Intelligent cerebral apoplexy disease identification method, system, equipment and medium
By combining deep learning algorithms with convolutional neural networks and recurrent neural networks to process multi-dimensional stroke data, the limitations of traditional methods in data processing are overcome, efficient and accurate stroke disease identification and prediction are achieved, and diagnostic efficiency and resource utilization are improved.
Patent Information
- Application Number
- CN202510524823.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-24
- Publication Date
- 2025-09-12
AI Technical Summary
Traditional data analysis methods are difficult to effectively process multi-dimensional, highly complex stroke disease data, cannot accurately identify correlation relationships, have high computational costs, are difficult to clean data, and perform poorly when dealing with nonlinear relationships and implicit features.
A deep learning algorithm is used, combined with convolutional neural networks and recurrent neural networks, to perform multimodal feature fusion. Image data and time-series medical record data are used to extract structured and unstructured data features from the health records of stroke patients. The model parameters are optimized through the cross-entropy loss function, and the fused feature values are generated and input into the classifier for prediction.
It improves the diagnostic accuracy and efficiency of stroke, provides a more accurate basis for diagnosis, reduces morbidity and mortality, optimizes the utilization of medical resources, and has important clinical value.
Smart Images
Figure CN120636837A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of data analysis technology, and more specifically relates to a method, system, device and medium for intelligent identification of stroke disease. Background Art
[0002] With the rapid advancement of medical technology, the healthcare industry is experiencing an unprecedented data explosion. The widespread use of data sources such as electronic health records (EHRs), genomic data, imaging data, and physiological parameters from wearable devices provides a rich information foundation for precision medicine and personalized treatment. However, the complexity, multidimensionality, and massive volume of this data pose significant challenges to traditional data analysis methods.
[0003] Traditional data analysis techniques typically achieve good results when processing single-dimensional or relatively simple data sets. However, their limitations become apparent when studying complex diseases such as stroke. As a disease with high disability and mortality rates, the pathogenesis of stroke involves the interaction of numerous biomarkers, genetic variations, lifestyle habits, environmental factors, and other factors. Therefore, to accurately identify and predict the occurrence of stroke, it is necessary to comprehensively consider data from multiple dimensions and uncover potential correlations.
[0004] However, traditional data analysis methods often struggle to effectively capture these complex relationships when processing multidimensional data. On the one hand, the sheer volume of data dramatically increases computational and time costs; on the other hand, the diversity and complexity of the data complicate data cleaning, preprocessing, and feature selection. Furthermore, traditional methods perform poorly when dealing with nonlinear relationships and implicit features, further limiting their application in stroke identification.
[0005] Given these challenges, there is an urgent need for an intelligent approach that can efficiently process multidimensional, highly complex data and accurately identify stroke associations. This approach needs to be able to automatically extract key features from the data and build accurate predictive models to achieve effective early warning and intervention for stroke. Summary of the Invention
[0006] In response to the above problems, the purpose of the present invention is to provide a method, system, device and medium for intelligent identification of stroke disease. By utilizing deep learning algorithms, especially convolutional neural networks and recurrent neural networks, multimodal feature fusion is performed to make full use of image data and time-series medical record data, which can help doctors predict stroke diseases in patients, improve the effectiveness of clinical decision-making, and achieve personalized medical care and resource optimization.
[0007] To achieve the above-mentioned purpose, the present invention is implemented through the following technical solutions: In a first aspect, the present invention provides an intelligent stroke disease identification method, comprising: Using big data technology to collect medical data from multiple sources, sorting them using the patient's ID number as a marker, extracting the health records of stroke patients, pre-processing the health records, sorting the health records, and extracting structured data and unstructured data; Analyze unstructured data through convolutional neural networks, extract spatial features, and optimize corresponding model parameters using cross-entropy loss function; Analyze structured data through recurrent neural networks, capture dynamic changes in data, and extract time series features; Perform multimodal fusion of spatial features and time series features to generate fused feature values; The fused feature values are input into the classifier, and the probability of the patient suffering a stroke is output.
