Intelligent medical platform based on big data and method thereof

Through a smart medical platform based on big data, semantic features encoding and correlation of patient's medical history and medical order content, the problem of information overload in complex diseases is solved, the rational judgment of medical orders is achieved, the medical error rate is reduced, and the Accuracy of medical decisions.

CN119007920AInactive Publication Date: 2024-11-22SECOND AFFILIATED HOSPITAL ZHEJIANG UNIV COLLEGE OF MEDICINE +1
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411183930.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Doctors are susceptible to information overload when facing complex conditions, resulting in inaccuracies in diagnosis and treatment decisions, increasing the rate of medical errors, and it is difficult for the existing technology to effectively assist doctors in judging the rationality of medical orders.

Method used

A smart medical platform based on big data was designed, and the patient's medical history and diagnosis were obtained through medical-related data collection module, patient's condition feature coding module, medical order content feature coding module, medical order disease correlation feature generation module and medical order reasonable identification result generation module. The results and historical medication use are used, semantic features are encoded and correlated to determine whether the medical order is reasonable.

Benefits of technology

By establishing a connection between the patient's condition and the content of the medical order in the same characteristic space, accurately determine whether the medical order is in line with the patient's actual situation, reduce the medical error rate, and improve the accuracy and safety of medical decisions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119007920A_ABST
    Figure CN119007920A_ABST
Patent Text Reader

Abstract

The invention provides an intelligent medical platform based on big data and a method thereof, and relates to the field of intelligent medical treatment, and the method comprises the steps: carrying out the semantic space feature extraction of the medical history, diagnosis result, examination result and historical medication condition of a patient, so as to obtain the semantic features of the illness state of the patient; meanwhile, semantic space feature extraction is carried out on the doctor's advice content to obtain semantic features of the doctor's advice content, then the relation between the patient's condition and the doctor's advice content is established in the same feature space, and then whether the doctor's advice is reasonable or not is judged through a classifier. In this way, whether the doctor's advice meets the actual conditions and requirements of the patient can be accurately judged, and therefore the medical error rate is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of smart medical care, and more specifically, to a big data-based smart medical care platform and method thereof. Background Art

[0002] With the aging of the population and the increase in medical needs, medical resources are becoming increasingly scarce. In order to optimize resource allocation and improve the efficiency of medical services, many hospitals are currently coping with the shortage of medical resources by building smart medical platforms. Smart medical platforms can promote the sharing and utilization of medical data, provide support for medical research and clinical practice, and promote the accumulation and dissemination of medical knowledge.

[0003] When doctors face complex medical conditions, they may be affected by information overload. In order to assist doctors in making more scientific decisions, reduce medical errors, and ensure patient safety, a smart medical solution based on big data is needed. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present application provides a big data-based smart medical platform and method thereof.

[0005] According to one aspect of the present application, a smart medical platform based on big data is provided, which includes:

[0006] Medical-related data collection module, used to obtain the patient's medical history, diagnosis results, historical medication, and doctor's orders;

[0007] A patient condition feature encoding module, used for performing semantic feature encoding on the patient's medical history, diagnosis results and historical medication conditions to obtain a multi-scale patient condition semantic understanding feature vector;

[0008] A medical order content feature encoding module, used for encoding the semantic features of the medical order content to obtain a medical order content semantic understanding feature vector;

[0009] A doctor's order-condition correlation feature generation module, used to correlate the multi-scale patient's condition semantic understanding feature vector with the doctor's order content semantic understanding feature vector to obtain a doctor's order-condition correlation feature vector;

[0010] The module for generating reasonable medical advice identification results is used to obtain the identification result of whether the medical advice is reasonable based on the medical advice-condition association feature vector.

[0011] According to another aspect of the present application, a smart medical method based on big data is provided, which includes:

[0012] Obtain the patient's medical history, diagnosis, medication history, and doctor's orders;

[0013] Performing semantic feature encoding on the patient's medical history, diagnosis results, and historical medication to obtain a multi-scale patient condition semantic understanding feature vector;

[0014] Performing semantic feature encoding on the medical order content to obtain a semantic understanding feature vector of the medical order content;

[0015] Associating the multi-scale patient condition semantic understanding feature vector with the doctor's order content semantic understanding feature vector to obtain a doctor's order-condition association feature vector;

[0016] Based on the doctor's advice-condition association feature vector, an identification result of whether the doctor's advice is reasonable is obtained.

