Medical archive management method and device, storage medium and program product

Abstract

The embodiment of the invention discloses a medical archive management method and device, a storage medium and a program product. The method comprises the following steps: collecting patient information, and establishing a medical file for each patient based on the patient information; in response to an input request of a medical worker for new information of a patient, collecting a voice signal of the medical worker and converting the voice signal into text information; matching the text information with the medical file, and querying a plurality of target fields; converting the text information into a value of a target field, and mapping the value into the medical file; and in response to a submission request of a medical worker, performing self-inspection on the medical file, and reminding the medical worker to correct the medical file when the key information is lost or the information is wrong. According to the method, the personalized medical archive of the patient can be created, and when new information is generated, medical personnel are assisted to input and self-check the new information in time, so that the updating of the medical archive is quickly completed, and the management of the medical archive is facilitated.

Description

Technical Field

[0001] The present disclosure relates to the field of medical health technology, and in particular to a medical archive management method, device, storage medium, and program product. Background Art

[0002] In traditional medical record management, manual records and paper documents are often relied on, which is not only inefficient but also prone to errors and difficult to meet the needs of the modern medical system. With the development of information technology, medical records are gradually relied on for management by medical systems. However, the medical system is composed of multiple subsystems, which may use different standards and formats to store data, resulting in fragmentation of patient information and difficulty in forming a complete medical record. In addition, as patients return for follow-up visits, new medical data will appear, and manual entry and other methods are difficult to meet the needs of updating medical records, which makes medical records in a state of information lag. Summary of the invention

[0003] In view of this, the embodiments of the present disclosure provide a medical record management method, device, storage medium, and program product, which can complete the creation of personalized medical records for patients, and when new information is generated, assist medical staff to enter and self-check new information in a timely manner, quickly complete the update of medical records, improve the timeliness, completeness and accuracy of medical records, and facilitate the management of medical records through intelligent means.

[0004] In a first aspect, the present disclosure provides a medical record management method, which adopts the following technical solution:

[0005] Collecting patient information and establishing a medical record for each patient based on the patient information;

[0006] In response to a medical staff's request to enter new patient information, the medical staff's voice signal is collected and converted into text information;

[0007] Matching the text information with the medical records to query a number of target fields;

[0008] Convert the text information into a value of a target field and map it into the medical record;

[0009] In response to the submission request of the medical staff, the medical file is self-checked, and when key information is missing or wrong, the medical staff is reminded to correct the medical file.

[0010] Optionally, the patient information includes medical images of the patient;

[0011] Standardize medical images;

[0012] According to the type of medical image and the detection part of the medical image, available processing technology is selected to pre-process the medical image;

[0013] The processed medical image is input into the image processing model, and the diagnosis result output by the image processing model is associated with the medical image before processing and stored in the medical file.

[0014] Optionally, collecting the voice signals of medical staff and converting them into text messages includes:

[0015] Train the preset Gaussian mixture model to obtain the initial parameter set;

[0016] Preprocessing and feature extraction of the speech signal to obtain speech features;

[0017] Based on the parameter set, the speech feature and the Q function, predict the next word and obtain the current Q value, where Q represents the state action value;

[0018] Using a temporal difference learning function to update the current Q value, and optimizing the Gaussian mixture model based on the update result;

[0019] Use the Riemann stochastic gradient descent function to obtain a new set of parameters;

[0020] Updating the Q function using a new set of parameters;

[0021] The next word is converted into text information.

[0022] Optionally, the expression of the Q function is as follows:

[0023]

[0024] Where Q(s,a) is the current Q value; s is the speech feature; a is the next word predicted; K is the number of Gaussian components; w k is the mixing weight of the kth Gaussian component; μ k is the mean vector of the kth Gaussian component; Σ k is the covariance matrix of the kth Gaussian component; N(μ k ,Σ k ) is the probability density function of the kth Gaussian component;

[0025] The expression of the temporal difference learning function is as follows:

[0026] δ t = r + γ max a′ Q(s′,a′)-Q(s,a);

[0027] Among them, δ tis the temporal difference error; r is the immediate reward; γ is the discount factor; s' is the next speech feature; a' is the word selected under the next speech feature; max a′ Q(s′,a′) is the maximum Q value under the next speech feature;

[0028] The expression of the Riemann stochastic gradient descent function is as follows:

[0029] θ′=Retr(θ-η·grad R Q(θ));

[0030] Among them, θ' is the new parameter set; θ is the current parameter set; Retr(·) is the retraction operation; η is the Riemann learning rate; grad R is the Riemann gradient; Q(θ) is the updated current Q value.