[0008] In an optional embodiment, the method uses big data technology to collect medical data from multiple sources, organizes the data using the patient's ID number as a marker, extracts the health records of stroke patients, organizes the health records, and extracts structured data and unstructured data, including: Using big data technology to collect medical data from multiple sources, including electronic health records, medical images, laboratory test results, wearable device data, sensor data, and medication use records; The medical data is sorted using the patient's ID number as a marker, and the physical examination data and examination reports of stroke patients over the years are organized in a chronological manner to form a health file for each patient; The health records are sorted and structured data and unstructured data are extracted. The structured data includes physiological parameters, and the unstructured data includes imaging data.
[0009] In an optional embodiment, the analyzing the unstructured data by using a convolutional neural network to extract spatial features includes: The image data is analyzed through the convolutional neural network, and the convolution kernel is used to perform convolution operation on the image data using the formula F=Conv(I,K), extract local features, generate feature map F, and apply the ReLU activation function F ReLU =max(0,F) generates the activated feature map F ReLU ; Where I is the pixel value of the image data, and K is the convolution kernel; Using formula F pool =max(F ReLU ) for the activated feature map F ReLUPerform pooling operation and output the pooled feature map F pool ; Take advantage of y CNN =σ(W·X+b)The feature map F after pooling pool Mapped to the classification result, output spatial feature y CNN , as the analysis value of stroke imaging data; Among them, W is the weight matrix, X is the feature map F after pooling pool , b is the bias term, and σ is the activation function.
[0010] In an optional embodiment, the cross entropy loss function includes:
[0011] in, is the binary representation of the actual label, indicating whether the patient has a stroke. is the probability predicted by the model, and N is the number of samples.
[0012] In an optional embodiment, analyzing structured data using a recurrent neural network to capture dynamic changes in data and extract time series features includes: Analyze the time series data of physiological parameters X=[x1,x2,…,x t ], capture the dynamic changes of data, and calculate the hidden state h by the following formula t , as the time series feature y RNN : h t =σ(W h ·h t-1 +W x ·x t +B) Among them, h t is the hidden state at time step t, h t-1 is the hidden state at the previous moment, x t is the physiological parameter at the current moment, W h and W x is the weight matrix and B is the bias term.
[0013] In an optional embodiment, the multimodal fusion of spatial features and time series features to generate fused feature values includes: Using the formula Z=α·y CNN +β·y RNN Perform weighted summation on spatial features and time series features to calculate the final fusion feature value Z; Among them, α and β are weighting coefficients.
[0014] In an optional embodiment, the classifier includes:
[0015] Among them, P is the probability of a patient suffering a stroke, W k is the weight of the classifier, Z is the fusion feature value, and K is the number of categories.
[0016] In a second aspect, the present application also provides an intelligent stroke disease identification system, including: A data preprocessing module is used to collect medical data from multiple sources using big data technology, organize it using the patient's ID number as a marker, extract the health records of stroke patients, preprocess the health records, organize the health records, and extract structured data and unstructured data; Unstructured data analysis module, which is used to analyze unstructured data through convolutional neural networks, extract spatial features, and optimize corresponding model parameters using the cross-entropy loss function; The structured data analysis module is used to analyze structured data through recurrent neural networks, capture dynamic changes in data, and extract time series features; Multimodal fusion module, used to perform multimodal fusion of spatial features and time series features to generate fusion feature values; The prediction module is used to input the fused feature value into the classifier and output the probability of the patient suffering a stroke.
[0017] In a third aspect, an embodiment of the present application further provides an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the program, the steps of the intelligent stroke disease identification method as described in any one of the above items are implemented.
[0018] In a fourth aspect, an embodiment of the present application further provides a storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the intelligent stroke disease identification method as described in any one of the above items are implemented.
