Lung-brain image processing method and apparatus, device, and storage medium

Abstract

Embodiments of the present application relate to the technical fields of AI and digital healthcare, and provide a lung-brain image processing method and apparatus, a device, and a storage medium. The method comprises: collecting medical image data and clinical feature data, wherein the medical image data comprises lung images and brain images, and the medical image data is obtained on the basis of positron emission tomography-computed tomography; screening for target clinical features from the clinical feature data; preprocessing the lung images to obtain a lung preliminary image; preprocessing the brain images to obtain a brain preliminary image; performing deep and shallow feature extraction on the lung preliminary image to obtain lung deep features and lung shallow features, and performing deep feature extraction on the brain preliminary image to obtain brain deep features; performing feature screening on the target clinical features, the lung deep features, the lung shallow features, and the brain deep features to obtain target features; and performing disease classification on the basis of the target features to obtain a target disease category. The embodiments of the present application can improve the accuracy and efficiency of disease classification.

Description

Lung-brain image processing method, device, equipment, and storage medium Technical Field

[0001] The present application relates to the fields of artificial intelligence and digital medical technology, and in particular to a lung-brain image processing method, apparatus, device, and storage medium. Background Art

[0002] Currently, there are some common problems with using clinical techniques such as biopsy to predict brain metastasis of non-small cell lung cancer, such as the difficulty of early detection and the uncertainty of prediction due to tumor cell heterogeneity. In addition, biopsy, as an invasive detection method, is not suitable for patients with poor physical conditions. In addition, due to the heterogeneity of tumor cells, traditional clinical detection methods have a certain degree of false positives. Moreover, the superficial features of the tumor may not accurately predict the high-dimensional difference features, and cannot well distinguish between highly similar primary tumors that may have already caused distant metastasis, resulting in inaccurate prediction results. Traditional methods for detecting brain metastasis of non-small cell lung cancer are usually based on lung CT images and brain MR images. Lung CT images and brain MR images are not acquired synchronously, resulting in high costs.

[0003] Summary of the Invention

[0004] The main purpose of the embodiments of the present application is to propose a lung-brain image processing method and apparatus, equipment, and storage medium, aiming to improve the accuracy and efficiency of identifying disease categories.

[0005] To achieve the above objectives, a first aspect of an embodiment of the present application provides a lung-brain image processing method, the method comprising:

[0006] Collecting medical image data and clinical characteristic data; wherein the medical image data includes lung images and brain images, and the medical image data is obtained based on positron emission tomography;

[0007] screening target clinical features from the clinical feature data;

[0008] Preprocessing the lung image to obtain a preliminary lung image;

[0009] Preprocessing the brain image to obtain a preliminary brain image;

[0010] Performing deep and shallow feature extraction on the preliminary lung image to obtain lung deep features and lung shallow features, and performing deep feature extraction on the preliminary brain image to obtain brain deep features;

[0011] Performing feature screening on the target clinical feature, the lung depth feature, the lung superficial feature, and the brain depth feature to obtain a target feature;

[0012] Diseases are classified according to the target characteristics to obtain target disease categories.

[0013] In some embodiments, the lung images include a lung function image sequence and a lung ecology image sequence, and preprocessing the lung images to obtain preliminary lung images includes:

[0014] Adjusting the resolution of the lung function image sequence and the resolution of the lung ecology image sequence to a target resolution to obtain a lung adjusted image;

[0015] Slice extraction is performed on the lung adjustment image to obtain a lung slice image; wherein the lung slice image includes a lung function slice image and a lung ecology slice image;

[0016] performing standardization processing on the lung ecology slice image to obtain a lung ecology standardized image;

[0017] The lung function slice image and the lung ecology standardized image are fused to obtain a preliminary lung image.

[0018] In some embodiments, performing image fusion based on the lung function slice image and the lung ecology standardized image to obtain a preliminary lung image includes:

[0019] Performing standardized uptake value conversion on pixels of the pulmonary function slice image to obtain a standardized pulmonary function image;

[0020] The lung function standardized image and the lung ecology standardized image are fused to obtain the preliminary lung image.

[0021] In some embodiments, the brain image includes a brain function image and a brain ecology image, and preprocessing the brain image to obtain a preliminary brain image includes:

[0022] performing standardization processing on the brain ecology image to obtain a brain ecology standardized image;

[0023] performing standardized uptake value conversion on pixels of the brain function image to obtain a standardized brain function image;

[0024] The brain function standardized image and the brain ecology standardized image are fused to obtain the preliminary brain image.

[0025] In some embodiments, the feature screening of the target clinical feature, the lung depth feature, the lung superficial feature, and the brain depth feature to obtain the target feature includes:

[0026] The target clinical feature, the lung depth feature, the lung superficial feature, and the brain depth feature are used as candidate features, and the candidate features are compared with preset reference labels to obtain feature difference data;

[0027] screening out selected features from the candidate features according to the feature difference data;

[0028] The target feature is screened out from the selected features based on a preset minimum absolute shrinkage and selection operator model.

[0029] In some embodiments, classifying the disease according to the target feature to obtain the target disease category includes:

[0030] Obtain at least two pre-trained target prediction models; wherein each of the target prediction models includes a target classifier, and the target classifier adopts a support vector machine;

[0031] Based on the target classifier, feature correlation analysis is performed on the target features to obtain the target disease category.

[0032] In some embodiments, screening out target clinical features from the clinical feature data includes:

[0033] Performing a significance assessment on the original clinical characteristics of the clinical characteristic data to obtain significance assessment data;

[0034] The original clinical features of the clinical feature data are screened based on the significance evaluation data to obtain the target clinical features.

