Pulmonary nodule distinguishing processing method and system based on CT image

Abstract

The invention relates to a pulmonary nodule distinguishing processing method based on a CT image, and the method comprises the following steps: carrying out the CT scanning of a user, obtaining lung image data, carrying out the three-dimensional reconstruction of the lung image data, and obtaining a three-dimensional image; performing pulmonary nodule attention recognition based on the three-dimensional image through a preset recognition algorithm to obtain a pulmonary nodule image; segmenting the pulmonary nodule image to obtain an internal structure and pulmonary nodule characteristics of the pulmonary nodule; performing risk assessment on the pulmonary nodule based on the internal structure and the pulmonary nodule characteristics to obtain a risk assessment level; if the risk assessment level exceeds a predetermined risk level, classifying the pulmonary nodules corresponding to the risk level exceeding the predetermined risk level to obtain a classification result of the pulmonary nodules; the technical problem that artificial recognition is difficult due to the fact that pulmonary nodules are diverse in form and low in contrast ratio is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of resolution and processing, and in particular to a resolution and processing method and system for lung nodules based on CT images. Background Art

[0002] In the field of modern medical imaging diagnosis, early detection and accurate classification of lung nodules are crucial for early intervention and treatment of lung cancer. Traditional CT image analysis methods often rely on manual reading by radiologists, a process that is not only time-consuming but also susceptible to subjective factors. Especially when faced with a large amount of CT image data, the risk of missed diagnosis and misdiagnosis increases significantly. In particular, the identification and evaluation of small nodules (less than 10 mm in diameter) is particularly difficult due to their diverse morphology and low contrast. Summary of the invention

[0003] The main purpose of the present invention is to provide a method and system for distinguishing and processing lung nodules based on CT images, which solves the technical problem that lung nodules are difficult to identify manually due to their diverse morphologies, low contrast, and poor quality.

[0004] To achieve the above object, the present invention provides a method for distinguishing and processing lung nodules based on CT images, comprising the following steps: Performing a CT scan on the user to obtain lung image data, and performing three-dimensional reconstruction on the lung image data to obtain a three-dimensional image; Using a preset recognition algorithm, the lung nodules are recognized based on the three-dimensional image to obtain a lung nodule image; Segmenting the pulmonary nodule image to obtain the internal structure and features of the pulmonary nodule; Performing risk assessment on the pulmonary nodule based on the internal structure and characteristics of the pulmonary nodule to obtain a risk assessment level; If the risk assessment level exceeds a predetermined risk level, the lung nodules corresponding to the risk level exceeding the predetermined risk level are classified to obtain a classification result of the lung nodules.

[0005] Further, the lung image data is three-dimensionally reconstructed to obtain a three-dimensional image, including: Using a preset image registration algorithm to perform spatial voxel alignment operations on lung CT scan images from different angles to obtain aligned images; wherein the aligned images are a set of spatially consistent and continuous image sequences; transforming the aligned image to obtain a three-dimensional surface model; Performing light and shadow processing on the three-dimensional surface model to obtain a light and shadow three-dimensional model; Rendering the three-dimensional surface model in different colors according to the lung tissue structure to obtain a three-dimensional rendering image; The transparency of the three-dimensional rendering image is adjusted to obtain a three-dimensional image.

[0006] Further, the lung nodules are recognized based on the three-dimensional image to obtain a lung nodule image, including: Performing multi-scale feature extraction on the three-dimensional image through a preset recognition algorithm to obtain a feature map of the potential location of the lung nodules; The feature map is weighted by an attention mechanism model to obtain a weighted feature map; Focusing on the pulmonary nodule region based on the weighted feature map, obtaining a candidate region of the pulmonary nodule; Performing time-series dynamic recognition on the candidate area through a preset long-short memory algorithm to obtain the active area of ​​the lung nodule; Removing non-pulmonary nodule structures from the active area of ​​the pulmonary nodules to obtain an image after the non-pulmonary nodule structures are removed; wherein the non-pulmonary nodule structures include pulmonary blood vessels, bronchi, and lymph nodes; The removed image is subjected to boundary processing to obtain a lung nodule image.

[0007] Furthermore, the candidate regions are dynamically identified in time series by using a preset long short memory algorithm to obtain the active region of the pulmonary nodules, including: Performing standardization processing on the candidate region to obtain a standardized region; wherein there are multiple standardized regions, and each of the standardized regions is different; Extracting time series features from the image frames of the standardized area in the time series, and performing feature vector conversion on the time series features to obtain feature conversion vectors; wherein the time series features include morphological changes, density changes, and spatial position movement of lung nodules; A time series matrix is ​​constructed based on the feature conversion vector, and the time series matrix is ​​input into a preset long short-term memory algorithm for dynamic analysis to obtain a dynamic change pattern of lung nodules over time; wherein the rows of the time series matrix represent the features at different time points, the columns of the time series matrix represent the vector dimensions, and the time series matrix represents a multi-dimensional feature profile of the evolution of lung nodules over time, and the dynamic change pattern includes morphological evolution and nodule motion trajectory; Performing probability calculation on the dynamic change pattern to obtain the activity probability distribution of the lung nodules; Based on the activity probability distribution, the candidate area is dynamically identified in time series to obtain the active area of ​​the lung nodule with significant dynamic changes in time series.

[0008] Furthermore, the pulmonary nodule image is segmented to obtain the internal structure and features of the pulmonary nodule, including: Performing three-dimensional segmentation on the pulmonary nodule image by using a preset segmentation algorithm to obtain an internal image of the pulmonary nodule; Extracting morphological features from the internal image to obtain morphological features; wherein the morphological features include the diameter, density distribution and sharpness of the pulmonary nodules; Based on the morphological characteristics, the internal structure of the pulmonary nodule is obtained; wherein the internal structure includes cavitation characteristics, calcification patterns, and internal density heterogeneity; Extracting texture features from the internal image to obtain texture features; wherein the texture features include gray-level co-occurrence matrix features; Performing Wavelet transformation on the texture features to obtain frequency features inside the pulmonary nodules; wherein the frequency features are microstructural changes of the pulmonary nodules; Anisotropy recognition is performed on the texture features to obtain the directionality of the image texture; wherein the directionality of the image texture is the directionality of the tissue arrangement of the lung nodules; A lung nodule feature is obtained based on the frequency feature and the directionality.

