Health medical data quality detection system and method

Abstract

The invention relates to the field of health and medical treatment, and particularly discloses a health and medical data quality detection system and method.The health and medical data quality detection method comprises the steps that firstly, a medical data table containing basic data of a doctor-seeing patient, a medical insurance data table and a doctor-seeing data table is obtained, then feature extraction and correlation analysis are conducted on the medical data table through the deep learning technology, and finally a classifier is used for classifying the medical data table; therefore, dependence on manual inspection is reduced, financial loss caused by data abnormality is further reduced, and reliability and efficiency of medical services are improved.

Description

Technical Field

[0001] The present application relates to the field of health care, and more specifically, to a health care data quality detection system and method. Background Art

[0002] In the field of medical data management, with the improvement of the social security system and the expansion of medical insurance coverage, the amount of medical data has increased dramatically, and it is particularly important to conduct quality inspections on health and medical data. However, the data generated by different medical institutions exist in different formats and storage methods, which increases the difficulty of data integration and quality control. Medical data is scattered in multiple tables and may be missing or wrong, which will affect the accounting and planning of medical insurance funds.

[0003] Existing assessment schemes for health and medical data quality mainly rely on manual work, which involves manually checking and evaluating the data in each data table. Errors are prone to occur during the check, resulting in inaccurate and inefficient assessments.

[0004] Therefore, a health care data quality detection system and method are desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides a health care data quality detection system and method, which first obtains a medical data table including basic data of patients, a medical insurance data table and a medical data table, and then uses deep learning technology to perform feature extraction and association analysis, and finally uses a classifier to determine whether to issue an abnormal medical data quality warning, thereby reducing reliance on manual inspections, thereby reducing financial losses caused by data anomalies, and improving the reliability and efficiency of medical services.

[0006] According to one aspect of the present application, a health care data quality detection system is provided, which includes:

[0007] The health and medical data acquisition module is used to obtain the medical data table including the basic data of the patients, the medical insurance data table and the medical treatment data table;

[0008] A health care data processing module, used to extract the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector from the medical data table containing the basic data of the patient, the medical insurance data table and the medical data table;

[0009] The medical data quality analysis module is used to determine whether to issue a warning of abnormal medical data quality based on the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector.

[0010] According to another aspect of the present application, a health care data quality detection method is provided, which includes:

[0011] Obtain a medical data table containing basic patient data, medical insurance data table, and medical treatment data table;

[0012] Extracting a patient medical data feature word granularity association feature vector and a patient medical data semantic feature vector from the medical data table including the basic data of the patient, the medical insurance data table and the medical treatment data table;

[0013] Based on the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector, it is determined whether to issue a medical data quality abnormality warning.

[0014] Compared with the prior art, the present application provides a health and medical data quality detection system and method, which first obtains a medical data table including basic data of patients, a medical insurance data table and a medical data table, and then uses deep learning technology to perform feature extraction and association analysis. Finally, a classifier is used to determine whether to issue an abnormal medical data quality warning, thereby reducing reliance on manual inspections, thereby reducing financial losses caused by data anomalies and improving the reliability and efficiency of medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 is a block diagram of a health care data quality detection system according to an embodiment of the present application.

[0017] Figure 2 4 is a block diagram of a health and medical data processing module in a health and medical data quality detection system according to an embodiment of the present application.

[0018] Figure 3 4 is a block diagram of a medical data semantic feature encoding unit in a health care data quality detection system according to an embodiment of the present application.

[0019] Figure 4 4 is a block diagram of a medical data quality analysis module in a health care data quality detection system according to an embodiment of the present application.

[0020] Figure 5 Flow chart of a health care data quality detection method according to an embodiment of the present application.

[0021] Figure 6 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0023] Figure 1 FIG. 1 is a block diagram of a health care data quality detection system according to an embodiment of the present application. Figure 1 As shown, the health and medical data quality detection system 100 according to the embodiment of the present application includes: a health and medical data acquisition module 110, which is used to acquire a medical data table including basic data of visiting patients, a medical insurance data table and a medical data table; a health and medical data processing module 120, which is used to extract a patient medical data feature word granularity-associated feature vector and a patient medical data semantic feature vector from the medical data table including the basic data of visiting patients, the medical insurance data table and the medical data table; a medical data quality analysis module 130, which is used to determine whether to issue a medical data quality abnormality warning based on the patient medical data feature word granularity-associated feature vector and the patient medical data semantic feature vector.

