Medical big data mining method and system based on cloud storage
By using a cloud-based medical big data mining method, convolutional neural networks and converter modules are employed to extract full-time semantic features of vital signs, solving the data integration and interoperability problems in traditional solutions and enabling efficient anomaly detection and personalized treatment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHENGZHOU LANBO ELECTRONIC TECH CO LTD
- Filing Date
- 2023-09-26
- Publication Date
- 2026-04-17
AI Technical Summary
Traditional medical big data mining solutions cannot fully utilize the correlations and interactions between different types of medical data, and face challenges in data integration and interoperability, making it difficult to achieve efficient anomaly detection and personalized treatment plans.
We employ a cloud-based medical big data mining method. By collecting, storing, and analyzing patients' clinical diagnostic data, we use convolutional neural networks and converter modules to extract full-time semantic association features of vital signs, and combine this with a classifier for anomaly detection.
It improves the accuracy and robustness of anomaly detection, helps doctors quickly and accurately diagnose diseases and provide personalized treatment plans, and improves the quality of medical services.
Smart Images
Figure CN121885209A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of big data mining, and more specifically, to a method and system for mining medical big data based on cloud storage. Background Technology
[0002] Medical big data refers to the large-scale, diverse datasets generated in the medical field. This data includes clinical data (such as patient medical records and diagnostic reports), medical imaging data (such as X-rays and CT scans), genomics data, and biosensor data. With the development of medical technology and the accumulation of medical data, medical big data mining has become an important research area. The mining and analysis of medical big data can provide valuable references and support for medical decision-making.
[0003] However, due to the wide range of sources of medical data, traditional medical big data mining solutions often only analyze specific, single types of data, failing to fully utilize the correlations and interactions between different data types. Furthermore, medical data comes from diverse sources and systems, possessing different formats and structures, and traditional medical big data mining solutions often face challenges in data integration and interoperability.
[0004] Therefore, a cloud-based medical big data mining solution is desired. Summary of the Invention
[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a method and system for mining medical big data based on cloud storage. It can automatically detect and assess patients' vital signs, improving the accuracy and robustness of anomaly detection, thereby helping doctors to quickly and accurately diagnose diseases and provide personalized treatment plans.
[0006] According to one aspect of this application, a method for mining medical big data based on cloud storage is provided, comprising:
[0007] Collect clinical diagnostic data of the medical subjects being analyzed. The clinical diagnostic data includes body temperature, pulse, respiratory rate, blood oxygen saturation, and blood pressure values at multiple predetermined time points within a predetermined time period.
[0008] The clinical diagnostic data of the analyzed medical subjects are stored in a cloud storage system;
[0009] The clinical diagnostic data of the analyzed medical subject is retrieved from the cloud storage system, and correlation analysis based on time and sample dimensions is performed on the clinical diagnostic data to obtain full-time semantic correlation features of vital signs; and
[0010] Based on the full-time semantic association features of vital signs, it is determined whether there are any abnormalities in the vital signs of the analyzed medical subject.
[0011] According to another aspect of this application, a cloud storage-based medical big data mining system is provided, comprising:
[0012] The data acquisition module is used to collect clinical diagnostic data of the medical object being analyzed. The clinical diagnostic data includes body temperature, pulse, respiratory rate, blood oxygen saturation and blood pressure values at multiple predetermined time points within a predetermined time period.
[0013] The storage module is used to store the clinical diagnostic data of the analyzed medical object in a cloud storage system;
[0014] The association analysis module is used to retrieve the clinical diagnostic data of the analyzed medical object from the cloud storage system, and perform association analysis on the clinical diagnostic data of the analyzed medical object based on time and sample dimensions to obtain full-time semantic association features of vital signs; and
[0015] The anomaly analysis module is used to determine whether there are any abnormalities in the vital signs of the analyzed medical object based on the full-time semantic association features of the vital signs.