[0017] This application has significant technical effects due to the adoption of the above technical solutions:

[0018] The big data-based smart medical platform and method provided by the present application extracts the semantic features of the patient's condition by extracting the features of the semantic space of the patient's medical history, diagnosis results, examination results, and historical medication, and extracts the semantic features of the doctor's order content by extracting the features of the semantic space, and then establishes the connection between the patient's condition and the doctor's order content in the same feature space, and then uses a classifier to judge whether the doctor's order is reasonable. In this way, it is possible to accurately judge whether the doctor's order meets the patient's actual situation and needs, thereby reducing the medical error rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] 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.

[0020] Figure 1 This is a system block diagram of a big data-based smart medical platform according to an embodiment of the present application.

[0021] Figure 2 Schematic diagram of the architecture of a big data-based smart medical platform according to an embodiment of the present application.

[0022] Figure 3 This is a block diagram of a patient condition characteristic encoding module in a big data-based smart medical platform according to an embodiment of the present application.

[0023] Figure 4 This is a block diagram of a module for generating reasonable identification results of traditional Chinese medicine orders in a smart medical platform based on big data according to an embodiment of the present application.

[0024] Figure 5 It is a flow chart of a smart medical method based on big data according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] 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.

[0026] With the development of society and the change of population structure, we are facing an increasingly severe challenge: the aging population and the surge in medical needs. This phenomenon has put tremendous pressure on the medical system, and the allocation of medical resources has become particularly tight. In order to meet this challenge, many medical institutions have begun to explore and implement smart medical platforms in order to optimize resource allocation and improve the efficiency and quality of medical services.

[0027] The construction of a smart medical platform can not only achieve the rational allocation of medical resources, but also promote the sharing and utilization of medical data through advanced information technology. The smart medical platform can greatly improve the convenience and accessibility of medical services. In addition, the smart medical platform can also provide strong data support for medical research and clinical practice, promote the accumulation and dissemination of medical knowledge, and accelerate innovation and development in the medical field.

[0028] In clinical practice, doctors often face complex conditions and huge amounts of information. In this case, doctors may be troubled by information overload, which affects the scientificity and accuracy of their decision-making. In order to assist doctors in making more accurate diagnosis and treatment decisions, reduce the rate of medical errors, and ensure patient safety, a smart medical solution based on big data is expected.

[0029] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also demonstrated a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks has provided new solutions and solutions for smart healthcare.

[0030] Figure 1 This is a system block diagram of a big data-based smart medical platform according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of a smart medical platform based on big data according to an embodiment of the present application. Figure 1 and Figure 2As shown, in the big data-based smart medical platform 100, it includes: a medical-related data acquisition module 110, which is used to obtain the patient's medical history, diagnosis results, historical medication conditions and medical order content; a patient condition feature encoding module 120, which is used to perform semantic feature encoding on the patient's medical history, diagnosis results and historical medication conditions to obtain a multi-scale patient condition semantic understanding feature vector; a medical order content feature encoding module 130, which is used to perform semantic feature encoding on the medical order content to obtain a medical order content semantic understanding feature vector; a medical order condition association feature generation module 140, which is used to associate the multi-scale patient condition semantic understanding feature vector with the medical order content semantic understanding feature vector to obtain a medical order-condition association feature vector; a medical order reasonable identification result generation module 150, which is used to obtain an identification result of whether the medical order is reasonable based on the medical order-condition association feature vector.

[0031] Specifically, in the technical solution of the present application, the patient's medical history, diagnosis results, historical medication and doctor's order content are first obtained. It should be understood that the patient's medical history and diagnosis results mainly include the patient's past disease experience, the type, severity, complications of the patient's disease, etc. The understanding of this information helps to determine whether the treatment plan in the doctor's order is accurate and effective for the current diagnosis, and the patient's historical medication covers the types, dosages, efficacy and adverse reactions of drugs used by the patient in the past, which can help understand the patient's tolerance and treatment effects of different drugs. The doctor's order content will include specific drug names, dosages, usage, courses of treatment and other drug treatment plans, and may also include recommendations for specific examinations of the patient. In general, the patient's medical history, diagnosis results and historical medication can provide a comprehensive understanding of the patient's condition. By analyzing the details of the doctor's order content, combined with the patient's medical history, diagnosis results and historical medication, it can help evaluate whether the doctor's order is in line with the patient's actual situation and condition characteristics, thereby reducing the rate of medical errors. In particular, in an embodiment of the present application, one possible implementation method for obtaining a patient's medical history, diagnosis results, historical medication, and medical advice content may be: obtaining patient-related data and medical advice content from a database of a smart medical platform.

