Medical image recognition method and apparatus, device and storage medium

Abstract

Embodiments of the present application relate to the technical field of medical image processing, and provide a medical image recognition method and apparatus, a device and a storage medium. The method comprises: acquiring a target image set, the target image set comprising a plurality of target whole-body tomographic images collected in a first preset time sequence; extracting organ tomographic images of a target organ from the target whole-body tomographic images; constructing an initial organ matrix for the target organ tomographic images of each target whole-body tomographic image; calculating a metabolic reference matrix and the initial organ matrix on the basis of a preset object metabolic difference sub-network, to obtain an organ metabolic difference matrix; and performing metabolic difference recognition on the basis of the organ metabolic difference matrix and the metabolic reference matrix, to obtain a target metabolic difference map that represents metabolic difference information between any two target organs in the target whole-body tomographic images. The embodiments of the present application can improve the recognition accuracy of images acquired by positron emission tomography, thereby improving the accuracy of determining metabolic conditions of individual objects.

Description

Medical image recognition method, device, equipment, and storage medium Technical Field

[0001] The present application relates to the field of medical image processing technology, and in particular to a medical image recognition method and apparatus, device, and storage medium. Background Art

[0002] Positron emission tomography (PET) is a medical imaging technique used to detect and diagnose a variety of diseases. Currently, when investigating systemic metabolic abnormalities at the individual level, PET images are often used to assist doctors in diagnosis, allowing them to detect abnormalities before they cause structural changes due to disease.

[0003] When identifying PET images, the related technology first determines the multi-organ Standardized Uptake Value (SUV) image corresponding to the PET image. By determining the lesion and the surrounding area in the SUV image, and combining the pre-trained organ metabolic network to analyze and predict the lesion and the surrounding area, a metabolic difference network diagram between the organs can be identified. At this time, the doctor can discover the metabolic abnormalities of the individual subject in advance based on the identified metabolic difference network diagram, and thus intervene in advance at the early stage of abnormal changes. However, the existing PET-based medical image recognition method has low recognition accuracy for PET images, which easily leads to incomplete metabolic representation of the individual subject corresponding to the image, thereby affecting the judgment of the metabolic status of the individual subject.

[0004] Summary of the Invention

[0005] The main purpose of the embodiments of the present application is to propose a medical image recognition method and apparatus, equipment, and storage medium that can more comprehensively characterize the metabolic status of individual subjects, improve the recognition accuracy of images acquired by positron emission tomography, and thus improve the accuracy of judging the metabolic status of individual subjects.

[0006] To achieve the above objectives, a first aspect of an embodiment of the present application provides a medical image recognition method, the method comprising:

[0007] Acquire a target image set of a target object, the target image set comprising a plurality of target whole-body tomographic images acquired in a first preset time sequence; the target whole-body tomographic images are images acquired based on positron emission tomography;

[0008] performing organ region extraction on the target whole-body tomographic image to obtain a target organ tomographic image of the target organ;

[0009] constructing an organ matrix for the target organ tomographic image of each target whole-body tomographic image to obtain an initial organ matrix;

[0010] According to a preset object metabolic difference subnetwork, a metabolic difference calculation is performed on the pre-acquired metabolic reference matrix and the initial organ matrix to obtain an organ metabolic difference matrix;

[0011] Metabolic difference identification is performed on the target whole-body tomographic image according to the organ metabolic difference matrix and the metabolic reference matrix to obtain a target metabolic difference map, which is used to represent metabolic difference information between any two target organs in the target whole-body tomographic image.

[0012] In some embodiments, the metabolic difference calculation is performed on the pre-acquired metabolic reference matrix and the initial organ matrix according to the preset subject metabolic difference subnetwork to obtain the organ metabolic difference matrix, including:

[0013] performing object organ correlation calculation on the initial organ matrix to obtain a first organ correlation matrix;

[0014] performing reference organ correlation calculation on the metabolic reference matrix to obtain a second organ correlation matrix;

[0015] Matrix difference calculation is performed on the first organ correlation matrix and the second organ correlation matrix to obtain the organ metabolism difference matrix.

[0016] In some embodiments, constructing an organ matrix for the target organ tomographic image of each target whole-body tomographic image to obtain an initial organ matrix includes:

[0017] performing image pixel mean processing on the target organ tomographic image to obtain a pixel mean of the target organ;

[0018] constructing a single organ matrix based on the pixel means of the plurality of object organs in the target whole-body tomographic image to obtain a first single organ initial matrix;

[0019] A whole-body organ matrix is ​​constructed for the first single-organ initial matrices of the plurality of target whole-body tomographic images to obtain the initial organ matrix.

[0020] In some embodiments, constructing an organ matrix for the target organ tomographic image of each target whole-body tomographic image to obtain an initial organ matrix includes:

[0021] performing organ metabolism quantitative analysis on the target organ tomographic image to obtain an organ metabolism quantitative analysis value;

[0022] constructing a single-organ matrix for the plurality of organ metabolic quantitative analysis values ​​of each target whole-body tomographic image to obtain a second single-organ initial matrix;

[0023] A whole-body organ matrix is ​​constructed for the plurality of the second single-organ initial matrices to obtain the initial organ matrix.

[0024] In some embodiments, the performing metabolic difference identification on the target whole-body tomographic image according to the organ metabolic difference matrix and the metabolic reference matrix to obtain a target metabolic difference map includes:

[0025] obtaining a number of reference objects according to the metabolic reference matrix;

[0026] Acquiring first matrix data from the organ metabolism difference matrix and acquiring second matrix data from the metabolism reference matrix;

[0027] Calculating the degree of metabolic abnormality based on the first matrix data, the second matrix data, and the number of reference objects to obtain target metabolic abnormality degree data; the target metabolic abnormality degree data is used to represent the degree of abnormal correlation between organs extracted from the target whole-body tomographic image;

[0028] constructing a target metabolic abnormality degree matrix according to the target metabolic abnormality degree data;

[0029] A metabolic difference map is constructed according to the target metabolic abnormality degree matrix to obtain the target metabolic difference map.

[0030] In some embodiments, constructing a metabolic difference map based on the target metabolic abnormality degree matrix to obtain the target metabolic difference map includes:

[0031] determining the organ number of the target organ according to the target metabolic abnormality degree matrix;

[0032] Acquire a candidate organ from the target organ, and acquire candidate metabolic abnormality degree data of the candidate organ from the target metabolic abnormality degree matrix, wherein the candidate metabolic abnormality degree data is used to represent the abnormal correlation degree between the candidate organ and other organs;

[0033] performing organ abnormality association calculation based on the plurality of candidate metabolic abnormality degree data of the candidate organs and the number of organs to obtain target organ abnormality association data;

[0034] The target metabolic difference map is constructed based on the target organ abnormality association data of the plurality of candidate organs.