[0016] Compared with existing technologies, the medical big data mining method and system based on cloud storage provided in this application first collects clinical diagnostic data of the medical subject being analyzed. This clinical diagnostic data includes body temperature, pulse, respiratory rate, blood oxygen saturation, and blood pressure values at multiple predetermined time points within a predetermined time period. Next, the clinical diagnostic data of the medical subject being analyzed is stored in a cloud storage system. Then, the clinical diagnostic data of the medical subject being analyzed is retrieved from the cloud storage system, and correlation analysis based on time and sample dimensions is performed on the clinical diagnostic data to obtain full-time semantic correlation features of vital signs. Finally, based on the full-time semantic correlation features of vital signs, it is determined whether there are any abnormalities in the vital signs of the medical subject being analyzed. This improves the accuracy and robustness of anomaly detection. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. The following drawings are not deliberately drawn to scale according to the actual size, but are intended to show the main idea of this application.
[0018] Figure 1 This is a flowchart of a cloud storage-based medical big data mining method according to an embodiment of this application.
[0019] Figure 2This is a schematic diagram of the architecture of a cloud storage-based medical big data mining method according to an embodiment of this application.
[0020] Figure 3 This is a flowchart of sub-step S130 of the cloud storage-based medical big data mining method according to an embodiment of this application.
[0021] Figure 4 This is a flowchart of sub-step S132 of the cloud storage-based medical big data mining method according to an embodiment of this application.
[0022] Figure 5 This is a block diagram of a cloud-based medical big data mining system according to an embodiment of this application.
[0023] Figure 6 This is an application scenario diagram of the cloud storage-based medical big data mining method according to an embodiment of this application. Detailed Implementation
[0024] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are also within the scope of protection of this application.
[0025] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.
[0026] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.
[0027] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously, as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.
[0028] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.
[0029] Cloud storage technology has been widely adopted in the medical field, providing a highly reliable, scalable, and flexible storage solution. By storing medical data in a cloud storage system, centralized management, sharing, and access of data can be achieved, meeting the storage needs of large-scale medical data. When storing medical data from different sources and in different formats in a cloud storage system, this data can be uniformly converted into structured or semi-structured data, and then cleaned, normalized, deduplicated, and imputed to improve data quality and consistency.
[0030] In the process of mining and analyzing medical big data, only the necessary medical data needs to be retrieved from the cloud storage system. Leveraging the advantages of cloud computing, efficient storage, management, and processing of medical big data can be achieved, thereby providing decision support and improving the quality of medical services. Specifically, pre-processed and stored medical data can be retrieved, and data processing and analysis algorithms can be introduced in the backend to analyze and evaluate patient medical data. This enables automated detection and judgment of patient vital signs, improving the accuracy and robustness of anomaly detection, thus helping doctors to quickly and accurately diagnose diseases and provide personalized treatment plans.
[0031] Figure 1 This is a flowchart of a cloud storage-based medical big data mining method according to an embodiment of this application. Figure 2 This is a schematic diagram of the architecture of a cloud-based medical big data mining method according to an embodiment of this application. Figure 1 and Figure 2 As shown, the medical big data mining method based on cloud storage according to an embodiment of this application includes the following steps: S110, collecting clinical diagnostic data of the medical object to be analyzed, the clinical diagnostic data including body temperature, pulse, respiratory rate, blood oxygen saturation and blood pressure values at multiple predetermined time points within a predetermined time period; S120, storing the clinical diagnostic data of the medical object to be analyzed in a cloud storage system; S130, retrieving the clinical diagnostic data of the medical object to be analyzed from the cloud storage system, and performing correlation analysis based on time dimension and sample dimension on the clinical diagnostic data of the medical object to be analyzed to obtain full-time semantic correlation features of vital signs; and S140, determining whether there are any abnormalities in the vital signs of the medical object to be analyzed based on the full-time semantic correlation features of vital signs.
[0032] Specifically, in the technical solution of this application, firstly, clinical diagnostic data of the medical subject being analyzed is collected. This clinical diagnostic data includes body temperature, pulse, respiratory rate, blood oxygen saturation, and blood pressure values at multiple predetermined time points within a predetermined time period. After collecting the clinical diagnostic data of the medical subject being analyzed, the data is stored in a cloud storage system. During the storage of this data, clinical diagnostic data from different sources and formats are uniformly converted into structured or semi-structured data, and operations such as cleaning, standardization, deduplication, and missing value imputation are performed to improve data quality and consistency.
