Construction and application of multidimensional medical big data fusion and disease prediction recognition engine

Abstract

The invention provides construction and application of a multi-dimensional medical big data fusion and disease prediction recognition engine, and aims to fuse features of multi-dimensional medical data according to dimension weights through a pre-constructed factor group: F-F-form (q) so as to generate a targeted feature data set M (qt). And further performing prediction model training based on the feature data set M (qt), and putting the prediction model into clinical disease prediction application. Therefore, the fusion features are identified and learned through the model on the dimension weight, so that the learning efficiency of the model on the features is improved, the construction of the prediction model is accelerated, the application period of a clinical model is shortened, targeted disease prediction is performed by using the biased model generated by fusion, and disease prediction guidance in the clinical dimension is achieved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical technology, and in particular to a method, application system and electronic equipment for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine. Background Art

[0002] Intelligent disease prediction refers to the use of big data, artificial intelligence and other technical means to predict the occurrence trend of diseases before they occur, so as to take corresponding prevention and control measures in advance. This is of great significance for public health management, optimal allocation of medical resources and patient health management.

[0003] In intelligent disease prediction, we first need to collect a large amount of health data and public health data, including demographics, environmental factors, living habits, medical records, etc. Then, through data cleaning and processing, we remove irrelevant and erroneous data to ensure the accuracy and completeness of the data. Next, we use big data technology and artificial intelligence algorithms, such as machine learning and deep learning, to analyze and model these data to predict the trend of disease occurrence. Finally, we visualize the prediction results to provide reference for decision makers.

[0004] The application scope of disease intelligent prediction is very wide, and it can be applied to the prediction of various diseases, such as infectious diseases, chronic diseases, cancer, etc. For example, by analyzing epidemic data, the spread trend and peak period of the epidemic can be predicted, providing a reference for the government to formulate epidemic prevention and control strategies; by analyzing the patient's medical records and living habits, the risk of the patient suffering from a certain disease can be predicted, providing a basis for doctors to formulate personalized prevention and treatment plans.

[0005] In the existing intelligent disease prediction solutions, such as the disease prediction models generated by machine learning algorithm training, the data features are mainly learned from medical data in one aspect of the clinic, such as respiratory data, blood gas data, etc., so it is impossible to integrate and identify multi-dimensional medical data features. Although some models can identify diseases for multiple data types, multiple model training processes are required to obtain intelligent models that can identify multi-dimensional data. However, it will take a long time to train the model, delaying the subsequent clinical use of the model.

[0006] In addition, existing clinical disease prediction models do not have directional predictions and can only perform feature recognition and prediction on disease data in a single dimension, which lacks specificity. Summary of the invention

[0007] In order to solve the technical problems existing in the prior art, the embodiments of the present invention provide a method, application system and electronic device for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine. The technical solution is as follows:

[0008] On the one hand, a method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine is provided, and the method comprises the following steps:

[0009] Constructing a factor group: FF-formula (q) for performing feature fusion on the medical data set M according to the dimension value weight δp, wherein the feature fusion factor: FF-formula is used to: perform feature fusion on the medical data on each dimension in the medical data set M with the dimension value weight δp as the center;

[0010] Collect medical data in various dimensions and generate corresponding medical data sets M;

[0011] Parsing the medical data set M to obtain a data structure of the medical data set M, and identifying a dimension value weight δp of the medical data on each dimension from the data structure;

[0012] Perform ranking calculation according to the dimension value weight δp, and identify the top-ranked medical data as the characteristic medical data Mq;

[0013] According to the dimension value weight δp(max) of the feature medical data Mq, the feature fusion factor: FF-formula(qt) satisfying the δp(max) is retrieved from the factor group: FF-formula(q) in the feature fusion algorithm library:

[0014] in:

[0015] δp(max)≤δq;

[0016] The medical data set M is subjected to feature fusion through the feature fusion factor: FF-formula (qt) to obtain a feature data set M (qt);

[0017] Based on a machine learning algorithm, the feature data set M(qt) is trained and learned to generate a corresponding disease prediction and recognition model.

[0018] As an optional embodiment of the present invention, optionally, a factor group for performing feature fusion on the medical data set M according to the dimension value weight δp is constructed: FF-formula (q), including:

[0019] Pre-constructing several feature fusion factors corresponding to the dimension value weight δp: FF-formula, wherein the feature fusion factor: FF-formula includes the following feature transformation fusion methods: feature linear fusion, feature nonlinear fusion, attention mechanism fusion or cascade fusion;

[0020] For each of the constructed feature fusion factors: FF-formula, a corresponding fusion factor δq is configured, wherein the fusion factor δq at least includes the dimension value weight δp possessed by the medical data on one dimension;

[0021] After the configuration is completed, each of the feature fusion factors: FF-formula determined by the plurality of fusion factors δq is obtained, and each of the feature fusion factors: FF-formula forms the factor group: FF-formula (q);

[0022] The factor group: FF-formula (q) is saved to the feature fusion algorithm library.

