Imaging system and method for heart and cerebral vessel detection

Abstract

The invention discloses an imaging system and method for heart and cerebral vessel detection, and relates to the technical field of medical image processing, and the method comprises the steps: monitoring the physiological state of a patient, recording the medical history of the patient, forming initial data, transmitting the initial data to a Hadoop data processing platform, carrying out the standardization processing of the data, and generating a data set; and importing the data set into a random forest algorithm model to select an imaging mode, and pre-scanning the data set based on the selected imaging mode. According to the invention, through integrating wearable device data acquisition, Hadoop data processing, random forest imaging mode selection, deep learning automatic segmentation and feature extraction and isolated forest anomaly detection, the data set can be obtained; the problem of insufficient precision of blood vessel segmentation and feature extraction is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly to an imaging system and method for cardiovascular and cerebrovascular detection. Background Art

[0002] In recent years, significant progress has been made in the diagnostic techniques for cardiovascular and cerebrovascular diseases. Traditional imaging techniques such as CT, MRI, and ultrasound imaging are widely used in clinical practice, providing static vascular structure information. With the development of artificial intelligence, machine learning methods such as convolutional neural networks (CNNs) and random forest algorithms have shown great potential in vascular segmentation, lesion detection, etc., improving the automation level of medical image processing. However, there are still many limitations in the existing technologies.

[0003] Traditional cardiovascular and cerebrovascular imaging methods rely on a single modality, making it difficult to comprehensively capture the dynamic changes and functional characteristics of blood vessels, resulting in inaccurate and incomplete diagnostic results. Data processing is concentrated on a single platform, lacking distributed processing capabilities, which affects the efficiency of large-scale real-time data analysis. Existing vascular segmentation algorithms have insufficient accuracy when faced with complex vascular structures, prone to false segmentation or missed segmentation phenomena. Summary of the Invention

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

[0005] Therefore, the present invention provides an imaging system and method for cardiovascular and cerebrovascular detection to solve the problem of insufficient accuracy in vascular segmentation and feature extraction in the existing technology.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an imaging method for cardiovascular and cerebrovascular detection, which includes,

[0008] Monitoring the physiological state of the patient and recording the patient's medical history to form initial data;

[0009] Transmitting the initial data to the Hadoop data processing platform for data normalization processing to generate a data set;

[0010] Importing the data set into a random forest algorithm model to select an imaging mode, pre-scanning the data set based on the selected imaging mode to obtain images of vascular structures and dynamic changes, and transmitting the images of vascular structures and dynamic changes to a distributed edge computing platform for preliminary processing;

[0011] Using a deep convolutional neural network model to automatically segment and extract features from the preliminarily processed images of vascular structures and dynamic changes to obtain segmented images of blood vessels and morphological, functional, and lesion characteristics of blood vessels;

[0012] Based on the segmented images of blood vessels and the morphological, functional, and pathological characteristics of blood vessels, the abnormal conditions of blood vessels are identified through a preset gradient boosting tree, and a diagnostic report is generated according to the identification results.

[0013] As a preferred embodiment of the imaging method for cardiovascular and cerebrovascular detection described in the present invention, wherein: monitoring the physiological state of the patient and recording the medical history of the patient to form initial data, the specific steps are as follows:

[0014] Select wearable devices and medical monitors to regularly collect the physiological data of the patient and transmit it to the central server;

[0015] Use the BERT model to parse the unstructured text data in the physiological data of the patient in the central server, extract diagnostic information, treatment information, medical history information, examination results, and symptom description information, and convert them into a structured format;

[0016] Match the real-time physiological data with the medical history data according to the patient ID, and arrange them in a time series dataset according to the timestamp;

[0017] Use Pandas to remove duplicates, KNN to fill missing values, and Z-score to remove outliers from the data in the arranged time series dataset, and verify the quality of the patient's physiological data through K-fold cross-validation, and finally form the initial data.

[0018] As a preferred embodiment of the imaging method for cardiovascular and cerebrovascular detection described in the present invention, wherein: transmitting the initial data to the Hadoop data processing platform to perform standardization processing on the data to generate a dataset, the specific steps are as follows:

[0019] Export the cleaned initial data as a Parquet file;

[0020] Use StandardScaler to perform standardization processing on the numerical fields in the Parquet file, and convert them into distribution fields with a mean of 0 and a standard deviation of 1;

[0021] Use SparkSQL to extract time series features from the distribution fields and store the dataset partitioned by patient ID;

[0022] Store the partitioned dataset in Parquet format in the HDFS directory.

[0023] As a preferred embodiment of the imaging method for cardiovascular and cerebrovascular detection described in the present invention, wherein: importing the dataset into a random forest algorithm model to select an imaging mode, pre-scanning the dataset based on the selected imaging mode, obtaining blood vessel structures and dynamic change images, and transmitting the blood vessel structures and dynamic change images to a distributed edge computing platform for preliminary processing, the specific steps are as follows:

[0024] Load the partitioned Parquet files from the HDFS directory and use Spark to load the data, generating a Spark DataFrame of patient IDs, time series features, normalized physiological data, and medical history data;

[0025] Extract the feature columns from the Spark DataFrame, combine them into feature vectors, and generate training and test datasets;

[0026] Use a random forest classifier to build a random forest algorithm model, set the number of trees, maximum depth, and feature subset selection strategy to automatic, and substitute the training dataset into the random forest algorithm model for training to generate a trained random forest algorithm model;

[0027] Use the trained model to predict the test set, generate a result dataset, and select an imaging mode;

[0028] Select and configure the corresponding imaging device according to the imaging mode predicted by the random forest model;

[0029] Use the selected imaging device to pre-scan the patient to generate a preliminary image dataset;

[0030] Transfer the preliminary image dataset to a distributed edge computing platform for preliminary processing of image denoising, enhancement, and segmentation;

[0031] Store the preliminarily processed image data in the edge computing platform.

[0032] As a preferred embodiment of the imaging method for cardiovascular and cerebrovascular detection described in the present invention, wherein: a deep convolutional neural network model is adopted, and the preliminarily processed vascular structure and dynamic change images are automatically segmented and feature extracted to obtain the segmented image of the blood vessels and the morphological, functional, and lesion characteristics of the blood vessels. The specific steps are as follows:

[0033] Load the preliminarily processed image data from the edge computing platform and use the encoder part of the U-Net model to extract multi-level features of the image;

[0034] Upsample the multi-level features through the decoder part to generate the segmented image of the blood vessels, and post-process the segmented image to remove noise and artifacts;

[0035] Cite F as the comprehensive score of the vascular features, and extract the morphological, functional, and lesion characteristics of the blood vessels from the segmented image. The formula is as follows:

[0036]

[0037] Where n is the number of vascular regions, A iis the area of the i-th vascular region, P i is the perimeter of the i-th vascular region, d i is the diameter of the i-th vascular region, μ d is the mean of the diameters of the vascular regions, d i is the perimeter of the i-th vascular region, σ d is the standard deviation of the diameters of the vascular regions;

[0038] Calculate the morphological features, functional features, pathological features of blood vessels, and image data of abnormal vascular regions based on the calculation results and store them in the edge computing platform.

[0039] As a preferred solution of the imaging method for cardiovascular and cerebrovascular detection described in the present invention, wherein: according to the segmented image of blood vessels and the morphological, functional, and pathological features of blood vessels, identify the abnormal conditions of blood vessels through a preset gradient boosting tree, and generate a diagnostic report according to the identification results. The specific steps are as follows:

[0040] Fuse and perform weighted fusion on the morphological, functional, and pathological features of blood vessels through a self-attention mechanism;

[0041] Based on the fused morphological, functional, and pathological feature vectors of blood vessels, use a graph neural network to detect the image data of abnormal vascular regions and extract the feature data of abnormal regions;

[0042] Select an isolation forest model to calculate the anomaly score for the feature data of abnormal regions;

[0043] Generate a diagnostic report according to the anomaly score and the image data of abnormal vascular regions.

[0044] As a preferred solution of the imaging method for cardiovascular and cerebrovascular detection described in the present invention, wherein: the specific steps for selecting an isolation forest model to calculate the anomaly score for the feature data of abnormal regions are as follows:

[0045] Use the Z-score normalization method to normalize the feature data of abnormal regions to generate a normalized feature data set;

[0046] Use an isolation forest model to train the normalized feature data set to generate isolation trees, calculate the anomaly score for each data point through the path length of the isolation trees, and normalize the anomaly score;

[0047] Compare the feature data of abnormal regions with a pathological feature library to determine the pathological type;

[0048] Generate a diagnostic report including the location of abnormal regions, pathological type, and anomaly score, and output it to doctors and patients.

[0049] In a second aspect, the present invention provides an imaging system for cardiovascular and cerebrovascular detection, including:

[0050] A monitoring module that monitors the patient's physiological state and records the patient's medical history to form initial data;

[0051] A preprocessing module that transmits the initial data to a Hadoop data processing platform for data standardization processing to generate a data set;

[0052] An imaging and pre-scanning module that imports the data set into a random forest algorithm model to select an imaging mode, pre-scans the data set based on the selected imaging mode, obtains vascular structure and dynamic change images, and transmits the vascular structure and dynamic change images to a distributed edge computing platform for preliminary processing;

[0053] An image processing module that uses a deep convolutional neural network model to automatically segment and extract features from the preliminarily processed vascular structure and dynamic change images to obtain a segmented image of the blood vessels and the morphological, functional, and lesion characteristics of the blood vessels;

[0054] A diagnostic analysis module that, based on the segmented image of the blood vessels and the morphological, functional, and lesion characteristics of the blood vessels, identifies abnormal conditions of the blood vessels through a preset gradient boosting tree, and generates a diagnostic report based on the identification result.

[0055] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the imaging method for cardiovascular and cerebrovascular detection as described in the first aspect of the present invention is implemented.

[0056] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the imaging method for cardiovascular and cerebrovascular detection as described in the first aspect of the present invention is implemented.

[0057] The beneficial effects of the present invention are as follows: By integrating wearable devices, the Hadoop platform, the random forest algorithm, the deep convolutional neural network, and the isolation forest model, multi-modal imaging, distributed data processing, and intelligent analysis are achieved. The wearable devices and medical monitors collect patients' physiological data, and in combination with the BERT model, the medical history data is parsed to form a structured initial data set. The Hadoop platform performs data standardization processing to improve processing efficiency and scalability. The random forest algorithm intelligently selects the optimal imaging mode, and the deep convolutional neural network automatically performs segmentation and feature extraction, significantly improving the accuracy of blood vessel segmentation and reducing mis-segmentation and missed segmentation in complex blood vessel structures. The gradient boosting tree and the isolation forest model are used for anomaly detection to generate accurate diagnostic reports. Each step works in coordination, significantly improving the accuracy of blood vessel segmentation and feature extraction, and avoiding the problem of insufficient accuracy of the blood vessel segmentation algorithm when dealing with complex blood vessel structures, which is prone to mis-segmentation or missed segmentation phenomena. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0059] Figure 1 It is a flowchart of the imaging method for cardiovascular and cerebrovascular detection in Embodiment 1.

[0060] Figure 2 It is a flowchart for calculating the anomaly score in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will provide a detailed description of the specific embodiments of the present invention with reference to the drawings in the specification.

[0062] In the following description, many specific details are set forth to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0063] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that excludes other embodiments.

[0064] Embodiment 1, refer to Figure 1 andFigure 2 , which is the first embodiment of the present invention. This embodiment provides an imaging method for cardiovascular and cerebrovascular detection, including the following steps:

[0065] S1. Monitor the patient's physiological state and record the patient's medical history to form initial data.

[0066] Furthermore, wearable devices and medical monitors are selected to regularly collect the patient's physiological data and transmit it to the central server.

[0067] Specifically, through the combination of wearable devices and medical monitors, the physiological data of patients can be comprehensively and continuously collected, avoiding the problems of discontinuous or missing data collection in traditional methods. The real-time data is transmitted to the central server, ensuring the centralized management and efficient utilization of data, and providing a reliable basis for subsequent data analysis and processing.

[0068] Furthermore, the BERT model is used to parse the unstructured text data in the patient's physiological data in the central server, extract diagnostic information, treatment information, medical history information, examination results, and symptom description information, and convert them into a structured format.

[0069] Specifically, through the natural language processing ability of the BERT model, key information such as diagnosis, treatment, and medical history can be efficiently extracted and converted into a structured format, avoiding the inefficiency and errors of manually processing unstructured data in traditional methods. The generation of structured data provides a high-quality data source for subsequent data matching and analysis.

[0070] Furthermore, the real-time physiological data is matched with the medical history data according to the patient ID and arranged in a time series dataset according to the timestamp.

[0071] Specifically, through matching by patient ID and arranging by timestamp, the real-time physiological data of patients can be organically combined with the historical medical data to form a complete time series dataset. This data integration method helps to comprehensively understand the patient's health status and disease evolution, and provides more comprehensive and accurate data support for subsequent data analysis and diagnosis.

[0072] Furthermore, the data in the arranged time series dataset is de-duplicated using Pandas, missing values are filled using KNN, and outliers are removed using Z-score. The quality of the patient's physiological data is cross-validated by K-fold, and finally the initial data is formed.

[0073] Specifically, by using Pandas to remove duplicates, KNN to fill in missing values, and Z-score to remove outliers, the noise and anomalies in the data can be effectively processed, ensuring the accuracy and consistency of the data. K-fold cross-validation further verifies the quality of the data, avoiding the problems of incomplete data cleaning or insufficient verification in traditional methods. The finally generated initial dataset has high quality and high reliability, providing a solid foundation for subsequent machine learning and diagnostic analysis.

[0074] It should be noted that through real-time data collection, intelligent parsing, data integration and time serialization, data cleaning and quality verification, the comprehensive and efficient processing of patients' physiological data and medical history data has been realized. Each step has been optimized for the deficiencies of traditional methods, ensuring the integrity, accuracy and availability of the data. The finally generated initial dataset provides high-quality data support for subsequent imaging mode selection, image processing and diagnostic analysis, significantly improving the accuracy and efficiency of cardiovascular and cerebrovascular detection.

[0075] S2. Transmit the initial data to the Hadoop data processing platform for data standardization processing to generate a dataset.

[0076] Furthermore, export the cleaned initial data as a Parquet file.

[0077] Specifically, by using the Parquet file format, the storage space and transmission time of the data can be significantly reduced, while maintaining the columnar storage structure of the data, which is convenient for subsequent efficient query and analysis. Compared with traditional CSV or JSON formats, Parquet files have higher performance and lower storage costs when dealing with large-scale data, providing a good data foundation for subsequent distributed processing.

[0078] Furthermore, use StandardScaler to standardize the numerical fields in the Parquet file and convert them into distribution fields with a mean of 0 and a standard deviation of 1.

[0079] Specifically, through standardization processing, the dimensional differences between different features can be eliminated, avoiding unnecessary impacts on model training caused by some features with too large numerical values. The standardized data better meets the input requirements of machine learning algorithms, improving the training efficiency and prediction accuracy of the model. Especially when involving multi-feature analysis, standardization processing can significantly improve the stability and generalization ability of the model.

[0080] Furthermore, use SparkSQL to extract time series features from the distribution fields and store the dataset partitioned by patient ID.

[0081] Specifically, by extracting time series features, the trend and pattern of the patient's physiological data changing over time can be captured, providing important feature inputs for subsequent time series analysis and prediction. Storing the dataset by patient ID can improve the query efficiency and management convenience of the data. Especially when dealing with large-scale data, partitioned storage can significantly reduce the data retrieval time and improve the overall performance of the system.

[0082] Furthermore, the partitioned dataset is stored in Parquet format in the HDFS directory.

[0083] Specifically, by storing the data in HDFS, distributed management and high availability of the data can be achieved, ensuring system stability and scalability in large-scale data processing scenarios. The distributed storage mechanism of HDFS can effectively cope with the growth of data volume, while the columnar storage structure of the Parquet format further improves the data query efficiency and processing performance. This storage method provides strong support for subsequent distributed computing and real-time analysis.

[0084] It should be noted that through efficient data storage, standardized processing, time series feature extraction, and distributed storage, the comprehensive optimization and management of the initial data are realized. Each step improves the deficiencies of traditional methods, ensuring the efficiency, comparability, and manageability of the data. The finally generated dataset provides high-quality data inputs for the subsequent random forest algorithm model training and imaging mode selection, significantly improving the overall performance and accuracy of the cardiovascular and cerebrovascular detection system.

[0085] S3. Import the dataset into the random forest algorithm model to select the imaging mode, pre-scan the dataset based on the selected imaging mode, obtain the blood vessel structure and dynamic change images, and transmit the blood vessel structure and dynamic change images to the distributed edge computing platform for preliminary processing.

[0086] Furthermore, load the partitioned Parquet file from the HDFS directory and use Spark to load the data to generate a SparkDataFrame of patient ID, time series features, standardized physiological data, and medical history data.

[0087] Specifically, by using Spark to load the Parquet file, the advantages of distributed computing can be fully utilized, significantly improving the speed of data loading and processing. The structured data format of SparkDataFrame is convenient for subsequent feature extraction and model training, avoiding the problem of low data loading efficiency in traditional methods and providing an efficient data processing foundation for subsequent machine learning models.

[0088] Furthermore, extract the feature columns from the SparkDataFrame to form feature vectors and generate the training dataset and the test set.

[0089] Specifically, by extracting the feature vectors and partitioning the dataset, the data independence and representativeness for model training and testing can be ensured. The generation of feature vectors enables the machine learning model to make full use of multi-dimensional features for training, avoiding the problems of insufficient feature selection or unreasonable dataset partitioning in traditional methods, and improving the training effect and generalization ability of the model.

[0090] Furthermore, use the random forest classifier to construct the random forest algorithm model, set the number of trees, the maximum depth, and the feature subset selection strategy to automatic, and substitute the training dataset into the random forest algorithm model for training to generate the trained random forest algorithm model.

[0091] Specifically, through the multi-tree integration and automatic feature subset selection of the random forest algorithm, the classification accuracy and robustness of the model can be effectively improved. The random forest algorithm performs excellently in dealing with high-dimensional data and complex feature relationships, avoiding the problem of overfitting that is prone to occur in traditional single decision tree models. The trained model can accurately capture the complex patterns in the data, providing a reliable prediction basis for subsequent imaging mode selection.

[0092] Furthermore, use the trained model to predict the test set, generate the result dataset, and select the imaging mode.

[0093] Specifically, through model prediction and imaging mode selection, the most suitable imaging mode can be dynamically selected according to the patient's physiological data and historical records, avoiding the problems of single imaging mode or inaccurate selection in traditional methods. The intelligent imaging mode selection significantly improves the targeting of imaging and the accuracy of diagnosis, providing the optimal imaging scheme for subsequent image acquisition and processing.

[0094] It should be noted that through efficient data loading, feature vector generation, random forest model training, and intelligent imaging mode selection, the full-process intelligent processing from data loading to imaging mode selection is realized. Each step is optimized for the deficiencies of traditional methods, ensuring the efficiency of data processing, the accuracy of model training, and the intelligence of imaging mode selection. The finally generated imaging mode provides the optimal solution for subsequent image acquisition and diagnostic analysis, significantly improving the overall performance and diagnostic accuracy of the cardiovascular and cerebrovascular detection system.

[0095] Furthermore, select and configure the corresponding imaging device according to the imaging mode predicted by the random forest model.

[0096] Specifically, through the prediction of the random forest model, the most suitable imaging mode can be dynamically selected according to the patient's physiological data and historical records, avoiding the problems of single imaging mode or inaccurate selection in traditional methods. The intelligent imaging mode selection significantly improves the pertinence of imaging and the accuracy of diagnosis, providing an optimal imaging scheme for subsequent image acquisition and processing.

[0097] Furthermore, use the selected imaging device to pre-scan the patient to generate a preliminary image dataset.

[0098] Specifically, by generating a preliminary image dataset through pre-scanning, the image information of blood vessel structures and dynamic changes can be quickly obtained, avoiding the problems of incomplete or low-quality image acquisition in traditional methods. The preliminary image dataset provides a high-quality data basis for subsequent image processing and analysis, significantly improving the efficiency and accuracy of image processing.

[0099] Furthermore, transfer the preliminary image dataset to a distributed edge computing platform for preliminary processing of image data such as denoising, enhancement, and segmentation.

[0100] Specifically, through the distributed edge computing platform, real-time processing can be performed close to the data source, reducing data transmission latency and improving the efficiency and real-time performance of image processing. Preliminary processing operations such as denoising, enhancement, and segmentation significantly improve the quality of image data, providing high-quality image input for subsequent deep learning models, and avoiding the problems of low efficiency or low quality in traditional image processing methods.

[0101] Furthermore, store the preliminarily processed image data in the edge computing platform.

[0102] Specifically, by storing the preliminarily processed image data in the edge computing platform, efficient management and quick access to data can be achieved, ensuring system stability and scalability in large-scale data processing scenarios. The distributed storage mechanism of the edge computing platform can effectively handle the growth of data volume, while improving the query efficiency and processing performance of data, providing strong support for subsequent deep learning model training and diagnostic analysis.

[0103] It should be noted that through intelligent imaging mode selection, pre-scan image acquisition, preliminary processing on the distributed edge computing platform, and image data storage, a full-process intelligent processing from imaging mode selection to preliminary image data processing is achieved. Each step has been optimized for the deficiencies of traditional methods, ensuring the accuracy of the imaging mode, the high quality of image acquisition, the real-time performance of image processing, and the efficiency of data storage. The finally generated preliminarily processed image data provides a high-quality data basis for subsequent deep learning model training and diagnostic analysis, significantly improving the overall performance and diagnostic accuracy of the cardiovascular and cerebrovascular detection system.

[0104] S4. Adopt a deep convolutional neural network model and perform automatic segmentation and feature extraction on the preliminarily processed vascular structure and dynamic change images to obtain the segmented images of blood vessels and the morphological, functional, and pathological features of blood vessels.

[0105] Furthermore, load the preliminarily processed image data from the edge computing platform and use the encoder part of the U-Net model to extract multi-level features of the images.

[0106] Specifically, through the encoder part of the U-Net model, multi-level features of the images can be extracted, capturing the detailed information of the vascular structure and avoiding the problem of insufficient feature extraction in traditional methods. The extraction of multi-level features provides a rich information basis for subsequent image segmentation and feature analysis, significantly improving the accuracy of segmentation and feature extraction.

[0107] Furthermore, perform upsampling on the multi-level features through the decoder part to generate the segmented images of blood vessels, and perform post-processing on the segmented images to remove noise and artifacts.

[0108] Specifically, through the upsampling and post-processing of the decoder part, high-quality segmented images can be generated, removing noise and artifacts and avoiding the problem of low-quality segmented images in traditional methods. The high-quality segmented images provide a reliable data basis for subsequent feature extraction and lesion detection, significantly improving the accuracy and reliability of lesion detection.

[0109] Furthermore, refer to F as the comprehensive score of vascular features, and extract the morphological, functional, and pathological features of blood vessels from the segmented images. The formula is as follows:

[0110]

[0111] where n is the number of vascular regions, A i is the area of the i-th vascular region, P i is the perimeter of the i-th vascular region, d i is the diameter of the i-th vascular region, μ d is the mean of the diameters of vascular regions, d i is the perimeter of the i-th vascular region, σ d is the standard deviation of the diameters of vascular regions.

[0112] Specifically, through the comprehensive scoring formula, the morphological, functional, and pathological features of blood vessels can be quantitatively evaluated, avoiding the problem of strong subjectivity in feature evaluation in traditional methods. The quantitative evaluation provides objective data support for subsequent lesion detection and diagnosis, significantly improving the accuracy and reliability of lesion detection.

[0113] Further, morphological features, functional features, lesion features of each blood vessel region, and image data of blood vessel abnormal regions are calculated based on the calculation results and stored in the edge computing platform.

[0114] Specifically, by storing blood vessel features in the edge computing platform, efficient data management and rapid access can be achieved, ensuring system stability and scalability in large-scale data processing scenarios. The distributed storage mechanism of the edge computing platform can effectively handle the growth of data volume, while improving data query efficiency and processing performance, providing strong support for subsequent lesion detection and diagnostic analysis.

[0115] It should be noted that through the deep convolutional neural network model, multi-level feature extraction, image segmentation and post-processing, blood vessel feature quantification evaluation, and data storage, accurate segmentation and feature extraction of blood vessel images are realized. Each step optimizes the deficiencies of traditional methods, ensuring the accuracy of image segmentation, the comprehensiveness of feature extraction, and the efficiency of data storage. The finally generated blood vessel features provide high-quality data support for subsequent lesion detection and diagnostic analysis, significantly improving the overall performance and diagnostic accuracy of the cardiovascular and cerebrovascular detection system.