[0033] When performing the clinical diagnostic data analysis, it is necessary to retrieve the clinical diagnostic data of the analyzed medical object from the cloud storage system, and arrange the clinical diagnostic data of the analyzed medical object into a vital sign time series correlation matrix according to the time dimension and the sample dimension. This is to integrate the distribution information of the clinical diagnostic data of the analyzed medical object in the time dimension and the sample dimension, so as to fully monitor the temporal co-change of the patient's vital signs and thus more accurately detect the vital signs of the analyzed medical object.
[0034] Then, considering that during the monitoring of vital signs of the analyzed medical subject, the clinical diagnostic data includes body temperature, pulse, respiratory rate, blood oxygen saturation, and blood pressure. These data not only exhibit their own temporal dynamic changes but also have temporal synergistic correlations. Furthermore, since the fluctuations and changes in each data point over time may be relatively weak, it is difficult to fully and effectively identify and characterize them using traditional feature extraction methods. Therefore, in the technical solution of this application, the vital sign temporal correlation matrix is further matrix-segmented to obtain multiple local vital sign temporal correlation matrices, in order to better capture the local temporal correlation characteristics of the vital sign data of the analyzed medical subject. In other words, by matrix-segmenting the vital sign temporal correlation matrix, the original matrix can be divided into multiple smaller sub-matrices, each representing vital sign data within a local time period. This facilitates more granular analysis and modeling of vital sign data within different time periods, thereby more accurately capturing potential anomalies or trends.
[0035] Furthermore, the multiple local temporal correlation matrices of vital signs are respectively subjected to feature mining through a local temporal feature extractor of vital signs based on a convolutional neural network model, so as to extract the local temporal correlation feature information between each clinical diagnostic data parameter of the analyzed medical object in a local time period, that is, the local temporal features of the vital signs of the analyzed medical object, thereby obtaining multiple local temporal correlation feature vectors of vital signs.
[0036] Furthermore, considering that the temporal synergistic correlation features between the various clinical diagnostic parameter data within a local time period have a temporally holistic correlation relationship throughout the entire predetermined time period, in order to enable full-time monitoring and analysis of the vital signs of the analyzed medical subject, the technical solution of this application further encodes the multiple local temporal correlation feature vectors of vital signs through a context encoder based on a converter module to extract the global contextual correlation feature information of the local temporal synergistic features of the analyzed medical subject's vital signs, thereby obtaining a full-time contextual semantic feature vector of vital signs.
[0037] Accordingly, such as Figure 3As shown, the clinical diagnostic data of the analyzed medical object is retrieved from the cloud storage system, and correlation analysis based on the time and sample dimensions is performed on the clinical diagnostic data of the analyzed medical object to obtain the full-time semantic correlation features of vital signs. This includes: S131, arranging the clinical diagnostic data of the analyzed medical object into a vital sign time-series correlation matrix according to the time and sample dimensions; S132, performing local semantic correlation analysis on the vital sign time-series correlation matrix to obtain multiple local vital sign time-series correlation features; and S133, performing context-related encoding on the multiple local vital sign time-series correlation features to obtain the full-time semantic correlation features of vital signs. It should be understood that the purpose of step S131 is to organize the clinical diagnostic data into a matrix, where rows represent the time dimension and columns represent the sample dimension. This allows the relationships between different time points and different samples to be represented in matrix form, providing a data foundation for subsequent analysis. In step S132, the temporal correlation matrix of vital signs is analyzed to identify local semantic correlations. These local temporal correlation features can help reveal patterns, trends, or anomalies in the vital sign data. By analyzing these local correlation features, more detailed information can be obtained, providing a basis for subsequent analysis and inference. In step S133, multiple local temporal correlation features are context-correlatedly encoded to obtain the full temporal semantic correlation features of vital signs. By integrating different local features, a more comprehensive and integrated set of vital sign features can be obtained, which helps in understanding and analyzing the physiological state and disease development trends of medical subjects. In summary, the purpose of these steps is to obtain clinical diagnostic data from the cloud storage system and extract the full temporal semantic correlation features of vital signs from it using correlation analysis methods. These features can be used for more in-depth data analysis, model building, and decision support, contributing to health monitoring, disease diagnosis, and treatment of medical subjects.