[0024] In the above-mentioned health and medical data quality detection system 100, the health and medical data acquisition module 110 is used to obtain a medical data table including basic patient data, a medical insurance data table and a medical data table. It should be understood that in the field of medical data management, with the continuous improvement of the social security system and the expansion of medical insurance coverage, the amount of medical data is growing rapidly, so it is becoming more and more critical to perform quality detection on health and medical data. However, due to the different data formats and storage methods generated by different medical institutions, data integration and quality control face greater challenges. Medical data are usually scattered in multiple tables and may be missing or erroneous, which in turn affects the accounting and planning of medical insurance funds. At present, the method of evaluating the quality of health and medical data mainly relies on manual review. Manually checking each data table is not only prone to errors, but also leads to inefficient evaluation process. Therefore, in the technical solution of the present application, by obtaining a medical data table containing the basic data of the patients, the medical insurance data table and the medical data table, and combining it with deep learning technology, it is automatically determined whether to issue an abnormal medical data quality warning, so as to reduce reliance on manual inspections, thereby reducing financial losses caused by data anomalies and improving the reliability and efficiency of medical services.

[0025] Specifically, the basic data of patients visiting the hospital contains the basic personal information of the patients, such as name, age, gender, contact information and medical history. Understanding this basic information helps to establish the basic files of patients and ensure the accuracy and consistency of the data. The medical insurance data table records the information related to medical insurance, including medical insurance card number, reimbursement record, expense category, etc. Medical insurance data is crucial for calculating medical insurance funds and formulating reasonable reimbursement policies. The medical data table contains detailed medical service data such as the patient's medical records, treatment process, and medication information. By integrating these three types of data, the health and medical data quality detection system can conduct a comprehensive analysis, identify anomalies, omissions or errors in the data, and ensure the accuracy and completeness of medical data. This not only improves the quality and reliability of the data, but also provides a scientific basis for the management of medical insurance funds and the optimization of medical services.

[0026] In the above-mentioned health and medical data quality detection system 100, the health and medical data processing module 120 is used to extract the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector from the medical data table containing the basic data of the patient, the medical insurance data table and the medical data table. It should be understood that the feature word granularity association feature vector provides the details and relationship perspective of the data, while the semantic feature vector provides the actual meaning and background information of the data. By combining the two, medical data analysis can not only improve the accuracy and efficiency of data processing, but also deeply understand the patient's health status and medical needs, thereby improving the overall quality and efficiency of medical services.

[0027] Figure 2 FIG. 1 is a block diagram of a health and medical data processing module in a health and medical data quality detection system according to an embodiment of the present application. Figure 2 As shown, in a specific embodiment of the present application, the health care data processing module 120 includes: a medical data word granularity feature encoding unit 121, which is used to perform word granularity feature encoding on the medical data table containing the basic data of the patient, the medical insurance data table and the medical data table to obtain the patient medical data feature word granularity associated feature vector; a medical data semantic feature encoding unit 122, which is used to perform semantic feature encoding on the medical data table containing the basic data of the patient, the medical insurance data table and the medical data table to obtain the patient medical data semantic feature vector.

[0028] It should be understood that the main purpose of word-granular feature encoding is to convert each word or term in the data into a quantifiable feature vector to capture and express the details and relationships of the data. Medical data usually involves complex terminology and coding systems, such as disease diagnosis codes, drug names, cost classifications, etc. By encoding these terms with word-granular features, they can be converted into standardized numerical representations, allowing computer systems to process and analyze these data. This encoding method ensures the structuring of the data and facilitates subsequent processing and analysis. In this way, text descriptions can be effectively converted into machine-readable feature vectors, thereby reducing human errors in the analysis process and improving the accuracy of data processing.

[0029] Furthermore, medical data involves a large number of professional terms and complex descriptions, such as diagnosis results, treatment plans, drug information, etc. These terms may have different meanings in different contexts, and traditional vocabulary encoding methods are difficult to capture these subtle differences at the semantic level. Through semantic feature encoding, the text information in medical data can be converted into feature vectors at the semantic level. These vectors can reflect the real relationship and context between terms, thereby improving the accuracy of data analysis.

[0030] In a specific embodiment of the present application, the medical data word granularity feature encoding unit 121 includes: segmenting the medical data table containing the basic data of the patient, the medical insurance data table and the medical data table to obtain multiple patient medical data word sequences; passing the multiple patient medical data word sequences through a converter-based patient medical data text convolutional neural network to obtain multiple patient medical data feature vectors; passing the multiple patient medical data feature vectors through a patient medical data feature word granularity association bidirectional long short-term memory model to obtain the patient medical data feature word granularity association feature vector.

[0031] It should be understood that medical data usually contains a large amount of free text descriptions, such as symptom descriptions, doctor's advice, medical records, etc. Direct analysis of these long texts can be very complex and inefficient. By segmenting the text into words or phrases, the information can be converted into a structured data form, making it easier for computer systems to process and analyze the information.