[0032] Figure 3 FIG. 1 is a block diagram of a patient condition characteristic encoding module in a big data-based smart medical platform according to an embodiment of the present application. Figure 3As shown, the patient condition feature encoding module 120 includes: a patient condition-related word embedding unit 121, which is used to segment the patient's medical history, diagnosis results and historical medication conditions and then pass them through a word embedding layer to obtain a plurality of patient condition-related attribute word embedding vectors; a first scale feature generation unit 122, which is used to extract contextual semantic features from the plurality of patient condition-related attribute word embedding vectors to obtain a first scale patient condition semantic understanding feature vector; a second scale feature generation unit 123, which is used to extract features from a two-dimensional input matrix of patient condition-related attribute words formed by arranging the plurality of patient condition-related attribute word embedding vectors to obtain a second scale patient condition semantic understanding feature vector; and a multi-scale feature fusion unit 124, which is used to fuse the first scale patient condition semantic understanding feature vector and the second scale patient condition semantic understanding feature vector to obtain the multi-scale patient condition semantic understanding feature vector.

[0033] In order to convert the original text data such as medical history, diagnosis results and historical medication into a form that can be understood and processed by computers, and to facilitate the extraction of semantic features from the original text data, so as to better understand the patient's condition description, the patient's medical history, diagnosis results and historical medication are segmented and then passed through the word embedding layer to obtain multiple patient condition-related attribute word embedding vectors. It is not difficult to understand that through word segmentation, the original long paragraph describing the condition can be turned into individual words. The word embedding layer is often used to map words to a low-dimensional space. Through word embedding, the segmented data is converted into a dense vector representation, which can capture the semantic relationship between words.

[0034] The embedding vectors of multiple patient condition-related attribute words reflect the position and meaning of each word in the semantic space, capture the semantic relationship between words, and can express the key information and semantic content in the patient's condition description. In order to comprehensively consider the information of all words in order to more comprehensively understand the patient's condition description, the multiple patient condition-related attribute word embedding vectors are passed through a context encoder to obtain a first-scale patient condition semantic understanding feature vector. The essential structure of the context encoder in this application is a converter-based context semantic encoder. The converter-based context semantic encoder is mainly composed of a multi-layer self-attention mechanism, which can effectively capture the dependency relationship and semantic information between multiple patient condition-related attribute word embedding vectors. Specifically, through the self-attention mechanism, the representation of the patient condition-related attribute word embedding vector can be adjusted according to the degree of correlation between each patient condition-related attribute word embedding vector and other patient condition-related attribute word embedding vectors in the input sequence, thereby better expressing the semantic relationship and context information between multiple word embedding vectors.

[0035] In an embodiment of the present application, an implementation method of passing the multiple patient condition-related attribute word embedding vectors through a context encoder to obtain a first-scale patient condition semantic understanding feature vector may be: arranging the multiple patient condition-related attribute word embedding vectors into a patient condition input vector; converting the patient condition input vector into a patient condition query vector and a patient condition key vector respectively through a learnable embedding matrix; calculating the product between the patient condition query vector and the transposed vector of the patient condition key vector to obtain a patient condition self-attention association matrix; normalizing the patient condition self-attention association matrix to obtain a patient condition standardized self-attention association matrix; inputting the patient condition standardized self-attention association matrix into a Softmax activation function for activation to obtain a patient condition self-attention feature matrix; multiplying the patient condition self-attention feature matrix with each patient condition-related attribute word embedding vector in the multiple patient condition-related attribute word embedding vectors as value vectors to obtain multiple patient condition semantic context feature vectors; and cascading the multiple patient condition semantic context feature vectors to obtain the first-scale patient condition semantic understanding feature vector.