[0023] As an optional embodiment of the present invention, optionally, medical data in various dimensions are collected and a corresponding medical data set M is generated, including:

[0024] From the medical database of the backend server, index the medical data in each dimension and import it into the preprocessing module;

[0025] In the preprocessing module, the medical data in each dimension is preprocessed, including:

[0026] Determine whether the medical data in each dimension contains disease medical data in the gene dimension:

[0027] If the disease medical data on the gene dimension is included, the disease medical data is subjected to genotyping or gene mutation classification processing, and the disease medical data is classified and processed according to the type of disease caused by the genotyping or gene mutation;

[0028] If not included, data cleaning and preprocessing will be performed in a unified format;

[0029] The medical data set M is constructed according to the preprocessed medical data in each dimension, and the medical data set M is imported into a background database for caching.

[0030] As an optional embodiment of the present invention, optionally, before parsing the medical data set M to obtain the data structure of the medical data set M and identifying the dimension value weight δp of the medical data on each dimension from the data structure, the method further includes:

[0031] The administrator enters the backend database and traverses the medical data in each dimension in the medical data set M;

[0032] For the medical data on each dimension traversed, the corresponding dimension value weight δp is configured, where:

[0033] If the disease medical data on the gene dimension is traversed, the dimension value weight δp configured for it is automatically changed to δp (max), where:

[0034] The δp(max) is the dimension value weight δp with weight ranking priority.

[0035] As an optional embodiment of the present invention, optionally, ranking calculation is performed according to the dimension value weight δp, and the top-ranked medical data is identified as the characteristic medical data Mq, including:

[0036] Ranking the dimension value weights δp of the medical data on each dimension and sorting them from large to small;

[0037] The medical data corresponding to the first dimension whose dimensional value weight δp is collected as the feature medical data Mq, and the dimensional value weight δp (max) of the feature medical data Mq is output to the feature fusion algorithm library.

[0038] As an optional embodiment of the present invention, optionally, according to the dimension value weight δp(max) of the feature medical data Mq, the feature fusion factor: FF-formula(qt) satisfying the δp(max) is retrieved from the factor group: FF-formula(q) in the feature fusion algorithm library, including:

[0039] Traverse each of the feature fusion factors: FF-formula in the factor group: FF-formula (q), and compare and search the dimension value weight δp (max) with the fusion factor δq of each feature fusion factor: FF-formula, and when the following conditions are found to be met:

[0040] δp(max)≤δq,

[0041] Mark the fusion factor δq that satisfies the feature fusion factor δp(max): FF-formula(qt);

[0042] The feature fusion factor: FF-formula (qt), the feature fusion algorithm library is called out, and wait for feature fusion.

[0043] As an optional implementation scheme of the present invention, optionally, based on a machine learning algorithm, the feature data set M(qt) is trained and learned to generate a corresponding disease prediction and recognition model, including:

[0044] Inputting the feature data set M(qt) into a preset machine learning model, and performing feature recognition and model training on the feature data set M(qt) based on a corresponding machine learning algorithm;

[0045] After the training is completed, the disease prediction and recognition model is generated, and the disease prediction and recognition model is clinically verified;

[0046] If the verification is passed, the disease prediction and recognition model will be deployed on the background server for clinical disease prediction.

[0047] On the other hand, an application system is provided, comprising:

[0048] A disease prediction and recognition model generated according to the method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine;

[0049] HIS system, used to collect patients’ clinical medical data and save patients’ clinical disease prediction results;

[0050] The backend server is used to call the disease prediction and recognition model, extract and recognize features of the clinical medical data, output the clinical disease prediction results of the patient and feed them back to the HIS system;

[0051] The disease prediction and recognition model and the HIS system are both deployed on the background server.

[0052] On the other hand, an electronic device is provided, comprising: a processor; a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, any one of the methods for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine is implemented.

[0053] On the other hand, a computer-readable storage medium is provided, in which at least one instruction is stored. The at least one instruction is loaded and executed by a processor to implement any one of the above-mentioned methods for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine.

[0054] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0055] Through the pre-constructed factor group: FF-formula (q), the features of multi-dimensional medical data are fused according to the dimension weights to generate a targeted feature data set M (qt), and further based on the feature data set M (qt) The prediction model is trained and put into clinical disease prediction applications. Therefore, the fusion features are identified and learned through the model on the dimension weights to improve the learning efficiency of the model for features, speed up the construction of the prediction model, shorten the application cycle of the clinical model, and use the biased model generated by the fusion to predict targeted diseases, which has a disease prediction orientation in the clinical dimension. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0057] Figure 1 It is a flow chart of a method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine provided by an embodiment of the present invention;

[0058] Figure 2 It is a schematic diagram of performing feature fusion according to calling factors based on dimension values ​​provided by an embodiment of the present invention;

[0059] Figure 3 It is a block diagram of an application system provided by an embodiment of the present invention;

[0060] Figure 4 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0062] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0063] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0064] In the embodiments of the present invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are consistent.