[0038] More specifically, such as Figure 4As shown, in step S132, local semantic association analysis is performed on the vital sign temporal correlation matrix to obtain multiple local temporal correlation features of vital signs, including: S1321, matrix segmentation of the vital sign temporal correlation matrix to obtain multiple local temporal correlation matrices of vital signs; and S1322, the multiple local temporal correlation matrices of vital signs are respectively processed by a vital sign local temporal feature extractor based on a convolutional neural network model to obtain multiple local temporal correlation feature vectors of vital signs as the multiple local temporal correlation features of vital signs. It is worth mentioning that a convolutional neural network (CNN) is a deep learning model mainly used to process data with a grid structure, such as images and temporal data. In the above description, the convolutional neural network is used as a local temporal feature extractor of vital signs to extract multiple local temporal correlation features of vital signs. Specifically, step S1321 divides the vital sign time-series correlation matrix into multiple sub-matrices, each corresponding to a local time-series correlation region. This decomposes the vital sign data into multiple local regions for analysis, capturing more granular correlation features. Step S1322 inputs each local time-series correlation matrix into a feature extractor based on a convolutional neural network. Through operations such as convolution and pooling, feature vectors for each local region are extracted. These feature vectors can represent the correlation features within that region, such as time series patterns and trends. In other words, the convolutional neural network model is used in this system to extract local time-series correlation features of vital sign data. By segmenting the vital sign time-series correlation matrix and using a convolutional neural network for feature extraction, correlation features in different regions can be captured, providing more informative feature representations for subsequent analysis and encoding.
[0039] More specifically, in step S133, context-related encoding is performed on the multiple local temporal correlation features of vital signs to obtain the full temporal semantic correlation features of vital signs. This includes: passing the multiple local temporal correlation feature vectors of vital signs through a context encoder based on a transformer module to obtain a full temporal contextual semantic feature vector of vital signs as the full temporal semantic correlation features of vital signs. It is worth noting that the transformer module is a deep learning model used for sequence modeling and self-attention mechanisms. The transformer module is used as a context encoder to encode multiple local temporal correlation feature vectors of vital signs into a full temporal contextual semantic feature vector of vital signs, thereby obtaining the full temporal semantic correlation features of vital signs. Specifically, multiple local temporal correlation feature vectors of vital signs are passed through a context encoder based on the transformer module. This step takes multiple local temporal correlation feature vectors as input and encodes them through the transformer module. The transformer module uses a self-attention mechanism to focus on and weight different elements in the input sequence, thereby capturing the contextual relationships between elements. Through multiple layers of self-attention and feedforward neural network layers, the transformer module can perform global context encoding on the sequence data. The full-temporal contextual semantic feature vector of vital signs is obtained as the full-temporal semantic association feature of the vital signs. After encoding processing by the converter module, each local temporal association feature vector of vital signs is integrated into a single full-temporal contextual semantic feature vector. This feature vector can be seen as a comprehensive representation of the entire vital sign data sequence, containing global semantic association information. In summary, the converter module in this system is used to perform contextual encoding on multiple local temporal association features of vital signs to obtain the full-temporal semantic association features of vital signs. Through a self-attention mechanism and a multi-layered neural network structure, the converter module can capture global association information in the vital sign data sequence, providing a more comprehensive and accurate feature representation for subsequent analysis and applications.
[0040] Subsequently, the full-temporal contextual semantic feature vector of vital signs is processed by a classifier to obtain a classification result. This classification result indicates whether there are any abnormalities in the vital signs of the analyzed medical subject. In other words, classification processing is performed using the global correlation feature information between the local temporal collaborative features of the analyzed medical subject's vital signs. This enables automatic monitoring and anomaly detection of the patient's vital signs, thereby achieving efficient storage, management, and processing of medical big data. This improves the accuracy and robustness of anomaly detection, helping doctors to quickly and accurately diagnose diseases and provide personalized treatment plans, thus providing decision support and improving the quality of medical services.
[0041] Accordingly, based on the full-time semantic association features of vital signs, determining whether there are abnormalities in the vital signs of the analyzed medical object includes: passing the full-time contextual semantic feature vector of vital signs through a classifier to obtain a classification result, wherein the classification result is used to indicate whether there are abnormalities in the vital signs of the analyzed medical object.