[0019] It can be seen from the above technical solutions that the present invention has the following advantages: The intelligent stroke identification method provided in this application achieves intelligent stroke identification through the comprehensive application of big data technology and deep learning algorithms, greatly improving the accuracy and efficiency of diagnosis. First, big data technology is used to collect and organize the health records of stroke patients from medical data from multiple sources. Through a sophisticated preprocessing step, the medical data is divided into structured and unstructured data, providing a solid foundation for subsequent analysis. During the data processing process, a convolutional neural network is used to analyze unstructured data such as image data. Through advanced techniques such as convolution operations, activation functions, and pooling operations, the spatial features of the images are efficiently extracted, providing strong support for the analysis of stroke image data. At the same time, the method also uses a recurrent neural network to analyze structured data such as physiological parameters, capturing the dynamic changes of the data and extracting time series features, further enriching the diagnostic information. On this basis, the spatial features and time series features are multimodally fused to generate fused feature values that contain more comprehensive diagnostic information, improving the accuracy of diagnosis. Finally, the fused feature values are input into a classifier, and the probability of the patient suffering a stroke is calculated using an advanced classification algorithm, providing doctors with a more accurate diagnosis basis. This method not only improves the accuracy and efficiency of diagnosis, but also provides strong technical support for the prevention and treatment of stroke, and has important clinical value and practical significance.
[0020] This application uses big data technology to effectively integrate information from multiple medical data sources, including electronic health records, medical imaging, and genomic data, to form a comprehensive and detailed patient health profile. This process not only improves the comprehensiveness and accuracy of the data but also, through sophisticated preprocessing steps, transforms complex medical data into structured and unstructured components, laying a solid foundation for subsequent analysis and identification. This data integration and processing method significantly improves the efficiency and accuracy of data analysis.
[0021] This application utilizes convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to analyze unstructured and structured data, respectively. Convolutional neural networks efficiently extract spatial features from imaging data, while RNNs capture dynamic changes in time series data such as physiological parameters. The combination of these two deep learning algorithms enables this method to more comprehensively extract characteristic information related to stroke, providing strong support for subsequent classification and identification.
[0022] This application combines the features extracted by convolutional neural networks and recurrent neural networks into a multimodal fusion, generating fused feature values that contain more comprehensive diagnostic information. This fusion approach not only improves diagnostic accuracy but also enables the method to more comprehensively reflect the patient's health status, providing doctors with more detailed diagnostic evidence.
[0023] This application uses a classifier to calculate the probability of a patient suffering a stroke, providing doctors with more accurate and rapid diagnostic results. This not only helps doctors take effective treatment measures in a timely manner, reducing the morbidity and mortality of stroke, but also improves the efficiency of medical resource utilization and reduces the burden on the medical system. Furthermore, this application can provide strong technical support for the prevention and treatment of stroke, and has important clinical value and practical significance. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for the description. 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.
[0025] Figure 1 A flow chart of the intelligent stroke identification method provided in this application.
[0026] Figure 2 This is a schematic diagram of the structure of the intelligent stroke identification system provided in this application.
[0027] Figure 3 This is a schematic diagram of the structure of the electronic device provided in this application. DETAILED DESCRIPTION
[0028] In the following detailed description of the specific steps of the intelligent stroke disease identification method, various embodiments of the present disclosure will be described in more detail. The present disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of the present disclosure to the specific embodiments disclosed herein, but rather the present disclosure should be understood to cover all adjustments, equivalents and / or alternatives that fall within the spirit and scope of the various embodiments of the present disclosure.
[0029] Hereinafter, the terms "include" or "may include" as used in various embodiments of the present disclosure indicate the presence of disclosed functions, operations, or elements, and do not limit the addition of one or more functions, operations, or elements. In addition, as used in various embodiments of the present disclosure, the terms "include," "have," and their cognates are intended only to indicate specific features, numbers, steps, operations, elements, components, or combinations of the foregoing, and should not be understood as excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or the possibility of adding one or more features, numbers, steps, operations, elements, components, or combinations of the foregoing.