[0009] Further, a risk assessment is performed on the pulmonary nodule based on the internal structure and the characteristics of the pulmonary nodule to obtain a risk assessment level, including: An internal structure matrix is ​​constructed based on the internal structure, and a lung nodule matrix is ​​constructed based on the lung nodule characteristics; wherein both the internal structure matrix and the lung nodule matrix are n-row and n-column matrices; Multiplying the internal structure matrix and the lung nodule matrix to obtain a first multiplication matrix; Performing spectral decomposition on the multiplication matrix to obtain an orthogonal matrix and a positive definite symmetric matrix; The positive diagonals of the orthogonal matrix and the positive definite symmetric matrix are extracted respectively, and the first positive diagonal and the second positive diagonal are obtained correspondingly; Multiply the first positive diagonal line and the second positive diagonal line to obtain a second multiplication matrix; Performing singular value decomposition on the second multiplication matrix to obtain a first orthogonal matrix, a diagonal matrix and a second orthogonal matrix; Calculate the eigenvalues ​​of the first orthogonal matrix, the diagonal matrix and the second orthogonal matrix respectively to obtain the first eigenvalue, the second eigenvalue and the third eigenvalue; Arrange and combine the first eigenvalue, the second eigenvalue, and the third eigenvalue to obtain a plurality of arrangement and combination matrices; Input each of the permutation and combination matrices into a preset matching algorithm for code recognition to obtain a code table with the highest similarity; wherein each of the code tables has a corresponding risk level; The risk of the pulmonary nodule is assessed based on the code with the highest similarity to obtain a risk assessment level.

[0010] Furthermore, the lung nodules corresponding to the predetermined risk level are classified to obtain the classification results of the lung nodules, including: Extracting the internal structure and features of the pulmonary nodules corresponding to the pulmonary nodules exceeding the predetermined risk level, and using the internal structure and features of the pulmonary nodules as the first node and the second node respectively; Calculate the distance between the first node and the second node, and use the distance as an edge; constructing a metagraph based on the first node, the second node and the edge; The meta-graph is input into a preset risk label algorithm for classification to obtain a classification result of the pulmonary nodules.

[0011] The present invention also provides a CT image-based lung nodule resolution and processing system, comprising: A reconstruction module, used to perform a CT scan on the user to obtain lung image data, and perform three-dimensional reconstruction on the lung image data to obtain a three-dimensional image; A recognition module, used to recognize lung nodules based on the three-dimensional image by using a preset recognition algorithm to obtain a lung nodule image; A segmentation module, used to segment the pulmonary nodule image to obtain the internal structure and features of the pulmonary nodule; An assessment module, used to perform risk assessment on the pulmonary nodule based on the internal structure and the characteristics of the pulmonary nodule to obtain a risk assessment level; The classification module is used to classify the lung nodules corresponding to the risk level exceeding the predetermined risk level if the risk assessment level exceeds the predetermined risk level, so as to obtain the classification result of the lung nodules.

[0012] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0013] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0014] The method for distinguishing and processing lung nodules based on CT images provided by the present invention comprises the following steps: performing a CT scan on a user to obtain lung image data, and performing three-dimensional reconstruction on the lung image data to obtain a three-dimensional image; performing attention recognition of lung nodules based on the three-dimensional image through a preset recognition algorithm to obtain a lung nodule image; segmenting the lung nodule image to obtain the internal structure and lung nodule characteristics of the lung nodules; performing risk assessment on the lung nodules based on the internal structure and lung nodule characteristics to obtain a risk assessment level; if the risk assessment level exceeds a predetermined risk level, classifying the lung nodules corresponding to the risk level exceeding the predetermined risk level to obtain a classification result of the lung nodules; the above steps solve the technical problem of difficulty in manual identification due to the diverse morphology, low contrast and difficulty in manual identification of lung nodules, avoid missed diagnosis or misdiagnosis due to nodules being too small or complex in morphology, and improve the accuracy of lung nodule detection and classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the steps of a method for distinguishing and processing lung nodules based on CT images in one embodiment of the present invention; Figure 2 It is a structural block diagram of a system for distinguishing and processing lung nodules based on CT images in one embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0016] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] like Figure 1 As shown, Figure 1 It is a schematic diagram of the steps of a method for distinguishing and processing lung nodules based on CT images in one embodiment of the present invention; In one embodiment of the present invention, a method for distinguishing and processing lung nodules based on CT images is provided, comprising the following steps: Step S1, performing a CT scan on the user to obtain lung image data, and performing three-dimensional reconstruction on the lung image data to obtain a three-dimensional image.

[0019] Specifically, in this step, the patient's lungs are first examined using medical imaging equipment such as a CT scanner. CT (Computed Tomography) scanning uses X-ray technology to scan the human lungs layer by layer from multiple angles, and then the computer system integrates and processes these two-dimensional tomographic image data. This process can capture the detailed anatomical structure of the lung tissue, including the presence of tiny nodules or other abnormal changes. The acquired lung image data is a series of digital cross-sectional images, which are rich in anatomical information, but in order to better understand and analyze the three-dimensional structure of the lungs, the next step is required - three-dimensional reconstruction. Three-dimensional reconstruction is an image post-processing technology that uses advanced image processing algorithms and computer graphics principles to convert and integrate a series of two-dimensional CT slice data into a continuous, interactive three-dimensional image model. In this process, technologies such as volume rendering or surface rendering are applied. The former can show the translucent effect of lung tissue and intuitively display the internal structure; the latter emphasizes the contours of the surface of nodules or organs to enhance morphological recognition. By adjusting different rendering parameters, such as transparency and color mapping, the structural features of the lungs, especially key lesion areas such as lung nodules, can be highlighted, providing an intuitive and accurate visual reference for subsequent diagnostic analysis. In short, the purpose of three-dimensional reconstruction of lung imaging data is to integrate and upgrade the originally scattered two-dimensional slices into three-dimensional views, thereby helping doctors or researchers to more intuitively and comprehensively evaluate the health of the lungs and lay the foundation for the identification, measurement and nature judgment of lung nodules.