[0042] More specifically, the vital signs full-time context semantic feature vector is passed through a classifier to obtain a classification result, which is used to indicate whether there are abnormalities in the vital signs of the analyzed medical object. This includes: using the fully connected layer of the classifier to fully connect and encode the vital signs full-time context semantic feature vector to obtain an encoded classification feature vector; and inputting the encoded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.
[0043] In other words, in the technical solution disclosed herein, the classifier's labels include "abnormal vital signs of the analyzed medical object" (first label) and "abnormal vital signs of the analyzed medical object" (second label). The classifier determines which label the full-time contextual semantic feature vector of the vital signs belongs to using a soft-maximum function. It is worth noting that the first label p1 and the second label p2 here do not contain artificially defined concepts. In fact, during the training process, the computer model does not have the concept of "whether the vital signs of the analyzed medical object are abnormal"; it simply has two classification labels and outputs the probability of the feature under these two labels, i.e., the sum of p1 and p2 is one. Therefore, the classification result of whether the vital signs of the analyzed medical object are abnormal is actually transformed into a binary probability distribution conforming to natural laws through the classification labels. Essentially, it uses the physical meaning of the natural probability distribution of the labels, rather than the linguistic meaning of "whether the vital signs of the analyzed medical object are abnormal."
[0044] As you can understand, the role of a classifier is to learn classification rules and classifiers using given categories and known training data, and then classify (or predict) unknown data. Logistic regression and SVM are commonly used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are needed to form the multi-class classification. However, this is prone to errors and inefficient. A commonly used multi-class classification method is the Softmax classification function.
[0045] Furthermore, in the technical solution of this application, the medical big data mining method based on cloud storage further includes a training step: for training the vital signs local temporal feature extractor based on the convolutional neural network model, the context encoder based on the converter module, and the classifier.
[0046] More specifically, the training steps include: acquiring training data, which includes training clinical diagnostic data of the analyzed medical object and true values indicating whether the vital signs of the analyzed medical object are abnormal; storing the training clinical diagnostic data of the analyzed medical object in the cloud storage system; retrieving the training clinical diagnostic data of the analyzed medical object from the cloud storage system and arranging the training clinical diagnostic data of the analyzed medical object into a training vital sign time-series correlation matrix according to the time dimension and the sample dimension; performing matrix partitioning on the training vital sign time-series correlation matrix to obtain multiple training vital sign local time-series correlation matrices; and respectively applying the multiple training vital sign local time-series correlation matrices to the vital sign local time-series correlation matrix based on the convolutional neural network model. The system employs a sequence feature extractor to obtain multiple training vital sign local temporal correlation feature vectors; it then passes these multiple training vital sign local temporal correlation feature vectors through a context encoder based on a converter module to obtain a training vital sign full temporal contextual semantic feature vector; finally, it passes the training vital sign full temporal contextual semantic feature vector through a classifier to obtain a classification loss function value; and finally, it trains the vital sign local temporal feature extractor, the context encoder based on the converter module, and the classifier based on the classification loss function value and through directional propagation of gradient descent, wherein, in each iteration of the training, the training vital sign full temporal contextual semantic feature vector undergoes weight space iterative recursive directional proposal optimization.
[0047] Specifically, in the technical solution of this application, after the multiple training vital sign local temporal correlation matrices are respectively processed by a vital sign local temporal feature extractor based on a convolutional neural network model, each training vital sign local temporal correlation feature vector can express the temporal-sample cross-dimensional correlation features in the local temporal-sample domain. Thus, after the multiple training vital sign local temporal correlation feature vectors are processed by a context encoder based on a converter module, the contextual correlation of the temporal-sample cross-dimensional correlation features between the local temporal-sample domain in the global temporal-sample domain is further extracted. This makes the training vital sign full temporal contextual semantic feature vector have cross-dimensional correlation feature representations in both the local and global temporal-sample domains. While this improves the cross-dimensional correlation feature representation effect of the training vital sign full temporal contextual semantic feature vector in the full temporal-sample domain, it also makes it difficult for the weight matrix of the classifier to converge relative to the class labels belonging to the predetermined domain dimension when it is classified by the classifier, thus affecting the training effect of the classifier.
[0048] Therefore, when classifying the training vital signs full-time context semantic feature vector using a classifier, the applicant of this application performs weight space iterative recursive directed proposal optimization on the training vital signs full-time context semantic feature vector at each iteration.
