a region shape identifier according to the region shape;
and generating the lesion area identifier according to the area identifier, the area position identifier and the area shape identifier.
7. The pneumonia image processing method according to claim 6,
the determining the area identifier according to the area comprises: if the area of the region is larger than a preset area, determining that the area identifier of the region is a first area identifier; if the area of the region is smaller than or equal to the preset area, determining that the area identifier of the region is a second area identifier;
the determining the area position identifier according to the area position includes: if the area position belongs to a preset position range, determining the area position identification as a first area position identification; if the area position does not belong to the preset position range, determining the area position identifier as a second area position identifier;
the determining the region shape identifier according to the region shape includes: if the region shape belongs to a preset shape set, determining that the region shape identifier is a first region shape identifier; and if the region shape does not belong to a preset shape set, determining that the region shape identifier is a second region shape identifier.
8. The pneumonia image processing method according to claim 6, wherein said generating the lesion region identification from the region area identification, the region position identification and the region shape identification includes:
judging whether a repeated mark exists in the area mark, the area position mark and the area shape mark;
if repeated marks exist in the area mark, the area position mark and the area shape mark, generating a first lesion area mark based on the repeated marks and the repeated times of the repeated marks;
generating a second lesion area identifier according to the area identifier, the area position identifier and other identifiers except the repeated identifier in the area shape identifier;
generating the lesion area identifier according to the first lesion area identifier, the second lesion area identifier and a third identifier generation algorithm.
9. The pneumonia image processing method according to claim 1, characterized in that the pneumonia image processing method further includes:
generating a third training data set based on the labeled pneumonia images;
training an initial third network model based on the third training data set to obtain a trained third network model; the third network model is an asymmetric network, and the trained third network model is used for labeling the pneumonia image;
when the initial third network model is trained, the initial third network model corresponds to a preset training time, and when the training time of the initial third network model reaches the preset training time, the training of the initial third network model is completed;
alternatively, the pneumonia image processing method further includes: testing the trained third network model based on a third test data set; and carrying out optimization training on the trained third network model based on the test result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211576953.3A CN115578394B (en) | 2022-12-09 | 2022-12-09 | Pneumonia image processing method based on asymmetric network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211576953.3A CN115578394B (en) | 2022-12-09 | 2022-12-09 | Pneumonia image processing method based on asymmetric network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115578394A CN115578394A (en) | 2023-01-06 |
| CN115578394B true CN115578394B (en) | 2023-04-07 |
Family
ID=84590094
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211576953.3A Active CN115578394B (en) | 2022-12-09 | 2022-12-09 | Pneumonia image processing method based on asymmetric network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115578394B (en) |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108665456B (en) * | 2018-05-15 | 2022-01-28 | 广州尚医网信息技术有限公司 | Method and system for real-time marking of breast ultrasound lesion region based on artificial intelligence |
| CN109919149A (en) * | 2019-01-18 | 2019-06-21 | 平安科技（深圳）有限公司 | Object mask method and relevant device based on object detection model |
| CN110473192B (en) * | 2019-04-10 | 2021-05-14 | 腾讯医疗健康(深圳)有限公司 | Digestive tract endoscope image recognition model training and recognition method, device and system |
| CN113392294B (en) * | 2020-10-15 | 2023-11-10 | 腾讯科技（深圳）有限公司 | Sample labeling method and device |
| CN115239945A (en) * | 2022-06-15 | 2022-10-25 | 中国医学科学院北京协和医院 | Image annotation reliability prediction method and device, electronic equipment and storage medium |
-
2022
- 2022-12-09 CN CN202211576953.3A patent/CN115578394B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115578394A (en) | 2023-01-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11004234B2 (en) | Method and apparatus for annotating point cloud data | |
| CN110363220B (en) | Behavior class detection method and device, electronic equipment and computer readable medium | |
| CN111428448B (en) | Text generation method, device, computer equipment and readable storage medium | |
| CN111415336B (en) | Image tampering identification method, device, server and storage medium | |
| CN112668453B (en) | Video identification method and related equipment | |
| CN110706121B (en) | Method and device for determining medical insurance fraud result, electronic equipment and storage medium | |
| CN113938408B (en) | Data traffic testing method and device, server and storage medium | |
| CN114639152A (en) | Multi-modal voice interaction method, device, equipment and medium based on face recognition | |
| CN111159241A (en) | Click conversion estimation method and device | |
| CN114418398A (en) | Scene task development method, device, equipment and storage medium | |
| CN113537207B (en) | Video processing method, training method and device of model and electronic equipment | |
| CN115578394B (en) | Pneumonia image processing method based on asymmetric network | |
| CN110795706B (en) | Hash-based verification method, equipment, storage medium and device | |
| CN112434651A (en) | Information analysis method and device based on image recognition and computer equipment | |
| CN112949305B (en) | Negative feedback information acquisition method, device, equipment and storage medium | |
| CN113628077B (en) | Method, terminal and readable storage medium for generating non-repeated questions | |
| CN113627576A (en) | Code scanning information detection method, device, equipment and storage medium | |
| CN114003784A (en) | Request recording method, device, equipment and storage medium | |
| CN111553476B (en) | Neural network training method, device and storage medium based on memory score | |
| CN114781149A (en) | Method and system for automatically acquiring scene element information | |
| CN114218574A (en) | Data detection method and device, electronic equipment and storage medium | |
| CN113918471A (en) | Test case processing method and device and computer readable storage medium | |
| CN113705468A (en) | Digital image identification method based on artificial intelligence and related equipment | |
| CN113723431A (en) | Image recognition method, image recognition device and computer-readable storage medium | |
| CN113420545A (en) | Abstract generation method, device, equipment and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |