A thoracic surgery image analysis method based on big data

Abstract

The present invention discloses a thoracic surgery image analysis method based on big data, which relates to the technical field of thoracic surgery image analysis, including collecting thoracic surgery images of patients and preprocessing them using a standardized protocol, automatically identifying and extracting key features from the preprocessed thoracic surgery images using a convolutional neural network (CNN), performing dimensionality reduction processing on the extracted key features using principal component analysis (PCA) technology, retaining representative information, integrating data from different imaging modes that have undergone dimensionality reduction processing, performing alignment using a mutual information maximization strategy, using a support vector machine (SVM) as a classification model, and classifying disease types according to the integrated data from different imaging modes, thereby improving the quality and efficiency of thoracic surgery image analysis through multimodal image fusion, feature selection and dimensionality reduction, and model training efficiency, promoting technological innovation in the field of medical image analysis, and bringing substantial improvements and enhancements to clinical practice.

Description

Technical Field

[0001] The present invention relates to the technical field of thoracic surgery image analysis, and in particular to a thoracic surgery image analysis method based on big data. Background Art

[0002] With the rapid development of medical imaging technology, especially the advancement of computed tomography and positron emission tomography imaging modes, the diagnosis and treatment of thoracic surgical diseases have been improved. In the past, image analysis mainly relied on the experience and judgment of radiologists. To overcome these limitations, automatic image analysis methods based on machine learning algorithms have gradually become a research hotspot, especially the application of support vector machines and convolutional neural network technologies, which enable the extraction of key features from large amounts of image data and classification.

[0003] Current technical solutions are still insufficient in multimodal image fusion, especially the lack of effective strategies for data integration of different imaging modes, resulting in low information utilization and failure to fully utilize their respective advantages, thus affecting the ultimate diagnostic accuracy. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a thoracic surgery image analysis method based on big data to solve the problem of lack of effective strategies for data integration of different imaging modes.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a thoracic surgery image analysis method based on big data, which comprises:

[0008] Thoracic surgical images of patients were acquired and preprocessed using standardized protocols;

[0009] Convolutional neural network (CNN) is used to automatically identify and extract key features from pre-processed thoracic surgical images.

[0010] Use principal component analysis (PCA) technology to reduce the dimensionality of the extracted key features and retain representative information;

[0011] Integrate data from different imaging modalities after dimensionality reduction and perform registration using a mutual information maximization strategy;

[0012] Support vector machine (SVM) was used as the classification model to classify the disease type based on the integrated data of different imaging modalities.

[0013] A real-time optimization feedback mechanism based on classification results adjusts model parameters according to the classification results after each classification is completed, and the model is retrained using the updated parameters;

[0014] By combining the retuned model with thoracic surgical images, an adaptive learning rate adjustment factor is applied to improve the quality and efficiency of thoracic surgical image analysis.

[0015] As a preferred embodiment of the big data-based thoracic surgery image analysis method of the present invention, wherein:

[0016] Thoracic surgical images of patients were acquired and preprocessed using a standardized protocol, specifically:

[0017] Thoracic surgical images of patients were acquired using CT scanners, with standardized imaging protocols and uniform scanning parameters.

[0018] The scanning parameters include layer thickness and scanning range;

[0019] Preprocess the thoracic surgery images;

[0020] The preprocessing includes removing noise and enhancing contrast.

[0021] As a preferred embodiment of the big data-based thoracic surgery image analysis method of the present invention, wherein:

[0022] The convolutional neural network (CNN) model is used to identify and extract key features from preprocessed thoracic surgical images, specifically:

[0023] The ResNet residual network was used as the architecture of the convolutional neural network (CNN) model, and the pre-processed thoracic surgery images were input into the CNN model for training.

[0024] Add multiple convolutional layers and pooling layers to the CNN model, use convolutional layers to extract key features, and use pooling layers to reduce the spatial dimensions of feature maps;

[0025] The key features of the thoracic surgery images include shape features, texture features, vascular structure, and tumor boundaries.