[0065] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0066] The dimension value weight of data refers to the weight or importance of a data item or data point relative to a certain dimension. In other words, it reflects the relative importance of the data point in that dimension.

[0067] Weight is a relative concept, which indicates the importance of a factor or indicator relative to other factors or indicators. In data analysis, dimension value weights are usually used to determine the contribution of different dimensions or attributes to the overall data interpretation.

[0068] For example, suppose there is a dataset with multiple dimensions, such as age, gender, income, etc. Each dimension has its own set of data points or values. Among these dimensions, some dimensions may be more important or more representative than others. By assigning a weight to each dimension, the dataset can be better understood and interpreted.

[0069] Weights can be determined in different ways, such as based on expert opinion, historical data, statistical analysis, etc. Once the weights are determined, they can be used to weight different dimensions in the dataset to get more accurate analysis results or predictions.

[0070] It should be noted that the selection and allocation of weights should be based on reasonable grounds and assumptions, and should take into account the characteristics of the data and the purpose of the analysis. Different weight allocations may lead to different results and interpretations, so weights should be selected with caution and fully verified and tested.

[0071] In this solution, the factors and dimension weights can be pre-set and configured by the administrator.

[0072] This solution identifies and predicts disease patterns by integrating multidimensional data such as genetic information, clinical indicators, and environmental factors.

[0073] Significance: Provide strong data support for early diagnosis and treatment of diseases and improve the accuracy of medical decision-making.

[0074] Main technologies: multi-source data fusion, high-dimensional data analysis, pattern recognition algorithm, and prediction model optimization.

[0075] Dimension values ​​involving genetic information, clinical treatment, and environmental factors are given priority, and their weight values ​​are greater than those of other medical data sets.

[0076] The embodiment of the present invention provides a method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine, which can be implemented by an electronic device, which can be a terminal or a server. Figure 1 The flowchart of the method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine is shown. The processing flow of the method may include the following steps:

[0077] The method comprises the following steps:

[0078] S1. Constructing a factor group: FF-formula (q) for performing feature fusion on the medical data set M according to the dimension value weight δp, wherein the feature fusion factor: FF-formula is used to perform feature fusion on the medical data on each dimension in the medical data set M with the dimension value weight δp as the center;

[0079] S2. Collect medical data in various dimensions and generate corresponding medical data sets M;

[0080] S3, parsing the medical data set M to obtain a data structure of the medical data set M, and identifying a dimension value weight δp of the medical data on each dimension from the data structure;

[0081] S4, performing ranking calculation according to the dimension value weight δp, and identifying the top-ranked medical data as the characteristic medical data Mq;

[0082] S5. According to the dimension value weight δp(max) of the feature medical data Mq, the feature fusion factor FF-formula(qt) satisfying δp(max) is retrieved from the factor group FF-formula(q) in the feature fusion algorithm library:

[0083] in:

[0084] δp(max)≤δq;

[0085] S6, performing feature fusion on the medical data set M through the feature fusion factor: FF-formula (qt) to obtain a feature data set M (qt);

[0086] S7. Based on a machine learning algorithm, the feature data set M(qt) is trained and learned to generate a corresponding disease prediction and recognition model.

[0087] In this solution, the medical data set needs to include various types of diseases in various departments and the corresponding clinical symptom data and treatment plans of the diseases. It can also include corresponding medical advice data and medication data, etc. The data structure and type in the data set can be determined from the clinical medical database based on the doctor's clinical work.

[0088] Combined with Figure 2 As shown, this solution mainly extracts the feature fusion factor: FF-formula (qt) that satisfies the dimension value weight δp (max) of the medical data set M on each dimension from the factor group: FF-formula (q), and uses the feature fusion factor: FF-formula (qt) to perform feature fusion on the medical data set M on the δp (max) dimension, construct the corresponding feature data set M (qt), and let the machine learning model perform training and recognition;

[0089] According to the above steps, for the medical data set M on other dimensions, the corresponding "feature fusion factor" is called according to the weight of its dimension value to perform feature extraction and fusion, and the model is trained to identify it;

[0090] By analogy, the model can complete the fusion and identification of the features of the medical data set M on all dimensions, and call the heat syndrome fusion factors in sequence according to the dimensional value weights of the medical data set M from large to small to process the data features, so that the model can identify and learn the fused features through the model on the dimensional weights, thereby improving the model's learning efficiency for features, accelerating the construction of the prediction model, shortening the application cycle of the clinical model, and using the biased model generated by the fusion to predict targeted diseases, which has a disease prediction orientation in the clinical dimension.