[0049] Accordingly, in a specific example, at each iteration of the training, the training vital signs full-time context semantic feature vector is subjected to weight space iterative recursive directed proposal optimization using the following optimization formula to obtain the optimized training vital signs full-time context semantic feature vector; wherein, the optimization formula is:
[0050]
[0051]
[0052]
[0053] Where M1 and M2 are the weight matrices of the previous and current iterations, respectively. In the first iteration, different initialization strategies are used to set M1 and M2 (e.g., M1 is set as the identity matrix while M2 is set as the mean diagonal matrix of the feature vectors to be classified). V c It is the training vital signs full-time contextual semantic feature vector. ⊙ represents matrix multiplication, and ⊙ represents positional multiplication. V' represents vector addition, and exp(·) represents the exponentiation operation of a vector, which means calculating the natural exponent function value raised to the power of the eigenvalues at each position in the vector. cThis represents the optimized training of the full-time contextual semantic feature vector of vital signs.
[0054] Here, the iterative and recursive directional proposal optimization of the weight space can be achieved by using the initial training vital sign full-time contextual semantic feature vector V to be classified. c As an anchor point, the corresponding full-time contextual semantic feature vector V of the training vital signs is obtained by iteratively applying the weight matrix within the weight space. c Anchor footprints are obtained under different domain spatial feature distribution dimensions through different spatial transformation directions. These footprints serve as oriented proposals for iterative recursion in the weight space, thereby improving the class confidence and local accuracy of the weight matrix convergence based on the predicted proposals. This enhances the training effect of the trained vital sign full-time context semantic feature vectors through the classifier. This enables efficient storage, management, and processing of medical big data, improves the accuracy and robustness of anomaly detection, helps doctors quickly and accurately diagnose diseases, provides personalized treatment plans, offers decision support, and improves the quality of medical services.
[0055] In summary, the cloud storage-based medical big data mining method based on the embodiments of this application has been clarified, which can improve the accuracy and robustness of anomaly detection.
[0056] Figure 5 This is a block diagram of a cloud-based medical big data mining system 100 according to an embodiment of this application. Figure 5 As shown, the cloud-based medical big data mining system 100 according to an embodiment of this application includes: a data acquisition module 110, used to acquire clinical diagnostic data of the medical object being analyzed, the clinical diagnostic data including body temperature, pulse, respiratory rate, blood oxygen saturation and blood pressure values at multiple predetermined time points within a predetermined time period; a storage module 120, used to store the clinical diagnostic data of the medical object being analyzed in a cloud storage system; an association analysis module 130, used to retrieve the clinical diagnostic data of the medical object being analyzed from the cloud storage system and perform association analysis on the clinical diagnostic data of the medical object being analyzed based on time dimension and sample dimension to obtain full-time semantic association features of vital signs; and an anomaly analysis module 140, used to determine whether there are any abnormalities in the vital signs of the medical object being analyzed based on the full-time semantic association features of vital signs.
[0057] In one example, in the aforementioned cloud-based medical big data mining system 100, the association analysis module 130 includes: a matrixing unit for arranging the clinical diagnostic data of the analyzed medical object into a vital sign temporal association matrix according to the time dimension and the sample dimension; a local semantic association analysis unit for performing local semantic association analysis on the vital sign temporal association matrix to obtain multiple vital sign local temporal association features; and a context association encoding unit for performing context association encoding on the multiple vital sign local temporal association features to obtain the vital sign full temporal semantic association features.
[0058] Here, those skilled in the art will understand that the specific functions and operations of each module in the cloud storage-based medical big data mining system 100 have been referenced above. Figures 1 to 4 The description of cloud storage-based medical big data mining methods is detailed here, and therefore, its repeated description will be omitted.
[0059] As described above, the cloud-based medical big data mining system 100 according to embodiments of this application can be implemented in various wireless terminals, such as servers with cloud-based medical big data mining algorithms. In one example, the cloud-based medical big data mining system 100 according to embodiments of this application can be integrated into a wireless terminal as a software module and / or hardware module. For example, the cloud-based medical big data mining system 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the cloud-based medical big data mining system 100 can also be one of many hardware modules of the wireless terminal.