[0032] Furthermore, processing multiple patient medical data word sequences through a transformer-based patient medical data text convolutional neural network (CNN) is an effective way to improve data representation and analysis capabilities. This process combines the advantages of transformers (such as BERT) and convolutional neural networks, providing strong support for in-depth analysis of medical data. Among them, the transformer-based model can capture the contextual relationship in the word sequence. Through the self-attention mechanism, the transformer can fully understand the meaning of words in different contexts, which is particularly important for medical data because medical terms and descriptions often have specific contextual dependencies. Among them, the convolutional neural network can effectively extract local features when processing word sequences. The convolution layer performs convolution operations on the text through a sliding window, which can capture local patterns of word combinations, such as the association between diseases and symptoms, descriptions of treatment plans, etc. This local feature extraction helps to identify key information in the text and improve the expressiveness of data features. Combined with the global context understanding of the transformer, the combination of the transformer and the convolutional neural network can take advantage of the advantages of both: the transformer captures the global context and the convolutional network extracts local features. This comprehensive processing can better cope with the complexity and diversity in medical data, thereby improving the expressiveness of the feature vector and the prediction accuracy of the model. Specifically, the embedding layer of the converter-based patient medical data text convolutional neural network is used to map each patient medical data word in the patient medical data word sequence into a word embedding vector to obtain a sequence of patient medical data word embedding vectors; and the converter-based Bert model of the converter-based patient medical data text convolutional neural network is used to perform global contextual semantic encoding on the sequence of patient medical data word embedding vectors to obtain multiple patient medical data feature vectors.

[0033] Furthermore, multiple patient medical data feature vectors are processed through the patient medical data feature word granularity association bidirectional long short-term memory (Bi-LSTM) model to obtain patient medical data feature word granularity association feature vectors, which combines the powerful sequence modeling ability of Bi-LSTM and the detailed capture of word granularity features, and can effectively capture the complex patterns and relationships in medical data. Among them, the Bi-LSTM model can simultaneously consider the forward and backward contextual information of the data. In medical data, the relationship between symptoms, diagnoses, and treatment plans is usually time series and context-related. The traditional LSTM model only considers forward information, while the Bi-LSTM can capture these complex relationships more comprehensively by processing data in both directions of the sequence. In addition, word granularity feature association can improve the fine-grained expression of features. Features in medical data often involve subtle terms and complex relationships between terms. By mapping feature vectors to word granularity feature associations, the Bi-LSTM model can capture these fine-grained connections. Bi-LSTM can process long time series data and adapt to complex dependencies in medical data. Medical records often involve data over a long period of time, such as long-term course of illness, treatment records, etc. Bi-LSTM can effectively process these long-term series data through its long-term and short-term memory mechanism, capture long-term dependent information, and thus improve the expressive power of feature vectors and the prediction accuracy of the model. The feature vector feature representation obtained can improve the accuracy and practicality of the analysis results and provide more reliable support for medical decision-making.

[0034] Figure 3 FIG. 1 is a block diagram of a medical data semantic feature encoding unit in a health care data quality detection system according to an embodiment of the present application. Figure 3 As shown, in a specific embodiment of the present application, the medical data semantic feature encoding unit 122 includes: a medical data preprocessing subunit 1221, which is used to perform data preprocessing on the medical data table including the basic data of the patient, the medical insurance data table and the medical data table to obtain the patient medical data word embedding matrix; a patient medical data convolution feature encoding subunit 1222, which is used to perform convolution feature encoding on the patient medical data word embedding matrix to obtain the patient medical data semantic feature vector.

[0035] It should be understood that medical data usually contains a large amount of unstructured and semi-structured information, such as free text medical records, diagnosis descriptions, and treatment plans. This information is difficult to use directly in machine learning models in its original form. Through data preprocessing, this information can be converted into a standardized format and word embedding operations can be performed to convert vocabulary into dense vector representations. The word embedding matrix can not only capture the semantic relationship between words, but also map high-dimensional vocabulary data to a low-dimensional vector space.

[0036] Furthermore, convolutional feature encoding is performed on the word embedding matrix of the patient's medical data to obtain the semantic feature vector of the patient's medical data, which can help to more accurately represent the semantic information of the medical data by extracting and capturing local patterns and features in the word embedding matrix. Among them, the convolutional neural network (CNN) can effectively extract local features when processing the word embedding matrix. The information in medical data often exists in specific vocabulary combinations, such as the symptom description of the disease, keywords in the diagnosis report, etc. Through the convolution operation, CNN can identify these local features, that is, local vocabulary patterns and combination relationships. This feature extraction helps to capture key information in medical texts and convert it into meaningful feature representations.