[0020] Step S2, using a preset recognition algorithm, focusing on the lung nodules based on the three-dimensional image to obtain a lung nodule image.

[0021] Specifically, in this step, a preset recognition algorithm is used, which is designed for effectively identifying lung nodules from three-dimensional lung images. The concept of "attention recognition" draws on the attention mechanism in the field of artificial intelligence, which means that the algorithm can automatically focus on the part of the image that is most relevant to the lung nodules and ignore other non-critical information, thereby improving the accuracy and efficiency of recognition. Specifically, the algorithm first conducts an in-depth analysis of the three-dimensional image, using deep learning models such as convolutional neural networks (CNN) or other advanced image processing technologies. These technologies can learn and extract the characteristics of lung nodules, such as shape, density, and edge characteristics. Through the layer-by-layer transmission and feature transformation of multi-layer neural networks, the algorithm can gradually extract the image features that best represent lung nodules. The role of the "attention mechanism" here is to guide the model to give priority to those areas that are most likely to contain lung nodules in the huge image data, which is similar to the process of the human eye looking for specific objects in a complex scene. The algorithm will dynamically enhance the importance of certain areas through weighting, scoring or other strategies, while suppressing irrelevant background information, thereby achieving accurate positioning of lung nodules. After this series of calculations and analyses, the algorithm ultimately outputs an image with lung nodules marked, also known as a "lung nodule image." In this type of image, the lung nodules are clearly marked or highlighted, making it easier for doctors to observe and analyze them. In this way, the originally complex three-dimensional lung image is simplified to an image that only highlights suspected nodules, greatly reducing the workload of doctors and improving the speed and accuracy of diagnosis.

[0022] Step S3, segmenting the lung nodule image to obtain the internal structure and features of the lung nodule.

[0023] Specifically, in this link, the goal is to accurately separate the internal structural details and unique features of the lung nodules themselves from the previously identified and marked lung nodule images. To achieve this goal, image segmentation technology is used, which is an image processing method that aims to divide the image into multiple regions, each of which corresponds to different objects or different parts of objects in the image. In specific operations, advanced image segmentation algorithms are first applied to the obtained lung nodule images. These algorithms include but are not limited to deep learning-based models, such as fully convolutional networks (FCNs), U-Nets, etc., which have learned how to distinguish the boundaries between nodules and the surrounding normal lung tissue through training. The segmentation process involves intensive classification of image pixels to ensure that the edges of the nodules are accurately defined, and the internal structures such as the cavities and calcifications of the nodules are also clearly distinguished. Through this fine segmentation, not only can the accurate outline of the lung nodules be obtained, but also its internal structural characteristics can be further analyzed, such as the shape of the nodules (regular or not), size, density distribution (uniform or uneven), whether there are cavities, and the proportion of cavities. These features are crucial for evaluating the nature of nodules, helping to distinguish benign and malignant nodules, and providing necessary information for subsequent risk assessment. In summary, the segmentation step of lung nodule images is automatically completed by a high-precision algorithm, which aims to accurately extract the detailed morphology and internal structure of lung nodules from complex lung images, providing clinicians with a more intuitive and quantitative diagnostic basis.

[0024] Step S4, performing risk assessment on the pulmonary nodule based on the internal structure and the characteristics of the pulmonary nodule to obtain a risk assessment level.

[0025] Specifically, in this step, a comprehensive risk assessment process is performed using the internal structure information obtained from the segmentation of lung nodules and various features of lung nodules. The purpose is to assign a risk level to each identified lung nodule to indicate its potential malignancy or growth dynamics. The core of risk assessment is to establish an assessment model, which is usually based on machine learning or statistical methods. The training data of the model comes from a large number of clinical cases in the past, which contain lung nodules with known results (such as benign or malignant confirmed by pathology). The model learns patterns extracted from internal structures (such as cavitation and calcification of nodules) and features (such as size, shape, edge smoothness, density uniformity, etc.) to establish an association with the final diagnosis result (risk level) of the nodule. In specific operations, the features extracted from the segmentation step are first quantized to form one or a group of digital indicators representing the characteristics of each nodule. Subsequently, these features are input into the pre-trained risk assessment model. The model includes logistic regression, support vector machine, random forest, and even deep learning models, which can output a probability value or direct risk level classification (such as low risk, medium risk, high risk) based on the combination of features. This assessment does not rely on a single feature, but is a comprehensive judgment made after considering all the information to ensure the comprehensiveness and accuracy of the assessment. The resulting risk assessment level is crucial to guiding subsequent medical decisions, such as whether further biopsies, regular follow-up or immediate treatment measures are needed.

[0026] Step S5: if the risk assessment level exceeds the predetermined risk level, classify the lung nodules corresponding to the risk level exceeding the predetermined risk level to obtain a classification result of the lung nodules.

[0027] Specifically, after completing the risk assessment of lung nodules, this step focuses on further classification of those nodules whose risk level exceeds a preset threshold. This preset risk level is usually set by clinical guidelines or professional consensus to distinguish high-risk lung nodules that require urgent attention or action from low-risk nodules that can continue to be observed.