[0060] Alternatively, in another example, the cloud-based medical big data mining system 100 and the wireless terminal can also be separate devices, and the cloud-based medical big data mining system 100 can connect to the wireless terminal via wired and / or wireless networks and transmit interactive information in accordance with an agreed data format.
[0061] Figure 6 This diagram illustrates an application scenario of the cloud-based medical big data mining method according to an embodiment of this application. Figure 6 As shown, in this application scenario, firstly, clinical diagnostic data of the medical subject being analyzed is collected (e.g., Figure 6 As shown in D), the clinical diagnostic data includes body temperature, pulse, respiratory rate, blood oxygen saturation, and blood pressure values at multiple predetermined time points within a predetermined time period. Then, the clinical diagnostic data of the analyzed medical subject is input into a server deployed with a cloud-based medical big data mining algorithm (e.g., ...). Figure 6As shown in S), the server is able to use the cloud-based medical big data mining algorithm to process the clinical diagnostic data of the analyzed medical object to obtain a classification result indicating whether the vital signs of the analyzed medical object are abnormal.
[0062] Furthermore, those skilled in the art will understand that aspects of this application can be described and illustrated through several patentable types or situations, including any new and useful combination of processes, machines, products, or substances, or any new and useful improvements thereof. Accordingly, aspects of this application can be implemented entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. All of the above hardware or software may be referred to as a “data block,” “module,” “engine,” “unit,” “component,” or “system.” Furthermore, aspects of this application may manifest as a computer product located on one or more computer-readable media, the product including computer-readable program code.
[0063] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in a common dictionary shall be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and not as having an idealized or highly formalized meaning, unless expressly defined herein.
[0064] The foregoing description is illustrative of the invention and should not be construed as limiting it. Although several exemplary embodiments of the invention have been described, those skilled in the art will readily understand that many modifications can be made to the exemplary embodiments without departing from the novel teachings and advantages of the invention. Therefore, all such modifications are intended to be included within the scope of the invention as defined in the claims. It should be understood that the foregoing description is illustrative of the invention and should not be construed as limiting it to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The invention is defined by the claims and their equivalents.
Claims
1.A cloud storage-based medical big data mining method, characterized by, include: Collect clinical diagnostic data of the medical subjects being analyzed. The clinical diagnostic data includes body temperature, pulse, respiratory rate, blood oxygen saturation, and blood pressure values at multiple predetermined time points within a predetermined time period. The clinical diagnostic data of the analyzed medical subjects are stored in a cloud storage system; The clinical diagnostic data of the analyzed medical object is retrieved from the cloud storage system, and the clinical diagnostic data of the analyzed medical object is subjected to correlation analysis based on the time dimension and sample dimension to obtain the full-time semantic correlation features of vital signs. as well as Based on the full-time semantic association features of vital signs, it is determined whether there are any abnormalities in the vital signs of the analyzed medical subject. 2.The cloud storage-based medical big data mining method of claim 1, wherein, The clinical diagnostic data of the analyzed medical object is retrieved from the cloud storage system, and correlation analysis based on the time and sample dimensions is performed on the clinical diagnostic data of the analyzed medical object to obtain the full-time semantic correlation features of vital signs, including: The clinical diagnostic data of the analyzed medical subjects are arranged into a vital signs time-series correlation matrix according to the time dimension and the sample dimension. Local semantic association analysis is performed on the aforementioned vital sign temporal association matrix to obtain multiple local temporal association features of vital signs; and Context-associative encoding is performed on the local temporal correlation features of the multiple vital signs to obtain the full temporal semantic correlation features of the vital signs. 3.The cloud storage-based medical big data mining method of claim 2, wherein, Local semantic association analysis was performed on the vital sign temporal association matrix to obtain multiple local temporal association features of vital signs, including: The vital signs temporal correlation matrix is partitioned to obtain multiple local temporal correlation matrices of vital signs; and The multiple local temporal correlation matrices of vital signs are respectively processed by a local temporal feature extractor based on a convolutional neural network model to obtain multiple local temporal correlation feature vectors of vital signs, which are used as the local temporal correlation features of the multiple vital signs. 