[0026] As a preferred embodiment of the big data-based thoracic surgery image analysis method of the present invention, wherein:

[0027] Using principal component analysis (PCA) technology, we perform dimensionality reduction on the extracted key features and retain representative information, specifically:

[0028] Standardize the key features extracted from thoracic surgery images;

[0029] The standardization process is to subtract the mean value of each key feature from its value and then divide it by its standard deviation, so that the mean of all features is 0 and the standard deviation is 1;

[0030] Calculate the covariance matrix based on the standardized key feature data, and extract eigenvalues ​​and eigenvectors from the covariance matrix;

[0031] The eigenvector corresponding to the largest eigenvalue is selected as the principal component, and the original thoracic surgical image is projected onto the principal component to obtain the thoracic surgical image after dimension reduction.

[0032] As a preferred embodiment of the big data-based thoracic surgery image analysis method of the present invention, wherein:

[0033] We use a mutual information maximization strategy to integrate thoracic surgical images from different imaging modalities that have undergone dimensionality reduction, and employ generative adversarial networks (GANs) to perform super-resolution reconstruction of thoracic surgical images. Specifically, we:

[0034] Integrate thoracic surgical images from different imaging modalities after dimensionality reduction;

[0035] The different imaging modes include CT scanning mode and PET scanning mode;

[0036] Images from different imaging modalities are spatially aligned for registration using a mutual information maximization strategy, combining the advantages of different imaging modalities to provide a comprehensive view of the condition.

[0037] Generative adversarial networks (GANs) are used to perform super-resolution reconstruction of thoracic surgical images and improve the quality of thoracic surgical images.

[0038] As a preferred embodiment of the big data-based thoracic surgery image analysis method of the present invention, wherein:

[0039] Support vector machine (SVM) is used as the classification model to classify the target disease type based on the integrated thoracic surgery images of different imaging modes. Specifically:

[0040] The SVM model was trained using integrated thoracic surgery images from different imaging modalities as the feature set;

[0041] Thoracic surgery images were input into the trained SVM model to classify disease types.

[0042] As a preferred embodiment of the big data-based thoracic surgery image analysis method of the present invention, wherein:

[0043] A real-time optimization feedback mechanism based on classification results adjusts model parameters according to the classification results after each classification is completed, and the model is retrained using the updated parameters. Specifically:

[0044] Evaluate the classification results of each batch and calculate the accuracy, precision, and recall of the classification results of the batch;

[0045] Adjust the penalty coefficient based on the gap between the existing classification results and the target classification results, and retrain the model using the re-adjusted penalty coefficient;

[0046] As a preferred embodiment of the big data-based thoracic surgery image analysis method of the present invention, wherein:

[0047] The retuned model was used in conjunction with thoracic surgical images and an adaptive learning rate adjustment factor was applied to improve the quality and efficiency of thoracic surgical image analysis. Specifically:

[0048] The latest thoracic surgical images after preprocessing, feature extraction and dimensionality reduction were input into the retrained model, and an adaptive learning rate adjustment factor was applied to improve the quality and efficiency of thoracic surgical image analysis.

[0049] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the big data-based thoracic surgery image analysis method as described in the first aspect of the present invention is implemented.

[0050] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the big data-based thoracic surgery image analysis method as described in the first aspect of the present invention.

[0051] The beneficial effects of the present invention are as follows: the present invention improves the quality and efficiency of thoracic surgery image analysis through multimodal image fusion, feature selection and dimensionality reduction, and support vector machine model training efficiency optimization. By integrating data from CT and PET imaging modes and adopting a mutual information maximization strategy for alignment, a comprehensive view of the disease is provided, and the accuracy of lesion detection is improved. The convolutional neural network CNN is used to automatically extract key features, and the principal component analysis PCA technology is used for dimensionality reduction processing, which reduces redundant data and enhances the generalization ability of the support vector machine model. The real-time optimization feedback mechanism ensures that the support vector machine model is always in the optimal state, accelerates the convergence speed and performance improvement of the support vector machine model, and provides clinicians with a more accurate and reliable diagnostic tool, which helps to detect diseases early and formulate effective treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 This is a flow chart of the big data-based thoracic surgery image analysis method in Example 1.