[0028] In specific operations, for those lung nodules whose risk assessment results exceed the predetermined standards, the system or doctor will conduct further analysis based on a set of detailed classification criteria or algorithms. This involves more detailed analysis of imaging features, dynamic tracking of nodule growth, or combining multimodal information such as serum marker detection and genetic testing. The goal of classification is to classify these high-risk nodules into specific categories as accurately as possible, such as benign nodules (such as inflammatory nodules, hamartomas), malignant-prone nodules (such as carcinoma in situ, microinvasive adenocarcinoma), or obvious malignant nodules (such as invasive adenocarcinoma, squamous cell carcinoma, etc.). Through such a classification process, clinicians can obtain more specific and targeted diagnostic information, providing direct guidance for subsequent treatment planning. For example, benign nodules require regular follow-up observation, while malignant or suspected malignant nodules require tissue biopsy confirmation and prompt arrangement of corresponding surgery or drug treatment. This strategy not only optimizes the allocation of medical resources, but also ensures that patients can receive timely and appropriate intervention measures.

[0029] In a specific embodiment, performing three-dimensional reconstruction on the lung image data to obtain a three-dimensional image includes: Using a preset image registration algorithm to perform spatial voxel alignment operations on lung CT scan images from different angles to obtain aligned images; wherein the aligned images are a set of spatially consistent and continuous image sequences; transforming the aligned image to obtain a three-dimensional surface model; Performing light and shadow processing on the three-dimensional surface model to obtain a light and shadow three-dimensional model; Rendering the three-dimensional surface model in different colors according to the lung tissue structure to obtain a three-dimensional rendering image; The transparency of the three-dimensional rendering image is adjusted to obtain a three-dimensional image.

[0030] Specifically, in this step, the detailed process of three-dimensional reconstruction of lung image data is divided into several key sub-steps, ensuring high-quality conversion from raw CT scan data to final three-dimensional images, so that doctors can perform more intuitive and accurate diagnostic analysis. The preset image registration algorithm is used to perform spatial voxel alignment operations on lung CT scan images from different angles: this process first solves the consistency problem between different scanning layers or sequences. Through the image registration algorithm, it can be ensured that even images taken in different positions or breathing stages can be perfectly aligned in space to form a seamless whole. The "aligned image" thus becomes a set of continuous image sequences that are precisely matched in space, laying a solid foundation for subsequent three-dimensional reconstruction. The aligned image is transformed to obtain a three-dimensional surface model: this transformation process uses computer graphics technology to convert two-dimensional tomographic image data into a geometric representation in three-dimensional space, namely a three-dimensional surface model. This step constructs the three-dimensional structure of the lungs by identifying and connecting common features in adjacent images, so that the three-dimensional morphology of structures such as nodules, airways, and blood vessels can be revealed. The three-dimensional surface model is subjected to light and shadow processing to obtain a light and shadow three-dimensional model: the light and shadow processing is to enhance the realism and detail readability of the model. By simulating the illumination of the light source and calculating the light and shade changes of each surface point, the surface of the model presents natural shadows and highlights, which helps doctors better understand the depth and level of the lung structure. The three-dimensional surface model is rendered in different colors according to the lung tissue structure: in order to further improve the interpretability of the image, different types of lung tissue (such as lung parenchyma, airway, blood vessels, etc.) will be assigned different colors. This color coding not only beautifies the image, but more importantly, it helps doctors quickly distinguish and identify different anatomical structures, improving the efficiency and accuracy of diagnosis. The transparency of the three-dimensional rendered image is adjusted to obtain a three-dimensional image: finally, by adjusting the transparency of the image, the internal structure can be observed through multiple layers of tissue while keeping the structure clearly visible. This is particularly important for observing the relationship between lung nodules and surrounding tissues and evaluating their location and possible impact range. Transparency adjustment enables doctors to analyze in depth details without losing the overall structural perception, completing the conversion process from raw data to high-availability three-dimensional images. In summary, this series of steps transforms complex lung CT data into easy-to-interpret, highly realistic three-dimensional images through high-tech image processing technology, greatly improving the accuracy and efficiency of lung nodule detection and analysis.

[0031] In a specific embodiment, the lung nodules are recognized based on the three-dimensional image by using a preset recognition algorithm to obtain a lung nodule image, including: Performing multi-scale feature extraction on the three-dimensional image through a preset recognition algorithm to obtain a feature map of the potential location of the lung nodules; The feature map is weighted by an attention mechanism model to obtain a weighted feature map; Focusing on the pulmonary nodule region based on the weighted feature map, obtaining a candidate region of the pulmonary nodule; Performing time-series dynamic recognition on the candidate area through a preset long-short memory algorithm to obtain the active area of ​​the lung nodule; Removing non-pulmonary nodule structures from the active area of ​​the pulmonary nodules to obtain an image after the non-pulmonary nodule structures are removed; wherein the non-pulmonary nodule structures include pulmonary blood vessels, bronchi, and lymph nodes; The removed image is subjected to boundary processing to obtain a lung nodule image.

[0032] Specifically, this step describes in detail an advanced lung nodule recognition and processing process, which aims to accurately identify and extract lung nodule images from three-dimensional CT images. The specific steps are as follows: Multi-scale feature extraction: A preset recognition algorithm is used to perform in-depth analysis of the three-dimensional image. This process involves extracting features at different scales to capture the various size and shape characteristics of the nodules. Through this multi-scale analysis, the algorithm can generate a "feature map" that highlights the characteristics of the potential location of the lung nodules and provides a focus for subsequent steps. Attention mechanism model weighting: The attention mechanism model is used to intelligently weight the feature map. This technology imitates the attention mechanism of the human visual system to important information, automatically strengthens the key area signals related to the lung nodules, and suppresses irrelevant background information. The resulting "weighted feature map" more clearly highlights the possible location of the lung nodules. Candidate region attention: Based on the weighted feature map, the algorithm can more accurately lock the candidate regions of the lung nodules. These regions represent the image segments that are most likely to contain lung nodules, laying the foundation for subsequent accurate recognition. Temporal dynamic recognition: For the candidate regions, the preset long short-term memory (LSTM) algorithm is used for temporal dynamic analysis. LSTM is good at processing time series data. By analyzing the image changes of nodules at different time points, it can identify the active areas of lung nodules with dynamic change characteristics, which helps to distinguish active and potentially malignant nodules from static or benign nodules. Removal of non-pulmonary nodule structures: The obtained images of the active areas of lung nodules are further processed to remove non-nodule structures such as pulmonary blood vessels, bronchi, and lymph nodes. This step ensures that only structures directly related to lung nodules are retained in the final image, reducing misjudgment and interference. Boundary processing: Finally, the image after removing the non-nodule structure is subjected to boundary optimization processing, such as smoothing edges and removing noise, to obtain a clear and accurate image of lung nodules. This image can be directly used for further diagnosis by doctors, improving the accuracy and diagnostic efficiency of nodule identification. Through this series of refined processing steps, not only the sensitivity and specificity of lung nodule identification are improved, but also the image quality is optimized, allowing doctors to more intuitively and accurately evaluate the nature of lung nodules, providing a powerful tool for the early diagnosis and treatment of lung cancer.