[0035] In some embodiments, before performing metabolic difference calculation on the pre-acquired metabolic reference matrix and the initial organ matrix according to the preset subject metabolic difference subnetwork to obtain the organ metabolic difference matrix, the method further includes: constructing the subject metabolic difference subnetwork, specifically including:

[0036] Acquiring a reference dataset and a sample dataset, wherein the sample dataset includes a sample subset of a sample subject, the sample subset includes a plurality of sample whole-body tomographic data acquired in a second preset time series, and the sample whole-body tomographic data includes sample whole-body tomographic images acquired at preset acquisition time points in the second preset time series; the reference dataset includes a reference subset of a reference subject, the reference subset includes a plurality of reference whole-body tomographic data acquired in the second preset time series, and the reference whole-body tomographic data includes reference whole-body tomographic images acquired at the preset acquisition time points;

[0037] Performing organ region extraction on the reference whole-body tomographic data to obtain a reference organ tomographic image of a reference organ; and constructing a reference sub-network based on the reference organ tomographic images of a plurality of reference whole-body tomographic data;

[0038] Constructing a perturbation object set according to the reference data set and the sample data set; the perturbation object set includes a perturbation subset, the perturbation subset includes a plurality of perturbed whole-body tomographic data acquired in the second preset time series, and the perturbed whole-body tomographic data includes perturbed whole-body tomographic images acquired at the preset acquisition time point;

[0039] performing organ region extraction on the perturbed whole-body tomographic data to obtain a perturbed organ tomographic image of the perturbed organ; and constructing a perturbation subnetwork based on the perturbed organ tomographic images of the perturbed whole-body tomographic data of the plurality of perturbation object sets;

[0040] A network difference calculation is performed on the reference subnetwork and the perturbation subnetwork, and the subject metabolic difference subnetwork is constructed according to the result of the network difference calculation.

[0041] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a medical image recognition device, comprising:

[0042] an acquisition module, configured to acquire a target image set of a target object, the target image set comprising a plurality of target whole-body tomographic images acquired in a first preset time sequence; the target whole-body tomographic images are images acquired based on positron emission tomography;

[0043] an extraction module, configured to extract an organ region from the target whole-body tomographic image to obtain a target organ tomographic image of the target organ;

[0044] a matrix construction module, configured to construct an organ matrix for the target organ tomographic image of each target whole-body tomographic image to obtain an initial organ matrix;

[0045] a difference calculation module, configured to perform metabolic difference calculation on the pre-acquired metabolic reference matrix and the initial organ matrix according to a preset object metabolic difference subnetwork, to obtain an organ metabolic difference matrix;

[0046] A graph construction module is used to identify metabolic differences in the target whole-body tomographic image based on the organ metabolic difference matrix and the metabolic reference matrix to obtain a target metabolic difference graph, wherein the target metabolic difference graph is used to represent metabolic difference information between any two target organs in the target whole-body tomographic image.

[0047] To achieve the above-mentioned objectives, a third aspect of the embodiments of the present application provides an electronic device, including:

[0048] at least one memory;

[0049] at least one processor;

[0050] at least one computer program;

[0051] The at least one computer program is stored in the at least one memory, and the at least one processor executes the at least one computer program to implement:

[0052] As described in the first aspect above.

[0053] To achieve the above-mentioned purpose, the fourth aspect of the embodiments of the present application proposes a computer-readable storage medium, which stores a computer program, and the computer program is used to enable a computer to execute the method described in the first aspect above.

[0054] This application proposes a medical image recognition method, apparatus, device, and storage medium. By incorporating time series information, this method further constructs a network representing the relationships and metabolism between multiple organs in an individual subject, enabling a more comprehensive representation of the individual's metabolic profile and improving the recognition accuracy of images acquired using positron emission tomography (PET). Specifically, a target image set of a target subject is acquired, comprising multiple target whole-body tomographic images acquired during a first preset time series. The target whole-body tomographic images are images acquired using PET. Organ regions are extracted from the target whole-body tomographic images to obtain organ tomographic images of the target organs. An organ matrix is ​​constructed for the target organ tomographic images of each target whole-body tomographic image to obtain an initial organ matrix. Based on a preset subject metabolic difference subnetwork, metabolic differences are calculated between a pre-acquired metabolic reference matrix and the initial organ matrix to obtain an organ metabolic difference matrix. The organ metabolic difference matrix represents the relationships and differences between the individual subject's metabolic profile and a reference metabolic profile. Metabolic difference recognition is performed on the target whole-body tomographic images based on the organ metabolic difference matrix and the metabolic reference matrix to obtain a target metabolic difference map. This target metabolic difference map represents metabolic difference information between any two target organs in the target whole-body tomographic images. Therefore, the present application provides rich time domain information in combination with dynamic images, which can more comprehensively characterize the metabolic status of individual subjects, improve the recognition accuracy of images acquired based on positron emission tomography, and thus improve the accuracy of judging the metabolic status of individual subjects. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] FIG1 is a schematic diagram of obtaining a whole-body image based on positron emission tomography provided in an embodiment of the present application;

[0056] FIG2 is a flow chart of a medical image recognition method provided by an embodiment of the present application;

[0057] FIG3 is a schematic diagram of extracting organ regions from a target whole-body tomographic image according to an embodiment of the present application;

[0058] FIG4 is a flow chart of step S230 in FIG2 ;

[0059] FIG5 is another flow chart of step S230 in FIG2 ;

[0060] FIG6 is a flow chart of constructing a target object metabolic difference network according to an embodiment of the present application;

[0061] FIG7 is a schematic diagram of a network construction of a target object metabolic difference network 710 provided in an embodiment of the present application;

[0062] FIG8 is a flow chart of step S240 in FIG2 ;

[0063] FIG9 is a flow chart of step S250 in FIG2 ;

[0064] FIG10 is a flow chart of step S950 in FIG9 ;

[0065] FIG11 is a schematic diagram showing a comparison of target metabolic difference maps constructed based on Ki or SUV parameters provided in an embodiment of the present application;

[0066] FIG12 is a schematic diagram of the structure of a medical image recognition device provided in an embodiment of the present application;

[0067] FIG13 is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0069] It should be noted that although the device schematics illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the device or the sequence in the flowcharts. The terms "first," "second," and so on, in the specification, claims, and drawings, are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

[0070] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0071] First, let’s analyze some of the terms used in this application:

[0072] Positron Emission Tomography (PET) is a medical imaging technique, specifically a nuclear medicine imaging technique that noninvasively obtains information about the internal metabolism of an object. PET can be used to detect and diagnose a variety of diseases, identifying abnormalities before disease-induced structural changes occur. It is an important diagnostic tool in early clinical screening.

[0073] Graph network: A data structure consisting of nodes and edges. Graph networks can be used to describe various complex relationships, such as connection networks, communication networks, and road networks. In this context, a graph network typically consists of a set of vertices (nodes) and the edges connecting these vertices.

[0074] The Standardized Uptake Value (SUV) is a commonly used quantitative parameter in medical imaging. It is used to assess the metabolic activity of tumors or other abnormal metabolic tissues in PET images. SUV is a standardized measurement that normalizes the uptake of radiolabeled substances in PET images as a function of injected dose and body weight, eliminating interpatient biodistribution variability and the effects of injected dose.

[0075] The Patlak model is a method used to analyze dynamic PET images. The Ki parameter in the Patlak model represents the rate constant for radiolabeled uptake by tissues and organs, and is crucial for quantitative analysis of PET images. The Ki parameter can also be used to assess the metabolic activity of tissues and organs, and is crucial for studying tumor biology and therapeutic efficacy.