[0054] Figure 2 Schematic diagram of classification optimization feedback in Example 1.

[0055] Figure 3 Schematic diagram of feature dimensionality reduction and multimodal integration in Example 1.

[0056] Figure 4 Schematic diagram of data preprocessing and feature extraction in Example 1. DETAILED DESCRIPTION

[0057] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0058] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0059] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0060] Example 1, with reference to Figures 1 to 4 This embodiment provides a thoracic surgery image analysis method based on big data, comprising the following steps:

[0061] S1. Thoracic surgical images of the patient were acquired and preprocessed using a standardized protocol, specifically:

[0062] Select a multi-slice spiral CT scanner with high resolution and low radiation dose, preset the parameters of the multi-slice spiral CT scanner, and set the thickness of the image slice after each scan to 1mm through the touch screen operation interface of the multi-slice spiral CT scanner. A smaller slice thickness can make the scanned image display with higher resolution and more detailed anatomical structure information, but it will also increase the scanning time and the amount of thoracic surgery images. For a slice thickness of 1mm, sufficient thoracic surgery image details can be provided to observe tiny lesions. For chest scans of the area from above the clavicle to the lower edge of the ribs, instruct the patient to lie supine with both arms raised and still to avoid motion artifacts affecting the quality of thoracic surgery images. The entire lungs and heart also need to be covered to ensure that no anatomical structures and lesions are missed. When it is necessary to observe the vascular structures of the chest, such as the aorta, pulmonary artery, and veins, inject iodine-based contrast agent into the patient's blood vessels for about 30 seconds before starting the scan to best display the vascular structure.

[0063] The multi-slice spiral CT scanner was started according to the above method to complete the tomographic scan from the upper clavicle to the lower edge of the ribs, obtain the thoracic surgical images, and perform denoising and contrast enhancement.

[0064] The non-local means filtering algorithm is applied. This algorithm reduces noise by comparing the similarity between different pixel blocks in thoracic surgery images obtained by multi-slice spiral CT scanner, which is expressed as,

[0065] ;

[0066] in, is the pixel value after filtering, is the target pixel position, is the reference pixel position, is the weight function;

[0067] The grayscale map is calculated using the cumulative distribution function CDF to enhance the overall contrast of the image, which is expressed as,

[0068] ;

[0069] in, represents grayscale mapping, is the cumulative distribution function CDF, The maximum grayscale value of thoracic surgery images is usually 255. Indicates grayscale;

[0070] Saving thoracic surgical images that have been denoised and contrast enhanced in DICOM standard format helps maintain data consistency and traceability.

[0071] S2. Use the convolutional neural network (CNN) model to identify and extract key features from the pre-processed thoracic surgery images. Specifically,

[0072] The ResNet50 residual network architecture is selected as the basic architecture of the convolutional neural network (CNN). The ResNet50 residual network architecture consists of multiple residual blocks. Each residual block contains several convolutional layers and a skip connection, which can effectively alleviate the gradient disappearance problem in deep networks.

[0073] The denoised and contrast-enhanced thoracic surgical images were input into the convolutional neural network (CNN) model in batches of 32. Each image was first scaled to a uniform size to ensure consistency in the training of the convolutional neural network (CCN) model.

[0074] Building on the ResNet50 residual network architecture, additional convolutional layers are added to further refine key feature extraction. A 3x3 convolutional layer with 64 filters is added before each residual block. To reduce the spatial dimension of the feature map and control overfitting, a maximum pooling layer is added after each residual block, with a pooling window size of 2x2 and a stride of 2. This step helps extract important key features, including shape and texture features, vascular structure, and tumor boundaries, while also reducing computational complexity.