[0037] In a specific embodiment of the present application, the medical data preprocessing subunit 1221 includes: passing the medical data table containing the basic data of the patient, the medical insurance data table and the medical data table through the patient medical data word embedding model based on the glove model to obtain a sequence of patient medical data word vectors; and arranging the sequence of patient medical data word vectors in two dimensions to obtain the patient medical data word embedding matrix.

[0038] It should be understood that the glove (Global Vectors for Word Representation) model is a word embedding method based on global vocabulary statistics, which can convert vocabulary into dense vector representations with semantic meaning. The glove model can generate high-quality word vectors by capturing the global co-occurrence information of vocabulary. In medical data, the meanings of terms such as disease names, symptom descriptions, and treatment plans can be very complex. The glove model can convert these terms into vectors with semantic information, thereby helping to understand and process medical data more accurately. The glove model can reflect the contextual association of vocabulary in the vector space by modeling global co-occurrence information. In this way, the word vectors generated by glove have good semantic expression capabilities, which can help the model more accurately identify and learn patterns and relationships in the data, thereby improving the accuracy and effectiveness of the analysis.

[0039] Furthermore, the two-dimensional arrangement can effectively organize and structure word vector sequences. In practical applications, medical data usually includes a series of records or descriptions, such as patient symptoms, diagnosis results, and treatment plans. These records can be represented as a series of word vectors. Arranging these word vectors in sequence into a two-dimensional matrix (rows correspond to different records or time steps, and columns correspond to the dimensions of the word vector) can provide a fixed structure for each record, so that the entire data set has a consistent format when input into the model, thereby improving the understanding and analysis of medical data.

[0040] In a specific embodiment of the present application, the patient medical data convolutional feature encoding subunit 1222 includes: passing the patient medical data word embedding matrix through a patient medical data text convolutional neural network as a feature extractor to obtain a patient medical data text local correlation feature map; passing the patient medical data text local correlation feature map through a patient medical data text non-local neural network to obtain a patient medical data text global correlation feature map; and performing global mean pooling on the patient medical data text global correlation feature map to obtain the patient medical data semantic feature vector.

[0041] It should be understood that, first of all, convolutional neural networks can automatically extract and learn local features. In medical data, the local structure of the text, such as specific word combinations or phrases, may contain useful information. CNN scans the word embedding matrix through the convolution layer, can identify these local patterns, and convert them into feature maps, which reflect the important local information in the text. This ability makes CNN particularly effective in processing medical texts with complex structures and contexts. Feature maps can capture semantic relationships and contextual information between words. The convolution operation not only recognizes local patterns, but also extracts different types of features through different convolution kernels. For example, some convolution kernels may focus on identifying keyword combinations for diseases, while others may focus on the description of symptoms. The extraction of these local features helps establish associations between words and enhances the understanding of the semantics of medical data. Local correlation feature maps help reduce the dimensionality of the data and improve the processing efficiency of the model. Specifically, each layer of the patient medical data text convolutional neural network used as a feature extractor performs convolution processing, mean pooling processing based on the local feature matrix and nonlinear activation processing on the input data in the forward pass of the layer, so that the last layer of the patient medical data text convolutional neural network used as a feature extractor outputs the patient medical data text local correlation feature map, wherein the input of the patient medical data text convolutional neural network used as a feature extractor is the patient medical data word embedding matrix.

[0042] Furthermore, non-local neural networks excel in processing global information and long-range dependencies, which are essential for capturing complex and comprehensive semantic relationships in medical data. Although convolutional neural networks (CNNs) extract local features through local convolution operations, these features may not fully capture the global information and long-range dependencies present in the data. For example, some important information in a patient's medical record may be distributed throughout the record, and non-local neural networks are able to capture this information by focusing on lexical relationships in a global scope, thereby providing a deep understanding of the entire text. Non-local neural networks then establish long-range associations through global feature maps to improve the overall understanding of the text. Next, through the global association feature map, the model is able to integrate the local information extracted from the local feature map to identify key patterns and relationships in the text. This integration not only enhances the expressiveness of the features, but also improves the model's ability to classify and predict medical data. Specifically, the local correlation feature map of the patient medical data text is subjected to a first point convolution process, a second point convolution process and a third point convolution process to obtain a first feature map, a second feature map and a third feature map; the position-weighted sum between the first feature map and the second feature map is calculated to obtain a fused feature map; the fused feature map is input into a Softmax function to map the feature values ​​of each position in the fused feature map to a probability space to obtain a normalized fused feature map; the position-point multiplication between the normalized fused feature map and the third feature map is calculated to obtain a re-fused feature map; the re-fused feature map is embedded with a Gaussian similarity function to obtain a global similarity feature map; the global similarity feature map is subjected to a fourth point convolution process to adjust the number of channels of the global similarity feature map to obtain a channel-adjusted global similarity feature map; the position-weighted sum of the channel-adjusted global similarity feature map and the local correlation feature map of the patient medical data text is calculated to obtain the global correlation feature map of the patient medical data text.