[0076] Pearson Correlation Analysis: A statistical method used to measure the strength and direction of the linear correlation between two continuous variables. Pearson Correlation Analysis measures the degree of linear relationship between two variables, with a value ranging from -1 to 1, where 1 indicates a perfect positive correlation, -1 indicates a perfect negative correlation, and 0 indicates no linear correlation.

[0077] Currently, when investigating systemic metabolic abnormalities at the individual level, PET images are often used to assist physicians in diagnosis, identifying abnormalities before disease-induced structural changes occur. Recent breakthroughs in imaging technology have led to the gradual introduction of whole-body PET imaging with a wider field of view, enabling simultaneous imaging of multiple organs throughout the body. However, most research and clinical applications focus solely on the lesion and surrounding areas, leaving a significant gap in understanding systemic organ connections. As shown in Figure 1, from a physiological perspective, the human body maintains homeostasis through dynamic, systemic interactions among multiple organs. Using 2-meter PET technology, a current whole-body PET metabolic image 110 can be captured, reflecting the energy requirements for interactions among the subject's various organs. When systemic feedback loops are poorly compensated, persistent disruptions to normal physiological homeostasis can lead to complex, chronic systemic diseases such as cancer, diabetes, arthritis, or cardiovascular disease. Therefore, multi-organ metabolic networks based on whole-body PET images have the potential to characterize these pathological conditions by analyzing parameter differences between disease and control groups, thereby enabling early detection of metabolic abnormalities and enabling early clinical diagnostic and therapeutic intervention.

[0078] When identifying PET images, the related technology first determines the multi-organ Standardized Uptake Value (SUV) image corresponding to the PET image. By determining the lesion and the surrounding area in the image, and combining the pre-trained organ metabolic network to analyze and predict the lesion and the surrounding area, a metabolic difference network diagram between the organs can be identified. At this time, the doctor can discover the metabolic abnormalities of the individual subject in advance based on the identified metabolic difference network diagram, so as to intervene in advance at the early stage of abnormal changes. However, the existing PET-based medical image recognition method is prone to incomplete metabolic characterization of the individual subject corresponding to the image, reducing the recognition accuracy of the PET image, and thus affecting the judgment of the metabolic status of the individual subject.

[0079] Based on this, the embodiments of the present application provide a medical image recognition method and apparatus, equipment, and storage medium that can more comprehensively characterize the metabolic status of individual subjects, improve the recognition accuracy of images acquired by positron emission tomography, and thus improve the accuracy of judging the metabolic status of individual subjects.

[0080] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0081] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0082] The medical image recognition method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The medical image recognition method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or as a server cluster or distributed system composed of multiple physical servers, or as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the medical image recognition method, etc., but is not limited to the above forms.

[0083] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0084] It should be noted that in each specific embodiment of the present application, when it comes to the need to perform relevant processing on data related to the identity or characteristics of the object based on the target whole-body tomographic data, sample whole-body tomographic data, reference whole-body tomographic data, etc., the permission or consent of the object will be obtained first, and the collection, use and processing of such data will comply with relevant laws, regulations and standards. In addition, when the embodiment of the present application needs to obtain the sensitive personal information of the object, the separate permission or consent of the object will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the separate permission or consent of the object, the necessary object-related data for the normal operation of the embodiment of the present application will be obtained.

[0085] Please refer to Figure 2, which is an optional flow chart of a medical image recognition method provided in an embodiment of the present application. In some embodiments of the present application, the method in Figure 2 may specifically include but is not limited to steps S210 to S250.

[0086] Step S210: Acquire a target image set of a target object, where the target image set includes a plurality of target whole-body tomographic images acquired in a first preset time sequence; the target whole-body tomographic images refer to images acquired based on positron emission tomography;

[0087] Step S220: extracting organ regions from the target whole-body tomographic image to obtain an organ tomographic image of the target organ;

[0088] Step S230, constructing an organ matrix for the target organ tomographic image of each target whole-body tomographic image to obtain an initial organ matrix;

[0089] Step S240: performing metabolic difference calculation on the pre-acquired metabolic reference matrix and the initial organ matrix according to the preset object metabolic difference sub-network to obtain an organ metabolic difference matrix;

[0090] Step S250: performing metabolic difference identification on the target whole-body tomographic image according to the organ metabolic difference matrix and the metabolic reference matrix to obtain a target metabolic difference map, which is used to represent metabolic difference information between any two target organs in the target whole-body tomographic image.

[0091] The medical image recognition method provided by the present application further constructs a network that characterizes the association and metabolism between multiple organs of an individual subject by introducing time series information, so as to more comprehensively characterize the metabolic status of the individual subject and improve the recognition accuracy of images collected by positron emission tomography. Compared with the construction method of static whole-body PET metabolic network, the present application provides rich time domain information by combining dynamic images, which can more comprehensively characterize the metabolic status of the individual subject. And the metabolic abnormalities of the individual subject are quantified in the form of abnormal intensity. Therefore, the present application uses the rich information on the time series to further improve the construction of the metabolic network, achieve the purpose of exploring the association of systemic diseases, and improve the recognition accuracy of images collected by positron emission tomography by combining the changes in the metabolic level of the target object over time, thereby improving the accuracy of judging the metabolic status of the individual subject.

[0092] In step S210 of some embodiments, the target object refers to an object used to judge the metabolic correlation and difference between various organs in the body. For example, the target object is a patient with a tumor. By judging the metabolic correlation and difference between various organs in the patient's body, the correlation between systemic diseases can be explored to assist doctors in conducting early clinical examinations of the patient. The target image set refers to a collection of multiple target whole-body tomographic images acquired on the target object in a first preset time sequence. The target whole-body tomographic image refers to a PET image acquired at an image acquisition time point in the first preset time sequence using PET technology. The first preset time series refers to a time series of dynamic whole-body tomographic images acquired for the target object. In actual applications, the target object may be a patient who has been discharged 30 days after hospitalization and has unexplained gastrointestinal bleeding.

[0093] For example, after administering a PET tracer to a target subject, a first preset time series is set from 10 to 60 minutes after administration, with an image sampling interval of 5 minutes. In this case, a target whole-body tomographic image is acquired every 5 minutes during the first preset time series, resulting in 10 target whole-body tomographic images of the target subject. These 10 target whole-body tomographic images of the target subject are then constructed into a target image set.

[0094] It should be noted that the image acquisition time points are determined based on the first preset time series and the image sampling interval, and can be flexibly adjusted according to actual needs, and are not specifically limited here. Therefore, by acquiring multiple target whole-body tomographic images in a time series containing multiple frames of images, the present application conducts multi-organ association and metabolic studies on the target subject in a dynamic time series, which can reflect the changes in the target subject's metabolic level over time, thereby improving the accuracy of the determination of the metabolic status of individual subjects.