[0075] Shape features refer to the geometric properties of objects in thoracic surgery images, such as edges and contours, which help identify the contours of organs such as the lungs and heart, and observe whether there are abnormal proliferations and lesions in thoracic surgery images;

[0076] Texture features describe repetitive patterns and local changes within thoracic surgical imaging areas, reflecting the microscopic structure within the tissue. They are used to identify lesions in lung nodules and tumors. For example, benign and malignant tumors may exhibit different texture patterns, which helps doctors make more accurate diagnoses.

[0077] Vascular structure refers to the distribution and connection of blood vessels shown in the image. Vascular structure is extremely important for surgical planning. For example, during lung cancer resection, a clear understanding of the location and distribution of blood vessels around the tumor can help surgeons design safer and more effective surgical plans and reduce the risk of bleeding.

[0078] Tumor margin refers to the boundary between a tumor and surrounding normal tissue. Accurately defining the tumor margin is critical for assessing tumor size and location and formulating treatment strategies. In addition, it can also be used to monitor treatment efficacy, such as observing changes in tumor volume after radiotherapy.

[0079] The extraction and mutual complementation of shape features, texture features, vascular structure, and tumor boundaries constitute the basis for interpreting thoracic surgical images.

[0080] S3. Use principal component analysis (PCA) technology to reduce the dimension of the extracted key features. Specifically,

[0081] The key features extracted from the convolutional layer of the CNN model include shape features, texture features, vascular structure, and tumor boundaries. The mean and standard deviation of each key feature are calculated and expressed as,

[0082] ;

[0083] ;

[0084] in, represents the mean value of the eigenvalue set, represents the number of samples of eigenvalues, Indicates the The characteristic values ​​of the samples, represents the standard deviation of the eigenvalue set;

[0085] Apply a standardization operation to each feature so that the mean of all features is 0 and the standard deviation is 1. The standardized feature value is expressed as,

[0086] ;

[0087] in, represents the standardized eigenvalue;

[0088] Arrange all standardized eigenvalues ​​in columns to form an n×m matrix, where n represents the number of samples and m represents the number of features.

[0089] Based on the standardized feature matrix, the covariance matrix is ​​calculated and expressed as,

[0090] ;

[0091] Where C represents the covariance matrix, Expressed as Transpose the matrix, The matrix representing the normalized eigenvalue combinations;

[0092] The goal of performing eigendecomposition on the covariance matrix is ​​to find the eigenvalues ​​and eigenvectors of the covariance matrix, expressed as,

[0093] ;

[0094] in, represents the eigenvalues ​​of the matrix, represents the eigenvector;

[0095] Obtain the eigenvalues ​​and corresponding eigenvectors. Sort by eigenvalue size and select the eigenvector corresponding to the largest eigenvalue as the principal component. The eigenvalue selection can be determined based on the cumulative contribution rate, for example, selecting the eigenvalue with a cumulative contribution rate of 80% to 95% as the principal component.

[0096] The standardized feature matrix is ​​projected onto the principal component to obtain the thoracic surgical image after dimension reduction.

[0097] S4. We use the mutual information maximization strategy to integrate thoracic surgical images from different imaging modes that have undergone dimensionality reduction, and use generative adversarial networks (GANs) to perform super-resolution reconstruction of thoracic surgical images. Specifically,

[0098] The thoracic surgical images of CT scans and PET scans obtained above were processed for dimensionality reduction using PCA. The data for each imaging modality was standardized and had the same number of samples and corresponding label information. A mutual information maximization strategy was employed to maximize the mutual information between the two images to find the optimal spatial transformation parameters for spatial alignment. Specifically, spatial transformation parameters such as translation, rotation, and scaling were optimized by setting a reasonable range, such as a translation range of ±10 pixels and a rotation angle of ±5 degrees. For each pair of CT and PET images to be registered, image processing software such as ITK-SNAP and SimpleITK was used to perform mutual information under different spatial transformations. Mutual information measures the amount of information shared between the two images. Spatial alignment was performed using the image processing software by performing a spatial transformation of the mutual information.