[0091] In this solution, the factor group in the feature fusion algorithm library: FF-formula (q) includes several feature fusion factors: FF-formula (qt) configured corresponding to each dimension.

[0092] Each feature fusion factor: FF-formula (qt) represents a feature fusion method. The specific function formula (feature fusion method, also called factor) is not described in detail in this embodiment. For example, there are mainly the following feature fusion methods:

[0093] 1. Additive fusion: Add different features to obtain new fusion features, which is suitable for the fusion of numerical features.

[0094] 2. Multiplication fusion: Multiply different features to obtain new fusion features, which is suitable for the fusion of numerical features.

[0095] 3. Splicing and fusion: directly splice different features together to form new fusion features, which is suitable for the fusion of categorical features.

[0096] 4. Linear fusion: Use linear combination to fuse different features and obtain new fusion features through weight distribution:

[0097] Linear fusion is a fusion method that performs weighted summation of features from different neural networks. Assuming there are two neural networks A and B, which extract feature vectors 9a and b respectively, linear fusion can be expressed as:

[0098] ,

[0099] Among them, c is the fused feature vector, w1 and w2 are weights, which can be obtained through training. In general, the weights need to meet the following conditions:

[0100] The sum of the weights is 1, i.e. w1+w2=1;

[0101] 5. Nonlinear fusion: Use nonlinear methods to fuse different features, such as combining polynomial features:

[0102] Nonlinear fusion is a fusion method that concatenates or stacks different neural network features and then transforms them through a nonlinear function.

[0103] Assuming there are two neural networks A and B, which extract feature vectors a and b respectively, the nonlinear fusion can be expressed as:

[0104] c = f([a, b]),

[0105] [a,b] means concatenating or stacking the feature vectors a and b, and f represents a nonlinear function, which can be ReLU, sigmoid9tanh, etc. The advantage of nonlinear fusion is that it can handle nonlinear relationships and performs better for complex tasks.

[0106] 6. Attention mechanism fusion;

[0107] 7. Cascade convolution fusion: Use convolution operations to fuse different features, which is suitable for feature fusion of image or sequence data.

[0108] These feature transformation and fusion methods have their own characteristics, and the appropriate method can be selected according to the specific application scenarios and requirements.

[0109] For the feature fusion method on each dimension, the set dimension weight of the medical data set on each dimension can be matched with the dimension value weight δp configured by each feature fusion factor: FF-formula (qt) in the corresponding library, and the feature fusion factor: FF-formula (qt) that meets the data feature fusion on each dimension can be retrieved.

[0110] Alternatively, the feature fusion algorithm library pre-orders each FF-formula (qt) in order according to the priority (also according to the δp configured for each FF-formula (qt)), and then retrieves and traverses the dimension values ​​of each medical data set, and traverses to the medical data set M ranked first in δp (max), and calls the first FF-formula (qt) in the library in order to fuse its features; for the medical data set of the second dimension, traverse the medical data set M ranked second in δp (max), and call the second FF-formula (qt) in order to fuse its features;

[0111] And so on.

[0112] This solution adopts the second method. Please refer to the subsequent description for details.

[0113] As an optional embodiment of the present invention, optionally, a factor group for performing feature fusion on the medical data set M according to the dimension value weight δp is constructed: FF-formula (q), including:

[0114] Pre-constructing several feature fusion factors corresponding to the dimension value weight δp: FF-formula, wherein the feature fusion factor: FF-formula includes the following feature transformation fusion methods: feature linear fusion, feature nonlinear fusion, attention mechanism fusion or cascade fusion;

[0115] For each of the constructed feature fusion factors: FF-formula, a corresponding fusion factor δq is configured, wherein the fusion factor δq at least includes the dimension value weight δp possessed by the medical data on one dimension;

[0116] After the configuration is completed, each of the feature fusion factors: FF-formula determined by the plurality of fusion factors δq is obtained, and each of the feature fusion factors: FF-formula forms the factor group: FF-formula (q);

[0117] The factor group: FF-formula (q) is saved to the feature fusion algorithm library.

[0118] For details of each FF-formula (q) in the feature fusion algorithm library, please refer to the feature fusion method settings in the above dimensions.

[0119] As an optional embodiment of the present invention, optionally, medical data in various dimensions are collected and a corresponding medical data set M is generated, including:

[0120] From the medical database of the backend server, index the medical data in each dimension and import it into the preprocessing module;

[0121] In the preprocessing module, the medical data in each dimension is preprocessed, including:

[0122] Determine whether the medical data in each dimension contains disease medical data in the gene dimension:

[0123] If the disease medical data on the gene dimension is included, the disease medical data is subjected to genotyping or gene mutation classification processing, and the disease medical data is classified and processed according to the type of disease caused by the genotyping or gene mutation;

[0124] If not included, data cleaning and preprocessing will be performed in a unified format;

[0125] According to the pre

[0126] The processed medical data in each dimension is used to construct the medical data set M, and the medical data set M is imported into the background database for caching.