[0035] To achieve the above-mentioned objectives, a second aspect of an embodiment of the present application provides a lung-brain image processing device, the device comprising:

[0036] A data acquisition module, configured to acquire medical image data and clinical characteristic data; wherein the medical image data includes lung images and brain images, and the medical image data is obtained based on positron emission tomography (PET);

[0037] A clinical feature screening module, configured to screen target clinical features from the clinical feature data;

[0038] A lung image preprocessing module, configured to preprocess the lung image to obtain a preliminary lung image;

[0039] A brain image preprocessing module, which preprocesses the brain image to obtain a preliminary brain image;

[0040] a feature extraction module, configured to perform deep and shallow feature extraction on the preliminary lung image to obtain lung deep features and lung shallow features, and perform deep feature extraction on the preliminary brain image to obtain brain deep features;

[0041] a target feature selection module, configured to perform feature screening on the target clinical feature, the lung depth feature, the lung superficial feature, and the brain depth feature to obtain a target feature;

[0042] The disease classification module is used to classify the disease according to the target characteristics to obtain the target disease category.

[0043] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the method described in the first aspect when executing the computer program.

[0044] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium and stores a computer program. When the computer program is executed by a processor, the method described in the first aspect is implemented.

[0045] The lung-brain image processing method, device, equipment, and storage medium proposed in the embodiments of the present application collect medical image data and clinical feature data, wherein the medical image data includes lung images and brain images, and the medical image data is obtained based on positron emission tomography. Target clinical features are screened out from the clinical feature data, the lung image is preprocessed to obtain a preliminary lung image, the brain image is preprocessed to obtain a preliminary brain image, the preliminary lung image is extracted for deep and shallow features to obtain lung depth features and lung shallow features, the preliminary brain image is extracted for deep features to obtain brain depth features, and the target clinical features, lung depth features, lung shallow features, and brain depth features are feature screened to obtain target features, thereby performing disease classification based on the target features to obtain a target disease category. The embodiments of the present application are based on PET-CT images and combine the deep and shallow features of the lungs and the depth features of the brain to perform disease classification, which can improve the accuracy and efficiency of disease classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] FIG1 is a flow chart of a lung-brain image processing method provided in an embodiment of the present application;

[0047] FIG2 is a flow chart of step 103 in FIG1 ;

[0048] FIG3 is a flow chart of step 204 in FIG2 ;

[0049] FIG4 is a flow chart of step 104 in FIG1 ;

[0050] FIG5 is a flow chart of step 106 in FIG1 ;

[0051] FIG6 is a flow chart of step 107 in FIG1 ;

[0052] FIG7 is a schematic diagram of the structure of a lung-brain image processing device provided in an embodiment of the present application;

[0053] FIG8 is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0054] 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.

[0055] 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.

[0056] 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.

[0057] First, let’s analyze some of the terms used in this application:

[0058] Artificial intelligence (AI) is a new technical discipline that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence. A branch of computer science, AI seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thinking. It also encompasses the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results.

[0059] Positron Emission Computed Tomography (PET): PET is a nuclear medicine instrument that uses positron radionuclide imaging, representing functional imaging, to examine tissue metabolism and provide molecular information on function and metabolism. The general principle of PET is that a substance, typically essential for biological metabolism, such as glucose, protein, nucleic acid, or fatty acid, is labeled with a short-lived radionuclide (such as F18 or carbon 11). After injection into the human body, the accumulation of the substance during metabolism is used to reveal metabolic activity, thereby achieving diagnostic purposes. PET uses annihilation radiation and positron collimation (or photon collimation) technology to noninvasively, quantitatively, and dynamically measure the spatial distribution, quantity, and dynamic changes of the PET imaging agent or its metabolites in vivo. This provides molecular-level imaging information on the biochemical, physiological, and functional metabolic changes caused by the interaction of the PET imaging agent with its target (such as receptors, enzymes, ion channels, antigenic determinants, and nucleic acids), providing data for clinical research.

[0060] Computed Tomography (CT): CT stands for computed tomography, a morphological imaging technique that can examine the structure of tissues and organs and provide precise anatomical localization of lesions. CT utilizes precisely collimated X-ray beams, gamma rays, and ultrasound waves, along with highly sensitive detectors, to scan a specific part of the human body in sections. CT offers fast scan times and clear images, making it suitable for diagnosing a wide range of diseases.

[0061] PET-CT: PET-CT is positron emission tomography, an imaging technology that integrates PET and CT. PET-CT is a technology that perfectly integrates functional imaging and morphological imaging, combining PET and CT together and using the same examination bed and an imaging workstation.

[0062] Ordered Subsets Expectation Maximization (OSEM): The OSEM algorithm has higher reconstructed image quality and shorter calculation time and is applied to PET image reconstruction algorithm.

[0063] Least absolute shrinkage and selection operator (LASSO); LASSO is a regression analysis technique used for variable selection and regularization. LASSO can also be used for feature selection.

[0064] Support Vector Machine (SVM): An SVM is a generalized linear classifier that performs binary classification using supervised learning. Its decision boundary is a maximum-margin hyperplane calculated for the learning sample. SVM uses a hinge loss function to calculate empirical risk and incorporates a regularization term to optimize structural risk. It is a sparse and robust classifier. It can perform nonlinear classification using kernel methods, a kernel learning approach.