[0031] Optionally, the text information is matched with the medical file to query several target fields, including:

[0032] Using word segmentation technology to convert the text information into a word sequence;

[0033] Performing vector conversion on each word in the word sequence to obtain multiple word vectors;

[0034] Based on the self-attention mechanism, obtain the dynamic score of each word vector;

[0035] Dynamically weight the dynamic scores to obtain multiple weighted word vectors;

[0036] Clustering the multiple weighted word vectors into a plurality of groups, and combining the word vectors of each group into a new vector;

[0037] The medical file includes multiple file pages, and all original fields on the current file page being used by the medical staff are extracted;

[0038] A similarity value between the new vector and each original field is obtained, and an original field whose similarity value is greater than a preset threshold is selected as a target field that matches the new vector.

[0039] Optionally, the calculation formula of the dynamic score is as follows:

[0040]

[0041] Among them, E i is the dynamic score of the i-th word vector; D ii' is the self-attention score between the i-th word vector and the i'th word vector, D ii' =AttentionScore(v i ,vi' ), AttentionScore(·) is the attention score function, v i is the i-th word vector, v i' is the i'th word vector, i∈(1,I), i'∈(1,I), and I is the total number of word vectors.

[0042] Optionally, the calculation formula of the weighted word vector is as follows:

[0043]

[0044] in, is the i-th weighted word vector; τ i is the dynamic weight of the i-th word vector, E i” is the dynamic score of the “i”-th word vector.

[0045] In a second aspect, the present disclosure also provides a medical record management system, which adopts the following technical solution:

[0046] A collection module, used to collect patient information and establish a medical file for each patient based on the patient information;

[0047] The collection module is used to respond to the medical staff's request to enter new patient information and collect YY+242247P

[0048] The voice signals of medical staff are converted into text messages;

[0049] A matching module, used to match the text information with the medical records and query a number of target fields;

[0050] A conversion module, used for converting the text information into a value of a target field and mapping it into the medical record;

[0051] The self-check module is used to perform a self-check on the medical file in response to a submission request from a medical staff, and to remind the medical staff to correct the medical file when key information is missing or wrong.

[0052] In a third aspect, the embodiments of the present disclosure further provide a computer device, which adopts the following technical solution:

[0053] The computer device comprises:

[0054] at least one processor; and,

[0055] a memory communicatively connected to the at least one processor; wherein,

[0056] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any of the medical record management methods described above.

[0057] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute any of the medical record management methods described above.

[0058] In a fifth aspect, an embodiment of the present disclosure further provides a computer program product, including a computer program / instruction, which implements the steps of any of the above methods when executed by a processor.

[0059] The medical record management method provided by the embodiment of the present disclosure can significantly reduce the time and workload of manual input by collecting the voice signals of medical staff and converting them into text information. In particular, in emergency situations, medical staff can record patient information more quickly, thereby improving the efficiency of diagnosis and treatment, and automatically matching the entered text information with the target field in the medical record, reducing the errors and cumbersome steps of manual operation, making the information entry more accurate and rapid. This automatic matching and mapping of the target field makes the management of medical records more standardized, which helps to establish a unified medical data management specification and improve the management level of medical institutions. Converting text information into the value of the target field ensures that the information is consistent with the standard format of the medical record, reduces format errors and information omissions that may occur when entering information, and automatically performs self-checks before medical staff submit medical records. When key information is found to be missing or wrong, the medical staff will be immediately reminded to make corrections, which helps to ensure the accuracy and completeness of the medical record, avoid treatment errors caused by information errors, and further improve the standardization and reliability of medical records.