[0036] In order to take into account the relationship and contextual information between the words expressing the patient's condition, so as to better capture the local and global information of the original text describing the patient's condition data, it is necessary to first arrange the multiple patient condition-related attribute word embedding vectors into a two-dimensional input matrix of patient condition-related attribute words. By constructing a two-dimensional matrix, the order relationship and contextual information between the original expression words can be retained, thereby improving the understanding and expression ability of the data. Then, the arranged patient condition-related attribute word two-dimensional input matrix is ​​passed through a multi-scale hybrid convolution semantic encoder to obtain a second-scale patient condition semantic understanding feature vector. In this application, the multi-scale hybrid convolution semantic encoder includes a parallel first convolution branch structure, a second convolution branch structure, a third convolution branch structure, and a fourth convolution branch structure, as well as a multi-scale fusion structure connected to the first to fourth convolution branch structures, wherein the first convolution branch uses a first convolution kernel having a first size, the second convolution branch uses a second convolution kernel having a first size and a first void rate, the third convolution branch uses a third convolution kernel having a first size and a second void rate, and the fourth convolution branch uses a fourth convolution kernel having a first size and a third void rate. Through the parallel structure of multiple convolution branches, the information contained in the vector can be extracted from different scales and angles, and finally the multi-scale fusion structure can be used to fuse the features extracted by each branch, further improving the richness and expressiveness of the semantic representation.

[0037] In an embodiment of the present application, after arranging the multiple patient condition-related attribute word embedding vectors into a two-dimensional input matrix of patient condition-related attribute words, a possible implementation method for obtaining a second-scale patient condition semantic understanding feature vector through a multi-scale mixed convolutional semantic encoder may be: using each mixed convolutional layer of the multi-scale mixed convolutional semantic encoder to respectively perform on the input data in the forward pass of the layer: using a first convolution kernel having a first size to perform convolution encoding on the two-dimensional input matrix of patient condition-related attribute words to obtain a first-scale feature map; using a second convolution kernel having a first void rate to perform convolution encoding on the two-dimensional input matrix of patient condition-related attribute words to obtain a second-scale feature map; using a third convolution kernel having a second void rate to perform convolution encoding on the two-dimensional input matrix of patient condition-related attribute words to obtain a third-scale feature map; using A fourth convolution kernel with a third dilation rate performs convolution encoding on the two-dimensional input matrix of attribute words related to the patient's condition to obtain a fourth-scale feature map, wherein the first convolution kernel, the second convolution kernel, the third convolution kernel and the fourth convolution kernel have the same size, and the second convolution kernel, the third convolution kernel and the fourth convolution kernel have different dilation rates; the first-scale feature map, the second-scale feature map, the third-scale feature map and the fourth-scale feature map are aggregated along the channel dimension to obtain an aggregated feature map; the aggregated feature map is pooled along the channel dimension for each feature matrix to generate a pooled feature map; the pooled feature map is activated to generate an activated feature map; wherein the output of the last layer of the multi-scale hybrid convolution semantic encoder is the second-scale patient condition semantic understanding feature vector.

[0038] The first-scale and second-scale patient condition semantic understanding feature vectors capture semantic information of different levels and scales respectively. In order to comprehensively consider semantic information of different levels, improve the understanding and analysis capabilities of the patient's condition, and thus provide more comprehensive and accurate support for medical decision-making, the first-scale patient condition semantic understanding feature vector and the second-scale patient condition semantic understanding feature vector are integrated in the technical solution of the present application to obtain a multi-scale patient condition semantic understanding feature vector.

[0039] In an embodiment of the present application, a possible implementation method of fusing the first-scale patient condition semantic understanding feature vector and the second-scale patient condition semantic understanding feature vector to obtain a multi-scale patient condition semantic understanding feature vector may be: fusing the first-scale patient condition semantic understanding feature vector and the second-scale patient condition semantic understanding feature vector using the following fusion formula to obtain the multi-scale patient condition semantic understanding feature vector; wherein the fusion formula is:

[0040] V c =Concat[V 1, V 2 ]

[0041] Among them, V 1 represents the semantic understanding feature vector of the patient's condition at the first scale, V 2 represents the semantic understanding feature vector of the patient's condition at the second scale, Concat[·] represents the cascade function, V c Represents the multi-scale patient condition semantic understanding feature vector.

[0042] In order to be able to more accurately capture the key information in the doctor's order, and then provide a more comprehensive basis for judging the rationality of the doctor's order, it is necessary to capture the key semantic information in the doctor's order content. In the technical solution of the present application, the doctor's order content is segmented and then passed through the doctor's order content semantic feature encoder to obtain the doctor's order content semantic understanding feature vector. The doctor's order content semantic feature encoder can extract the key features in the doctor's order, such as the name of the drug, dosage, frequency and other important information, which helps to accurately understand the doctor's order content, thereby supporting subsequent medical decision-making processing. The structure of the doctor's order content semantic feature encoder in the present application is a converter-based content semantic feature encoder including a word embedding layer. The doctor's order content is segmented, and the segmented data is converted into a word embedding vector representation. The words can be mapped to the same semantic vector space, and the converter structure is used to encode the input doctor's order content keyword embedding vector sequence to capture the dependency between the vectors in the input sequence data. This can more comprehensively capture the semantic information of the doctor's order content and help to accurately understand the doctor's order intent.