[0065] Currently, there are some common problems with using clinical techniques such as biopsy to predict brain metastasis of non-small cell lung cancer, such as the difficulty of early detection and the uncertainty of prediction due to tumor cell heterogeneity. In addition, biopsy, as an invasive detection method, is not suitable for patients with poor physical conditions. In addition, due to the heterogeneity of tumor cells, traditional clinical detection methods have a certain degree of false positives. Moreover, the superficial features of the tumor may not accurately predict the high-dimensional difference features, and cannot well distinguish between highly similar primary tumors that may have already caused distant metastasis, resulting in inaccurate prediction results. Traditional methods for detecting brain metastasis of non-small cell lung cancer are usually based on lung CT images and brain MR images. Lung CT images and brain MR images are not acquired synchronously, resulting in high costs.

[0066] Based on this, the embodiments of the present application provide a lung-brain image processing method, apparatus, device, and storage medium, aiming to improve the accuracy and efficiency of disease classification.

[0067] The lung-brain image processing method, apparatus, device, and storage medium provided in the embodiments of the present application are specifically illustrated through the following examples. First, the lung-brain image processing method in the embodiments of the present application is described.

[0068] 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.

[0069] 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.

[0070] The method for processing lung-brain images provided in the embodiment of the present application relates to the field of artificial intelligence technology. The method for processing lung-brain images 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 method for processing lung-brain images, etc., but is not limited to the above forms.

[0071] This application is applied to medical scenarios, and the relevant medical images include objects of the type of lesion, that is, the part of the body where the disease occurs. Medical images refer to images of internal tissues, such as the stomach, abdomen, heart, knee, and brain, obtained non-invasively for medical treatment or medical research. Examples include CT, MRI (Magnetic Resonance Imaging), US (ultrasonic), X-ray images, electroencephalograms, and images generated by medical instruments using optical photography.

[0072] 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.

[0073] It should be noted that in each specific embodiment of the present application, when it comes to the need to perform relevant processing based on data related to the user's identity or characteristics, such as user information, clinical characteristics, etc., the user's permission or consent 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 user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0074] FIG1 is an optional flowchart of a lung-brain image processing method provided in an embodiment of the present application. The method in FIG1 may include but is not limited to steps 101 to 107 .

[0075] Step 101: Collect medical image data and clinical characteristic data; wherein the medical image data includes lung images and brain images, and the medical image data is obtained based on positron emission tomography (PET);

[0076] Step 102, screening target clinical features from the clinical feature data;

[0077] Step 103, preprocessing the lung image to obtain a preliminary lung image;

[0078] Step 104, preprocessing the brain image to obtain a preliminary brain image;

[0079] Step 105: performing deep and shallow feature extraction on the preliminary lung image to obtain lung deep features and lung shallow features, and performing deep feature extraction on the preliminary brain image to obtain brain deep features;

[0080] Step 106, performing feature selection on the target clinical features, lung depth features, lung superficial features, and brain depth features to obtain target features;

[0081] Step 107: classify the disease according to the target features to obtain the target disease category.

[0082] Steps 101 to 107 shown in the embodiment of the present application are performed by collecting medical image data and clinical feature data, wherein the medical image data includes lung images and brain images, and the medical image data is obtained based on positron emission tomography. Target clinical features are screened out from the clinical feature data, the lung image is preprocessed to obtain a preliminary lung image, the brain image is preprocessed to obtain a preliminary brain image, deep and shallow features are extracted from the preliminary lung image to obtain lung depth features and lung shallow features, deep features are extracted from the preliminary brain image to obtain brain depth features, and feature selection is performed on the target clinical features, lung depth features, lung shallow features and brain depth features to obtain target features, thereby performing disease classification based on the target features to obtain a target disease category. The embodiment of the present application is based on PET-CT images and combines lung deep and shallow features and brain depth features for disease classification, which can improve the accuracy and efficiency of disease classification.

[0083] In some embodiments, step 101 involves obtaining medical image data based on positron emission tomography (PET-CT). The medical image data includes lung images and brain images. The lung image is an image of the patient's lungs, and the brain image is an image of the patient's brain. Both the lung image and the brain image are PET-CT images. In one application scenario, a whole-body PET-CT scan of a non-small cell lung cancer patient is performed to obtain medical image data. More specifically, for example, a whole-body PET-CT scan of the brain and entire body of a non-small cell lung cancer patient is performed. The scanning parameters for a brain CT image can be set to: tube voltage 120 kV, X-ray tube current 175 mA, exposure time 1000 mS, and slice thickness 5 mm. The scanning parameters for a whole-body CT image can be set to: tube voltage 120 kV, X-ray tube current 75 mA, exposure time 500 mS, and slice thickness 5 mm. Scanning parameters for brain and whole-body PET images were: acquisition time 50-60 minutes after 18F-FDG tracer injection, reconstruction algorithm OSEM (Ordered Subset Expectation Maximization), and 5-mm slice thickness. OSEM is a positron emission tomography (PET) image reconstruction algorithm.

[0084] In one application scenario, the tumor contours in the PET-CT images of the lungs were pre-delineated and reviewed by two experienced radiologists. The radiologists could then delineate the region of interest (ROI) where the lung tumor was located (without excluding the necrotic area). In addition, the radiologists could also delineate the region of interest (ROI) in the PET-CT images of the brain.

[0085] Clinical characteristic data is used to characterize the patient's clinical characteristics. Clinical characteristic data can represent basic patient information, such as age, weight, and gender. Clinical characteristic data can also characterize the patient's lesion status, such as TNM stage, tumor location, pathological type, PS (Performance Status) score, and EGFR (epidermal growth factor receptor) mutation type. TNM is used to stage tumors, where T represents the primary tumor, N represents regional lymph nodes, and M represents distant metastasis. The tumor stage is determined by comprehensively determining the size and extent of invasion of the tumor (T), the location and number of lymph nodes (N), and the presence of distant metastasis (M). The PS score can be used to evaluate the overall behavior and daily living ability of cancer patients. Clinical characteristic data includes multiple raw clinical characteristics, which can include age, weight, or gender. Raw clinical characteristics can also include TNM characteristics, tumor location characteristics, pathological type characteristics, PS score characteristics, or EGFR mutation type characteristics.