[0043] Furthermore, global mean pooling can effectively integrate the information of the entire feature map into a single vector. The feature map contains global correlation features extracted from non-local neural networks, which can represent the complex relationships and patterns of the entire medical text. However, the size of the feature map is usually large, which is not convenient for direct use in subsequent model processing and analysis. Through global mean pooling, the values ​​of each feature map position in the feature map are averaged to obtain a fixed-length semantic feature vector, which makes subsequent model processing more efficient and consistent. Global mean pooling helps to reduce the dimensionality of the data and reduce the computational complexity. The original global feature map may contain a large amount of feature information. Through mean pooling, these features are compressed into a smaller vector, which significantly reduces the computational burden and memory requirements of the model. This dimensionality reduction not only improves the processing speed, but also simplifies the model training process, allowing the model to converge faster.

[0044] In the above-mentioned health care data quality detection system 100, the medical data quality analysis module 130 is used to determine whether to issue a medical data quality abnormality warning based on the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector. It should be understood that the feature word granularity associated feature vector captures the relationship between local details and keywords in the medical data. This fine-grained feature extraction can identify key elements such as specific symptoms, disease names and drugs in the text. The semantic feature vector provides a global understanding of the entire medical text and can capture the overall trends and semantic relationships in the data. By combining the feature word granularity associated feature vector and the semantic feature vector for abnormal warning, the system can determine whether the data is abnormal from both the details and the overall level to improve the accuracy of abnormality detection.

[0045] Figure 4 FIG. 1 is a block diagram of a medical data quality analysis module in a health care data quality detection system according to an embodiment of the present application. Figure 4 As shown, in a specific embodiment of the present application, the medical data quality analysis module 130 includes: a medical data quality feature association unit 131, which is used to perform shift compensation fusion based on local structure collapse on the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector to obtain a medical data quality warning judgment feature vector; a medical data quality abnormality warning judgment unit 132, which is used to pass the medical data quality warning judgment feature vector through a classifier to obtain a classification result, and the classification result is used to determine whether to issue a medical data quality abnormality warning.

[0046] It should be understandable that by combining local detail features with global semantic features, the system can issue early warnings at the early stages of data quality problems, thereby making medical data processing more reliable and effective.

[0047] In particular, in the technical solution of the present application, the feature vector of the patient medical data feature word granularity association is extracted from the word sequence of the patient medical data through a bidirectional long short-term memory model (Bi-LSTM), which focuses on the contextual information and local vocabulary relationship at the word level. Among them, Bi-LSTM can capture the dynamic relationship between words and their impact on the overall semantics, but its processing scope is usually limited to the context and order at the word level. Relatively speaking, the semantic feature vector of the patient medical data is obtained by text convolutional neural network and global mean pooling method, which deal with the global structure and long-distance dependency of the text. The text convolutional neural network focuses on capturing local patterns, while the global mean pooling summarizes the global information of the entire data. In this way, different feature expressions have significant differences in dimension and representation from word granularity features, resulting in feature fine-grained dimension offset. In the process of merging the feature vector of the patient medical data feature word granularity association with the semantic feature vector of the patient medical data, if these features are simply fused, it may lead to information loss or inaccurate mapping, so that the original structural relationship cannot be effectively retained. For example, word-granular features may have high accuracy in details, but they are diluted by the generalized information of global features during the fusion process, resulting in the collapse of local structure, that is, the loss of local associations and detail information of features, affecting the quality of the final warning judgment. In this way, when the classifier uses these fused features for prediction, the loss or distortion of local information is not effectively compensated, which will lead to a reduction in the classifier's ability to identify abnormal medical data quality, making it impossible to accurately issue warnings. Therefore, in the technical solution of the present application, a medical data quality warning judgment feature vector is obtained by performing a shift compensation fusion based on local structure collapse on the patient medical data feature word-granularity association feature vector and the patient medical data semantic feature vector.