[0060] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the following preferred embodiments are specifically cited and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0062] Figure 1A flowchart of a medical records management method provided by an embodiment of the present disclosure;

[0063] Figure 2 A flowchart of a speech-to-text conversion method provided in an embodiment of the present disclosure;

[0064] Figure 3 A flowchart of a target field determination method provided by an embodiment of the present disclosure;

[0065] Figure 4 A block diagram of the medical records management system provided by the embodiment of the present disclosure;

[0066] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0067] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0068] It should be clear that the following embodiments of the present disclosure are described by specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other in the absence of conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0069] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms, and any specific structures and / or functions described herein are merely illustrative. Based on this disclosure, those skilled in the art should understand that an aspect described herein may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described herein may be used to implement the device and / or practice the method. In addition, one or more of the aspects described herein may be used.

[0070] Other structure and / or functionality besides those described above may be used to implement this apparatus and / or practice this method.

[0071] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The drawings only show components related to the present disclosure rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0072] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0073] Reference Figure 1 The present disclosure provides a medical record management method, comprising the following steps:

[0074] S1: Collect patient information and establish a medical record for each patient based on the patient information;

[0075] S2: In response to the medical staff's request to enter new patient information, the medical staff's voice signal is collected and converted into text information;

[0076] S3: Match the text information with the medical records and query several target fields;

[0077] S4: Convert text information into target field values ​​and map them into medical records;

[0078] S5: In response to the submission request from the medical staff, the medical file is self-checked, and when key information is missing or incorrect, the medical staff is reminded to correct the medical file.

[0079] The medical record management method provided by the present disclosure can significantly reduce the time and workload of manual input by collecting the voice signals of medical staff and converting them into text information. In particular, in emergency situations, medical staff can record patient information more quickly, thereby improving the efficiency of diagnosis and treatment, automatically matching the entered text information with the target field in the medical record, reducing the errors and cumbersome steps of manual operation, making the information entry more accurate and rapid, and this automatic matching and mapping of the target field makes the management of medical records more standardized, which helps to establish a unified medical data management specification and improve the management level of medical institutions. Converting text information into the value of the target field ensures that the information is consistent with the standard format of the medical record, reduces the format errors and information omissions that may occur when entering information, and automatically performs self-checking before the medical staff submits the medical record. When it is found that key information is missing or wrong, the medical staff will be immediately reminded to make corrections, which helps to ensure the accuracy and completeness of the medical record, avoid treatment errors caused by information errors, and further improve the standardization and reliability of the medical record.

[0080] In summary, this medical record management method not only improves the efficiency of information entry and management through automation, standardization and intelligent means, but also ensures the accuracy and completeness of information, enhances data security, promotes the sharing and collaboration of medical information, and ultimately promotes the overall quality and efficiency of medical services.

[0081] S1: Collect patient information and establish a medical record for each patient based on the patient information;

[0082] In S1, the basic information of patients, including name, gender, date of birth, ID number, contact information, etc., is collected through the hospital information system (HIS). When patients receive services, such as registration, medical consultation, examination, medication, hospitalization, surgery, etc., the information system automatically collects and updates the patients' medical information.

[0083] Using the patient's ID number, medical record number, biometric information and other core data as identification tags, a medical file is created for each patient in the hospital information system. The medical record number is a unique number assigned to the patient by the medical institution when the patient visits the hospital.

[0084] When collecting medical images of patients, a preset image processing model is used to analyze the medical images to obtain the patient's diagnosis results, and the medical images before processing are associated with the diagnosis results, and then included in the medical files as part of the medical files.

[0085] Specifically, the collected medical images are standardized and preprocessed, wherein the preprocessing includes adjusting the image size (e.g., 224x224 pixels), contrast enhancement, noise removal, and other operations. Preprocessing refers to selecting available processing techniques to process the medical images according to the type of medical images and the detection site of the medical images. For example, if the detection site is the brain and the medical image is a magnetic resonance imaging (MRI), a skull stripping algorithm is used to remove the skull and only retain the brain tissue.