[0086] In some embodiments, step 102 may include, but is not limited to:

[0087] Performing a significance evaluation on the original clinical characteristics of the clinical characteristic data to obtain significance evaluation data;

[0088] The original clinical features of the clinical feature data are screened based on the significance evaluation data to obtain the target clinical features.

[0089] Specifically, the original clinical features of the clinical feature data are evaluated for feature significance. The resulting significance evaluation data is used to characterize whether the original clinical features of the clinical feature data are significant. The significance evaluation data is represented by a probability P. The original clinical features are screened based on the significance evaluation data, and original clinical features that meet the requirements are selected as target clinical features. In one embodiment, original clinical features with a probability P < 0.05 are evaluated as significant and selected as target clinical features. The target clinical features may have an impact on brain metastasis prediction results.

[0090] In one application scenario, the age feature is evaluated for feature significance, and the obtained probability P is 0.04, indicating that the age feature is significant. The age feature is then used as the target clinical feature, and the age feature may affect the prediction results of brain metastasis.

[0091] In an actual scenario, medical image data of 226 patients were collected, including 109 males and 117 females with a median age of 58.7 years and a range of 32-89 years. These 226 patients underwent PET / CT whole-body scans, and the age characteristics, weight characteristics, gender characteristics, TNM characteristics, tumor location characteristics, pathological type characteristics, PS score characteristics, and EGFR mutation type characteristics of these 226 patients were collected as original clinical feature evaluations, and the original clinical feature evaluations with a probability P<0.05 were selected as target clinical features.

[0092] The lung images include a lung function image sequence and a lung ecology image sequence. Please refer to FIG2 . In some embodiments, step 103 may include but is not limited to:

[0093] Step 201, adjusting the resolution of the lung function image sequence and the resolution of the lung ecology image sequence to a target resolution to obtain a lung adjusted image;

[0094] Step 202: Slice the lung adjustment image to obtain a lung slice image; wherein the lung slice image includes a lung function slice image and a lung ecology slice image;

[0095] Step 203, performing standardization processing on the lung ecology slice image to obtain a lung ecology standardized image;

[0096] Step 204 : performing image fusion based on the lung function slice image and the lung ecology standardized image to obtain a preliminary lung image.

[0097] In some embodiments, the lung image includes a lung function image and a lung ecology image, the lung function image is a PET image, the lung ecology image is a CT image, the lung function image includes a lung function image sequence, and the lung ecology image includes a lung ecology image sequence. Through step 201, the resolution of the lung function image sequence and the resolution of the lung ecology image sequence are adjusted to the same target resolution, so that the lung function image and the lung ecology image have the same size. The image obtained by fusing the lung function image and the lung ecology image with the same size is the lung adjusted image, that is, the lung adjusted image is obtained by fusing the PET image and the CT image with the same size.

[0098] In some embodiments, step 202 involves performing slice extraction on the lung adjustment image to ensure that each patient's slice image contains complete tumor information. The lung function slice image is a PET image, and the lung ecology slice image is a CT image. During step 202, the PET and CT images of the lung adjustment image are in a registration state.

[0099] Image registration is the process of matching and overlaying two or more images acquired at different times, using different sensors (imaging devices), or under different conditions (such as weather, illumination, camera position, and angle). The registration process involves first extracting features from the two images to obtain feature points; then performing a similarity measurement to find matching feature point pairs; then, using these matched feature point pairs to determine image space coordinate transformation parameters; and finally, performing image registration based on these coordinate transformation parameters.

[0100] In some embodiments, step 203 of normalizing the lung ecology slice image may include normalizing the lung ecology slice image to a window level of 40 HU and a window width of 300 HU. The purpose of step 203 is to normalize the pixel data of the lung ecology slice image so that the image features have equal status.

[0101] Referring to FIG. 3 , in some embodiments, step 204 may include, but is not limited to:

[0102] Step 301, performing standardized uptake value conversion on the pixels of the lung function slice image to obtain a lung function standardized image;

[0103] Step 302 : Fusing the standardized lung function image and the standardized lung ecology image to obtain a preliminary lung image.

[0104] In some embodiments, step 301 converts the pixels of the lung function slice image from grayscale values ​​to standardized uptake values ​​(SUVs). Through step 301, the influence of the patient's individual weight and tracer dosage on the prediction result in the subsequent step 107 can be weakened.

[0105] In step 302 of some embodiments, the lung function standardized image obtained in step 301 and the lung ecology standardized image obtained in step 203 are fused to obtain a preliminary lung image.

[0106] Through the above steps 201 to 204, the lung image is preprocessed to facilitate the subsequent step 105 of extracting deep and shallow features from the preliminary lung image. During the lung image preprocessing process, the lung function image and the lung ecology image are made to have the same size, and slice extraction is performed on the lung adjustment image to ensure that each patient's slice image contains complete tumor information. The pixel data of the lung ecology slice image is normalized to ensure that the image features have equal status. The pixels of the lung function slice image are converted from grayscale values ​​to standardized uptake values ​​to weaken the influence of the patient's individual weight and tracer dose on the prediction results in the subsequent step 107.