[0095] In step S220 of some embodiments, the target organ tomographic image refers to a PET image of a single organ obtained by performing organ region extraction on the region where each organ is located in the target whole-body tomographic image. As shown in Figure 1, the organs of the human body include the lungs, liver, stomach, spleen, small intestine, brain, heart, pancreas, kidneys, large intestine, etc., and the metabolic conditions of these organs can be captured in a target whole-body tomographic image. As shown in Figure 3, organ region extraction is performed on the region where each pre-selected organ (such as the brain, heart, lungs, liver, stomach, spleen, pancreas, and kidneys) is located in a target whole-body tomographic image to obtain a corresponding target organ tomographic image. Therefore, the present application performs organ region extraction on each target whole-body tomographic image to obtain multiple target organ tomographic images corresponding to each target whole-body tomographic image.

[0096] For example, let's consider the region of interest (ROI) for an organ in a PET image, where the image contains m organs (m is a positive integer greater than 1). Based on the target whole-body tomographic image, images of the m organ regions are segmented as target organ tomographic images. In other words, if the target image set includes 10 target whole-body tomographic images of a target subject, and m is 8, organ region extraction is performed on each of the 10 target whole-body tomographic images to obtain target organ tomographic images of the eight target organs corresponding to each target whole-body tomographic image, resulting in a total of 80 target organ tomographic images.

[0097] In the above embodiments, the present application can effectively expand the number of target organ tomographic images of the target object by conducting multi-organ association and metabolism research on the target object in a dynamic time series, and use the rich information in the time series to improve the accuracy of judging the metabolic status of individual objects.

[0098] In step 230 of some embodiments, the perturbation subnetwork (denoted as ptbNet) refers to a graph network constructed by adding the data of each sample subset to the reference data set one by one, based on the impact of the data of the sample subset on the health baseline level determined based on the reference data set. The perturbation subnetwork is constructed based on the covariance network structure (i.e., the Pearson correlation coefficient is calculated for the values ​​corresponding to different organs in the matrix to determine the connection strength between different organs represented by the correlation coefficient). The initial organ matrix refers to a matrix constructed according to a preset perturbation subnetwork, which is used to represent the degree of influence of the target object on the health baseline level. The initial organ matrix is ​​a higher dimensional matrix determined by dynamically considering the metabolic transformation of the target object.

[0099] It should be noted that the perturbation subnetwork consists of first nodes and first edges. Each first node represents an organ in a different region, and the first edges represent the connection strength between the first nodes, as represented by the correlation coefficient. A larger correlation coefficient indicates a stronger connection, as indicated by a darker black edge in the figure. The connection strength represented by the first edges is determined based on the target subject's impact on the baseline health level.

[0100] It should be noted that the data corresponding to a target whole-body tomographic image can be represented as a t×m matrix, where each row t represents a time node (i.e., a time point in a first preset time series) and each column m represents an organ region. In this case, constructing an organ matrix for target organ tomographic images in multiple target whole-body tomographic images to obtain an initial organ matrix includes: calculating the Pearson correlation coefficient for multiple target organ tomographic images in each target whole-body tomographic image to obtain a first single-organ initial matrix; and calculating the Pearson correlation coefficient for multiple first single-organ initial matrices to obtain an initial organ matrix.

[0101] It should be noted that the matrix size of the initial organ matrix is ​​m×m, where m represents the number of organ regions. That is, each matrix value in the initial organ matrix represents the correlation strength between the two organs at the corresponding horizontal and vertical positions. The calculation process of the Pearson correlation coefficient is shown in the following formula (1). s =cov(X,Y) / (σ X σ Y ) Formula (1)

[0102] Among them, r s represents the Pearson correlation coefficient obtained based on the matrix variables X and Y, cov(X,Y) represents the covariance of the matrix variables X and Y, σ X represents the standard deviation of the matrix variable X, σ Y Represents the standard deviation of the matrix variable Y.

[0103] Please refer to Figure 4, which is a specific flow chart of step S230 provided in an embodiment of the present application. In some embodiments of the present application, step S230 may specifically include but is not limited to steps S410 to S430.

[0104] Step S410: performing pixel mean processing on the target organ tomographic image to obtain the target organ pixel mean;

[0105] Step S420: constructing a single organ matrix based on the pixel means of multiple object organs in the target whole-body tomographic image to obtain a first single organ initial matrix;

[0106] Step S430: construct a whole-body organ matrix for the first single-organ initial matrix of the plurality of target whole-body tomographic images to obtain an initial organ matrix.

[0107] In some embodiments, in steps S410 to S430, the target organ pixel mean refers to the value obtained by averaging each pixel in the target organ tomographic image. A single-organ matrix is ​​constructed using the pixel means of multiple target organs in the target whole-body tomographic image to obtain a first single-organ initial matrix. In this case, the matrix size of the first single-organ initial matrix is ​​1×m. Subsequently, a whole-body organ matrix is ​​constructed by calculating the Pearson correlation coefficient between the columns of the first single-organ initial matrices for the multiple target whole-body tomographic images to obtain an initial organ matrix. This initial organ matrix incorporates dynamic time series information, i.e., it is transformed into higher-dimensional data.

[0108] For example, the dimensions of a three-dimensional whole-body tomographic image are 192*192*673, indicating that the number of pixels on the X-axis (horizontal direction) is 192, the number of pixels on the Y-axis (vertical direction) is 192, and the number of pixels on the Z-axis (depth direction, used to indicate the number of layers or slices in the image) is 673. Because the present application incorporates dynamic sequence information, the target whole-body tomographic image can be converted from three dimensions to four dimensions, for example, 192*192*673*number of frames, where the number of frames is the number of target whole-body tomographic images acquired in the first preset time sequence.

[0109] Please refer to Figure 5, which is another specific flow chart of step S230 provided in an embodiment of the present application. In some embodiments of the present application, step S230 may further include but is not limited to steps S510 to S530.

[0110] Step S510: performing organ metabolism quantitative analysis on the target organ tomographic image to obtain an organ metabolism quantitative analysis value;

[0111] Step S520: constructing a single-organ matrix for the multiple organ metabolic quantitative analysis values ​​of each target whole-body tomographic image to obtain a second single-organ initial matrix;

[0112] Step S530: construct a whole-body organ matrix based on the multiple second single-organ initial matrices to obtain an initial organ matrix.

[0113] It should be noted that this application can directly use PET images for image recognition, or it can construct the target object metabolic difference network of this application through semi-quantitative and quantitative parameter images such as SUV and Patlak Ki. Compared with the existing technology that is based only on SUV images, resulting in the disadvantage of incomplete metabolic characterization, this application integrates different aspects of metabolic information by considering different semi-quantitative and quantitative parameter images, and at the same time combines the rich time domain information provided by dynamic images to improve the diversity and comprehensiveness of individual organ metabolic networks.

[0114] In steps S510 to S530 of some embodiments, the preset quantitative analysis function is a function constructed based on Patlak Ki. Therefore, the present application can perform organ metabolism quantitative analysis on the target organ tomographic image according to the preset quantitative analysis function to obtain the organ metabolism quantitative analysis value. The organ metabolism quantitative analysis value refers to the value obtained after the Patlak Ki parameter calculation for each target organ tomographic image. At this time, a single organ matrix is ​​constructed for the multiple organ metabolism quantitative analysis values ​​in the target whole-body tomographic image to obtain a second single organ initial matrix. The construction method of the second single organ initial matrix is ​​similar to that of the above-mentioned first single organ initial matrix, except that the pixel mean of the object organ is converted into the organ metabolism quantitative analysis value for calculation, which will not be repeated here. Afterwards, based on the multiple second single organ initial matrices of the multiple target whole-body tomographic images, the whole-body organ matrix is ​​constructed by calculating the Pearson correlation coefficient between the columns of the multiple second single organ initial matrices to obtain the initial organ matrix.