[0099] Each pixel value in a CT image represents the X-ray absorption at that pixel, typically expressed in Hounsfield units (HU). For example, water absorbs 0 HU, while bone may absorb over +1000 HU. Each pixel value in a PET image reflects the concentration of the radiotracer at that pixel, which is typically correlated with metabolic activity. PET image pixel values ​​are often quantified using the Standard Uptake Value (SUV).

[0100] The CT and PET pixel values ​​at the same spatial location are directly added together, and the resulting value is used as the pixel value at the corresponding location in the new image. Similarly, the CT and PET pixel values ​​at all corresponding locations are added together using the above method and combined to form a new fused image. Taking into account the advantages of different imaging modalities, the feature vectors of CT and PET thoracic surgical images can also be concatenated to form a new comprehensive feature vector.

[0101] Using a medical image viewer, such as 3D Slicer, to load the fused thoracic surgery images can provide a more comprehensive view of the condition than a single imaging modality alone. For example, CT images provide detailed anatomical information, while PET images reveal metabolic activity. Combining the two can help doctors better understand the physiological state of the lesion area, providing a more comprehensive understanding of the condition. When thoracic surgery image quality is found to be substandard (e.g., low resolution, high noise, etc.), the GANs enhancement process is initiated, using the trained GANs model to perform super-resolution reconstruction on the selected thoracic surgery images. Specifically, for each thoracic surgery image to be enhanced, the generator network first generates a high-resolution version with a new perspective. The discriminator network then evaluates the image's authenticity. This process is iterated until a satisfactory result is achieved. The GANs model generates realistic medical images, improving low-quality and partially missing image data.

[0102] In addition to CT scanning and PET scanning, other imaging modalities, such as MRI and ultrasound, can also be considered to form a more comprehensive multimodal imaging data set. Different imaging modes provide complementary information. The thoracic surgical images obtained by combining multiple imaging modes can provide doctors with a more three-dimensional and comprehensive view of the disease, which helps to more accurately judge the disease status and formulate treatment plans.

[0103] S5. Use support vector machine (SVM) as a classification model to classify the target disease type based on the integrated thoracic surgery images of different imaging modes. Specifically,

[0104] Based on the good performance of support vector machines in high-dimensional space, the linear kernel support vector machine is selected as the classification model. The linear kernel function in the support vector machine model is simple in form and has high computational efficiency, which is suitable for medical image analysis.

[0105] Use cross-validation to determine the optimal penalty coefficient. Specifically, in a 5-fold cross-validation process, try a series of penalty coefficients, such as 0.01, 0.1, 1, 10, 100, and select the penalty coefficient that gives the highest accuracy on the validation set.

[0106] The fused comprehensive feature vectors were randomly divided into training and test sets, with a ratio of 80% training and 20% test, to ensure that samples of each target disease type were reasonably distributed in the training and test sets. The training set was then fed into the support vector machine model for training.

[0107] After preprocessing, dimensionality reduction, and fusion, thoracic surgical images are used as samples and fed into a trained support vector machine model. The SVM model then determines which class (positive or negative) each new sample belongs to based on its learned rules. Specifically, the SVM model calculates a score for each new sample and uses this score to determine its class. For example, if the score is greater than a certain threshold, the sample is considered to be in the positive class; otherwise, it is considered to be in the negative class. If the SVM model predicts a positive result, the disease type predicted by the SVM is consistent with the target disease type; if it is a negative result, it is inconsistent with the target disease type. This interpretation helps doctors understand the patient's condition or lesion. If necessary, the prediction results can be annotated on the original thoracic surgical image to help doctors better understand the basis for the SVM model's decision. For example, the lesion area can be marked in the image and the predicted classification result can be displayed. This can be achieved using professional medical image viewing software.