[0033] In a specific embodiment, the candidate regions are dynamically identified in time series by using a preset long short-term memory algorithm to obtain the active region of the pulmonary nodules, including: Performing standardization processing on the candidate region to obtain a standardized region; wherein there are multiple standardized regions, and each of the standardized regions is different; Extracting time series features from the image frames of the standardized area in the time series, and performing feature vector conversion on the time series features to obtain feature conversion vectors; wherein the time series features include morphological changes, density changes, and spatial position movement of lung nodules; A time series matrix is ​​constructed based on the feature conversion vector, and the time series matrix is ​​input into a preset long short-term memory algorithm for dynamic analysis to obtain a dynamic change pattern of lung nodules over time; wherein the rows of the time series matrix represent the features at different time points, the columns of the time series matrix represent the vector dimensions, and the time series matrix represents a multi-dimensional feature profile of the evolution of lung nodules over time, and the dynamic change pattern includes morphological evolution and nodule motion trajectory; Performing probability calculation on the dynamic change pattern to obtain the activity probability distribution of the lung nodules; Based on the activity probability distribution, the candidate area is dynamically identified in time series to obtain the active area of ​​the lung nodule with significant dynamic changes in time series.

[0034] Specifically, this step is to conduct an in-depth time series analysis of the candidate regions to identify the active areas of lung nodules with actual pathological significance. The specific operations are as follows: Standardization: First, each candidate region is standardized to eliminate the inconsistency caused by image acquisition conditions (such as differences in scanning equipment and changes in patient position) to ensure that the comparison is carried out under the same conditions. Each standardized region is adjusted to ensure that its parameters such as size, brightness, and contrast are consistent, although they may represent different nodule locations or morphologies. Time series feature extraction and transformation: Then, for a series of image frames of each standardized region over time, key time series features are extracted, including but not limited to changes in the morphology (size, shape), density (transparency), and slight movements of the spatial position of lung nodules. These features are converted into feature vectors, that is, a series of numerical values ​​that can quantify the state of nodules at different time points for further analysis. Constructing a time series matrix and LSTM analysis: These feature vectors are arranged in chronological order to form a time series matrix. Each row of the matrix corresponds to a feature vector at a time point, and each column represents a specific feature dimension. This matrix comprehensively reflects the multi-dimensional feature evolution of lung nodules over time. This matrix is ​​input into the preset long short-term memory (LSTM) algorithm. LSTM can capture long-term dependencies and analyze the dynamic change pattern of lung nodules, including how the morphology evolves over time and the specific movement trajectory of the nodules in the lungs. Probability calculation and dynamic identification: Based on the results of LSTM analysis, the probability distribution of lung nodule activity in each candidate area is further calculated, that is, the probability that the nodule shows significant dynamic changes. High-probability areas indicate that the nodules have active signs such as growth, morphological changes, or position movement, which usually require special clinical attention. Based on these probability distributions, those lung nodule activity areas with significant temporal dynamic changes are finally screened out as the target of focus and further diagnosis. In summary, this process combines image processing, feature extraction, time series analysis, and machine learning techniques to accurately identify clinically significant dynamic behaviors of lung nodules from complex lung imaging data and provide doctors with reliable auxiliary diagnostic information.

[0035] In a specific embodiment, the pulmonary nodule image is segmented to obtain the internal structure and features of the pulmonary nodule, including: Performing three-dimensional segmentation on the pulmonary nodule image by using a preset segmentation algorithm to obtain an internal image of the pulmonary nodule; Extracting morphological features from the internal image to obtain morphological features; wherein the morphological features include the diameter, density distribution and sharpness of the pulmonary nodules; Based on the morphological characteristics, the internal structure of the pulmonary nodule is obtained; wherein the internal structure includes cavitation characteristics, calcification patterns, and internal density heterogeneity; Extracting texture features from the internal image to obtain texture features; wherein the texture features include gray-level co-occurrence matrix features; Performing Wavelet transformation on the texture features to obtain frequency features inside the pulmonary nodules; wherein the frequency features are microstructural changes of the pulmonary nodules; Anisotropy recognition is performed on the texture features to obtain the directionality of the image texture; wherein the directionality of the image texture is the directionality of the tissue arrangement of the lung nodules; A lung nodule feature is obtained based on the frequency feature and the directionality.

[0036] Specifically, in this series of detailed steps, the processing and analysis of lung nodule images aims to deeply explore the internal structure and key features of lung nodules to support subsequent diagnosis and evaluation. The following is a detailed explanation of each step: 3D segmentation: First, the lung nodule image is processed in 3D using a preset segmentation algorithm. This process can accurately separate the 3D morphology of the lung nodule from the surrounding lung tissue. The internal image obtained provides a clear focus area for subsequent feature extraction.