[0115] It should be noted that in some other embodiments, step S230 further includes: performing a semi-quantitative analysis of organ metabolism on the target organ tomographic image according to a semi-quantitative analysis function (e.g., a function constructed based on the SUV) to obtain a semi-quantitative analysis value of organ metabolism. Constructing a single-organ matrix based on the semi-quantitative analysis values ​​of organ metabolism in the target whole-body tomographic image to obtain a third single-organ initial matrix. Constructing a whole-body organ matrix based on the third single-organ initial matrix of the multiple target whole-body tomographic images to obtain an initial organ matrix.

[0116] It should be noted that this application integrates metabolic information from different aspects by considering images constructed with different semi-quantitative and quantitative parameters, that is, jointly constructing a metabolic difference network of the target object, thereby better achieving the diversity and comprehensiveness of the constructed metabolic difference network of the target object.

[0117] In some embodiments, before step 240, the medical image recognition method provided by the present application further includes: constructing a target metabolic difference subnetwork. It should be noted that the perturbation subnetwork, reference subnetwork, and target metabolic difference subnetwork of the present application can construct a target target metabolic difference network. Please refer to Figure 6, which is a specific flow chart of constructing a target metabolic difference subnetwork provided by an embodiment of the present application. In some embodiments of the present application, the steps of constructing a target metabolic difference subnetwork may specifically include but are not limited to steps S610 to S650.

[0118] Step S610: Acquire a reference dataset and a sample dataset, wherein the sample dataset includes a sample subset of a sample subject, the sample subset includes a plurality of sample whole-body tomographic data acquired in a second preset time series, and the sample whole-body tomographic data includes sample whole-body tomographic images acquired at preset acquisition time points in the second preset time series; and the reference dataset includes a reference subset of a reference subject, the reference subset includes a plurality of reference whole-body tomographic data acquired in the second preset time series, and the reference whole-body tomographic data includes reference whole-body tomographic images acquired at preset acquisition time points.

[0119] Step S620: extracting organ regions from the reference whole-body tomographic data to obtain a reference organ tomographic image of the reference organ; and constructing a reference sub-network based on the reference organ tomographic images of the multiple reference whole-body tomographic data;

[0120] Step S630: constructing a perturbation object set based on the reference dataset and the sample dataset; the perturbation object set includes a perturbation subset, the perturbation subset includes a plurality of perturbed whole-body tomographic data acquired in a second preset time series, and the perturbed whole-body tomographic data includes perturbed whole-body tomographic images acquired at preset acquisition time points;

[0121] Step S640: performing organ region extraction on the perturbed whole-body tomographic data to obtain a perturbed organ tomographic image of the perturbed organ; and constructing a perturbation subnetwork based on the perturbed organ tomographic images of the perturbed whole-body tomographic data of the plurality of perturbation object sets;

[0122] Step S650: performing network difference calculation on the reference sub-network and the perturbation sub-network, and constructing the subject metabolic difference sub-network according to the result of the network difference calculation.

[0123] In step S610 of some embodiments, the reference data set refers to a data set of a healthy control group consisting of subjects without any disease records. The sample data set refers to a set of subjects constructed for constructing a metabolic difference network of the target object. The sample data set may include data of subjects diagnosed with lung cancer and different lesion sites, and may also include data of subjects without any disease records. The second preset time series refers to a time series of dynamic whole-body tomographic images collected for the objects in the reference data set and the sample data set. The sample subset refers to the PET data of each subject (i.e., the sample object) in the sample data set. The sample whole-body tomographic image has the same meaning as the above-mentioned target whole-body tomographic image, but is used here as a network construction, and will not be repeated here. The reference subset refers to the PET data of each subject (i.e., the reference object) in the reference data set. The reference whole-body tomographic image has the same meaning as the above-mentioned target whole-body tomographic image, but the objects of the image source are different, and it is used here as a network construction, and will not be repeated here.

[0124] In step S620 of some embodiments, as shown in Figure 7, the target object metabolic difference network 710 is constructed by a perturbation subnetwork 711, a reference subnetwork 712, and a subject metabolic difference subnetwork 713. First, a reference subnetwork 712 is constructed from a reference data set (i.e., a healthy control group, such as 24 healthy subjects), denoted as refNET. Specifically, organ region extraction is performed on the reference whole-body tomographic data to obtain a reference organ tomographic image; a reference subnetwork is constructed by calculating each organ region and based on the reference organ tomographic images in multiple reference whole-body tomographic data. A reference metabolic network refNET is constructed from a reference data set (i.e., a healthy group, such as 24 healthy subjects). The reference subnetwork is obtained by calculating the partial Pearson correlation coefficient between each organ region in the reference whole-body tomographic data. The reference subnetwork adopts a covariance network structure, including a second node and a second connection edge. The second nodes of the network correspond to organs in different regions, the second connection edges connect different second nodes, and the second connection edges represent the connection strength between the second nodes represented by the correlation coefficient. The larger the correlation coefficient, the stronger the connection. Generally speaking, the constructed reference subnetwork has the common features of inter-organ connectivity in all reference subsets.

[0125] In steps S630 and S640 of some embodiments, after obtaining the reference subnetwork refNET, the sample subsets in the sample data set are added to the healthy control group one by one to establish multiple perturbation data sets containing objects corresponding to the sample subsets. At this time, the perturbation data set includes perturbation subsets, and each perturbation subset includes a sample subset and a reference data set. The number of perturbation subsets in the perturbation data set can be the number of sample subsets. That is, one subject is added to the reference data set (i.e., the healthy control group, such as containing 24 healthy subjects) to form a new group consisting of 25 subjects to construct a new perturbation subset. For example, n represents the number of subjects in the reference data set, and the number of subjects included in the perturbation subset is n+1. The multiple perturbation whole-body tomographic data in the perturbation subset are fused by considering different semi-quantitative and quantitative parameters in the above-mentioned embodiment to construct a perturbation subnetwork. That is, the perturbed whole-body tomographic data may replace the target whole-body tomographic data in step S230 to construct a new structural covariance network matrix, and a perturbation subnetwork may be constructed based on the matrix, which is recorded as ptbNET.

[0126] In some embodiments, in step S650, the perturbation subnetwork ptbNET is subtracted from the reference subnetwork refNET to obtain a subject metabolic difference subnetwork, denoted as resNET. In other words, the subject metabolic difference subnetwork is constructed by subtracting the matrix output by the perturbation subnetwork and the matrix output by the reference subnetwork.

[0127] The subject metabolic difference subnetwork of the embodiment of the present application is constructed based on the perturbation subnetwork (i.e., constructed by adding one subject at a time) and the reference subnetwork. It is constructed based on individual subjects and can more accurately judge the difference between individual subjects and healthy baseline levels, improve the recognition accuracy of PET images, and thus improve the accuracy of judging the metabolic status of individual subjects.