4.The cloud storage-based medical big data mining method of claim 3, wherein, Context-associative encoding is performed on the local temporal correlation features of the multiple vital signs to obtain the full temporal semantic correlation features of the vital signs, including: The multiple local temporal correlation feature vectors of vital signs are passed through a context encoder based on a converter module to obtain the full temporal contextual semantic feature vectors of vital signs, which are then used as the full temporal semantic correlation features of vital signs. 5.The cloud storage-based medical big data mining method of claim 4, wherein, Based on the full-time semantic association features of vital signs, determine whether there are any abnormalities in the vital signs of the analyzed medical subject, including: The full-time contextual semantic feature vector of vital signs is passed through a classifier to obtain a classification result, which is used to indicate whether there are abnormalities in the vital signs of the analyzed medical object. 6.The cloud storage-based medical big data mining method of claim 5, wherein, It also includes a training step: for training the vital signs local temporal feature extractor based on the convolutional neural network model, the context encoder based on the converter module, and the classifier. 7.The cloud storage-based medical big data mining method of claim 6, wherein, The training steps include: Acquire training data, which includes training clinical diagnostic data of the medical subjects being analyzed, and the actual values of whether the vital signs of the medical subjects being analyzed are abnormal. The training clinical diagnostic data of the analyzed medical subjects are stored in the cloud storage system; The training clinical diagnostic data of the analyzed medical object is retrieved from the cloud storage system, and the training clinical diagnostic data of the analyzed medical object is arranged into a training vital sign time series correlation matrix according to the time dimension and the sample dimension. The training vital signs temporal correlation matrix is partitioned to obtain multiple training vital signs local temporal correlation matrices; The multiple training vital sign local temporal correlation matrices are respectively passed through the vital sign local temporal feature extractor based on the convolutional neural network model to obtain multiple training vital sign local temporal correlation feature vectors; The multiple local temporal correlation feature vectors of training vital signs are passed through the context encoder based on the converter module to obtain the full temporal contextual semantic feature vectors of training vital signs. The training vital signs full-time contextual semantic feature vector is passed through the classifier to obtain the classification loss function value; and Based on the classification loss function value and through gradient descent directional propagation, the vital sign local temporal feature extractor based on the convolutional neural network model, the context encoder based on the converter module, and the classifier are trained. In each iteration of the training, the trained vital sign full temporal context semantic feature vector is subjected to weight space iterative recursive directed proposal optimization. 8.The cloud storage-based medical big data mining method of claim 7, wherein, In each iteration of the training, the training vital signs full-time context semantic feature vector is subjected to weight space iterative recursive directed proposal optimization using the following optimization formula to obtain the optimized training vital signs full-time context semantic feature vector. The optimization formula is as follows: where M1 and M2 are the weight matrix of last and current iteration respectively, V c is the training vital signs full time series context semantic feature vector, denotes matrix multiplication, and denotes point-wise multiplication, denotes vector addition, and exp(·) denotes the exponential operation of a vector, which means calculating the natural exponential function value with the feature value of each position in the vector, V' c denotes the optimized training vital signs full time series context semantic feature vector. 9.A cloud storage-based medical big data mining system, characterized in that, include: The data acquisition module is used to collect clinical diagnostic data of the medical object being analyzed. The clinical diagnostic data includes body temperature, pulse, respiratory rate, blood oxygen saturation and blood pressure values at multiple predetermined time points within a predetermined time period. The storage module is used to store the clinical diagnostic data of the analyzed medical object in a cloud storage system; The association analysis module is used to retrieve the clinical diagnostic data of the analyzed medical object from the cloud storage system, and perform association analysis on the clinical diagnostic data of the analyzed medical object based on the time dimension and the sample dimension to obtain the full-time semantic association features of vital signs. as well as The anomaly analysis module is used to determine whether there are any abnormalities in the vital signs of the analyzed medical object based on the full-time semantic association features of the vital signs. 10.The cloud storage based medical big data mining system according to claim 9, wherein, The correlation analysis module includes: The matrix unit is used to arrange the clinical diagnostic data of the analyzed medical object into a vital sign time-series correlation matrix according to the time dimension and the sample dimension. A local semantic association analysis unit is used to perform local semantic association analysis on the vital sign temporal association matrix to obtain multiple local temporal association features of vital signs; and The context association encoding unit is used to perform context association encoding on the local temporal association features of the multiple vital signs to obtain the full temporal semantic association features of the vital signs.