[0127] The methods for extracting medical data features from medical data sets in various dimensions mainly include the following:

[0128] 1. Statistical feature extraction: By performing basic statistical analysis on the data, statistical indicators such as mean, standard deviation, maximum, and minimum are extracted as features. This method is simple and efficient and is suitable for the study of certain clinical indicators.

[0129] 2. Frequency domain feature extraction: Convert medical signals to the frequency domain and use methods such as Fourier transform to extract frequency domain features, such as frequency components and power spectrum density. This method has wide applications in signal processing and waveform classification.

[0130] 3. Time series feature extraction: Many medical data are collected in the form of time series, such as electrocardiograms, vital signs, etc. Time series feature extraction can effectively extract information from time series by calculating indicators such as the mean, standard deviation, slope, etc. within the sliding window.

[0131] 4. Principal Component Analysis (PCA): PCA is a commonly used dimensionality reduction method that reduces the dimensionality of data to improve the interpretation and readability of data while ensuring that important features are correctly retained. In many medical data analyses, classifiers and detectors often use PCA to analyze data sets.

[0132] 5. Independent Component Analysis (ICA): ICA is another data analysis method that can extract independent components separately. In the medical field, this method has been successfully used in diagnostic systems such as electroencephalograms and magnetic resonance imaging, because the independent components of the data are often useful information about the underlying structure of the data.

[0133] 6. Kernel principal component analysis (KPCA): KPCA is an extension of the linear PCA method, originally used to deal with large nonlinear data structures. In many application scenarios that require the analysis of nonlinear data, the KPCA method can significantly improve prediction accuracy and generalization ability. In the medical field, this method is also widely used.

[0134] 7. Entity feature extraction, there are mainly the following methods for entity feature recognition:

[0135] 7.1 Rule-based method: mainly implemented by string template matching. This method relies on rule templates manually constructed by linguistic experts, and uses features including statistical information, punctuation, keywords, indicator words and direction words, position words (such as the last word), center words, etc., with pattern and string matching as the main means. This method usually relies on the establishment of knowledge bases and dictionaries.

[0136] 7.2 Statistical-based method: Identify by predicting the probability of the entity’s part of speech. This method treats entity recognition and word segmentation problems as sequence labeling problems. The predicted label is not only related to the current predicted sequence label, but also to the previous predicted label. There is a strong interdependence between the predicted labels.

[0137] 7.3 Deep learning-based methods: Word classification through neural network sequence models. For example, BiLSTM+CRF is a common model structure that uses neural networks to model text context and successfully overcomes some limitations of traditional methods.

[0138] 7.4 Attention-based method: Recognition is performed by studying the connection between words through the attention model.

[0139] In this embodiment, a deep learning model is preferred to identify the various entity features in the medical data set, such as disease type and corresponding clinical symptom data, symptom type and diagnosis scheme, which can be initially identified by a CNN deep learning model and then manually labeled and checked by a doctor.

[0140] Regardless of the fusion method used, the following points need to be noted:

[0141] 1. Data preprocessing: Before data fusion, data from each data source needs to be preprocessed, including data cleaning and data standardization, to ensure data quality and consistency.

[0142] 2. Feature alignment: Since the features of different data sources may have different dimensions and distributions, feature alignment is required, that is, converting the features of different data sources to the same dimension and distribution for fusion.

[0143] 3. Fusion strategy: It is necessary to select an appropriate fusion strategy based on the specific application scenario and data characteristics. For example, in some cases, it may be necessary to emphasize the importance of genetic information, while in other cases, it may be necessary to emphasize the importance of clinical indicators or environmental factors.

[0144] As an optional embodiment of the present invention, optionally, before parsing the medical data set M to obtain the data structure of the medical data set M and identifying the dimension value weight δp of the medical data on each dimension from the data structure, the method further includes:

[0145] The administrator enters the backend database and traverses the medical data in each dimension in the medical data set M;

[0146] For the medical data on each dimension traversed, the corresponding dimension value weight δp is configured, where:

[0147] If the disease medical data on the gene dimension is traversed, the dimension value weight δp configured for it is automatically changed to δp (max), where:

[0148] The δp(max) is the dimension value weight δp with weight ranking priority.

[0149] As an optional embodiment of the present invention, optionally, ranking calculation is performed according to the dimension value weight δp, and the top-ranked medical data is identified as the characteristic medical data Mq, including:

[0150] Ranking the dimension value weights δp of the medical data on each dimension and sorting them from large to small;

[0151] The medical data corresponding to the first dimension whose dimensional value weight δp is collected as the feature medical data Mq, and the dimensional value weight δp (max) of the feature medical data Mq is output to the feature fusion algorithm library.