[0037] Morphological feature extraction: Based on the internal image obtained, the morphological features of the nodules are analyzed and extracted, such as diameter, density distribution (reflecting the density of the nodules) and sharpness (edge ​​clarity). These are directly related to the size, texture and boundary characteristics of the nodules, and are key indicators for evaluating the basic morphological properties of the nodules. Internal structure recognition: Based on the morphological features, the internal structure of the pulmonary nodules is further identified, including cavitation characteristics (the presence of internal cavities), calcification patterns (the morphology of calcium salt deposition) and internal density heterogeneity. These structural features are crucial for determining the nature of the nodules (benign or malignant). Texture feature extraction: The texture features of the internal images are extracted through gray-level co-occurrence matrix (GLCM) analysis. This statistical method can describe the relationship between image pixels and reflect the roughness and complexity of the nodule surface. It is an important tool for evaluating the surface texture and structural variation of the nodule. Wavelet transform: The extracted texture features are subjected to wavelet transform, which is a powerful signal analysis tool that can reveal the frequency characteristics of the nodules, that is, the changes in microstructure, and help identify subtle structural differences inside the nodules, which is of great value in understanding the biological behavior of nodules. Anisotropy recognition: By analyzing the directionality of texture features, the arrangement directionality of the image texture inside the lung nodule is identified. This information can reflect the arrangement pattern of the tissue inside the nodule, which is very useful for distinguishing certain specific types of nodules (such as fibrosis, arrangement of tumor tissue). Comprehensive analysis of lung nodule features: Finally, based on the above-mentioned frequency characteristics (microstructure changes) and directional information, a comprehensive feature description of the lung nodules is obtained. These features combined not only provide macroscopic morphological information of the nodules, but also include an in-depth understanding of the microstructure and tissue characteristics, providing clinicians with a wealth of diagnostic basis to help determine the nature and potential disease status of lung nodules.

[0038] In a specific embodiment, risk assessment of the pulmonary nodule is performed based on the internal structure and the characteristics of the pulmonary nodule to obtain a risk assessment level, including: An internal structure matrix is ​​constructed based on the internal structure, and a lung nodule matrix is ​​constructed based on the lung nodule characteristics; wherein both the internal structure matrix and the lung nodule matrix are n-row and n-column matrices; Multiplying the internal structure matrix and the lung nodule matrix to obtain a first multiplication matrix; Performing spectral decomposition on the multiplication matrix to obtain an orthogonal matrix and a positive definite symmetric matrix; The positive diagonals of the orthogonal matrix and the positive definite symmetric matrix are extracted respectively, and the first positive diagonal and the second positive diagonal are obtained correspondingly; Multiply the first positive diagonal line and the second positive diagonal line to obtain a second multiplication matrix; Performing singular value decomposition on the second multiplication matrix to obtain a first orthogonal matrix, a diagonal matrix and a second orthogonal matrix; Calculate the eigenvalues ​​of the first orthogonal matrix, the diagonal matrix and the second orthogonal matrix respectively to obtain the first eigenvalue, the second eigenvalue and the third eigenvalue; Arrange and combine the first eigenvalue, the second eigenvalue, and the third eigenvalue to obtain a plurality of arrangement and combination matrices; Input each of the permutation and combination matrices into a preset matching algorithm for code recognition to obtain a code table with the highest similarity; wherein each of the code tables has a corresponding risk level; The risk of the pulmonary nodule is assessed based on the code with the highest similarity to obtain a risk assessment level.

[0039] Specifically, this step describes a complex algorithmic process for risk assessment based on the internal structure and characteristics of lung nodules. The specific operations are as follows: Constructing matrices: First, based on the extracted internal structural information of lung nodules (such as cavitation, calcification, density heterogeneity) and lung nodule features (such as morphology, density, texture, etc.), two n-row and n-column matrices are constructed, namely the internal structure matrix and the lung nodule feature matrix. These two matrices are intended to quantify the structure and properties of the nodules and provide mathematical representation for subsequent analysis. Matrix multiplication: Multiply the internal structure matrix with the lung nodule feature matrix to generate a new matrix, called the first multiplication matrix. This operation is intended to integrate structural and feature information and create a matrix that comprehensively characterizes the characteristics of the nodules. Spectral decomposition: Spectral decomposition is performed on the first multiplication matrix to obtain an orthogonal matrix and a positive definite symmetric matrix. Spectral decomposition is an important tool in linear algebra that can reveal the intrinsic structure of a matrix. Here, it is used to separate the main features of the matrix. Extracting positive diagonals: Extracting the positive diagonal elements from the orthogonal matrix and the positive definite symmetric matrix to form the first positive diagonal and the second positive diagonal. The positive diagonal elements usually contain important information of the matrix, such as eigenvalues. Second multiplication matrix: Multiply the first positive diagonal and the second positive diagonal to obtain a new matrix, namely the second multiplication matrix. This step is intended to further condense and transform the previously obtained information. Singular value decomposition: Perform singular value decomposition on the second multiplication matrix to obtain three matrices: the first orthogonal matrix, the diagonal matrix, and the second orthogonal matrix. Singular value decomposition is another matrix decomposition method that can reveal the rank, main direction, and scale of the matrix. Calculate eigenvalues: Calculate the eigenvalues ​​of the three matrices obtained from the singular value decomposition, namely the first eigenvalue, the second eigenvalue, and the third eigenvalue. The eigenvalue is the scalar corresponding to the eigenvector of the matrix, which is often used to describe certain essential properties of the matrix. Permutation and combination and code recognition: The obtained eigenvalues ​​are permuted and combined to generate multiple permutation and combination matrices, each of which represents a specific arrangement of eigenvalues. These permutation matrices are input into the preset matching algorithm, and the most similar code to the known risk level code table is found through code recognition technology. Each code table corresponds to a specific risk level. Risk assessment: According to the code table with the highest similarity, the risk assessment of pulmonary nodules is performed, and the risk assessment level is finally obtained. This level reflects the potential malignancy or clinical concern of lung nodules, providing a basis for subsequent medical decisions. In summary, this solution deeply mines the internal structure and characteristic information of lung nodules through a series of complex mathematical operations and matrix operations, and achieves accurate assessment of lung nodule risk in a highly structured and mathematical way.