[0116] S5. According to the segmented image of the blood vessel and the morphological, functional, and lesion features of the blood vessel, a preset gradient boosting tree is used to identify abnormal conditions of the blood vessel, and a diagnostic report is generated based on the identification result.

[0117] Further, the morphological, functional, and lesion features of the blood vessel are weighted and fused through a self-attention mechanism.

[0118] Specifically, through the self-attention mechanism, weighted fusion can be performed according to the importance of different features, avoiding the problems of insufficient feature fusion or unreasonable weight allocation in traditional methods. The weighted fusion feature vector can more comprehensively reflect the abnormal conditions of the blood vessel, significantly improving the accuracy and reliability of subsequent anomaly detection.

[0119] Further, based on the fused morphological, functional, and lesion feature vectors of the blood vessel, a graph neural network is used to detect the image data of the blood vessel abnormal region and extract the feature data of the abnormal region.

[0120] Specifically, through the powerful graph structure analysis ability of the graph neural network, the complex relationships in the blood vessel abnormal region can be effectively captured, avoiding the problem of inaccurate detection of abnormal regions in traditional methods. The graph neural network can process the topological information of the blood vessel structure, significantly improving the accuracy of abnormal region detection and the comprehensiveness of feature extraction, providing high-quality data support for subsequent abnormal score calculation.

[0121] Furthermore, an isolation forest model is selected to calculate the anomaly score for the feature data of the abnormal region.

[0122] Specifically, through the isolation forest model, the anomaly points in the feature data of the abnormal region can be effectively identified, avoiding the problem of strong subjectivity in anomaly scoring in traditional methods. The isolation forest model performs excellently in processing high-dimensional data, can accurately calculate the anomaly score, significantly improves the objectivity and accuracy of the abnormal region assessment, and provides a reliable quantitative basis for the subsequent generation of diagnostic reports.

[0123] Furthermore, a diagnostic report is generated based on the anomaly score and the image data of the vascular abnormal region.

[0124] Specifically, through the comprehensive analysis of the anomaly score and the detection results, a detailed diagnostic report can be generated, avoiding the problems of low efficiency or incomplete information in the generation of diagnostic reports in traditional methods. The diagnostic report provides clear lesion information and quantitative evaluation results for doctors, significantly improves the efficiency and accuracy of diagnosis, and provides a reliable basis for the treatment of patients.

[0125] It should be noted that through gradient boosting tree anomaly recognition, self-attention mechanism feature fusion, graph neural network abnormal region detection, isolation forest anomaly score calculation, and diagnostic report generation, the intelligent recognition and diagnosis of vascular anomalies are realized. Each step optimizes the deficiencies of traditional methods, ensuring the accuracy of anomaly recognition, the comprehensiveness of feature fusion, the precision of abnormal region detection, the objectivity of anomaly scoring, and the efficiency of diagnostic report generation. The finally generated diagnostic report provides a reliable diagnostic basis for doctors, significantly improving the overall performance and diagnostic accuracy of the cardiovascular and cerebrovascular detection system.

[0126] S6. Select an isolation forest model to calculate the anomaly score for the feature data of the abnormal region.

[0127] Furthermore, the Z-score standardization method is used to standardize the feature data of the abnormal region to generate a standardized feature data set.

[0128] Specifically, through Z-score standardization, the dimensionality differences between different features can be eliminated, avoiding unnecessary impacts on model training caused by some features with too large numerical values. The standardized data better meets the input requirements of the isolation forest model, improves the training efficiency and prediction accuracy of the model. Especially in the case of multi-feature analysis, the standardization process can significantly enhance the stability and generalization ability of the model.

[0129] Furthermore, the isolation forest model is used to train the standardized feature data set to generate isolation trees, and the anomaly score of each data point is calculated through the path length of the isolation trees and the anomaly scores are normalized.

[0130] Specifically, through the training of the isolation forest model and the calculation of path length, the anomaly degree of each data point can be accurately evaluated, avoiding the problem of strong subjectivity in anomaly scoring in traditional methods. The normalized anomaly score makes the result more intuitive and easy to understand, providing a reliable basis for subsequent lesion type judgment and diagnostic report generation.

[0131] Furthermore, the characteristic data of the abnormal region is compared with the lesion feature library to determine the lesion type.

[0132] Specifically, through comparison with the lesion feature library, the lesion type of the abnormal region can be accurately identified, avoiding the problem of inaccurate lesion type judgment in traditional methods. The intelligent lesion type judgment provides accurate data support for subsequent diagnostic report generation, significantly improving the accuracy and reliability of lesion detection.

[0133] Furthermore, a diagnostic report including the location of the abnormal region, the lesion type, and the anomaly score is generated and output to doctors and patients.

[0134] Specifically, through automatic generation of the diagnostic report, detailed diagnostic information including the location of the abnormal region, the lesion type, and the anomaly score can be quickly provided, avoiding the problem of low efficiency in diagnostic report generation in traditional methods. The automatic output of the diagnostic report provides timely and accurate diagnostic basis for doctors and patients, significantly improving the efficiency and accuracy of diagnosis.