[0048] Among them, the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector are subjected to shift compensation fusion based on local structure collapse to obtain the medical data quality warning judgment feature vector, including: calculating the position difference between the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector to obtain the shift compensation differential feature vector; calculating the position point multiplication between the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector to obtain the shift compensation position point multiplication feature vector; calculating the position addition between the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector to obtain the shift compensation position addition feature vector; adding the shift compensation differential feature vector, the The shift-compensated positional dot-product feature vector and the shift-compensated positional addition feature vector are input into a one-dimensional convolutional layer and maximum pooling is performed to obtain a shift-compensated convolutional coding pooling feature vector; the square root of the two norm of the shift-compensated convolutional coding pooling feature vector is weighted and added to the one norm of the shift-compensated convolutional coding pooling feature vector to obtain a shift compensation factor; the shift-compensated positional dot-product feature vector and the square root of the length of the feature vector associated with the patient's medical data feature word granularity are added by position to obtain a shift-compensated intermediate vector; the shift compensation factor is used as the first weighting coefficient and the second weighted hyperparameter is used as the second weighting coefficient to calculate the position-weighted sum of the shift-compensated positional addition feature vector and the shift-compensated intermediate vector to obtain a medical data quality warning judgment feature vector.

[0049] Among them, the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector are subjected to shift compensation fusion based on local structure collapse to obtain the medical data quality warning judgment feature vector, including:

[0050]

[0051] Among them, V 1 Represents the granularity-related feature vector of patient medical data feature words, V 2 represents the semantic feature vector of patient medical data, represents positional subtraction, ⊙ represents positional multiplication, represents positional addition, conv1D represents a one-dimensional convolutional layer, MaxPool represents maximum pooling, V μ represents the shift-compensated convolutional coding pooling feature vector, ||·|| 1 represents a norm, ||·|| 2 represents the bi-norm, α represents the first weighted hyperparameter, τ represents the shift compensation factor, L represents the length of the granularity-related feature vector of the patient's medical data feature word, β represents the second weighted hyperparameter, V fRepresents the medical data quality warning judgment feature vector.

[0052] In the technical solution of the present application, in the process of fusing the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector, the feature fine-grained shift compensation between the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector is taken into consideration, which may cause the fused medical data quality warning judgment feature vector to have local structural collapse. Based on this, in the technical solution of the present application, the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector are fused with shift compensation based on local structural collapse.

[0053] Specifically, based on the restricted expression of the foreground-background structure distinction of the high-dimensional manifold of the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector by the amplitude of the semantic difference vector between the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector, the key feature fine-grained correspondence between the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector is modeled, and then active fine-grained correspondence adjustment propagation is performed through the back-propagation displacement compensation information between the corresponding feature values ​​of the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector, so as to avoid the imbalance at the feature vector level of the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector in a focused manner, thereby synchronizing the micro dimensions of the two feature vectors, ensuring that during the fusion process, the information between the feature vectors can be effectively aligned and combined, thereby avoiding abnormal accumulation of local structures.

[0054] Furthermore, the medical data quality warning judgment feature vector integrates local and global medical data features, which contain potential abnormal information in the data. By learning to extract useful patterns and rules from these features, the classifier can classify data points as "normal" or "abnormal" and automatically identify problems in the data. This automated processing can significantly improve the efficiency of anomaly detection and reduce the need for manual intervention. In addition, the classifier can be trained using historical data to improve the accuracy of anomaly detection. By training on a large amount of annotated medical data, the classifier can learn the pattern differences between normal and abnormal data. This training enables the classifier to accurately judge whether there is an abnormality when facing new data. Using a trained classifier can effectively improve the accuracy of the early warning system and reduce false alarms and missed reports.

[0055] In summary, the embodiment of the present application first obtains a medical data table including basic patient data, a medical insurance data table, and a medical data table, and then uses deep learning technology to perform feature extraction and association analysis. Finally, a classifier is used to determine whether to issue an abnormal medical data quality warning, thereby reducing reliance on manual inspections, and further reducing financial losses caused by data anomalies, thereby improving the reliability and efficiency of medical services.

[0056] As described above, the health and medical data quality detection system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the health and medical data quality detection system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the health and medical data quality detection system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the health and medical data quality detection system 100 can also be one of the many hardware modules of the terminal device.

[0057] Alternatively, in another example, the health care data quality detection system 100 and the terminal device may also be separate devices, and the health care data quality detection system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0058] Figure 5 FIG. 1 is a flow chart of a method for detecting quality of health and medical data according to an embodiment of the present application. Figure 5 As shown, according to the health care data quality detection method of the embodiment of the present application, it includes: S110, obtaining a medical data table including basic data of visiting patients, a medical insurance data table and a medical treatment data table; S120 extracting a patient medical data feature word granularity-associated feature vector and a patient medical data semantic feature vector from the medical data table including the basic data of visiting patients, the medical insurance data table and the medical treatment data table; S130, judging whether to issue a medical data quality abnormality warning based on the patient medical data feature word granularity-associated feature vector and the patient medical data semantic feature vector.

[0059] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned health and medical data quality detection method have been referred to above. Figures 1 to 4 The description of the health care data quality detection system has been described in detail, and therefore, its repeated description will be omitted.