[0040] In a specific embodiment, the lung nodules corresponding to the risk level exceeding the predetermined level are classified to obtain the classification result of the lung nodules, including: Extracting the internal structure and features of the pulmonary nodules corresponding to the pulmonary nodules exceeding the predetermined risk level, and using the internal structure and features of the pulmonary nodules as the first node and the second node respectively; Calculate the distance between the first node and the second node, and use the distance as an edge; constructing a metagraph based on the first node, the second node and the edge; The meta-graph is input into a preset risk label algorithm for classification to obtain a classification result of the pulmonary nodules.

[0041] Specifically, this step is the process of classifying lung nodules that exceed the predetermined risk level. The specific operations are as follows: Target nodule selection: First, select those nodules whose risk level exceeds the preset threshold from all evaluated lung nodules. These nodules are considered to have higher clinical attention value and need further detailed classification. Feature extraction: For each selected high-risk lung nodule, extract its internal structure information (first node) and lung nodule features (second node) respectively. The internal structure involves the details of the nodule's composition, such as cavitation, fibrosis, etc., while the lung nodule features include size, shape, edge clarity, etc. Construct distance as edge: Next, calculate the distance between the first node (internal structure) and the second node (lung nodule features). This "distance" is an abstract concept, which can be understood as a quantitative indicator to measure the difference between the two feature sets, which is used to reflect the closeness or difference between the internal structure of the nodule and its overall features. Construct a metagraph: Use the above two nodes (internal structure and lung nodule features) and the "distance" between them as edges to construct a metagraph. In graph theory, a metagraph is a complex data structure that can simultaneously represent different types of nodes and their relationships. The edges (distances) here connect the two key attributes of the nodules to form an integrated network representation. Risk label algorithm classification: The constructed metagraph is input into a preset algorithm model. This algorithm is specially designed to process such metagraph data. By analyzing the metagraph structure, it can identify different patterns or feature combinations and then classify the lung nodules. The classification results reflect the pathological type, growth trend or other clinically relevant categories of the nodules. Obtaining classification results: After being processed by the above algorithm, each lung nodule that exceeds the predetermined risk level will be assigned a specific classification label. These labels help doctors understand the characteristics of each nodule and provide guidance for formulating treatment plans, monitoring disease progression or further diagnostic examinations. In short, this solution achieves a fine classification of these nodules by constructing and analyzing the metagraph of high-risk lung nodules, thereby deepening the understanding of nodule characteristics and providing more detailed information support for clinical decision-making.

[0042] The above describes the method for distinguishing and processing lung nodules based on CT images in an embodiment of the present invention. The following describes the system for distinguishing and processing lung nodules based on CT images in an embodiment of the present invention. Figure 2 In one embodiment of the present invention, a system for distinguishing and processing lung nodules based on CT images includes: A reconstruction module 21 is used to perform a CT scan on the user to obtain lung image data, and to perform three-dimensional reconstruction on the lung image data to obtain a three-dimensional image; The recognition module 22 is used to recognize the lung nodules based on the three-dimensional image by using a preset recognition algorithm to obtain a lung nodule image; A segmentation module 23 is used to segment the pulmonary nodule image to obtain the internal structure and features of the pulmonary nodule; An assessment module 24, configured to perform risk assessment on the pulmonary nodule based on the internal structure and the characteristics of the pulmonary nodule to obtain a risk assessment level; The classification module 25 is used to classify the lung nodules corresponding to the risk level exceeding the predetermined risk level if the risk assessment level exceeds the predetermined risk level, so as to obtain the classification result of the lung nodules.

[0043] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the above method embodiment, which will not be repeated here.

[0044] Reference Figure 3 The present invention also provides a computer device in an embodiment, wherein the internal structure of the computer device can be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0045] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0046] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0047] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0048] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0049] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for distinguishing and processing lung nodules based on CT images, characterized in that: The following steps are involved: Performing a CT scan on the user to obtain lung image data, and performing three-dimensional reconstruction on the lung image data to obtain a three-dimensional image; Using a preset recognition algorithm, the lung nodules are recognized based on the three-dimensional image to obtain a lung nodule image; Segmenting the pulmonary nodule image to obtain the internal structure and features of the pulmonary nodule; Performing risk assessment on the pulmonary nodule based on the internal structure and characteristics of the pulmonary nodule to obtain a risk assessment level; If the risk assessment level exceeds a predetermined risk level, the lung nodules corresponding to the risk level exceeding the predetermined risk level are classified to obtain a classification result of the lung nodules.

2. The method for distinguishing and processing lung nodules based on CT images according to claim 1, characterized in that: Performing three-dimensional reconstruction on the lung image data to obtain a three-dimensional image includes: Using a preset image registration algorithm to perform spatial voxel alignment operations on lung CT scan images from different angles to obtain aligned images; wherein the aligned images are a set of spatially consistent and continuous image sequences; transforming the aligned image to obtain a three-dimensional surface model; Performing light and shadow processing on the three-dimensional surface model to obtain a light and shadow three-dimensional model; Rendering the three-dimensional surface model in different colors according to the lung tissue structure to obtain a three-dimensional rendering image; The transparency of the three-dimensional rendering image is adjusted to obtain a three-dimensional image.

3. The method for distinguishing and processing lung nodules based on CT images according to claim 1, characterized in that: By using a preset recognition algorithm, the lung nodules are recognized based on the three-dimensional image to obtain a lung nodule image, including: Performing multi-scale feature extraction on the three-dimensional image through a preset recognition algorithm to obtain a feature map of the potential location of the lung nodules; The feature map is weighted by an attention mechanism model to obtain a weighted feature map; Focusing on the pulmonary nodule region based on the weighted feature map, obtaining a candidate region of the pulmonary nodule; Performing time-series dynamic recognition on the candidate area through a preset long-short memory algorithm to obtain the active area of ​​the lung nodule; Removing non-pulmonary nodule structures from the active area of ​​the pulmonary nodules to obtain an image after the non-pulmonary nodule structures are removed; wherein the non-pulmonary nodule structures include pulmonary blood vessels, bronchi, and lymph nodes; The removed image is subjected to boundary processing to obtain a lung nodule image.