[0135] It should be noted that through feature data standardization, anomaly score calculation and normalization, lesion type judgment, and diagnostic report generation, the intelligent detection and diagnosis of abnormal regions are realized. Each step has been optimized for the deficiencies of traditional methods, ensuring the accuracy of anomaly detection, the precision of lesion type judgment, and the timeliness of diagnostic report generation. The finally generated diagnostic report provides a reliable diagnostic basis for doctors and patients, significantly improving the overall performance and diagnostic accuracy of the cardiovascular and cerebrovascular detection system.

[0136] This embodiment also provides a spectrum axis control adjustment system for assisting myopia prevention and control, including:

[0137] A monitoring module that monitors the patient's physiological state and records the patient's medical history to form initial data;

[0138] A preprocessing module that transmits the initial data to the Hadoop data processing platform for data standardization processing to generate a data set;

[0139] The imaging and pre-scanning module imports the dataset into the random forest algorithm model to select the imaging mode, pre-scans the dataset based on the selected imaging mode, obtains the vascular structure and dynamic change images, and transmits the vascular structure and dynamic change images to the distributed edge computing platform for preliminary processing;

[0140] The image processing module uses a deep convolutional neural network model to automatically segment and extract features from the preliminarily processed vascular structure and dynamic change images, obtaining the segmented images of blood vessels and the morphological, functional, and lesion features of blood vessels;

[0141] The diagnostic analysis module, based on the segmented images of blood vessels and the morphological, functional, and lesion features of blood vessels, identifies abnormal conditions of blood vessels through a preset gradient boosting tree, and generates a diagnostic report according to the identification results.

[0142] This embodiment also provides a computer device applicable to the case of an imaging method for cardiovascular and cerebrovascular detection, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the imaging method for cardiovascular and cerebrovascular detection proposed in the above embodiment.

[0143] The computer device can be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device 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 and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad set on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0144] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the imaging method for cardiovascular and cerebrovascular detection 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 (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, magnetic disk or optical disc.

[0145] In summary, the present invention realizes multi-modal imaging, distributed data processing and intelligent analysis by integrating wearable devices, Hadoop platform, random forest algorithm, deep convolutional neural network and isolation forest model. The wearable device and medical monitor collect the physiological data of patients, and combine with the BERT model to parse the medical history data to form a structured initial data set. The Hadoop platform performs data standardization processing to improve the processing efficiency and scalability. The random forest algorithm intelligently selects the optimal imaging mode, and the deep convolutional neural network automatically segments and extracts features, significantly improving the blood vessel segmentation accuracy and reducing mis-segmentation and missed segmentation in complex blood vessel structures. The gradient boosting tree and isolation forest model perform anomaly detection to generate an accurate diagnosis report. Each step works together to significantly improve the accuracy of blood vessel segmentation and feature extraction, avoiding the problem of insufficient accuracy of the blood vessel segmentation algorithm when facing complex blood vessel structures, and being prone to mis-segmentation or missed segmentation.

[0146] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. An imaging method for cardiovascular and cerebrovascular detection, characterized in that: include, Monitor the patient's physiological status and record the patient's medical history to form initial data; The initial data is transferred to the Hadoop data processing platform to standardize the data and generate a data set; Import the data set into the random forest algorithm model to select an imaging mode, pre-scan the data set based on the selected imaging mode, obtain the vascular structure and dynamic change images, and transmit the vascular structure and dynamic change images to the distributed edge computing platform for preliminary processing; A deep convolutional neural network model is used to automatically segment and extract features from the preliminarily processed vascular structure and dynamic change images to obtain segmented images of the blood vessels as well as the morphological, functional and pathological characteristics of the blood vessels. According to the segmented images of blood vessels and the morphological, functional and pathological characteristics of blood vessels, the abnormal conditions of blood vessels are identified through the preset gradient boosting tree, and a diagnosis report is generated based on the identification results.

2. The imaging method for cardiovascular and cerebrovascular detection according to claim 1, characterized in that: The monitoring of the patient's physiological state and recording of the patient's medical history constitute initial data, and the specific steps are: Wearable devices and medical monitors are used to regularly collect patient physiological data and transmit them to the central server; Use the BERT model to parse unstructured text data of patient physiological data in the central server, extract diagnosis information, treatment information, medical history information, examination results and symptom description information and convert them into a structured format; Match real-time physiological data with medical history data by patient ID and arrange them by timestamp to form a time series data set; Pandas was used to remove duplicates from the arranged time series data set, KNN was used to fill missing values, and Z-score was used to remove outliers. The quality of the patient's physiological data was cross-validated through K-fold to finally form the initial data.

3. The imaging method for cardiovascular and cerebrovascular detection according to claim 2, characterized in that: The initial data is transferred to the Hadoop data processing platform to perform standardization processing on the data to generate a data set. The specific steps are: Export the cleaned initial data as a Parquet file; Use StandardScaler to standardize the numeric fields in the Parquet file and convert them into distribution fields with a mean of 0 and a standard deviation of 1. Use SparkSQL to extract time series features from the distribution field and store the dataset partitioned by patient ID; The partitioned dataset is stored in Parquet format in an HDFS directory.