[0107] The brain image includes a brain function image and a brain ecology image. Referring to FIG. 4 , step 104 in some embodiments may include, but is not limited to:

[0108] Step 401, performing standardization processing on the brain ecology image to obtain a brain ecology standardized image;

[0109] Step 402 , performing standardized uptake value conversion on the pixels of the brain function image to obtain a standardized brain function image;

[0110] Step 403 : Fusing the standardized brain function image and the standardized brain ecology image to obtain a preliminary brain image.

[0111] Through the above steps 401 to 403 , the brain image is preprocessed to improve the accuracy of deep feature extraction of the preliminary brain image in step 105 .

[0112] In some embodiments, step 401 of the present invention normalizes the brain ecology image using a similar principle to the normalization of the lung ecology slice image in step 203. Specifically, the brain ecology image is normalized using a window level of 40 HU and a window width of 300 HU. The purpose of step 401 is to normalize the pixel data of the brain ecology slice image so that the image features have equal status.

[0113] In some embodiments, step 402 converts the pixels of the brain function image to standardized uptake values ​​(SEVs) similarly to the SEVs performed on the pixels of the lung function image in step 301. Specifically, the pixels of the brain function slice image are converted from grayscale values ​​to SEVs. Step 402 can mitigate the effects of the patient's individual weight and tracer dosage on the prediction results in the subsequent step 107.

[0114] In step 403 of some embodiments, the preliminary brain image obtained by fusing the standardized brain function image (PET image) and the standardized brain ecology image (CT image) is a PET-CT fusion image.

[0115] In the above steps 401 to 403 , the difference from the lung image processing method is that the resolution adjustment in step 201 and the slice extraction in step 202 are not performed on the brain image.

[0116] Through the above steps 401 to 403, the brain image is preprocessed to facilitate deep feature extraction of the preliminary brain image in the subsequent step 105. During the preprocessing of the brain image, the pixel data of the brain ecology slice image is normalized so that the image features have equal status, and the pixels of the brain function slice image are converted from grayscale values ​​to standardized uptake values ​​to weaken the influence of the patient's individual weight and tracer dose on the prediction results in the subsequent step 107.

[0117] In step 105 of some embodiments, performing depth and shallow feature extraction on the preliminary lung image includes:

[0118] Perform deep feature extraction on the preliminary lung image to obtain the lung depth feature;

[0119] Perform shallow feature extraction on the preliminary lung image to obtain the shallow features of the lung;

[0120] Deep feature extraction is performed on the preliminary brain image to obtain brain depth features.

[0121] In one embodiment, both deep and shallow feature extraction were performed on the preliminary lung image, while only deep feature extraction was performed on the preliminary brain image. In one application scenario, shallow lung features were extracted from the preliminary lung image and its derivatives (Laplacian of Gaussian function (LoG) and wavelet transform image) using traditional radiomics methods (e.g., the PyRadiomics package). The Sigma values ​​of the derivative images were 1.0, 3.0, 5.0, and 7.0, respectively.

[0122] In one embodiment, deep feature extraction is performed on the preliminary lung image and shallow feature extraction is performed on the preliminary lung image through a 3D network. For example, a C3D network can be used to extract deep features from the preliminary lung image and deep features from the preliminary brain image. In one application scenario, the C3D network has 8 convolutions (3×3×3), 5 maximum pooling layers and 3 fully connected layers. The last fully connected layer is the output unit, and the output unit has a total of 1000 features. Before inputting the C3D network, the pixel values ​​of the PET image and the CT image need to be normalized and interpolated. The number of slices of the preliminary lung image is 24, and the number of slices of the preliminary brain image is 33. In another application scenario, a 3D convolutional network (Res18, R3D) can be used to perform shallow feature extraction on the preliminary lung image.

[0123] Since the PET-CT images of each patient are sorted in chronological order, which is consistent with the learning principle of the 3D convolutional network, this application uses a 3D network to process the PET-CT images, which is more reasonable than a 2D network.

[0124] Referring to FIG. 5 , in some embodiments, step 106 may include, but is not limited to:

[0125] Step 501: Target clinical features, lung depth features, lung superficial features, and brain depth features are selected as candidate features, and the candidate features are compared with preset reference labels to obtain feature difference data;

[0126] Step 502, screening selected features from candidate features based on feature difference data;

[0127] Step 503 : Filter out target features from the selected features based on a preset minimum absolute shrinkage and selection operator model.

[0128] Through the above steps 501 to 503 , target features with significant differences can be determined to determine the most potential transfer prediction features, so as to facilitate the prediction in step 107 .

[0129] In some embodiments, steps 501 and 502 are performed on target clinical features, lung depth features, superficial lung features, and brain depth features as candidate features. These candidate features are then compared with a preset reference label in a loop for significant differences, thereby filtering selected features from the candidate features based on the feature difference data. In one application scenario, the reference label is a brain metastasis label, and the feature difference data probability P is obtained. The candidate features are compared with the brain metastasis label in a loop for P value calculation, and features with a P value < 0.05 are selected as selected features. Typically, the selected features are considered to have significant differences.

[0130] In some embodiments, in step 503, a least absolute shrinkage and selection operator (LASSO) with 5-fold cross-validation can be used to screen a target feature from the selected features. The target feature can determine the most potential metastasis prediction feature to facilitate prediction in step 107. In one application scenario, the Mann-Whitney U test (Mann-Whitney rank sum test) is used to measure the significance of the difference between the target feature and the brain metastasis signature.

[0131] In actual application scenarios, significant target clinical features (P value less than 0.05), lung depth features, lung superficial features, and brain depth features are linearly connected, and feature selection is performed to remove redundant features and identify target features with strong brain metastasis prediction potential.