[0086] The processed medical image is input into the image processing model to output the patient's diagnosis result, wherein the image processing model uses a convolutional neural network (such as ResNet50) as the backbone network for extracting image features, and the backbone network includes 5 stages. The first stage includes a Conv2d layer and a maximum pooling layer, the second stage includes 3 residual blocks, each residual block includes 3 convolutional layers, the third stage includes 4 residual blocks, each residual block includes 3 convolutional layers, the fourth stage includes 6 residual blocks, each residual block includes 3 convolutional layers, and the fifth stage includes 3 residual blocks, each residual block includes 3 convolutional layers.

[0087] A neuron interaction module is added between two adjacent residual blocks (ResNet) to allow features at different levels to interact and exchange information, thereby enhancing the feature extraction capability. The neuron interaction module refers to adding an attention mechanism between the second residual block and the third residual block in the third stage, so that the network can focus on important areas of the image, improve the accuracy of diagnosis, and improve the detection effect of lesions. In addition, the second convolutional layer of the first residual block in the third stage is cross-connected with the first convolutional layer of the first residual block in the fourth stage, and the third convolutional layer of the first residual block in the fourth stage is cross-connected with the second convolutional layer of the first residual block in the fifth stage, so as to achieve the interaction between low-level features and high-level features and enhance the feature extraction capability.

[0088] Auxiliary prediction branches are added after the 3rd residual block in the 2nd stage, the 4th residual block in the 3rd stage, the 6th residual block in the 4th stage, and the 3rd residual block in the 5th stage, respectively. Each auxiliary prediction branch can generate a preliminary diagnosis result (normal / mild / moderate / severe).

[0089] The main prediction branch and each auxiliary prediction branch of the image processing model use the cross entropy loss function, but the cross entropy loss function used by the auxiliary prediction branch has a smaller weight. The total loss function is: L total =0.5*L main +0.2*L aux2 +0.15*L aux3 +0.15*L aux4 , where L total is the total loss function, L main is the loss function of the main prediction branch, L aux2 is the loss function of the second auxiliary prediction branch (i.e., added after the fourth residual block in the third stage), L aux3 is the loss function of the third auxiliary prediction branch (i.e., added after the sixth residual block in the fourth stage), L aux4 is the loss function of the 4th auxiliary prediction branch (i.e., added after the 3rd residual block in the 5th stage).

[0090] The image processing model is trained using the Adam optimizer with an initial learning rate of 0.001. In the forward propagation, the output of the main prediction branch and the output of all auxiliary prediction branches are calculated simultaneously. In the back propagation, the gradient information of all prediction branches is used to update the network parameters. Since medical images are usually large, the gradient accumulation technique is used to update the parameters every 4 small batches (16 medical images per batch), which is equivalent to using a batch size of 64 images. Dynamic batch normalization is implemented to adapt to the image differences produced by X-ray equipment in different hospitals. A larger DenseNet201 is trained as the teacher model, and the backbone network is used as the student model. The soft labels of the teacher model are used to guide the training of the student model. After the training is completed, the auxiliary prediction branches of the student model are removed to obtain the image processing model. The image processing model is used for actual deployment. In use, the diagnostic results output by the image processing model include the probability distribution of the disease type. Appropriate probability thresholds are set for different disease types. When the probability of the output disease type is greater than the corresponding probability threshold, the patient is judged to have the corresponding category of disease, and the severity of the disease is determined according to the size of the probability. Use TensorRT to optimize the performance of the image processing model and improve the inference speed. Collect feedback from medical staff on the model and fine-tune the image processing model every month using newly collected data and feedback to keep the image processing model up to date and accurate.

[0091] Compared with traditional methods, the above method reduces training time by about 30% and achieves 94% accuracy on the test set, which is 5% higher than the baseline model. The processing time for each medical image is less than 100 milliseconds, which meets real-time diagnosis needs and can adapt to X-ray equipment in different hospitals to maintain stable performance.