[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0031] See also Figure 1 FIG. 1 is a flow chart of a method for intelligently identifying a stroke disease in a specific embodiment. The method includes: S1: Utilize big data technology to collect medical data from multiple sources, organize them using the patient's ID number as a marker, extract the health records of stroke patients, preprocess the health records, organize the health records, and extract structured data and unstructured data.
[0032] For example, first, big data technology is used to collect medical data from multiple sources, and patient data with a stroke condition are screened based on the electronic medical record diagnosis. The medical-related data should at least include: clinical symptoms and corresponding disease severity data recorded in several electronic health records (EHRs), medical images (such as X-rays, MRIs, etc.), laboratory test results, wearable devices, sensor data, drug use records, etc.
[0033] Then, for the collected medical data, the patient's ID number is used as the identifier ID, and the physical examination data and examination reports of the years before and after the stroke are uniformly organized in a chronological manner. This is equivalent to each stroke patient having his or her own health file after data preprocessing, which contains both structured physiological parameters and unstructured image data.
[0034] S2: Analyze unstructured data through convolutional neural networks, extract spatial features, and optimize the corresponding model parameters using the cross-entropy loss function.
[0035] For example, the image data is analyzed by a convolutional neural network, and the image data is convolved using the convolution kernel using the formula F=Conv(I,K) to extract local features and generate a feature map F.
[0036] Where I is the pixel value of the image data, typically a two-dimensional matrix from a brain CT or MRI scan; K is the convolution kernel, used to extract local features. This method can quickly extract image features, filter out non-feature areas, and reduce the space occupied by the image data.
[0037] Furthermore, the ReLU (Rectified Linear Unit) activation function is used to increase the nonlinearity of the model, which helps the model capture the complex characteristics of diseased tissue. The specific method is as follows: Apply the ReLU activation function F ReLU =max(0,F) generates the activated feature map F ReLU .
[0038] The corresponding process is: Apply the ReLU activation function to the feature map F after the convolution operation. Replace all negative values with 0, retain positive values, and increase the nonlinearity of the model. Output the activated feature map F ReLU .
[0039] The next step is to use pooling to reduce the size of the feature map. By selecting the maximum value of the local window, the most important features are retained to help reduce the amount of calculation. The specific method is: using formula F pool =max(F ReLU ) for the activated feature map F ReLU Perform pooling operation and output the pooled feature map F pool .
[0040] The corresponding data processing process is: first, the feature map F after ReLU activation ReLU Perform pooling operation. Select the maximum value of the local window, reduce the size of the feature map, and retain the most important features. Finally, output the pooled feature map F pool .
[0041] In the next step, we need to map the CNN output to the classification result (such as whether there is a stroke) in the following way: Take advantage of y CNN =σ(W·X+b)The feature map F after pooling pool Mapped to the classification result, output spatial feature y CNN , as the analysis value of stroke imaging data; Among them, W is the weight matrix, X is the feature map F after pooling pool , b is the bias term, and σ is the activation function (such as Sigmoid), which is used to output the classification result. Data analysis shows that a bias value of 2.8904 generally yields high data accuracy. The output is processed by the activation function to obtain the final classification result. This yields the analysis value for the stroke imaging data.
[0042] The corresponding data processing process is: the pooled feature map F pool As input feature X. It is linearly transformed by weight matrix W and bias term b. Apply activation function σ to get the final classification result y CNN .
[0043] In addition, to further analyze the model's accuracy in predicting probabilities, we employed the mathematical tool cross-entropy loss. The convexity of the cross-entropy loss function helps optimization algorithms (such as gradient descent) find the global optimal solution. Through backpropagation, the gradient information of cross-entropy can effectively guide model parameter updates, accelerate the training process, improve convergence, and better adapt the model to complex medical data. The specific function is as follows:
[0044] in, is a binary representation of the actual label (such as 0 or 1), indicating whether the patient has had a stroke. is the probability predicted by the model, and N is the number of samples. By minimizing the cross-entropy loss, the model can more accurately predict the risk of stroke in patients, thereby improving the effectiveness of early detection.
[0045] In stroke analysis, the cross-entropy loss function, as a core optimization metric for classification tasks, has multiple implications for measuring model performance, handling imbalanced data, optimizing decision boundaries, providing interpretable results, and accelerating model convergence. By continuously minimizing cross-entropy, the model can gradually improve its ability to identify and predict stroke risk, providing a powerful tool for early intervention in clinical applications.
[0046] S3: Analyze structured data through recurrent neural networks to capture dynamic changes in data and extract time series features.
[0047] For example, when using recurrent neural networks (RNNs) to analyze stroke, they are often applied to time series data, such as patient physiological parameters (such as blood pressure and heart rate) and other long-term monitoring data. RNNs have time-dependent properties and can capture dynamic changes in data, making them important for early stroke prediction. The following is the specific implementation process for this step: Time series data is typically used as input. Typical data includes a patient's physiological indicators (such as blood pressure, heart rate, oxygen saturation, etc.) and past medical records. Each time step t represents a data record, which is input into the hidden layer of the RNN.
[0048] The time series data of physiological parameters X=[x1,x2,…,xt] are analyzed by recurrent neural network to capture the dynamic changes of the data. The hidden state ht is calculated as the time series feature y by the following formula: RNN : h t =σ(W h ·h t-1 +W x ·x t +B) Among them, h t is the hidden state at time step t, h t-1 is the hidden state at the previous moment, x t is the physiological parameter at the current moment, W h and W x is the weight matrix, which is used for the hidden state of the previous moment and the current input respectively; B is the bias term, which is initialized to a random number between 0.41 and 0.966 during the calculation process through calculation optimization. t Represents the data at time step t. x t is the input at the current time t, such as the blood pressure or heart rate at the current time.
[0049] S4: Multimodal fusion of spatial features and time series features to generate fused feature values.
[0050] For example, the final stroke intelligent recognition method needs to consider the imaging features and physiological features of stroke comprehensively, so it is necessary to perform a weighted sum operation on the two vectors obtained above. Specifically, using the formula Z=α•y CNN +β•y RNN Perform weighted summation on spatial features and time series features to calculate the final fusion feature value Z; α and β are weighting coefficients, indicating the importance of different features. Generally, the sum of these two coefficients should be 1. Based on theoretical data reliability and the relevance of stroke disease, image data reliability should be greater than comprehensive physiological features. Therefore, we initialized α and β to approximately 0.55 and 0.45, respectively. When verifying correlations based on actual data, we ultimately set α and β to 0.6148 and 0.3852, respectively, to maximize stroke recognition accuracy.
[0051] S5: Input the fused feature value into the classifier and output the probability of the patient suffering a stroke.
[0052] For example, the fused feature value Z is fed into the classifier for final prediction. The classifier outputs the probability of the patient having a stroke or being healthy, as follows:
[0053] Among them, P is the probability of a patient suffering a stroke, W k is the classifier weight, Z is the fused feature value, and K is the number of categories (e.g., healthy vs. stroke). Through multimodal feature fusion, the importance weights of CNN and RNN features are adjusted, combining information from image and time series data to form the final fused representation. This method can be used in multimodal stroke prediction tasks.
[0054] In this embodiment, by acquiring medical data from multiple sources, effectively processing large-scale and high-dimensional data, and utilizing deep learning algorithms, especially convolutional neural networks (CNNs) and recurrent neural networks (RNNs), to perform multimodal feature fusion and fully utilize image data and time-series medical record data, it can help doctors predict stroke in patients, improve the effectiveness of clinical decision-making, and achieve personalized medical care and resource optimization.
[0055] Specifically, the intelligent stroke identification method disclosed in this embodiment leverages big data technology to comprehensively integrate multiple medical data sources, including electronic health records, medical images, genomic data, laboratory test results, wearable device data, sensor data, and medication usage records, to create a comprehensive patient health profile. During data processing, advanced algorithms are first used to meticulously organize the health profile. Complex medical data is uniquely identified by the patient's ID number and organized in a time series format to ensure data accuracy and completeness. Subsequently, the organized data is classified to extract structured data (such as physiological parameters) and unstructured data (such as medical images). In the feature extraction stage, this method innovatively combines the advantages of convolutional neural networks (CNNs) and recurrent neural networks (RNNs). CNNs utilize advanced techniques such as convolution, ReLU activation functions, and pooling to efficiently extract spatial features from image data. RNNs utilize a recurrent connection structure to capture the dynamic changes in time series data, such as physiological parameters, and extract time series features. The application of this comprehensive technical means not only improves the efficiency and accuracy of data processing, but also enables the method to more comprehensively extract characteristic information related to stroke, providing strong support for subsequent classification and identification, and significantly improving the accuracy and efficiency of intelligent identification of stroke diseases.
[0056] like Figure 2 As shown, the following is an embodiment of the intelligent recognition system for stroke disease provided by the embodiment of the present disclosure. The system and the intelligent recognition method for stroke disease in the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiment of the intelligent recognition system for stroke disease, please refer to the embodiment of the above-mentioned intelligent recognition method for stroke disease.
[0057] A stroke disease intelligent recognition system includes: a data preprocessing module, an unstructured data analysis module, a structured data analysis module, a multimodal fusion module and a prediction module.
[0058] The data preprocessing module is used to use big data technology to collect medical data from multiple sources, organize them using the patient's ID number as a marker, extract the health records of stroke patients, preprocess the health records, organize the health records, and extract structured data and unstructured data.
[0059] The unstructured data analysis module is used to analyze unstructured data through convolutional neural networks, extract spatial features, and optimize the corresponding model parameters using the cross-entropy loss function.
[0060] The structured data analysis module is used to analyze structured data through recurrent neural networks, capture dynamic changes in data, and extract time series features.
[0061] The multimodal fusion module is used to perform multimodal fusion of spatial features and time series features to generate fused feature values.
[0062] The prediction module is used to input the fused feature value into the classifier and output the probability of the patient suffering a stroke.
[0063] The intelligent stroke disease identification system provided in this embodiment integrates multiple medical data sources through the use of big data technology and divides the data into two categories: structured and unstructured through careful preprocessing. Convolutional neural networks are used to efficiently extract the spatial features of medical images, while recurrent neural networks capture the time series changes of physiological parameters. This comprehensive technical approach not only ensures the accuracy and completeness of data processing, but also comprehensively mines characteristic information related to stroke, significantly improving the accuracy and efficiency of intelligent stroke disease identification. The application of this innovative method provides doctors with more accurate and rapid diagnostic results, helps reduce the incidence and mortality of stroke, and has important clinical value and practical significance.
[0064] Figure 3 A schematic diagram of the hardware structure of an electronic device for implementing various embodiments of the present invention.
[0065] The intelligent identification method for stroke disease provided in the embodiments of the present application can be applied to electronic devices. Those skilled in the art will understand that the electronic device structure involved in the embodiments of the present invention does not constitute a limitation on the electronic device, and the electronic device may include more or fewer components than shown, or combine certain components, or arrange components differently. In the embodiments of the present invention, electronic devices include but are not limited to laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present application described and / or required herein.
[0066] The electronic device may include a processor, an external memory interface, an internal memory, a universal serial bus (USB) interface, a charging management module, a power management module, a battery, a wireless communication module, an audio module, a speaker, a microphone, a sensor module, a button, a camera, a display, and a SIM card interface, etc.
[0067] A processor may include one or more processing units, such as a central processing unit (CPU), an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, a memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural-network processing unit (NPU). Different processing units may be independent devices or integrated into one or more processors.
[0068] The processor can be the nerve center and command center of the electronic device. The controller can generate operation control signals based on the instruction opcode and timing signal to complete the control of instruction fetching and execution.
[0069] The processor may also include a memory for storing instructions and data. In some embodiments, the memory in the processor is a cache memory. This memory can store instructions or data that the processor has just used or is reusing. If the processor needs to use the instruction or data again, it can directly call it from the memory. This avoids repeated accesses, reduces processor latency, and thus improves system efficiency.
[0070] The external memory interface can be used to connect an external memory card, such as a MicroSD card, to expand the storage capacity of an electronic device. The external memory card communicates with the processor through the external memory interface, enabling data storage. For example, files such as music and videos can be stored on the external memory card.
[0071] Internal memory can be used to store computer-executable program code, which includes instructions. The processor executes the instructions stored in the internal memory to perform various functional applications and data processing of the electronic device. The internal memory can include a program storage area and a data storage area. The internal memory can include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, universal flash storage (UFS), etc.
[0072] The wireless communication function of an electronic device can be implemented through an antenna, a wireless communication module, a modem processor, and a baseband processor.
[0073] Wireless communication modules can provide wireless communication solutions for electronic devices, including wireless local area networks (WLAN) (such as wireless fidelity (Wi-Fi) networks), Bluetooth (BT), global navigation satellite system (GNSS), frequency modulation (FM), near field communication (NFC), infrared technology (IR), etc.
[0074] Electronic devices can implement audio functions through audio modules, speakers, receivers, microphones, headphone jacks, and application processors.
[0075] Electronic devices can achieve shooting functions through ISP, camera, video codec, GPU, display and application processor.
[0076] Electronic devices can achieve display functions through GPU, display screen and application processor.
[0077] A GPU is a microprocessor for image processing that connects the display screen to the application processor. The GPU performs mathematical and geometric calculations for graphics rendering. A processor may include one or more GPUs, which execute program instructions to generate or modify display information.
[0078] The display screen is used to display images, videos, etc. The display screen includes a display panel.
[0079] The above-mentioned electronic device realizes the intelligent identification method of stroke disease in this application by integrating multiple medical data sources, carefully preprocessing classification data, using convolutional neural networks to extract spatial features and recurrent neural networks to capture time series changes, thereby achieving the beneficial effect of comprehensively improving the accuracy and efficiency of intelligent identification of stroke disease.
[0080] The storage medium provided in this application stores a program product that can implement an intelligent stroke disease identification method.
[0081] Intelligent identification methods for stroke diseases include: Using big data technology to collect medical data from multiple sources, sorting them using the patient's ID number as a marker, extracting the health records of stroke patients, pre-processing the health records, sorting the health records, and extracting structured data and unstructured data; Analyze unstructured data through convolutional neural networks, extract spatial features, and optimize corresponding model parameters using cross-entropy loss function; Analyze structured data through recurrent neural networks, capture dynamic changes in data, and extract time series features; Perform multimodal fusion of spatial features and time series features to generate fused feature values; The fused feature values are input into the classifier, and the probability of the patient suffering a stroke is output. In some possible embodiments, the intelligent stroke disease identification method disclosed herein can be implemented in the form of a program product, which includes program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps described in the above "Exemplary Method" section of this specification according to various exemplary embodiments of the present disclosure.
[0082] The storage medium of the present disclosure can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.
[0083] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for intelligent identification of stroke disease, characterized in that: include: Using big data technology to collect medical data from multiple sources, sorting them using the patient's ID number as a marker, extracting the health records of stroke patients, preprocessing the health records, sorting the health records, and extracting structured data and unstructured data; Analyze unstructured data through convolutional neural networks, extract spatial features, and optimize corresponding model parameters using cross-entropy loss function; Analyze structured data through recurrent neural networks, capture dynamic changes in data, and extract time series features; Perform multimodal fusion of spatial features and time series features to generate fused feature values; The fused feature values are input into the classifier, and the probability of the patient suffering a stroke is output.