[0043] In an embodiment of the present application, after the medical order content is segmented, it is passed through a medical order content semantic feature encoder to obtain a medical order content semantic understanding feature vector. An implementation method may be: perform word segmentation on the medical order content to obtain a plurality of medical order content information keywords; pass the plurality of medical order content information keywords through an embedding layer to convert each of the plurality of medical order content information keywords into a medical order content information keyword embedding vector to obtain a sequence of medical order content information keyword embedding vectors, wherein the embedding layer uses a learnable embedding matrix to embed-code each of the medical order content information keywords; input the sequence of the medical order content information keyword embedding vectors into a converter-based content semantic feature encoder to obtain a plurality of medical order content information feature vectors; and cascade the plurality of medical order content information feature vectors to obtain the medical order content semantic understanding feature vector.

[0044] The multi-scale patient condition semantic understanding feature vector captures the semantic information of the patient's condition description at different scales, including the understanding of the patient's condition at different levels and angles; the doctor's order content semantic understanding feature vector captures the semantic information of the doctor's order content, reflecting the doctor's treatment recommendations and medication requirements for the patient. In order to evaluate the rationality of the doctor's order, it is necessary to consider the relationship between the patient's condition and the corresponding doctor's order, that is, to associate the multi-scale patient condition semantic understanding feature vector with the doctor's order content semantic understanding feature vector to obtain the doctor's order-condition association feature vector. By associating the doctor's order content with the patient's condition characteristics, more comprehensive information can be provided to determine whether the doctor's order is reasonable, which can effectively improve the accuracy and rationality of medical decision-making, improve medical quality, and reduce medical risks.

[0045] Figure 4 The block diagram of the module for generating reasonable identification results of Chinese medicine orders in the smart medical platform based on big data according to the embodiment of the present application is shown in FIG. Figure 4 As shown, the medical order rationality identification result generation module 150 includes: a feature optimization unit 151, used to predict the predetermined class probability label of the medical order-condition association feature vector based on probabilistic activation to obtain an optimized medical order-condition association feature vector; a feature vector classification unit 152, used to pass the optimized medical order-condition association feature vector through a medical order identification classifier to obtain a classification result, and the classification result is used to indicate whether the medical order is reasonable.

[0046] In particular, in the technical solution of the present application, if the data collection of the patient's medical history, diagnosis results, historical medication and doctor's order content is incomplete or erroneous, it will directly affect the accuracy of word segmentation and word embedding, and then affect the quality of the feature vector and the performance of the classifier. The word segmentation process may cause information loss due to insufficient recognition of professional terms, abbreviations or new words. The word embedding layer may not be able to fully capture the semantic information in the medical text. The context encoder and the multi-scale mixed convolution semantic encoder may not be able to fully understand the complex contextual relationships and nuances in the medical text. If the encoder's training data is insufficient or the model parameters are improperly set, it may not be possible to accurately extract the semantic features of the patient's condition, which will lead to the doctor's order-condition association feature vector. When the doctor's order identification classifier is used for classification judgment, there will be a class harmony offset relative to the predetermined class label of the doctor's order identification classifier, affecting the classification iteration effect. In order to improve the accuracy of doctor's order identification and the iterative effect of the classifier, in the technical solution of the present application, the doctor's order-condition association feature vector is predicted based on the probability label of the predetermined class activated by probability to obtain an optimized doctor's order-condition association feature vector.

[0047] Specifically, in an embodiment of the present application, a possible implementation method of performing a probability label prediction of a predetermined class based on probabilistic activation on the medical order-condition association feature vector to obtain an optimized medical order-condition association feature vector may be: performing a probabilistic activation based on a Sigmoid function on the medical order-condition association feature vector to obtain a probabilistic medical order-condition association feature vector; inputting the medical order-condition association feature vector into the medical order identification classifier to obtain an initial class label probability value; calculating the difference between the feature value of each position in the probabilistic medical order-condition association feature vector and the initial class label probability value to obtain a difference feature vector, and passing the difference feature vector through a mask function to obtain an initial weight vector for viewpoint metric prediction. Calculate the autocorrelation matrix of the viewpoint metric prediction initial weight vector, divide the autocorrelation matrix of the viewpoint metric prediction initial weight vector by the square root of the length of the viewpoint metric prediction initial weight vector, and then multiply it by the viewpoint metric prediction initial weight vector through the softmax function to obtain the viewpoint metric prediction attention-oriented weight vector; multiply the viewpoint metric prediction attention-oriented weight vector and the medical advice-condition association feature vector by position point to obtain the first weighted medical advice-condition association feature vector; multiply the predetermined weight hyperparameter by the medical advice-condition association feature vector, and then add it to the first weighted medical advice-condition association feature vector by position to obtain the optimized medical advice-condition association feature vector.