[0152] Combined with Figure 2 As shown, the corresponding dimension value weight δp configured for the medical data on each dimension is determined by the administrator based on the department to which the model is applied. For example, for cardiovascular and cerebrovascular departments, the clinical data of cardiovascular and cerebrovascular departments is configured first, and the weight value is ranked first.

[0153] The medical data corresponding to the dimension with the first dimension value weight δp (that is, the dimension value weight δp (max) found this time, the maximum weight; the next time it is the dimension corresponding to the second largest weight) will be matched with the corresponding FF-formula (qt) for feature fusion according to its δp (max) (it is also configured with an adapted dimension weight value, so the attribute values ​​of each FF-formula (qt) can be traversed and retrieved from the library to find the FF-formula (qt) matching δp (max)).

[0154] As an optional embodiment of the present invention, optionally, according to the dimension value weight δp(max) of the feature medical data Mq, the feature fusion factor: FF-formula(qt) satisfying the δp(max) is retrieved from the factor group: FF-formula(q) in the feature fusion algorithm library, including:

[0155] Traverse each of the feature fusion factors: FF-formula in the factor group: FF-formula (q), and compare and search the dimension value weight δp (max) with the fusion factor δq of each feature fusion factor: FF-formula, and when the following conditions are found to be met:

[0156] δp(max)≤δq,

[0157] Mark the fusion factor δq that satisfies the feature fusion factor δp(max): FF-formula(qt);

[0158] The feature fusion factor: FF-formula (qt), the feature fusion algorithm library is called out, and wait for feature fusion.

[0159] δp(max) is within the range of dimensional values ​​that FF-formula(qt) is adapted to: δq, which means that the dimensional value weight δq that FF-formula(qt) is adapted to for feature fusion is at least not less than δp(max), indicating that FF-formula(qt) can perform feature fusion on the medical data set corresponding to δp(max). If δp(max) is less than δq, it means that FF-formula(qt) is not suitable for its feature fusion processing.

[0160] As an optional implementation scheme of the present invention, optionally, based on a machine learning algorithm, the feature data set M(qt) is trained and learned to generate a corresponding disease prediction and recognition model, including:

[0161] Inputting the feature data set M(qt) into a preset machine learning model, and performing feature recognition and model training on the feature data set M(qt) based on a corresponding machine learning algorithm;

[0162] After the training is completed, the disease prediction and recognition model is generated, and the disease prediction and recognition model is clinically verified;

[0163] If the verification is passed, the disease prediction and recognition model will be deployed on the background server for clinical disease prediction.

[0164] This solution enables the model to have priority recognition performance on the dimension value weights of medical data on each dimension, which is convenient for making biased and guiding disease predictions on big data, and for predicting diseases according to the main disease medical decision-making dimensions of each department (for example, the patient's disease is mainly caused by the disease types belonging to multiple departments, but cardiovascular and cerebrovascular diseases are the main influencing dimensions, and other causes such as gastrointestinal, rheumatic, etc. are secondary influencing dimensions. Therefore, the model prioritizes feature recognition and fusion of medical data on the cardiovascular and cerebrovascular dimensions, so as to predict the clinical diagnosis of cardiovascular and cerebrovascular diseases for patients' diseases in the first place).

[0165] A machine learning model is a model that automatically learns and improves algorithms based on data. These models extract useful information and patterns from data and then perform tasks such as prediction, classification, and clustering on new data. Here are some common machine learning models:

[0166] 1. Supervised learning model: This type of model uses pre-labeled data to train the model. Common supervised learning algorithms include classification and regression algorithms. Classification algorithms are used to determine the category of entities, objects, or events represented in the data, while regression algorithms are used to identify the relationship between multiple variables represented in the data set. The disadvantage of supervised learning models is that the data needs to be labeled, which usually requires a lot of manpower and material resources.

[0167] 2. Unsupervised learning model: This type of model does not require human intervention, but relies entirely on machine automation to learn the model based on previously unlabeled data. Unsupervised learning models can automatically identify the internal structure of data and classify the data.

[0168] 3. Semi-supervised learning model: This type of model refers to a model that contains labeled data and unlabeled technical data in the learning data. The model is trained by both the supervision information of labeled data and the information of unlabeled data.

[0169] 4. Reinforcement Learning Model: Reinforcement learning, also known as evaluation learning, refers to the use of algorithms that can obtain reward signals and adjust the model based on interactions with the environment.

[0170] These models have a wide range of applications in their respective fields, such as natural language processing, image recognition, speech recognition, recommendation systems, etc. The choice of model depends on the specific problem and data characteristics.

[0171] In this solution, the model is not limited and can be selected by the user. CNN can also be used for model training.