[0060] Below, reference Figure 6 To describe an electronic device according to an embodiment of the present application.

[0061] Figure 6is a block diagram of an electronic device according to an embodiment of the present application. Figure 6 , which shows a schematic diagram of the structure of an electronic device suitable for implementing a terminal device or a server in an embodiment of the present application.

[0062] like Figure 6 As shown, the electronic device includes one or more processors, a communication unit, etc. The one or more processors are, for example: one or more central processing units (CPUs), and / or one or more graphics processing units (GPUs), etc. The processor can perform various appropriate actions and processes according to executable instructions stored in a read-only memory (ROM) or executable instructions loaded from a storage part to a random access memory (RAM). The communication unit may include but is not limited to a network card, and the network card may include but is not limited to an IB (Infiniband) network card. The processor may communicate with a read-only memory and / or a random access memory to execute executable instructions, and is connected to the communication unit through a bus and communicates with other target devices through the communication unit, thereby completing the operation corresponding to any method provided in the embodiments of the present application, for example, the primary shard manager shards the features in the feature queue; for any shard, the primary shard manager stores the any shard in the corresponding shard feature library, and records the mapping relationship between the any shard and the corresponding shard feature library in a mapping table; in response to a failure of the primary shard manager, a shard manager is reselected from at least one slave shard manager as the new primary shard manager; the new primary shard manager obtains shard information of the shard currently being written from all shard feature libraries, and recreates the mapping relationship of the shard currently being written based on the shard information of the shard currently being written in all shard feature libraries.

[0063] In addition, various programs and data required for the operation of the device can also be stored in the RAM. The CPU, ROM and RAM are connected to each other via a bus. In the case of RAM, the ROM is an optional module. The RAM stores executable instructions, or writes executable instructions into the ROM at runtime, and the executable instructions enable the processor to perform operations corresponding to any of the above methods of the present invention. The input / output (I / O) interface is also connected to the bus. The communication unit can be integrated or can be configured to have multiple submodules (e.g., multiple IB network cards) and be linked on the bus.

[0064] The following components are connected to the I / O interface: an input section including a keyboard, a mouse, etc.; an output section including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section including a hard disk, etc.; and a communication section including a network interface card such as a LAN card, a modem, etc. The communication section performs communication processing via a network such as the Internet. A drive is also connected to the I / O interface as needed. Removable media, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., are installed on the drive as needed so that a computer program read therefrom is installed into the storage section as needed.

[0065] If necessary, Figure 6 The architecture shown is only an optional implementation method. In the specific practice, the above can be modified according to actual needs. Figure 6 The number and type of components can be selected, deleted, added or replaced; different functional components can be set up by separate settings or integrated settings, for example, the GPU and CPU can be set up separately or the GPU can be integrated on the CPU, the communication unit can be set up separately or integrated on the CPU or GPU, etc. These replaceable implementations all fall within the protection scope disclosed in the present invention.

[0066] In addition, an embodiment of the present invention further provides a computer program, including computer-readable code. When the computer-readable code runs on a device, a processor in the device executes instructions for implementing each step in the data processing method of any of the above embodiments of the present invention.

[0067] In addition, an embodiment of the present invention further provides a computer-readable storage medium for storing computer-readable instructions, which, when executed, implement the operations of each step in the data processing method of any of the above embodiments of the present invention.

[0068] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0069] The method and apparatus of the present invention may be implemented in many ways. For example, the method and apparatus of the present invention may be implemented by software, hardware, firmware, or any combination of software, hardware, firmware. The above order of steps for the method is only for illustration, and the steps of the method of the present invention are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present invention may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present invention. Thus, the present invention also covers a recording medium storing a program for executing the method according to the present invention.

[0070] The description of the present invention is given for the purpose of illustration and description, and is not intended to be exhaustive or to limit the invention to the disclosed form. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for particular uses.

Claims

1. A health care data quality detection system, characterized in that: include: The health and medical data acquisition module is used to obtain the medical data table including the basic data of the patients, the medical insurance data table and the medical treatment data table; A health care data processing module, used to extract the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector from the medical data table containing the basic data of the patient, the medical insurance data table and the medical data table; The medical data quality analysis module is used to determine whether to issue a warning of abnormal medical data quality based on the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector.

2. The health care data quality detection system according to claim 1, characterized in that: The health and medical data processing module includes: A medical data word granularity feature encoding unit, used for performing word granularity feature encoding on the medical data table including the basic data of the patient, the medical insurance data table and the medical data table to obtain a patient medical data feature word granularity associated feature vector; The medical data semantic feature encoding unit is used to perform semantic feature encoding on the medical data table including the basic data of the patient, the medical insurance data table and the medical data table to obtain the patient medical data semantic feature vector.