4. The method for distinguishing and processing lung nodules based on CT images according to claim 3, characterized in that: The candidate regions are dynamically identified in time series by using a preset long-short memory algorithm to obtain active regions of pulmonary nodules, including: Performing standardization processing on the candidate region to obtain a standardized region; wherein there are multiple standardized regions, and each of the standardized regions is different; Extracting time series features from the image frames of the standardized area in the time series, and performing feature vector conversion on the time series features to obtain feature conversion vectors; wherein the time series features include morphological changes, density changes, and spatial position movement of lung nodules; A time series matrix is ​​constructed based on the feature conversion vector, and the time series matrix is ​​input into a preset long short-term memory algorithm for dynamic analysis to obtain a dynamic change pattern of lung nodules over time; wherein the rows of the time series matrix represent the features at different time points, the columns of the time series matrix represent the vector dimensions, and the time series matrix represents a multi-dimensional feature profile of the evolution of lung nodules over time, and the dynamic change pattern includes morphological evolution and nodule motion trajectory; Performing probability calculation on the dynamic change pattern to obtain the activity probability distribution of the lung nodules; Based on the activity probability distribution, the candidate area is dynamically identified in time series to obtain the active area of ​​the lung nodule with significant dynamic changes in time series.

5. The method for distinguishing and processing lung nodules based on CT images according to claim 1, characterized in that: The pulmonary nodule image is segmented to obtain the internal structure and features of the pulmonary nodule, including: Performing three-dimensional segmentation on the pulmonary nodule image by using a preset segmentation algorithm to obtain an internal image of the pulmonary nodule; Extracting morphological features from the internal image to obtain morphological features; wherein the morphological features include the diameter, density distribution and sharpness of the pulmonary nodules; Based on the morphological characteristics, the internal structure of the pulmonary nodule is obtained; wherein the internal structure includes cavitation characteristics, calcification patterns, and internal density heterogeneity; Extracting texture features from the internal image to obtain texture features; wherein the texture features include gray-level co-occurrence matrix features; Performing Wavelet transformation on the texture features to obtain frequency features inside the pulmonary nodules; wherein the frequency features are microstructural changes of the pulmonary nodules; Anisotropy recognition is performed on the texture features to obtain the directionality of the image texture; wherein the directionality of the image texture is the directionality of the tissue arrangement of the lung nodules; A lung nodule feature is obtained based on the frequency feature and the directionality.

6. The method for distinguishing and processing lung nodules based on CT images according to claim 1, characterized in that: Performing a risk assessment on the pulmonary nodule based on the internal structure and the characteristics of the pulmonary nodule to obtain a risk assessment level includes: An internal structure matrix is ​​constructed based on the internal structure, and a lung nodule matrix is ​​constructed based on the lung nodule characteristics; wherein both the internal structure matrix and the lung nodule matrix are n-row and n-column matrices; Multiplying the internal structure matrix and the lung nodule matrix to obtain a first multiplication matrix; Performing spectral decomposition on the multiplication matrix to obtain an orthogonal matrix and a positive definite symmetric matrix; The positive diagonals of the orthogonal matrix and the positive definite symmetric matrix are extracted respectively, and the first positive diagonal and the second positive diagonal are obtained correspondingly; Multiply the first positive diagonal line and the second positive diagonal line to obtain a second multiplication matrix; Performing singular value decomposition on the second multiplication matrix to obtain a first orthogonal matrix, a diagonal matrix and a second orthogonal matrix; Calculate the eigenvalues ​​of the first orthogonal matrix, the diagonal matrix and the second orthogonal matrix respectively to obtain the first eigenvalue, the second eigenvalue and the third eigenvalue; Arrange and combine the first eigenvalue, the second eigenvalue, and the third eigenvalue to obtain a plurality of arrangement and combination matrices; Input each of the permutation and combination matrices into a preset matching algorithm for code recognition to obtain a code table with the highest similarity; wherein each of the code tables has a corresponding risk level; The risk of the pulmonary nodule is assessed based on the code with the highest similarity to obtain a risk assessment level.

7. The method for distinguishing and processing lung nodules based on CT images according to claim 1, characterized in that: The pulmonary nodules that exceed the predetermined risk level are classified to obtain the classification results of the pulmonary nodules, including: Extracting the internal structure and features of the pulmonary nodules corresponding to the pulmonary nodules exceeding the predetermined risk level, and using the internal structure and features of the pulmonary nodules as the first node and the second node respectively; Calculate the distance between the first node and the second node, and use the distance as an edge; constructing a metagraph based on the first node, the second node and the edge; The meta-graph is input into a preset risk label algorithm for classification to obtain a classification result of the pulmonary nodules.

8. A CT image-based lung nodule identification and processing system, characterized in that: include: A reconstruction module, used to perform a CT scan on the user to obtain lung image data, and perform three-dimensional reconstruction on the lung image data to obtain a three-dimensional image; A recognition module, used to recognize lung nodules based on the three-dimensional image by using a preset recognition algorithm to obtain a lung nodule image; A segmentation module, used to segment the pulmonary nodule image to obtain the internal structure and features of the pulmonary nodule; An assessment module, used to perform risk assessment on the pulmonary nodule based on the internal structure and the characteristics of the pulmonary nodule to obtain a risk assessment level; The classification module is used to classify the lung nodules corresponding to the risk level exceeding the predetermined risk level if the risk assessment level exceeds the predetermined risk level, so as to obtain the classification result of the lung nodules.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

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

Images