[0128] It should be noted that after constructing the subject metabolic difference subnetwork, the perturbation subnetwork and the reference subnetwork, the target subject metabolic difference network can be constructed. The present application can also set a threshold (such as 0.2, 0.3, etc.) to eliminate weak correlations that may come from noise.

[0129] This application introduces time series information to further construct a target subject metabolic difference network that characterizes the connections and metabolism between multiple organs of an individual subject. Compared to the construction method of static whole-body PET metabolic network, this application integrates different aspects of metabolic information by considering images constructed using different semi-quantitative and quantitative parameters. At the same time, combined with the rich time domain information provided by dynamic images, the diversity and comprehensiveness of the target subject metabolic difference network are improved, the recognition accuracy of PET images is improved, and thus the accuracy of judging the metabolic status of individual subjects is improved.

[0130] In step 240 of some embodiments, the reference subnetwork is a network constructed based on a reference dataset (i.e., a healthy control group), and the metabolic reference matrix refers to a matrix constructed based on the reference dataset and is used to represent the healthy baseline level of the organ. The subject metabolic difference subnetwork has been described in detail in the above embodiments and will not be repeated here. The organ metabolic difference matrix refers to the matrix obtained by performing a matrix subtraction between the metabolic reference matrix and the initial organ matrix.

[0131] Please refer to Figure 8, which is a specific flow chart of step S240 provided in an embodiment of the present application. In some embodiments of the present application, step S240 may further include but is not limited to steps S810 to S830.

[0132] Step S810, performing object organ correlation calculation on the initial organ matrix to obtain a first organ correlation matrix;

[0133] Step S820: performing reference organ correlation calculation on the metabolic reference matrix to obtain a second organ correlation matrix;

[0134] Step S830: Calculate the matrix difference between the first organ correlation matrix and the second organ correlation matrix to obtain an organ metabolism difference matrix.

[0135] In steps S810 to S830 of some embodiments, when performing metabolic difference calculation on the metabolic reference matrix and the initial organ matrix, in combination with formula (1), the object organ correlation calculation is first performed on the initial organ matrix to obtain a first organ correlation matrix. Each column in the initial organ matrix represents the information of a single organ combined with time information. Performing object organ correlation calculation on the initial organ matrix means performing correlation calculation on the data of different columns in the initial organ matrix to obtain the connection strength represented by the correlation coefficient between different organs. Performing reference organ correlation calculation on the metabolic reference matrix means performing correlation calculation on the data of different columns in the initial organ matrix to obtain the connection strength represented by the correlation coefficient between different organs in the healthy control group (i.e., the reference data set). Afterwards, matrix difference calculation is performed on the first organ correlation matrix and the second organ correlation matrix to obtain an organ metabolic difference matrix. The organ metabolic difference matrix at this time is used to represent the deviation value between different organs of the target object and the normal reference value.

[0136] In step 250 of some embodiments, the target metabolic difference map is used to represent metabolic difference information between different organs in the target whole-body tomographic image. The target metabolic difference map may be a structure constructed based on the Z-score map.

[0137] Please refer to Figure 9, which is a specific flow chart of step S250 provided in an embodiment of the present application. In some embodiments of the present application, step S250 may further include but is not limited to steps S910 to S950.

[0138] Step S910, obtaining the number of reference objects according to the metabolic reference matrix;

[0139] Step S920: obtaining first matrix data from the organ metabolism difference matrix and obtaining second matrix data from the metabolism reference matrix;

[0140] Step S930: Calculate the degree of metabolic abnormality based on the first matrix data, the second matrix data, and the number of reference objects to obtain target metabolic abnormality degree data; the target metabolic abnormality degree data is used to represent the degree of abnormal correlation between organs extracted from the target whole-body tomographic image;

[0141] Step S940: constructing a target metabolic abnormality degree matrix according to the target metabolic abnormality degree data;

[0142] Step S950: construct a metabolic difference map according to the target metabolic abnormality degree matrix to obtain a target metabolic difference map.

[0143] In steps S910 to S940 of some embodiments, the first matrix data refers to the matrix column data in the organ metabolic difference matrix. The second matrix data refers to the matrix column data in the metabolic reference matrix. The target metabolic abnormality degree data refers to the numerical value calculated based on the ZCC function for the corresponding numerical values ​​in the organ metabolic difference matrix and the metabolic reference matrix. The target metabolic abnormality degree matrix refers to the matrix calculated based on the ZCC function for the first matrix data, the second matrix data, and the number of reference subjects.

[0144] It should be noted that the calculation process of each target metabolic abnormality degree data in the target metabolic abnormality degree matrix is ​​shown in the following formula (2).

[0145] Among them, if is the set containing all regions, and i≠j,ZCC j,i Represents the target metabolic abnormality data between organs i and j of the target object. In this case, resNET represents the value associated with organs i and j in the organ metabolic difference matrix. In this case, refNET 2 represents the value associated with organ i and organ j in the metabolic reference matrix; N refers to the number of subjects in a perturbation subset (such as n+1, where n is the number of subjects in the reference dataset), μ represents the mean value calculated based on the organ metabolic difference matrix, and σ represents the standard deviation calculated based on the organ metabolic difference matrix.

[0146] In some embodiments, in step S950, after determining the target metabolic abnormality degree matrix, a target metabolic difference graph is constructed based on the degree of metabolic change between organs corresponding to each value in the matrix. In the target metabolic difference graph, nodes represent organs, and edges between nodes represent the degree of metabolic change.

[0147] It should be noted that the target metabolic abnormality matrix essentially represents the number of standard deviations between each data point and the sample mean, and can be used to measure the degree of abnormal connectivity between all regions. Each residual network and the target metabolic difference map constructed based on the target metabolic abnormality matrix contain m(m-1) / 2 (m represents the number of organs in the image segmentation) undirected edges connecting all regions. Each edge represents the degree of metabolic change and reflects the deviation from the normal reference value of the healthy control group. Each value in the target metabolic abnormality matrix can be regarded as a statistic in the Z test for comparing significance levels. For example, when using a 95% confidence interval to filter out edges without significant differences, a target metabolic difference subnetwork that is significantly different from the reference subnetwork and can reflect individual metabolic information can be obtained.

[0148] Please refer to Figure 10, which is a specific flow chart of step S950 provided in an embodiment of the present application. In some embodiments of the present application, step S950 may further include but is not limited to steps S1010 to S1040.

[0149] Step S1010: determining the number of target organs according to the target metabolic abnormality degree matrix;

[0150] Step S1020: Obtain candidate organs from the target organs, and obtain candidate metabolic abnormality degree data of the candidate organs from the target metabolic abnormality degree matrix, where the candidate metabolic abnormality degree data is used to represent the degree of abnormal correlation between the candidate organs and other organs;

[0151] Step S1030: performing organ abnormality association calculation based on the multiple candidate metabolic abnormality degree data of the candidate organs and the number of organs to obtain target organ abnormality association data;

[0152] Step S1040: constructing a target metabolic difference map based on target organ abnormality association data of multiple candidate organs.