[0108] S6. A real-time optimization feedback mechanism based on classification results. After each classification is completed, the classification model parameters are adjusted according to the classification results, and the classification model is retrained using the updated parameters. Specifically,

[0109] If the SVM model predicts a positive result, the disease type predicted by the SVM for that sample is consistent with the target disease type; if it predicts a negative result, it indicates that the disease type is inconsistent with the target disease type. The accuracy rate is calculated by dividing the total number of samples by the total number of samples. Specifically, it measures the proportion of samples whose predicted disease type is consistent with the target disease type. The precision rate is the proportion of samples predicted as positive, such as those with lesions, that are actually positive. The recall rate is the proportion of samples that are correctly predicted as positive among all samples that are actually positive.

[0110] Compare the performance metrics of the current support vector machine model, including accuracy, precision, and recall, with ideal performance metrics. Ideal performance metrics are set based on domain knowledge and experimental results. Based on the difference between the current support vector machine model performance metrics and the ideal performance metrics, adjust the penalty coefficient in the support vector machine model. For example, if the precision of the support vector machine model is too low, increase the penalty coefficient to reduce false positives. Conversely, if the recall rate is low, reduce the penalty coefficient to improve the recall rate. Here, we combine the above penalty coefficient adjustment method with the actual situation in life to adjust the penalty coefficient in the support vector machine model. Use the adjusted penalty coefficient and the training set as input to the support vector machine model for training.

[0111] S7. Use the re-adjusted model in combination with thoracic surgical images and apply an adaptive learning rate adjustment factor to improve the quality and efficiency of thoracic surgical image analysis. Specifically,

[0112] New thoracic surgery imaging data is preprocessed and feature extracted according to the steps mentioned above. The feature vectors of the CT and PET thoracic surgery images are concatenated to form a new comprehensive feature vector. The retrained support vector machine model is used to predict the new samples. For each new sample, if the score is greater than 0, it is predicted to belong to the positive class; if it is less than 0, it belongs to the negative class.

[0113] By retraining and reapplying the support vector machine model, an adaptive learning rate adjustment factor was introduced. This factor is designed to improve the generalization ability of the support vector machine model by increasing the learning rate of samples close to the decision boundary. The adaptive learning rate adjustment factor continuously adjusts itself without interrupting the workflow, improving the learning efficiency and classification accuracy of the support vector machine model. By comparing the performance of the new and old support vector machine models in thoracic surgery image analysis, documenting improvements to the support vector machine model, and establishing a feedback mechanism to collect feedback from doctors and patients, the support vector machine model is adjusted accordingly. This feedback can help identify potential issues with the support vector machine model and guide future improvements, providing clinicians with more accurate diagnostic information and facilitating the development of more effective personalized treatment plans.

[0114] This embodiment also provides a computer device suitable for the case of a thoracic surgery image analysis method based on big data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the thoracic surgery image analysis method based on big data proposed in the above embodiment.

[0115] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0116] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the big data-based thoracic surgery image analysis method proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0117] In summary, the present invention improves the quality and efficiency of thoracic surgery image analysis through: multimodal image fusion, feature selection and dimensionality reduction, and support vector machine model training efficiency optimization. By integrating the data of CT and PET imaging modes and adopting the mutual information maximization strategy for alignment, it provides a comprehensive view of the disease and improves the accuracy of lesion detection. It uses convolutional neural network CNN to automatically extract key features and performs dimensionality reduction through principal component analysis PCA technology, which reduces redundant data and enhances the generalization ability of the support vector machine model. It optimizes the feedback mechanism in real time to ensure that the support vector machine model is always in the optimal state, accelerates the convergence speed and performance improvement of the support vector machine model, and provides clinicians with a more accurate and reliable diagnostic tool, which helps to detect diseases early and formulate effective treatment plans.