[0132] This application uses ONLY_TRAD, C3D_Lung, and C3D_LungBrain to investigate whether different tumor information from different locations has different effects on brain metastasis prediction. Specifically, this application analyzes deep features of the lung and brain, or shallow features of the lung, to predict brain metastasis and compares which features are most effective in predicting brain metastasis. This application also uses C3D_LungBrain, Res18_LungBrain, and R3D_LungBrain to evaluate the stability of deep lung-brain features in identifying stroke-related tumors, specifically to investigate the stability of using deep lung-brain features to predict brain metastasis.

[0133] This application performs vector linear fusion on the target clinical features, lung depth features, lung superficial features and brain depth features of patients with non-small cell lung cancer, and establishes 5 models, as shown in Table 1 below, wherein the ONLY_TRAD model is constructed based on lung superficial features, the C3D_Lung model is constructed based on lung superficial features and lung depth features, the C3D_LungBrain model is constructed based on lung superficial features, lung depth features and brain depth features, the Res18_LungBrain model is constructed based on lung superficial features, lung depth features and brain depth features, and the R3D_LungBrain model is constructed based on lung superficial features, lung depth features and brain depth features. In practical application scenarios, two sets of comparative experiments were conducted using the five models described above. First, the ONLY_TRAD model, the C3D_Lung model, and the C3D_LungBrain model were used to investigate whether different tumor information from different locations had different effects on brain metastasis prediction. Second, the C3D_LungBrain model, the Res18_LungBrain model, and the R3D_LungBrain model were used to evaluate the stability of lung-brain depth features (i.e., lung depth features and brain depth features) in identifying stroke-related tumors. In Table 1 below, "**" indicates that features with a P < 0.05 were selected as target clinical features. Furthermore, experimental verification showed that the C3D_LungBrain model had the best prediction results.

[0134] Table 1

[0135] Referring to FIG. 6 , in some embodiments, step 107 may include, but is not limited to:

[0136] Step 601: Obtain at least two pre-trained target prediction models; wherein each target prediction model includes a target classifier, and the target classifier adopts a support vector machine;

[0137] Step 602: Perform feature correlation analysis on the target features based on the target classifier to obtain the target disease category.

[0138] In step 601 of one embodiment, five target prediction models are used. The five target prediction models are the five models mentioned above: ONLY_TRAD model, C3D_Lung model, C3D_LungBrain model, Res18_LungBrain model, and R3D_LungBrain model. A support vector machine is introduced into each target prediction model.

[0139] In step 602 of one embodiment, the target disease category is used to characterize the presence of brain metastasis. Feature correlation analysis of the target features is performed using a support vector machine to obtain the target disease category to predict the presence of brain metastasis. Classifier performance is evaluated using metrics such as the receiver operating characteristic curve (ROC) and its area under the curve (AUC), F1 score, and accuracy. The F1 score is a machine learning evaluation metric for classification models and is the harmonic mean of precision and recall. All models are evaluated by integrating clinical data that influence metastasis prediction (P < 0.05).

[0140] In the embodiment of the present application, through steps 601 and 602, the disease category can be predicted to predict whether brain metastasis occurs.

[0141] Currently, the use of clinical techniques such as biopsy to predict brain metastasis in non-small cell lung cancer faces several common challenges, including the difficulty of early detection and uncertainty in prediction due to tumor cell heterogeneity. Furthermore, biopsy, as an invasive test, is unsuitable for patients in poor physical condition. Furthermore, due to tumor cell heterogeneity, traditional clinical testing methods are prone to false positives. Furthermore, superficial tumor features may not accurately predict high-dimensional differential features, making it difficult to distinguish highly similar primary tumors that may have already undergone distant metastasis, leading to inaccurate predictions.

[0142] Traditional methods for detecting brain metastases in non-small cell lung cancer are typically based on lung CT images and brain MR images, which are not acquired synchronously, resulting in high costs. This application seamlessly combines PET and CT technologies based on PET-CT images, greatly improving the accuracy of clinical diagnosis and reducing costs, including precise positioning and qualitative assessment. PET-CT images also overcome the problem of synchronous scanning. Furthermore, this application uses a 3D CNN algorithm and multimodal radiomics data to improve the accuracy of predicting the occurrence of brain metastases in patients with non-small cell lung cancer based on PET-CT images. This application also combines the patient's clinical characteristics, superficial lung features, and deep lung-brain features of PET-CT images, achieving superior performance in predicting brain metastases in patients with non-small cell lung cancer. This overcomes the shortcomings of biopsy and traditional radiomics, providing a more effective and reliable solution so that doctors can develop personalized treatment plans for patients earlier. In this embodiment, the use of lung-brain images before metastasis occurs to predict the probability of brain metastasis after treatment is demonstrated, demonstrating a certain degree of foresight. This application enables quantitative mapping of medical images and a deeper understanding of shallow and deep tumor characteristics. This is a non-invasive method. Compared with invasive detection methods such as traditional biopsies, non-invasive methods have the potential to identify brain metastasis-related features with statistical significance, thereby helping to develop targeted treatment plans for patients with non-small cell lung cancer.

[0143] The present embodiment extracts and selects the most discriminative shallow and deep features from PET-CT images to identify potential brain metastasis-related tumors, and calculates the significance of the final retained features, as well as the ROC curve and AUC value of the prediction model. Finally, different deep networks are used to extract deep features. The present embodiment compares the prediction results using different deep models to evaluate the robustness of deep features in predicting metastasis.