[0153] In some embodiments, in steps S1010 and S1020, since the target metabolic abnormality degree matrix is ​​an m*m matrix, the number m of target organs extracted from the target whole-body tomographic image can be determined based on the number of rows and columns in the target metabolic abnormality degree matrix. A candidate organ refers to an organ represented by each target organ tomographic image extracted from the target whole-body tomographic image. Candidate metabolic abnormality degree data refers to metabolic abnormality degree data associated with a candidate organ in the target metabolic abnormality degree matrix. In other words, the candidate metabolic abnormality degree data is used to indicate the degree of abnormal correlation between the candidate organ and other organs.

[0154] In step S1030 of some embodiments, under normal circumstances, when the metabolic homeostasis of a target object changes, the connection edges in its target metabolic difference graph will also change accordingly. Moreover, the change in the degree of abnormality can be visualized by drawing a connectivity graph from the Z-score graph constructed based on the ZCC function, that is, the target metabolic difference graph of the present application. For the target metabolic difference graph, the degree of this abnormal change can be quantified by defining the abnormality strength (STR) of each node (i.e., organ). For the jth organ, it performs organ abnormality association calculation based on multiple candidate metabolic abnormality degree data and the number of organs of the candidate organ, and the process of obtaining target organ abnormality association data is shown in the following formula (3). STR j =∑|ZCC j,i | / (m-1) Formula (3)

[0155] Among them, if is the set containing all regions, and i≠j; STR j Refers to the target organ abnormality association data of the jth organ, ZCC j,i Represents the target metabolic abnormality data between organ i and organ j of the target object, and m represents the total number of candidate organs. Thus, the present application can construct a subnetwork of object metabolic differences between each patient (i.e., target object) and a healthy control group (i.e., reference data set) at the individual level, and quantify this difference by means of abnormal intensity. The construction method of this metabolic network has better extensibility. The present invention analyzes and compares semi-quantitative and quantitative parameter images such as SUV and Ki to obtain more comprehensive results.

[0156] In step S1040 of some embodiments, a target metabolic difference map is constructed based on the target organ abnormality association data for multiple candidate organs. As shown in FIG7 , at the output of the subject metabolic difference subnetwork 713 , a difference map is constructed based on the organ metabolic difference matrix and the metabolic reference matrix to obtain a target metabolic difference map. This target metabolic difference map is a Z-score map to visualize the results. The visualized difference maps in FIG7 show the metabolic difference maps of the target subject in Case 1, Case 2, and Case 3, respectively.

[0157] It should be noted that, as shown in Figure 11, the present application can construct a target metabolic difference map based on Ki (preset quantitative analysis function) or SUV parameter (semi-quantitative parameter), and the target object metabolic difference network constructed based on semi-quantitative or quantitative parameters can be compared with the results between patients with lung adenocarcinoma and lung squamous cell carcinoma in the healthy group, showing obvious differences between different groups and a certain degree of similarity within the group. The nodes on each target metabolic difference map in Figure 11 are different organs (i.e., the English meaning corresponding to the organ can be used, which is not the focus of the solution, so it is not specifically explained, and one English only represents one organ), and the edge connecting the organ refers to the difference between the organ's correlation and health. Therefore, in a healthy state, there is no connecting edge on the difference map that clearly indicates the degree of difference. In a diseased state, the darker the color of the connecting edge, the more obvious the difference. In addition, as shown in Figure 11, the target metabolic difference map constructed based on the Ki parameter takes into account more data details, so the connecting edge between different organs is darker in color, the higher the difference intensity, and the more obvious the difference.

[0158] It should be noted that the medical image recognition method provided in this application can be applied not only to systemic disease associations but also to extract features from organ metabolic networks using appropriate methods such as graph theory or artificial intelligence to establish clinical intervention and treatment models for systemic chronic diseases. Furthermore, in addition to using whole-body PET images, this target subject metabolic difference network construction method can also be applied to imaging technologies such as functional magnetic resonance imaging after appropriate modification.

[0159] The medical image recognition method provided in the embodiment of the present application fuses metabolic information from different aspects by considering different semi-quantitative and quantitative parameter images. At the same time, combined with the rich time domain information provided by the dynamic image, a series of values ​​for each organ in the time series can be obtained, reflecting the changes in the metabolic level of the object over time, thereby improving the diversity and comprehensiveness of the target object metabolic difference network. The metabolic difference results are visualized according to the target metabolic difference graph to make their meaning clearer. Therefore, the present application uses the rich information on the time series to further improve the construction of the metabolic network, realizes the purpose of exploring the association of systemic diseases, and improves the recognition accuracy of PET images by combining the changes in the metabolic level of the target object over time, thereby improving the accuracy of judging the metabolic situation of individual objects.

[0160] Referring to FIG. 12 , an embodiment of the present application further provides a medical image recognition device that can implement the above-mentioned medical image recognition method. The device includes:

[0161] An acquisition module 1210 is configured to acquire a target image set of a target object, the target image set comprising a plurality of target whole-body tomographic images acquired in a first preset time sequence; the target whole-body tomographic images are images acquired based on positron emission tomography;

[0162] An extraction module 1220 is configured to extract an organ region from a target whole-body tomographic image to obtain a target organ tomographic image of the target organ;

[0163] A matrix construction module 1230 is configured to construct an organ matrix for a target organ tomographic image of each target whole-body tomographic image to obtain an initial organ matrix;

[0164] a difference calculation module 1240 for performing metabolic difference calculation on the pre-acquired metabolic reference matrix and the initial organ matrix according to a preset object metabolic difference subnetwork to obtain an organ metabolic difference matrix;

[0165] The map construction module 1250 is used to identify metabolic differences in the target whole-body tomographic image based on the organ metabolic difference matrix and the metabolic reference matrix to obtain a target metabolic difference map. The target metabolic difference map is used to represent the metabolic difference information between any two target organs in the target whole-body tomographic image.

[0166] The specific implementation of the medical image recognition device in the embodiment of the present application is basically the same as the specific implementation of the above-mentioned medical image recognition method, and will not be repeated here.

[0167] An embodiment of the present application further provides an electronic device comprising: at least one memory; at least one processor; and at least one computer program. The at least one computer program is stored in the at least one memory, and the at least one processor executes the at least one computer program to implement the above-described medical image recognition method. The electronic device can be any intelligent terminal, including a tablet computer and an in-vehicle computer.

[0168] Please refer to FIG13 , which illustrates a hardware structure of an electronic device according to another embodiment. The electronic device includes:

[0169] The processor 1310 can be implemented using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0170] The memory 1320 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1320 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1320 and is called by the processor 1310 to execute the medical image recognition method of the embodiments of this application.

[0171] Input / output interface 1330, used to implement information input and output;

[0172] Communication interface 1340, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0173] bus 1350 , which transmits information between the various components of the device (e.g., processor 1310 , memory 1320 , input / output interface 1330 , and communication interface 1340 );

[0174] The processor 1310 , the memory 1320 , the input / output interface 1330 , and the communication interface 1340 are communicatively connected to each other within the device via a bus 1350 .

[0175] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program, and the computer program is used to enable a computer to execute and implement the above-mentioned medical image recognition method.