[0048] In an embodiment of the present application, another expression of performing a probability label prediction of a predetermined class based on probabilistic activation on the medical advice-condition association feature vector to obtain an optimized medical advice-condition association feature vector may be: processing the medical advice-condition association feature vector with the following formula to obtain the optimized medical advice-condition association feature vector; wherein the formula is:

[0049] V cp =sigmoid(V co )

[0050]

[0051] Among them, V co represents the doctor's advice-condition association feature vector, sigmoid represents the activation function, V cp represents the probabilistic doctor's advice-condition association feature vector, P represents the initial class label probability value, represents positional subtraction, mask represents the mask function, x represents the variable of the input mask function, θ represents the predetermined threshold, V e represents the initial weight vector for viewpoint metric prediction, represents matrix multiplication, T represents the transpose of the vector, L represents the length of the initial weight vector for viewpoint metric prediction, Softmax represents the normalized exponential function, V ea represents the viewpoint metric prediction attention-oriented weight vector, α represents the predetermined weight hyperparameter, ⊙ represents the position point multiplication, Indicates addition by position, V' co Represents the optimized doctor's advice-condition association feature vector.

[0052] That is, in the technical solution of the present application, the medical order-condition association feature vector is predicted to have a predetermined class probability label based on probabilistic activation. First, the medical order-condition association feature vector is subjected to probabilistic activation based on the Sigmoid function to obtain a probabilistic medical order-condition association feature vector, and the medical order-condition association feature vector is input into the medical order identification classifier to obtain an initial class label probability value, and the viewpoint metric prediction contribution of each position in the standardized probabilistic medical order-condition association feature vector to the predetermined class probability label is initialized through inverse error analysis and masking.

[0053] Then, the viewpoint space of the reverse optimization is expanded by predicting the autocorrelation attention representation of the initial weight vector with the viewpoint metric, and using its sub-dimensional consistency as the attention-guided weight, the reverse consistency attention mechanism of the sub-dimensional distribution is applied to the feature vector as the medical order-condition association. In this way, the viewpoint metric prediction based on post-machine learning is used to perform learning predictive matching between the forward feature manifold and the backward class probability distribution, so that the optimized medical order-condition association feature vector has a relatively better distribution consistency with respect to the class probability distribution of the medical order identification classifier, so as to improve its classification judgment accuracy through the medical order identification classifier.

[0054] Finally, the optimized doctor's order-condition association feature vector is passed through a doctor's order identification classifier to obtain a classification result, and the classification result is used to indicate whether the doctor's order is reasonable. The doctor's order identification classifier is a trained machine learning model, which can learn the association pattern between the doctor's order information and the patient's condition related information during training, and then judge the rationality of the doctor's order. The classification result is used to indicate the rationality of the doctor's order, that is, to determine whether the doctor's order is consistent with the patient's condition. By evaluating and judging the rationality of the doctor's order, unreasonable doctor's orders can be discovered and corrected in a timely manner, the quality of medical care can be improved, and the safety and health of patients can be guaranteed.

[0055] In an embodiment of the present application, the optimized medical order-condition association feature vector is passed through a medical order identification classifier to obtain a classification result. The classification result is used to indicate whether the medical order is reasonable. One implementation method may be: use the fully connected layer of the medical order identification classifier to fully connect encode the optimized medical order-condition association feature vector to obtain a fully connected encoded feature vector; input the fully connected encoded feature vector into the Softmax classification function of the medical order identification classifier to obtain the probability value of the optimized medical order-condition association feature vector belonging to each classification label, and the classification labels include those for indicating that the medical order is reasonable and those for indicating that the medical order is unreasonable; and determine the classification label corresponding to the largest of the probability values ​​as the classification result.