4. The imaging method for cardiovascular and cerebrovascular detection according to claim 3, characterized in that: The data set is imported into the random forest algorithm model to select an imaging mode, the data set is pre-scanned based on the selected imaging mode, the vascular structure and dynamic change images are obtained, and the vascular structure and dynamic change images are transmitted to the distributed edge computing platform for preliminary processing. The specific steps are: Load the partitioned Parquet files from the HDFS directory and use Spark to load the data to generate a SparkDataFrame of patient ID, time series features, standardized physiological data, and medical history data; Extract feature columns from SparkDataFrame and combine them into feature vectors to generate training and test data sets; Use the random forest classifier to build a random forest algorithm model, set the number of trees, maximum depth, and feature subset selection strategy to automatic, and substitute the training data set into the random forest algorithm model for training to generate a trained random forest algorithm model; Use the trained model to make predictions on the test set, generate the resulting dataset and select the imaging mode; Select and configure the corresponding imaging equipment according to the imaging mode predicted by the random forest model; Pre-scanning the patient using the selected imaging device to generate a preliminary image dataset; The preliminary image data set is transmitted to the distributed edge computing platform to perform preliminary processing such as denoising, enhancement and segmentation of the image data; The preliminarily processed image data is stored in the edge computing platform.

5. The imaging method for cardiovascular and cerebrovascular detection according to claim 4, characterized in that: The method uses a deep convolutional neural network model to automatically segment and extract features of the vascular structure and dynamic change images after preliminary processing to obtain the segmented images of the blood vessels as well as the morphological, functional and pathological characteristics of the blood vessels. The specific steps are: Load the preliminarily processed image data from the edge computing platform and use the encoder part of the U-Net model to extract multi-level features of the image; The decoder part upsamples the multi-level features to generate a segmented image of the blood vessels, and the segmented image is post-processed to remove noise and artifacts; F is used as a comprehensive score of vascular features to extract the morphological, functional and lesion characteristics of the blood vessels from the segmented image. The formula is as follows: Where n is the number of vascular regions, A i is the area of ​​the ith blood vessel region, P i is the perimeter of the ith vascular region, d i is the diameter of the ith blood vessel region, μ d is the mean diameter of the vascular area, d i is the perimeter of the ith blood vessel region, σ d is the standard deviation of the diameter of the vascular area; The calculation result F is used to calculate the morphological characteristics, functional characteristics, vascular pathological characteristics and image data of abnormal vascular areas of each vascular region and store them on the edge computing platform.

6. The imaging method for cardiovascular and cerebrovascular detection according to claim 5, characterized in that: The method of identifying abnormal conditions of blood vessels by using a preset gradient boosting tree based on the segmented image of blood vessels and the morphological, functional and pathological characteristics of blood vessels, and generating a diagnosis report based on the identification results, specifically comprises the following steps: The morphological, functional and lesion characteristics of the blood vessels are weightedly fused through the self-attention mechanism; Based on the fused vascular morphology, functionality, and lesion feature vectors, a graph neural network is used to detect image data of abnormal vascular areas and extract feature data of abnormal areas. The isolation forest model is used to calculate the anomaly score for the feature data of the abnormal area; A diagnostic report is generated based on the abnormality score and image data of the vascular abnormality area.

7. The imaging method for cardiovascular and cerebrovascular detection according to claim 6, characterized in that: The isolation forest model is used to calculate the abnormal score of the feature data of the abnormal area, and the specific steps are: Use the Z-score standardization method to standardize the feature data of the abnormal area and generate a standardized feature data set; The standardized feature data set is trained using the isolation forest model to generate an isolation tree. The anomaly score of each data point is calculated by the path length of the isolation tree and the anomaly score is normalized. Compare the characteristic data of the abnormal area with the lesion characteristic library to determine the lesion type; A diagnostic report containing the location of the abnormal area, lesion type and abnormality score is generated and output to doctors and patients.

8. An imaging system for cardiovascular and cerebrovascular detection, based on the imaging method for cardiovascular and cerebrovascular detection according to any one of claims 1 to 7, characterized in that: include, The monitoring module monitors the patient's physiological status and records the patient's medical history to form the initial data; The preprocessing module transfers the initial data to the Hadoop data processing platform to standardize the data and generate a data set; The imaging and pre-scanning module imports the data set into the random forest algorithm model to select the imaging mode, pre-scans the data set based on the selected imaging mode, obtains the vascular structure and dynamic change images, and transmits the vascular structure and dynamic change images to the distributed edge computing platform for preliminary processing; The image processing module uses a deep convolutional neural network model to automatically segment and extract features from the vascular structure and dynamic change images after preliminary processing, obtaining the segmented images of the blood vessels as well as the morphological, functional and pathological characteristics of the blood vessels; The diagnostic analysis module identifies abnormal conditions of blood vessels based on the segmented images of blood vessels and the morphological, functional and pathological characteristics of the blood vessels through a preset gradient boosting tree, and generates a diagnostic report based on the identification results.

9. 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 imaging method for cardiovascular and cerebrovascular detection described in 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 imaging method for cardiovascular and cerebrovascular detection described in any one of claims 1 to 7 are implemented.

Images