3. The health care data quality detection system according to claim 2, characterized in that: The medical data word granularity feature encoding unit comprises: Segmenting the medical data table including the basic data of the patients, the medical insurance data table and the medical treatment data table to obtain a plurality of patient medical data word sequences; Passing the plurality of patient medical data word sequences through a converter-based patient medical data text convolutional neural network to obtain a plurality of patient medical data feature vectors; The plurality of patient medical data feature vectors are passed through a patient medical data feature word granularity association bidirectional long short-term memory model to obtain the patient medical data feature word granularity association feature vector.

4. The health care data quality detection system according to claim 3, characterized in that: The medical data semantic feature encoding unit comprises: A medical data preprocessing subunit, used for preprocessing the medical data table including the basic data of the patients, the medical insurance data table and the medical data table to obtain a patient medical data word embedding matrix; The patient medical data convolutional feature encoding subunit is used to perform convolutional feature encoding on the patient medical data word embedding matrix to obtain the patient medical data semantic feature vector.

5. The health care data quality detection system according to claim 4, characterized in that: The medical data preprocessing subunit comprises: The medical data table including the basic data of the patients, the medical insurance data table and the medical data table is embedded in the patient medical data word embedding model based on the glove model to obtain a sequence of patient medical data word vectors; The sequence of the patient medical data word vectors is arranged in two dimensions to obtain the patient medical data word embedding matrix.

6. The health care data quality detection system according to claim 5, characterized in that: The patient medical data convolution feature encoding subunit comprises: Passing the patient medical data word embedding matrix through a patient medical data text convolutional neural network as a feature extractor to obtain a patient medical data text local correlation feature map; Passing the local correlation feature map of the patient medical data text through the patient medical data text non-local neural network to obtain a global correlation feature map of the patient medical data text; Perform global mean pooling on the global correlation feature map of the patient medical data text to obtain the semantic feature vector of the patient medical data.

7. The health care data quality detection system according to claim 6, characterized in that: The medical data quality analysis module comprises: A medical data quality feature association unit, used for performing shift compensation fusion based on local structure collapse on the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector to obtain a medical data quality early warning judgment feature vector; The medical data quality abnormality warning judgment unit is used to pass the medical data quality warning judgment feature vector through a classifier to obtain a classification result, and the classification result is used to determine whether to issue a medical data quality abnormality warning.

8. The health care data quality detection system according to claim 7, characterized in that: The medical data quality feature association unit includes: Calculating the position difference between the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector to obtain a shift-compensated differential feature vector; Calculating the positional point multiplication between the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector to obtain a shift-compensated positional point multiplication feature vector; Calculate the positional addition between the patient medical data feature word granularity associated feature vector and the patient medical data semantic feature vector to obtain a shift-compensated positional addition feature vector; Inputting the shift-compensated differential feature vector, the shift-compensated positional dot-product feature vector and the shift-compensated positional addition feature vector into a one-dimensional convolution layer and performing maximum pooling to obtain a shift-compensated convolutional coding pooling feature vector; Weighting the square root value of the two-norm of the shift-compensated convolutional coding pooling feature vector and adding the square root value to the one-norm of the shift-compensated convolutional coding pooling feature vector to obtain a shift compensation factor; Adding the positional dot product of the displacement compensation feature vector and the square root of the length of the feature vector associated with the feature word granularity of the patient's medical data according to the positional sum to obtain a displacement compensation intermediate vector; Using the shift compensation factor as the first weighting coefficient and the second weighted hyperparameter as the second weighting coefficient, the position-weighted sum of the shift compensation position-added feature vector and the shift compensation intermediate vector is calculated to obtain the medical data quality warning judgment feature vector.

9. A method for detecting the quality of health care data, characterized in that: include: Obtain a medical data table containing basic patient data, medical insurance data table, and medical treatment data table; Extracting a patient medical data feature word granularity association feature vector and a patient medical data semantic feature vector from the medical data table including the basic data of the patient, the medical insurance data table and the medical treatment data table; Based on the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector, it is determined whether to issue a medical data quality abnormality warning.

10. The health care data quality detection method according to claim 9, characterized in that: Extracting the patient medical data feature word granularity association feature vector and the patient medical data semantic feature vector from the medical data table including the basic data of the patient, the medical insurance data table and the medical data table includes: Performing word granularity feature encoding on the medical data table including the basic data of the patients, the medical insurance data table and the medical data table to obtain a patient medical data feature word granularity associated feature vector; The medical data table including the basic data of the patient, the medical insurance data table and the medical treatment data table is semantically encoded to obtain a semantic feature vector of the patient's medical data.

Images