[0144] Referring to FIG. 7 , an embodiment of the present application further provides a lung-brain image processing device that can implement the above-mentioned lung-brain image processing method. The device includes:

[0145] A data acquisition module, configured to acquire medical image data and clinical characteristic data; wherein the medical image data includes lung images and brain images, and the medical image data is obtained based on positron emission tomography (PET);

[0146] A clinical feature screening module, used to screen target clinical features from clinical feature data;

[0147] A lung image preprocessing module, used for preprocessing the lung image to obtain a preliminary lung image;

[0148] A brain image preprocessing module preprocesses the brain image to obtain a preliminary brain image;

[0149] The feature extraction module is used to extract deep and shallow features from the preliminary lung image to obtain lung deep features and lung shallow features, and to extract deep features from the preliminary brain image to obtain brain deep features;

[0150] The target feature selection module is used to screen the target clinical features, lung depth features, lung superficial features, and brain depth features to obtain the target features;

[0151] The disease classification module is used to classify diseases according to target features and obtain the target disease category.

[0152] In some embodiments, the clinical feature screening module is specifically used to implement:

[0153] Performing a significance evaluation on the original clinical characteristics of the clinical characteristic data to obtain significance evaluation data;

[0154] The original clinical features of the clinical feature data are screened based on the significance evaluation data to obtain the target clinical features.

[0155] In some embodiments, the lung image preprocessing module is specifically configured to implement:

[0156] Adjusting the resolution of the lung function image sequence and the resolution of the lung ecology image sequence to the target resolution to obtain a lung adjustment image;

[0157] Slice extraction is performed on the lung adjustment image to obtain a lung slice image; wherein the lung slice image includes a lung function slice image and a lung ecology slice image;

[0158] Performing standardization processing on lung ecology slice images to obtain lung ecology standardized images;

[0159] The lung function slice image and the lung ecology standardized image are fused to obtain the preliminary lung image.

[0160] Specifically, the lung image preprocessing module can be used to implement the above steps 201 to 204, which will not be described in detail here.

[0161] In some embodiments, the lung image preprocessing module is configured to perform image fusion based on the lung function slice image and the lung ecology standardized image to obtain a preliminary lung image, specifically including:

[0162] Performing standardized uptake value conversion on the pixels of the lung function slice image to obtain a lung function standardized image;

[0163] The lung function standardized image and the lung ecology standardized image are fused to obtain the preliminary lung image.

[0164] Specifically, the lung image preprocessing module can be used to implement the above steps 301 to 302, which will not be described in detail here.

[0165] In some embodiments, the brain image preprocessing module is specifically configured to implement:

[0166] Performing standardization processing on the brain ecology image to obtain a brain ecology standardized image;

[0167] Performing standardized uptake value conversion on pixels of the brain function image to obtain a standardized brain function image;

[0168] The standardized brain function images and the standardized brain ecology images were fused to obtain a preliminary brain image.

[0169] Specifically, the lung image preprocessing module can be used to implement the above steps 401 to 403, which will not be described in detail here.

[0170] In some embodiments, the target feature selection module is used to implement:

[0171] The target clinical features, lung depth features, lung superficial features, and brain depth features are taken as candidate features, and the candidate features are compared with the preset reference labels to obtain feature difference data;

[0172] Filtering selected features from candidate features based on feature difference data;

[0173] Target features are filtered out from selected features based on the preset minimum absolute shrinkage and selection operator model.

[0174] Specifically, the target feature selection module can be used to implement the above steps 501 to 503, which will not be described in detail here.

[0175] In some embodiments, the disease classification module is used to implement:

[0176] Obtain at least two pre-trained target prediction models; wherein each target prediction model includes a target classifier, and the target classifier adopts a support vector machine;

[0177] Based on the target classifier, feature correlation analysis is performed on the target features to obtain the target disease category.

[0178] Specifically, the disease classification module can be used to implement the above steps 601 to 602, which will not be described in detail here.

[0179] The specific implementation of the lung-brain image processing device is basically the same as the specific embodiment of the lung-brain image processing method described above, and will not be repeated here.

[0180] The present application also provides an electronic device comprising a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned lung-brain image processing method. The electronic device can be any smart terminal, such as a tablet computer or an in-vehicle computer.

[0181] Please refer to FIG8 , which illustrates a hardware structure of an electronic device according to another embodiment. The electronic device includes:

[0182] The processor 801 may be implemented as 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.

[0183] The memory 802 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 802 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 802 and is called by the processor 801 to execute the lung-brain image processing method of the embodiments of this application.

[0184] Input / output interface 803, used to implement information input and output;

[0185] Communication interface 804, 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.);

[0186] Bus 805 , which transmits information between various components of the device (e.g., processor 801 , memory 802 , input / output interface 803 , and communication interface 804 );

[0187] The processor 801 , the memory 802 , the input / output interface 803 and the communication interface 804 are connected to each other in communication within the device via a bus 805 .

[0188] An embodiment of the present application further provides a storage medium, which is a computer-readable storage medium and stores a computer program. When the computer program is executed by a processor, the above-mentioned lung-brain image processing method is implemented.

[0189] 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.

[0190] The lung-brain image processing method, apparatus, equipment, and storage medium provided in the embodiments of the present application collect medical image data and clinical feature data, wherein the medical image data includes lung images and brain images, and the medical image data is obtained based on positron emission tomography. A preliminary lung image is obtained by preprocessing the lung image, and a preliminary brain image is obtained by preprocessing the brain image. The preliminary lung image is subjected to deep and shallow feature extraction to obtain lung depth features and lung shallow features, and the preliminary brain image is subjected to deep feature extraction to obtain brain depth features. Feature selection is performed on the target clinical features, lung depth features, lung shallow features, and brain depth features to obtain target features, and then disease classification is performed based on the target features to obtain the target disease category. The embodiments of the present application are based on PET-CT images and combine the deep and shallow features of the lungs and the depth features of the brain to perform disease classification, which can improve the accuracy and efficiency of disease classification.