[0092] In S2, medical records include multiple file pages, each of which carries different types of information, such as basic personal information, clinical diagnosis records, treatment plans, drug prescriptions, examination reports, etc. In order to protect patient privacy and ensure data security, each file page is set with different access rights, which are defined based on the roles and responsibilities of medical staff. This strict permission control mechanism ensures that only authorized medical staff can access and modify the file pages. At the same time, all operation records will be automatically recorded by the system for easy auditing and tracing, thereby enhancing data security. Through a unified medical record management system, medical staff from different departments can more easily share and access patient information, thereby promoting cross-departmental collaboration and communication and improving overall diagnosis and treatment efficiency. In telemedicine scenarios, the system can support medical staff to quickly obtain patients' historical medical records and provide more accurate remote diagnosis and treatment recommendations.

[0093] When medical staff need to view or edit a patient's medical records, they must first pass identity authentication, such as using a username and password, or more secure biometric technology such as fingerprint scanning or facial recognition. Once verified, the corresponding file page is automatically launched based on the medical staff's role and permissions.

[0094] The system provides a trigger mechanism. When the current file page is activated, that is, when the medical staff opens a specific file page to view or update the patient information, options such as "automatic entry", "view only", and "self-editing" are displayed in the system interface. If the medical staff selects the "view only" option, the medical staff is only granted the right to view the current file page; if the medical staff selects the "self-editing" option, the medical staff is granted the right to view and manually edit the current file page, and supports the medical staff to edit using devices such as a mouse and keyboard; if the medical staff selects the "automatic entry" option, the medical staff is granted the right to view, voice edit, and manual edit the current file page, triggering the medical staff's request to enter new patient information. In response to the entry request, a microphone or other recording device is used to collect the medical staff's voice signal.

[0095] Reference Figure 2 The flow chart of the speech-to-text method shown in the figure shows that the method for converting a speech signal into text information includes the following steps:

[0096] S21: training a preset Gaussian mixture model to obtain an initial parameter set;

[0097] S22: Preprocessing and feature extraction of speech signals to obtain speech features;

[0098] S23: Based on the parameter set, speech features and Q function, predict the next word and obtain the current Q value;

[0099] S24: using a temporal difference learning function to update the current Q value, and optimizing the Gaussian mixture model based on the update result;

[0100] S25: Use the Riemann stochastic gradient descent function to obtain a new set of parameters;

[0101] S26: Update the Q function using a new parameter set;

[0102] S27: Convert the next word into text information.

[0103] In S21, the parameter set includes the mixing weights, mean vectors and covariance matrix of each Gaussian component of the Gaussian mixture model (GMM). The Gaussian component refers to the Gaussian distribution, the mixing weight represents the proportion of the Gaussian component in the entire data set, the mean vector represents the center position of the Gaussian component, and the covariance matrix represents the degree of dispersion of the data points in the Gaussian component.

[0104] In S22, preprocessing includes denoising, segmentation and endpoint detection. Denoising refers to the use of various filters to remove background noise to improve the clarity of the speech signal. Segmentation refers to dividing the continuous speech stream into smaller units, such as sentences or words. Endpoint detection refers to determining the beginning and end of the speech segment for subsequent processing. Features are extracted from the preprocessed speech signal to form speech features. The speech features can specifically be Mel-frequency cepstral coefficients (MFCCs), or a pre-trained language model (such as the BERT model) is used to extract features from the preprocessed speech signal, that is, the preprocessed speech signal is converted into a high-dimensional vector, namely, speech features, so that medical terms and common expressions can be specially encoded to enhance domain specificity.

[0105] In S23, the expression of Q-function (action-value function) is as follows:

[0106]

[0107] Where Q(s,a) is the current Q value (where Q refers to Q-value, indicating the quality value or state action value); s is the speech feature, indicating the current state; a is the predicted next word, indicating the action taken in the current state; K is the number of Gaussian components; w k is the mixing weight of the kth Gaussian component; μ k is the mean vector of the kth Gaussian component; Σ k is the covariance matrix of the kth Gaussian component; N(μ k ,Σ k ) is the probability density function of the kth Gaussian component.