[0176] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0177] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0178] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0179] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0180] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0181] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0182] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0183] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0184] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0185] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0186] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0187] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A medical image recognition method, characterized in that, The method includes: Obtaining a target image set of a target object, where the target image set includes a plurality of target whole-body tomographic images collected in a first preset time series; the target whole-body tomographic image refers to an image collected based on positron emission tomography; Extracting an organ region from the target whole-body tomographic image to obtain a target organ tomographic image of the target organ; Constructing an organ matrix for the target organ tomographic image of each target whole-body tomographic image to obtain an initial organ matrix; Performing a metabolic difference calculation on a pre-obtained metabolic reference matrix and the initial organ matrix according to a preset object metabolic difference sub-network to obtain an organ metabolic difference matrix; Performing a metabolic difference recognition on the target whole-body tomographic image according to the organ metabolic difference matrix and the metabolic reference matrix to obtain a target metabolic difference map, where the target metabolic difference map is used to represent the metabolic difference information between any two target organs in the target whole-body tomographic image.

2. The method according to claim 1, characterized in that, The performing a metabolic difference calculation on a pre-obtained metabolic reference matrix and the initial organ matrix according to a preset object metabolic difference sub-network to obtain an organ metabolic difference matrix includes: Performing an object organ correlation calculation on the initial organ matrix to obtain a first organ correlation matrix; Performing a reference organ correlation calculation on the metabolic reference matrix to obtain a second organ correlation matrix; Performing a matrix difference calculation on the first organ correlation matrix and the second organ correlation matrix to obtain the organ metabolic difference matrix.

3. The method according to claim 1, characterized in that, The constructing an organ matrix for the target organ tomographic image of each target whole-body tomographic image to obtain an initial organ matrix includes: Performing an image pixel mean processing on the target organ tomographic image to obtain an object organ pixel mean; Constructing a single-organ matrix for the plurality of object organ pixel means in the target whole-body tomographic image to obtain a first single-organ initial matrix; Constructing a whole-body organ matrix for the first single-organ initial matrix of the plurality of target whole-body tomographic images to obtain the initial organ matrix.

4. The method according to claim 1, characterized in that, The constructing an organ matrix for the target organ tomographic image of each target whole-body tomographic image to obtain an initial organ matrix includes: Performing an organ metabolic quantitative analysis on the target organ tomographic image to obtain an organ metabolic quantitative analysis value; Constructing a single-organ matrix for the plurality of organ metabolic quantitative analysis values of each target whole-body tomographic image to obtain a second single-organ initial matrix; Constructing a whole-body organ matrix for the plurality of second single-organ initial matrices to obtain the initial organ matrix.

5. The method according to claim 1, characterized in that, The performing a metabolic difference recognition on the target whole-body tomographic image according to the organ metabolic difference matrix and the metabolic reference matrix to obtain a target metabolic difference map includes: Obtaining the number of reference objects according to the metabolic reference matrix; Obtaining first matrix data from the organ metabolic difference matrix and obtaining second matrix data from the metabolic reference matrix; Calculate the degree of metabolic abnormality based on the first matrix data, the second matrix data, and the number of reference objects to obtain target metabolic abnormality degree data; the target metabolic abnormality degree data is used to represent the data of the abnormal association degree between organs extracted from the target whole-body tomographic image; Construct a target metabolic abnormality degree matrix based on the target metabolic abnormality degree data; Construct a target metabolic difference map according to the target metabolic abnormality degree matrix to obtain the target metabolic difference map.

6. The method according to claim 5, characterized in that, The constructing a target metabolic difference map according to the target metabolic abnormality degree matrix to obtain the target metabolic difference map includes: Determine the number of organs of the target organ according to the target metabolic abnormality degree matrix; Obtain candidate organs in the target organ, and obtain candidate metabolic abnormality degree data of the candidate organs from the target metabolic abnormality degree matrix, where the candidate metabolic abnormality degree data is used to represent the data of the abnormal association degree between the candidate organs and other organs; Perform organ abnormal association calculation according to the multiple candidate metabolic abnormality degree data of the candidate organs and the number of organs to obtain target organ abnormal association data; Construct the target metabolic difference map according to the target organ abnormal association data of the multiple candidate organs.

7. The method according to any one of claims 1 to 6, characterized in that, Before calculating the metabolic difference between the pre-obtained metabolic reference matrix and the initial organ matrix according to the preset object metabolic difference sub-network to obtain the organ metabolic difference matrix, the method further includes: constructing the object metabolic difference sub-network, specifically including: Obtain a reference data set and a sample data set, where the sample data set includes a sample subset of sample objects, and the sample subset includes a plurality of sample whole-body tomographic data collected in a second preset time series, and the sample whole-body tomographic data includes sample whole-body tomographic images collected at a preset acquisition time point in the second preset time series; the reference data set includes a reference subset of reference objects, and the reference subset includes a plurality of reference whole-body tomographic data collected in the second preset time series, and the reference whole-body tomographic data includes reference whole-body tomographic images collected at the preset acquisition time point; Extract the organ regions from the reference whole-body tomographic data to obtain reference organ tomographic images of the reference organs; and construct a reference sub-network according to the reference organ tomographic images of the plurality of reference whole-body tomographic data; Construct a perturbed object set according to the reference data set and the sample data set; the perturbed object set includes a perturbed subset, and the perturbed subset includes a plurality of perturbed whole-body tomographic data collected in the second preset time series, and the perturbed whole-body tomographic data includes perturbed whole-body tomographic images collected at the preset acquisition time point; Extract the organ regions from the perturbed whole-body tomographic data to obtain perturbed organ tomographic images of the perturbed organs; and construct a perturbed sub-network according to the perturbed organ tomographic images of the plurality of perturbed whole-body tomographic data of the perturbed object set; Perform network difference calculation on the reference sub-network and the perturbed sub-network, and construct the object metabolic difference sub-network according to the result of the network difference calculation.

8. An apparatus for medical image recognition, characterized in that, The device includes: An acquisition module, configured to acquire a target image set of a target object, where the target image set includes a plurality of target whole-body tomographic images acquired in a first preset time series; the target whole-body tomographic image refers to an image acquired based on positron emission tomography; An extraction module, configured to extract an organ region from the target whole-body tomographic image to obtain a target organ tomographic image of a target organ; A matrix construction module, configured to construct an organ matrix for the target organ tomographic image of each target whole-body tomographic image to obtain an initial organ matrix; A difference calculation module, configured to perform a metabolic difference calculation on a pre-acquired metabolic reference matrix and the initial organ matrix according to a preset object metabolic difference sub-network to obtain an organ metabolic difference matrix; A graph construction module, configured to perform a metabolic difference recognition on the target whole-body tomographic image according to the organ metabolic difference matrix and the metabolic reference matrix to obtain a target metabolic difference graph, where the target metabolic difference graph is used to represent the metabolic difference information between any two target organs in the target whole-body tomographic image.

9. An electronic device, characterized in that, Comprising: At least one memory; At least one processor; At least one computer program; The at least one computer program is stored in the at least one memory, and the at least one processor executes the at least one computer program to implement: The method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and the computer program is used to cause a computer to execute: The method according to any one of claims 1 to 7.

Images