[0191] This application seamlessly combines PET and CT technologies based on PET-CT images, greatly improving the accuracy of clinical diagnosis and reducing costs. At the same time, PET-CT images overcome the problem of synchronous scanning. In addition, this application uses 3D CNN algorithms and multimodal radiomics data to improve the accuracy of predicting the occurrence of brain metastases in patients with non-small cell lung cancer based on PET-CT images. This application also combines the patient's clinical characteristics, the shallow lung features, and the deep lung-brain features of PET-CT images, and performs better in predicting brain metastases in patients with non-small cell lung cancer. It also makes up for the shortcomings of biopsy and traditional imaging omics, providing a more effective and reliable solution so that doctors can make personalized treatment plans for patients earlier.

[0192] In addition, the embodiment of the present application uses lung-brain images before metastasis to predict the probability of brain metastasis after treatment of patients, which has a certain degree of foresight. The present application can quantitatively draw medical images and gain a deeper understanding of shallow features and deep tumor features. This is a non-invasive method. Compared with invasive detection methods such as traditional biopsy, non-invasive methods have the potential to identify brain metastasis-related features with statistical significance, which helps to formulate targeted treatment plans for patients with non-small cell lung cancer. In addition, the embodiment of the present application compares the prediction results by using different depth models to evaluate the robustness of deep features in predicting metastasis.

[0193] 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.

[0194] 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.

[0195] 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.

[0196] 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.

[0197] 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.

[0198] 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.

[0199] 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.

[0200] 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.

[0201] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, 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.

[0202] 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.

[0203] 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 method for processing lung and brain images, characterized in that, The method includes: Collecting medical image data and clinical feature data; wherein, the medical image data includes lung images and brain images, and the medical image data is obtained based on positron emission tomography; Screening out target clinical features from the clinical feature data; Preprocessing the lung images to obtain preliminary lung images; Preprocessing the brain images to obtain preliminary brain images; Performing deep and shallow feature extraction on the preliminary lung images to obtain lung deep features and lung shallow features, and performing deep feature extraction on the preliminary brain images to obtain brain deep features; Performing feature screening on the target clinical features, the lung deep features, the lung shallow features and the brain deep features to obtain target features; Performing disease classification according to the target features to obtain target disease categories.

2. The method according to claim 1, characterized in that, The lung images include a lung function image sequence and a lung ecological image sequence. The preprocessing the lung images to obtain preliminary lung images includes: Adjusting the resolutions of the lung function image sequence and the lung ecological image sequence to a target resolution to obtain adjusted lung images; Performing slice extraction on the adjusted lung images to obtain lung slice images; wherein, the lung slice images include lung function slice images and lung ecological slice images; Performing standardization processing on the lung ecological slice images to obtain standardized lung ecological images; Fusing the lung function slice images and the standardized lung ecological images to obtain preliminary lung images.

3. The method according to claim 2, characterized in that, The fusing the lung function slice images and the standardized lung ecological images to obtain preliminary lung images includes: Performing standardized uptake value conversion on the pixels of the lung function slice images to obtain standardized lung function images; Fusing the standardized lung function images and the standardized lung ecological images to obtain the preliminary lung images.

4. The method according to claim 1, characterized in that, The brain images include brain function images and brain ecological images. The preprocessing the brain images to obtain preliminary brain images includes: Performing standardization processing on the brain ecological images to obtain standardized brain ecological images; Performing standardized uptake value conversion on the pixels of the brain function images to obtain standardized brain function images; Fusing the standardized brain function images and the standardized brain ecological images to obtain the preliminary brain images.

5. The method according to any one of claims 1 to 4, characterized in that, The performing feature screening on the target clinical features, the lung deep features, the lung shallow features and the brain deep features to obtain target features includes: Taking the target clinical features, the lung deep features, the lung shallow features and the brain deep features as candidate features, and performing difference comparison between the candidate features and a preset reference label to obtain feature difference data; Screening out selected features from the candidate features according to the feature difference data; Screening out the target features from the selected features based on a preset least absolute shrinkage and selection operator model.

6. The method according to any one of claims 1 to 4, characterized in that, Obtaining at least two pre-trained target prediction models; wherein, each target prediction model includes a target classifier, and the target classifier uses a support vector machine; Performing feature correlation analysis on the target features based on the target classifier to obtain the target disease category.

7. The method according to any one of claims 1 to 4, characterized in that, The screening of the target clinical features from the clinical feature data includes: Performing a significance evaluation on the original clinical features of the clinical feature data to obtain significance evaluation data; Screening the original clinical features of the clinical feature data based on the significance evaluation data to obtain the target clinical features.

8. A device for processing lung and brain images, characterized in that, The device includes: A data acquisition module for acquiring medical image data and clinical feature data; wherein, the medical image data includes lung images and brain images, and the medical image data is obtained based on positron emission tomography. A clinical feature screening module for screening the target clinical features from the clinical feature data. A lung image preprocessing module for preprocessing the lung images to obtain preliminary lung images. A brain image preprocessing module for preprocessing the brain images to obtain preliminary brain images. A feature extraction module for extracting deep and shallow features from the preliminary lung images to obtain lung deep features and lung shallow features, and extracting deep features from the preliminary brain images to obtain brain deep features. A target feature selection module for screening the target clinical features, the lung deep features, the lung shallow features, and the brain deep features to obtain target features. A disease classification module for classifying diseases based on the target features to obtain the target disease category.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, the method according to any one of claims 1 to 7 is implemented.

Images