[0108] Substitute the parameters and speech features in the parameter set into the expression of the Q function, predict the next word and calculate the current Q value.

[0109] In S24, the expression of the temporal difference learning function is as follows:

[0110] δ t = r + γ max a′ Q(s′,a′)-Q(s,a);

[0111] Among them, δ t is the temporal difference error; r is the immediate reward, which is usually calculated based on the accuracy of the prediction, for example, if the prediction is correct, a positive reward is given; if it is wrong, a negative reward is given; γ is a discount factor used to measure the current value of future rewards; s' is the next speech feature; a' is the word selected under the next speech feature; max a′ Q(s′,a′) is the maximum Q value under the next speech feature.

[0112] The temporal difference error is calculated through the temporal difference learning function, and the current Q value is updated based on the temporal difference error, which reflects the immediate reward and expected future return after taking a specific action, and improves the Gaussian mixture model's estimation of the action value, so that the intelligent agent can make better decisions. By continuously updating the Q value, the intelligent agent can learn the long-term rewards of taking different actions in different states, and then optimize its strategy to obtain the maximum cumulative reward.

[0113] Among them, the expression for updating the current Q value is as follows:

[0114] Q(s,a)←Q(s,a)+α·δ t ;

[0115] Among them, α is the TD learning rate, which is used to control the amplitude of Q value update.

[0116] In S25, the expression of the Riemann stochastic gradient descent function is as follows:

[0117] θ′=Retr(θ-η·grad R Q(θ));

[0118] Among them, θ' is the new parameter set, including the new mixing weights, mean vectors and covariance matrices of each Gaussian component; θ is the current parameter set; Retr(·) is a retraction operation, which is an operation that maps the vector in the tangent space back to the manifold itself; η is the Riemann learning rate, which is used to control the step size of parameter update on the Riemann manifold; grad R is the Riemann gradient, which refers to the gradient on the Riemann manifold; Q(θ) is the updated current Q value, which refers to the updated Q value obtained after adopting the current parameter set.

[0119] In S26, the new parameter set is substituted into the expression of the Q function to update the Q function so that it can be more accurate in the next prediction.

[0120] In S27, a pre-trained language model (such as a BERT model) is used to convert the next word into text information.

[0121] Among them, the training of the Gaussian mixture model adopts the reinforcement learning method, that is, feature extraction is performed based on the data set, and then S23 to S26 are repeated until convergence or a predefined number of iterations is reached. The performance of the Gaussian mixture model is regularly evaluated on the validation set, and early stopping is performed to prevent overfitting.

[0122] The above method converts the voice signals of medical staff into text information through continuous prediction and conversion, realizing an efficient speech-to-text process. In this process, temporal difference (TD) learning and Riemann stochastic gradient descent (RSGD) can work together to not only update the GMM parameters to improve the clustering or probability estimation ability of GMM, but also update the Q function to improve the decision-making process, ultimately achieving accurate speech-to-text conversion.

[0123] In S3, refer to Figure 3 A flowchart of a target field determination method is shown, where the method of "matching text information with medical records to query a number of target fields" includes the following steps:

[0124] S31: Use word segmentation technology to convert text information into word sequences;

[0125] S32: Perform vector conversion on each word in the word sequence to obtain multiple word vectors;

[0126] S33: Based on the self-attention mechanism, obtain the dynamic score of each word vector;

[0127] S34: dynamically adjust the weight of the dynamic score to obtain multiple weighted word vectors;

[0128] S35: clustering multiple weighted word vectors into several groups, and combining the word vectors of each group into a new vector;

[0129] S36: The medical file includes multiple file pages, and all original fields on the current file page being used by the medical staff are extracted;

[0130] S37: Obtain a similarity value between the new vector and each original field, and select an original field whose similarity value is greater than a preset threshold as a target field that matches the new vector.

[0131] In S31, the text information is T, and the word sequence is {C1, C2, ..., C i ,…,C I}, where C i is the i-th word, i∈(1,I), and I is the total number of words in the word sequence.