[0118] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A thoracic surgery image analysis method based on big data, characterized by: include, Thoracic surgical images of patients were acquired and preprocessed using standardized protocols; Convolutional neural network (CNN) is used to identify and extract key features from pre-processed thoracic surgical images. Use principal component analysis (PCA) technology to reduce the dimensionality of the extracted key features. The specific steps are as follows: Standardize the key features extracted from thoracic surgery images; The standardization process is to subtract the mean value of each key feature from its value and then divide it by its standard deviation, so that the mean of all features is 0 and the standard deviation is 1; Calculate the covariance matrix based on the standardized key feature data, and extract eigenvalues ​​and eigenvectors from the covariance matrix; Select the eigenvector corresponding to the largest eigenvalue as the principal component, project the original thoracic surgical image onto the principal component, and obtain the thoracic surgical image after dimension reduction; We used a mutual information maximization strategy to integrate thoracic surgical images from different imaging modalities that had undergone dimensionality reduction, and used generative adversarial networks (GANs) to perform super-resolution reconstruction of thoracic surgical images. The specific steps were as follows: Integrate thoracic surgical images from different imaging modalities after dimensionality reduction; The different imaging modes include CT scanning mode and PET scanning mode; Images from different imaging modalities are spatially aligned for registration using a mutual information maximization strategy, combining the advantages of different imaging modalities to provide a comprehensive view of the condition. Generative Adversarial Networks (GANs) are used to perform super-resolution reconstruction of thoracic surgical images and improve the quality of thoracic surgical images. Support vector machine (SVM) was used as the classification model to classify the target disease type based on the integrated thoracic surgery images of different imaging modes. A real-time optimization feedback mechanism based on classification results adjusts the classification model parameters according to the classification results after each classification is completed, and the classification model is retrained using the updated parameters; By combining the re-adjusted classification model with thoracic surgical images, an adaptive learning rate adjustment factor is applied to improve the quality and efficiency of thoracic surgical image analysis.

2. The big data-based thoracic surgery image analysis method according to claim 1, wherein: Thoracic surgical images of patients were acquired and preprocessed using a standardized protocol, specifically: Thoracic surgical images of patients were acquired using CT scanners, with standardized imaging protocols and uniform scanning parameters. The scanning parameters include layer thickness and scanning range; Preprocess the thoracic surgery images; The preprocessing includes removing noise and enhancing contrast.

3. The big data-based thoracic surgery image analysis method according to claim 2, wherein: The convolutional neural network (CNN) model is used to identify and extract key features from preprocessed thoracic surgical images, specifically: The ResNet residual network was used as the architecture of the convolutional neural network (CNN) model, and the pre-processed thoracic surgery images were input into the CNN model for training. Add multiple convolutional layers and pooling layers to the CNN model, use convolutional layers to extract key features, and use pooling layers to reduce the spatial dimensions of feature maps; The key features of thoracic surgery imaging include: Shape features, texture features, vascular structure, and tumor boundaries.

4. The big data-based thoracic surgery image analysis method according to claim 3, wherein: Support vector machine (SVM) was used as the classification model to classify the disease types based on the integrated thoracic surgery images of different imaging modes. Specifically: The SVM model was trained using integrated thoracic surgery images from different imaging modalities as the feature set; Thoracic surgery images were input into the trained SVM model to classify disease types.

5. The big data-based thoracic surgery image analysis method according to claim 4, wherein: Based on the real-time optimization feedback mechanism of classification results, after each classification is completed, the classification model parameters are adjusted according to the classification results, and the classification model is retrained using the updated parameters. Specifically: Evaluate the classification results of each batch and calculate the accuracy, precision, and recall of the classification results of the batch; Adjust the penalty coefficient based on the gap between the existing classification results and the target classification results, and retrain the model using the re-adjusted penalty coefficient.

6. The big data-based thoracic surgery image analysis method according to claim 5, wherein: The retuned model was used in conjunction with thoracic surgical images and an adaptive learning rate adjustment factor was applied to improve the quality and efficiency of thoracic surgical image analysis. Specifically: The latest thoracic surgical images after preprocessing, feature extraction and dimensionality reduction are input into the retrained classification model, and an adaptive learning rate adjustment factor is applied to improve the quality and efficiency of thoracic surgical image analysis.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the big data-based thoracic surgery image analysis method according to any one of claims 1 to 6 are implemented.

8. 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 big data-based thoracic surgery image analysis method according to any one of claims 1 to 6 are implemented.

Images