2. The intelligent stroke identification method according to claim 1, characterized in that: The method uses big data technology to collect medical data from multiple sources, organizes them using the patient's ID number as a marker, extracts the health records of stroke patients, organizes the health records, and extracts structured data and unstructured data, including: Using big data technology to collect medical data from multiple sources, including electronic health records, medical images, laboratory test results, wearable device data, sensor data, and medication use records; The medical data is sorted using the patient's ID number as a marker, and the physical examination data and examination reports of stroke patients over the years are organized in a chronological manner to form a health file for each patient; The health records are sorted and structured data and unstructured data are extracted. The structured data includes physiological parameters, and the unstructured data includes imaging data.
3. The intelligent stroke identification method according to claim 2, characterized in that: The unstructured data is analyzed by a convolutional neural network to extract spatial features, including: The image data is analyzed through the convolutional neural network, and the convolution kernel is used to perform convolution operation on the image data using the formula F=Conv(I,K), extract local features, generate feature map F, and apply the ReLU activation function F ReLU =max(0,F) generates the activated feature map F ReLU ; Where I is the pixel value of the image data, and K is the convolution kernel; Using formula F pool =max(F ReLU ) for the activated feature map F ReLU Perform pooling operation and output the pooled feature map F pool ; Take advantage of y CNN =σ(W·X+b)The feature map F after pooling pool Mapped to the classification result, output spatial feature y CNN , as the analysis value of stroke imaging data; Among them, W is the weight matrix, X is the feature map F after pooling pool , b is the bias term, and σ is the activation function.
4. The intelligent stroke identification method according to claim 3, characterized in that: The cross entropy loss function includes: in, is the binary representation of the actual label, indicating whether the patient has a stroke. is the probability predicted by the model, and N is the number of samples.
5. The intelligent stroke identification method according to claim 4, characterized in that: The structured data is analyzed through a recurrent neural network to capture the dynamic changes of the data and extract time series features, including: Analyze the time series data of physiological parameters X=[x1,x2,…,x t ], capture the dynamic changes of data, and calculate the hidden state h by the following formula t , as the time series feature y RNN : h t =σ(W h ·h t-1 +W x ·x t +B) Among them, h t is the hidden state at time step t, h t-1 is the hidden state at the previous moment, x t is the physiological parameter at the current moment, W h and W x is the weight matrix and B is the bias term.
6. The intelligent stroke identification method according to claim 5, characterized in that: The fusion feature values generated by multimodal fusion of spatial features and time series features include: Using the formula Z=α·y CNN +β·y RNN Perform weighted summation on spatial features and time series features to calculate the final fusion feature value Z; Among them, α and β are weighting coefficients.
7. The intelligent stroke identification method according to claim 6, characterized in that: The classifier includes: Among them, P is the probability of a patient suffering a stroke, W k is the weight of the classifier, Z is the fusion feature value, and K is the number of categories.
8. A stroke disease intelligent identification system, characterized in that: The system adopts the intelligent identification method for stroke disease according to any one of claims 1 to 7; The system comprises: A data preprocessing module is used to collect medical data from multiple sources using big data technology, organize it using the patient's ID number as a marker, extract the health records of stroke patients, preprocess the health records, organize the health records, and extract structured data and unstructured data; Unstructured data analysis module, which is used to analyze unstructured data through convolutional neural networks, extract spatial features, and optimize corresponding model parameters using the cross-entropy loss function; The structured data analysis module is used to analyze structured data through recurrent neural networks, capture dynamic changes in data, and extract time series features; Multimodal fusion module, used to perform multimodal fusion of spatial features and time series features to generate fusion feature values; The prediction module is used to input the fused feature value into the classifier and output the probability of the patient suffering a stroke.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the intelligent stroke identification method according to any one of claims 1 to 7 are implemented.
10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent stroke identification method according to any one of claims 1 to 7 are implemented.