[0132] In S32, a word embedding method is used to convert each word in the word sequence into a word vector. Specifically, v i =WordEmbedding(C i ), WordEmbedding(·) is the word embedding function, v i is the i-th word vector.

[0133] In S33, the calculation formula of the dynamic score is as follows:

[0134]

[0135] Among them, E i is the dynamic score of the i-th word vector; D ii' is the self-attention score between the i-th word vector and the i'th word vector; i'∈(1,I), since the total number of word vectors is equal to the total number of words in the word sequence, I is also the total number of word vectors.

[0136] D ii' =AttentionScore(v i ,v i' );

[0137] Among them, AttentionScore(·) is the attention score function; v i' is the i'th word vector.

[0138] In S34, the calculation formula for dynamically adjusting the dynamic score is as follows:

[0139]

[0140] Among them, τ i is the dynamic weight of the i-th word vector; E i” is the dynamic score of the i”th word vector; i”∈(1,I).

[0141] The calculation formula of weighted word vector is as follows:

[0142]

[0143] in, is the i-th weighted word vector, that is, the representation of the i-th word vector after dynamic weight adjustment.

[0144] In S35, a clustering algorithm (such as K-means clustering algorithm) is used to cluster all weighted word vectors, and all weighted word vectors are divided into several groups according to the clustering results. The word vectors in each group are combined to construct a new vector.

[0145] In S36, optical character recognition (OCR) technology is used to extract all original fields in the current archive page.

[0146] In S37, a similarity algorithm such as cosine similarity or Jaccard similarity is used to calculate the similarity value between the new vector and each original field. If the similarity value is greater than a preset threshold, the corresponding original field is determined to be the target field.

[0147] In S4, according to the method of S3 above, the target field matching each new vector is queried, the new vector is converted into the value of the target field and mapped into the medical file, thereby completing the automatic filling of the current file page.

[0148] Here are some examples to illustrate the above:

[0149] The text information is: Patient Zhang San, it is recommended to monitor blood sugar level every day. After the text information is segmented and transformed into vectors, the word vector includes v 患者 、v 张三 、v 建议 、v 每日 、v 监测 and v 血糖水平 After the self-attention mechanism and dynamic weight adjustment, v 患者 and v 张三 Combined into the first new word vector, v 建议 、v 每日 、v 监测 and v 血糖水平 Combined into the second new word vector. The current archive page includes original fields such as "name", "current symptoms", and "treatment suggestions". Through similar value matching, "name" is determined as the target field of the first new word vector, and "treatment suggestions" is determined as the target field of the second new word vector. The two new word vectors are converted into the values ​​of the corresponding target fields and then automatically entered into the corresponding positions in the current archive page.

[0150] In S5, when the medical staff completes the entry of the current file page, they click the submit button or issue a voice command such as "submit", which triggers the medical staff's submission request. In response to the request, the system performs a self-inspection of the medical file.

[0151] There are two ways to self-check. The first is to use a predefined rule engine to check the integrity and accuracy of key information to determine whether key information is missing or incorrect. Information errors include format errors, numerical errors, word errors, etc. For example, for basic patient information, check whether it contains the patient's name, gender, age, contact information, etc. For medical records, check whether there are diagnosis results, treatment plans, medication records, etc. For dates and times, ensure that all date and time information is complete and reasonable.

[0152] The second is to use pre-trained machine learning models to detect possible errors or missing key information. Specifically, a large amount of medical file data is collected, and which files are missing or erroneous are manually marked. A classification model is trained using the marked data, and the trained classification model is deployed to the system. After triggering the submission request of the medical staff, the detection is automatically completed. If there is no missing key information or information error, the submission of the current file page is allowed and the current file page is saved to complete the update of the medical file. If there is missing key information or information error, an error prompt is issued to remind the medical staff to make changes until the medical staff makes the changes correctly, or the medical staff selects the "Submit Still" button after viewing the error prompt, then the submission of the current file page is allowed, and the content in the current file page is analyzed to optimize the system.