[0172] Through model training, data features in various dimensions can be identified and learned, enabling the model to have the ability to identify multi-dimensional data, and to perform priority and guidance identification according to dimensional weights, which is more inclined to application in the department where the current doctor is located.

[0173] Therefore, this model identifies and predicts disease patterns by integrating multidimensional data such as genetic information, clinical indicators, and environmental factors. It can provide strong data support for early diagnosis and treatment of diseases and improve the accuracy of medical decision-making.

[0174] On the other hand, an application system is provided, comprising:

[0175] A disease prediction and recognition model 320 generated according to the method for building a multi-dimensional medical big data fusion and disease prediction and recognition engine;

[0176] HIS system 330, used to collect clinical medical data of patients and save clinical disease prediction results of patients;

[0177] The backend server 310 is used to call the disease prediction and recognition model, extract and recognize features of the clinical medical data, output the clinical disease prediction results of the patient and feed them back to the HIS system;

[0178] The disease prediction and recognition model and the HIS system are both deployed on the background server.

[0179] Figure 3This is a block diagram of an application system according to an exemplary embodiment, which is used for a method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine. Figure 3 The device includes a background server 310, a disease prediction and recognition model 320, and a HIS system 330.

[0180] The examination data of clinical patients can be stored in the HIS system 330. The background can call the disease prediction and recognition model 320 to identify and predict the examination data of clinical patients in the department, give priority to identification according to the medical data dimension of the department, and first output the disease diagnosis and policy results of the disease type to which the department belongs. For example, based on clinical examination data and symptoms, predict the patient's disease type and possible applicable diagnosis and medical advice, as well as medication type and dosage, to assist doctors in issuing orders.

[0181] Please understand it in combination with the above methods for details.

[0182] Figure 4 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 4 As shown, the electronic device may include the above Figure 3 Optionally, the electronic device 410 may include a first processor 2001.

[0183] Optionally, the electronic device 410 may further include a memory 2002 and a transceiver 2003 .

[0184] The first processor 2001, the memory 2002 and the transceiver 2003 may be connected via a communication bus.

[0185] Combine the following Figure 4 The components of the electronic device 410 are described in detail:

[0186] The first processor 2001 is the control center of the electronic device 410, and may be a processor or a general term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), or may be application specific integrated circuits (ASICs), or may be one or more integrated circuits configured to implement the embodiments of the present invention, such as one or more microprocessors (digital signal processors, DSPs), or one or more field programmable gate arrays (field programmable gate arrays, FPGAs).

[0187] Optionally, the first processor 2001 can execute various functions of the electronic device 410 by running or executing a software program stored in the memory 2002 and calling data stored in the memory 2002 .

[0188] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 4 CPU0 and CPU1 are shown in FIG.

[0189] In a specific implementation, as an embodiment, the electronic device 410 may also include multiple processors, such as Figure 4 The first processor 2001 and the second processor 2004 are shown in FIG. Each of these processors may be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0190] The memory 2002 is used to store the software program for executing the solution of the present invention, and is controlled to be executed by the first processor 2001. The specific implementation method can refer to the above method embodiment, which will not be repeated here.

[0191] Optionally, the memory 2002 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 may be integrated with the first processor 2001, or may exist independently and access the first processor 2001 through the interface circuit ( Figure 4 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0192] The transceiver 2003 is used to communicate with a network device or a terminal device.

[0193] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 4 The receiver is used to implement a receiving function, and the transmitter is used to implement a sending function.

[0194] Optionally, the transceiver 2003 may be integrated with the first processor 2001, or may exist independently and communicate with the first processor 2001 through the interface circuit ( Figure 4 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0195] It should be noted that Figure 4 The structure of the electronic device 410 shown in the figure does not constitute a limitation on the router, and the actual knowledge structure recognition device may include more or fewer components than those shown in the figure, or combine certain components, or arrange the components differently.

[0196] In addition, the technical effects of the electronic device 410 can refer to the technical effects of the method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine described in the above method embodiment, and will not be repeated here.

[0197] It should be understood that the first processor 2001 in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0198] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0199] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0200] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0201] In the present invention, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0202] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0203] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0204] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0205] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0206] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0207] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0208] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0209] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine, characterized in that: The steps include: Constructing a factor group: FF-formula (q) for performing feature fusion on the medical data set M according to the dimension value weight δp, wherein the feature fusion factor: FF-formula is used to: perform feature fusion on the medical data on each dimension in the medical data set M with the dimension value weight δp as the center; Collect medical data in various dimensions and generate corresponding medical data sets M; Parsing the medical data set M to obtain a data structure of the medical data set M, and identifying a dimension value weight δp of the medical data on each dimension from the data structure; Perform ranking calculation according to the dimension value weight δp, and identify the top-ranked medical data as the characteristic medical data Mq; According to the dimension value weight δp(max) of the feature medical data Mq, the feature fusion factor: FF-formula(qt) satisfying the δp(max) is retrieved from the factor group: FF-formula(q) in the feature fusion algorithm library: in: δp(max)≤δq; The medical data set M is subjected to feature fusion through the feature fusion factor: FF-formula (qt) to obtain a feature data set M (qt); Based on a machine learning algorithm, the feature data set M(qt) is trained and learned to generate a corresponding disease prediction and recognition model.