[0056] In summary, the big data-based smart medical platform 100 according to the embodiment of the present application is explained, which extracts the semantic features of the patient's condition by performing feature extraction in the semantic space on the patient's medical history, diagnosis results, examination results, and historical medication conditions, and extracts the semantic features of the doctor's order content by performing feature extraction in the semantic space on the doctor's order content, and then establishes a connection between the patient's condition and the doctor's order content in the same feature space, and then uses a classifier to determine whether the doctor's order is reasonable. In this way, it is possible to accurately determine whether the doctor's order meets the patient's actual situation and needs, thereby reducing the medical error rate.

[0057] As described above, the smart medical platform 100 based on big data according to the embodiment of the present application can be implemented in various terminal devices, such as a server for smart medical care based on big data. In one example, the smart medical platform 100 based on big data according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the smart medical platform 100 based on big data can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the smart medical platform 100 based on big data can also be one of the many hardware modules of the terminal device.

[0058] Alternatively, in another example, the big data-based smart medical platform 100 and the terminal device may also be separate devices, and the big data-based smart medical platform 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.

[0059] Figure 5 FIG. 1 is a flow chart of a smart medical method based on big data according to an embodiment of the present application. Figure 5As shown, in the smart medical method based on big data, it includes: S110, obtaining the patient's medical history, diagnosis results, historical medication conditions and medical order content; S120, performing semantic feature encoding on the patient's medical history, diagnosis results and historical medication conditions to obtain a multi-scale patient condition semantic understanding feature vector; S130, performing semantic feature encoding on the medical order content to obtain a medical order content semantic understanding feature vector; S140, associating the multi-scale patient condition semantic understanding feature vector with the medical order content semantic understanding feature vector to obtain a medical order-condition association feature vector; S150, based on the medical order-condition association feature vector, obtaining an identification result of whether the medical order is reasonable.

[0060] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned smart medical method based on big data have been referred to above. Figures 1 to 4 The description of the big data-based smart medical platform has been introduced in detail, and therefore, its repeated description will be omitted.

[0061] In summary, the big data-based smart medical method based on the embodiment of the present application is explained, which extracts the semantic features of the patient's condition by performing feature extraction in the semantic space on the patient's medical history, diagnosis results, examination results, and historical medication conditions, and extracts the semantic features of the doctor's order content by performing feature extraction in the semantic space on the doctor's order content, and then establishes a connection between the patient's condition and the doctor's order content in the same feature space, and then uses a classifier to determine whether the doctor's order is reasonable. In this way, it is possible to accurately determine whether the doctor's order meets the patient's actual situation and needs, thereby reducing the medical error rate.

[0062] In the several embodiments provided in this application, it should be understood that the disclosed methods, systems, etc. can be implemented in other ways. For example, the system embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0063] The modules described as separate components may or may not be physically separated, and the components shown as modules 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 modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0064] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0065] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application.

[0066] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present application is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present application. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0067] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The second and other words are used to indicate names, but not to indicate any particular order.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical solution of the present application may be modified or replaced by equivalents without departing from the spirit of the technical solution of the present application.

Claims

1. A smart medical platform based on big data, characterized in that: include: Medical-related data collection module, used to obtain the patient's medical history, diagnosis results, historical medication, and doctor's orders; A patient condition feature encoding module, used for performing semantic feature encoding on the patient's medical history, diagnosis results and historical medication conditions to obtain a multi-scale patient condition semantic understanding feature vector; A medical order content feature encoding module, used for encoding the semantic features of the medical order content to obtain a medical order content semantic understanding feature vector; A doctor's order-condition correlation feature generation module, used to correlate the multi-scale patient's condition semantic understanding feature vector with the doctor's order content semantic understanding feature vector to obtain a doctor's order-condition correlation feature vector; The module for generating reasonable medical advice identification results is used to obtain the identification result of whether the medical advice is reasonable based on the medical advice-condition association feature vector.

2. The big data-based smart medical platform according to claim 1, characterized in that: The patient condition characteristic coding module comprises: A patient condition-related word embedding unit is used to segment the patient's medical history, diagnosis results, and historical medication conditions and pass them through a word embedding layer to obtain multiple patient condition-related attribute word embedding vectors; A first scale feature generating unit, configured to extract contextual semantic features from the embedding vectors of the plurality of patient condition-related attribute words to obtain a first scale patient condition semantic understanding feature vector; A second scale feature generating unit is used to extract features from a two-dimensional input matrix of patient condition related attribute words formed by arranging the plurality of patient condition related attribute word embedding vectors to obtain a second scale patient condition semantic understanding feature vector; A multi-scale feature fusion unit is used to fuse the first-scale patient condition semantic understanding feature vector and the second-scale patient condition semantic understanding feature vector to obtain the multi-scale patient condition semantic understanding feature vector.