2. The method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine according to claim 1, characterized in that: Construct a factor group for feature fusion of the medical data set M according to the dimension value weight δp: FF-formula (q), including: Pre-constructing several feature fusion factors corresponding to the dimension value weight δp: FF-formula, wherein the feature fusion factor: FF-formula includes the following feature transformation fusion methods: feature linear fusion, feature nonlinear fusion, attention mechanism fusion or cascade fusion; For each of the constructed feature fusion factors: FF-formula, a corresponding fusion factor δq is configured, wherein the fusion factor δq at least includes the dimension value weight δp possessed by the medical data on one dimension; After the configuration is completed, each of the feature fusion factors: FF-formula determined by the plurality of fusion factors δq is obtained, and each of the feature fusion factors: FF-formula forms the factor group: FF-formula (q); The factor group: FF-formula (q) is saved to the feature fusion algorithm library.

3. The method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine according to claim 1, characterized in that: Collect medical data in various dimensions and generate corresponding medical data sets M, including: From the medical database of the backend server, index the medical data in each dimension and import it into the preprocessing module; In the preprocessing module, the medical data in each dimension is preprocessed, including: Determine whether the medical data in each dimension contains disease medical data in the gene dimension: If the disease medical data on the gene dimension is included, the disease medical data is subjected to genotyping or gene mutation classification processing, and the disease medical data is classified and processed according to the type of disease caused by the genotyping or gene mutation; If not included, data cleaning and preprocessing will be performed in a unified format; The medical data set M is constructed according to the preprocessed medical data in each dimension, and the medical data set M is imported into a background database for caching.

4. The method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine according to claim 1, characterized in that: Before parsing the medical data set M to obtain the data structure of the medical data set M and identifying the dimension value weight δp of the medical data on each dimension from the data structure, the method further includes: The administrator enters the backend database and traverses the medical data in each dimension in the medical data set M; For the medical data on each dimension traversed, the corresponding dimension value weight δp is configured, where: If the disease medical data on the gene dimension is traversed, the dimension value weight δp configured for it is automatically changed to δp (max), where: The δp(max) is the dimension value weight δp with weight ranking priority.

5. The method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine according to claim 2, characterized in that: The ranking calculation is performed according to the dimension value weight δp, and the top-ranked medical data is identified as the characteristic medical data Mq, including: Ranking the dimension value weights δp of the medical data on each dimension and sorting them from large to small; The medical data corresponding to the first dimension whose dimensional value weight δp is collected as the feature medical data Mq, and the dimensional value weight δp (max) of the feature medical data Mq is output to the feature fusion algorithm library.

6. The method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine according to claim 5, characterized in that: According to the dimension value weight δp(max) of the feature medical data Mq, the feature fusion factor: FF-formula(qt) satisfying the δp(max) is retrieved from the factor group: FF-formula(q) in the feature fusion algorithm library, including: Traverse each of the feature fusion factors: FF-formula in the factor group: FF-formula (q), and compare and search the dimension value weight δp (max) with the fusion factor δq of each feature fusion factor: FF-formula, and when the following conditions are found to be met: δp(max)≤δq, Mark the fusion factor δq that satisfies the feature fusion factor δp(max): FF-formula(qt); The feature fusion factor: FF-formula (qt), the feature fusion algorithm library is called out, and wait for feature fusion.

7. The method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine according to claim 6, characterized in that: Based on the machine learning algorithm, the feature data set M(qt) is trained and learned to generate a corresponding disease prediction and recognition model, including: Inputting the feature data set M(qt) into a preset machine learning model, and performing feature recognition and model training on the feature data set M(qt) based on a corresponding machine learning algorithm; After the training is completed, the disease prediction and recognition model is generated, and the disease prediction and recognition model is clinically verified; If the verification is passed, the disease prediction and recognition model will be deployed on the background server for clinical disease prediction.

8. The application system is characterized in that, include: A disease prediction and recognition model generated by the method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine according to any one of claims 1 to 7; HIS system, used to collect patients’ clinical medical data and save patients’ clinical disease prediction results; The backend server is used to call the disease prediction and recognition model, extract and recognize features of the clinical medical data, output the clinical disease prediction results of the patient and feed them back to the HIS system; The disease prediction and recognition model and the HIS system are both deployed on the background server.

9. An electronic device, characterized in that include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement a method for constructing a multi-dimensional medical big data fusion and disease prediction and recognition engine as described in any one of claims 1-7 when executing the executable instructions.

Images