3. The big data-based smart medical platform according to claim 2, characterized in that: The first scale feature generating unit is used to: pass the embedding vectors of the plurality of patient condition related attribute words through a context encoder to obtain the first scale patient condition semantic understanding feature vector.

4. The big data-based smart medical platform according to claim 3, characterized in that: The second-scale feature generating unit is used to: arrange the multiple patient condition-related attribute word embedding vectors into a patient condition-related attribute word two-dimensional input matrix and then pass it through a multi-scale mixed convolution semantic encoder to obtain the second-scale patient condition semantic understanding feature vector.

5. The big data-based smart medical platform according to claim 4, characterized in that: The patient condition feature encoding module is used to: segment the medical order content and pass it through a medical order content semantic feature encoder to obtain the medical order content semantic understanding feature vector.

6. The big data-based smart medical platform according to claim 5, characterized in that: The context encoder is a converter-based context semantic encoder, and the medical order content semantic feature encoder is a converter-based content semantic feature encoder including a word embedding layer.

7. The big data-based smart medical platform according to claim 6, characterized in that: The multi-scale hybrid convolutional semantic encoder includes a first convolution branch structure, a second convolution branch structure, a third convolution branch structure and a fourth convolution branch structure in parallel, and a multi-scale fusion structure connected to the first to fourth convolution branch structures, wherein the first convolution branch uses a first convolution kernel having a first size, the second convolution branch uses a second convolution kernel having a first size and a first void ratio, the third convolution branch uses a third convolution kernel having a first size and a second void ratio, and the fourth convolution branch uses a fourth convolution kernel having a first size and a third void ratio.

8. The big data-based smart medical platform according to claim 7, characterized in that: The module for generating reasonable medical order identification results includes: A feature optimization unit, used for predicting the predetermined class probability label based on probabilistic activation on the medical advice-condition association feature vector to obtain an optimized medical advice-condition association feature vector; The feature vector classification unit is used to pass the optimized medical advice-condition association feature vector through a medical advice identification classifier to obtain a classification result, and the classification result is used to indicate whether the medical advice is reasonable.

9. The big data-based smart medical platform according to claim 8, characterized in that: The feature optimization unit is used to: Performing probabilistic activation based on the Sigmoid function on the doctor's advice-condition association feature vector to obtain a probabilistic doctor's advice-condition association feature vector; Inputting the medical advice-condition association feature vector into the medical advice identification classifier to obtain an initial class label probability value; Calculating the difference between the feature value of each position in the probabilistic doctor's advice-condition association feature vector and the initial class label probability value to obtain a difference feature vector, and passing the difference feature vector through a mask function to obtain an initial weight vector for viewpoint metric prediction; Calculating the autocorrelation matrix of the viewpoint metric prediction initial weight vector, dividing the autocorrelation matrix of the viewpoint metric prediction initial weight vector by the square root of the length of the viewpoint metric prediction initial weight vector, and then multiplying it by the viewpoint metric prediction initial weight vector through a softmax function to obtain a viewpoint metric prediction attention-oriented weight vector; Multiplying the viewpoint metric prediction attention-oriented weight vector by the doctor's advice-condition association feature vector by position point to obtain a first weighted doctor's advice-condition association feature vector; The predetermined weight hyperparameter is multiplied by the medical advice-condition association feature vector, and then the resultant is added to the first weighted medical advice-condition association feature vector by position to obtain the optimized medical advice-condition association feature vector.

10. A smart medical method based on big data, characterized in that: include: Obtain the patient's medical history, diagnosis, medication history, and doctor's orders; Performing semantic feature encoding on the patient's medical history, diagnosis results, and historical medication to obtain a multi-scale patient condition semantic understanding feature vector; Performing semantic feature encoding on the medical order content to obtain a semantic understanding feature vector of the medical order content; Associating the multi-scale patient condition semantic understanding feature vector with the doctor's order content semantic understanding feature vector to obtain a doctor's order-condition association feature vector; Based on the doctor's advice-condition association feature vector, an identification result of whether the doctor's advice is reasonable is obtained.

Citation Information

Cited By

  • Data analysis system and method for intraperitoneal hyperthermic perfusion chemotherapy

    CN119541768A

  • Oral patient postoperative doctor advice and follow-up visit automatic classification and management system

    CN122158098A