[0153] Furthermore, the medical records are checked regularly. When information is missing or wrong, the medical records are automatically updated to ensure the integrity and accuracy of the data. Regular checks can be set to once a week or a month. The specific frequency is determined according to the frequency of use and importance of the medical records. During the test, the system automatically scans all the information in the medical records. If the information is correct, the patient's relevant materials are queried based on the core data such as the patient's ID number, medical record number, and biometric information, and the missing information is extracted from the materials. For example, if the patient's age is missing, try to infer a reasonable age from the ID number. For the detected erroneous information, it is automatically corrected or marked as pending confirmation. For example, if unreasonable date or time information is detected, it is automatically adjusted or required to be confirmed by medical staff. The system will generate detailed update prompts, list all the information that needs to be corrected or supplemented, and send them to relevant medical staff. The prompts can be conveyed through system notifications, emails, or text messages. The system also provides a real-time feedback mechanism, allowing medical staff to check and confirm the updated content in time after receiving the update prompt. After confirmation, the update will be automatically applied to the medical records to ensure the latest and accuracy of the archive data.

[0154] In summary, this method integrates various medical information of patients, including text descriptions, image data, historical data, etc., to form a complete medical file, optimize the outpatient screening record process, ensure the accuracy and completeness of patient data, and use advanced data encryption and security technologies to ensure the storage and transmission security of data, so that patient data is not illegally obtained or leaked, and protect patient privacy. It realizes the intelligent management of patient outpatient screening files, including text entry, image upload, historical data query and other functions, reduces tedious manual operations, reduces the workload of medical staff, improves work efficiency, and provides reliable data support for subsequent diagnosis and treatment. The system can count and analyze the patient's historical data, provide a reference for doctors, and help doctors develop more scientific and effective treatment plans. Moreover, this method is not only applicable to hospital outpatient departments, but also can be applied to other fields that need to manage a large number of patient files, such as physical examination centers, community medical centers, etc., to help these institutions better manage patient file information, which will help improve the quality and level of overall medical services, thereby better meeting the needs and expectations of patients.

[0155] Reference Figure 4 The present disclosure provides a medical record management system, comprising:

[0156] A collection module 101 is used to collect patient information and establish a medical file for each patient based on the patient information;

[0157] The collection module 102 is used to respond to the medical staff's request for inputting new patient information, collect the medical staff's voice signal and convert it into text information;

[0158] A matching module 103 is used to match the text information with the medical records and query a number of target fields;

[0159] A conversion module 104, for converting text information into a value of a target field and mapping it into a medical record;

[0160] The self-check module 105 is used to perform a self-check on the medical file in response to the submission request of the medical staff, and remind the medical staff to correct the medical file when key information is missing or the information is wrong.

[0161] The various variations and specific examples of the medical record management method provided above are also applicable to the medical record management system provided by the present disclosure. Through the above detailed description of the medical record management method, those skilled in the art can clearly understand the implementation method of the medical record management system. For the sake of brevity of the specification, it will not be described in detail here.

[0162] The computer device according to the embodiment of the present disclosure includes a memory and a processor. The memory is used to store non-temporary computer-readable instructions. Specifically, the memory may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may, for example, include a read-only memory (ROM), a hard disk, a flash memory, etc.

[0163] The processor may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the computer device to perform desired functions. In one embodiment of the present disclosure, the processor is used to run the computer-readable instructions stored in the memory, so that the computer device performs all or part of the steps of the medical record management method of each embodiment of the present disclosure.

[0164] Those skilled in the art should be able to understand that in order to solve the technical problem of how to obtain a good user experience, the present embodiment may also include well-known structures such as a communication bus and an interface, and these well-known structures should also be included in the protection scope of the present disclosure.

[0165] like Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure is shown, which is a schematic diagram of the structure of a computer device suitable for implementing the embodiment of the present disclosure. Figure 5 The computer device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0166] like Figure 5 As shown, the computer device may include a processor (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) or a program loaded from a storage device into a random access memory (RAM). In the RAM, various programs and data required for the operation of the computer device are also stored. The processor, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0167] Typically, the following devices can be connected to the I / O interface: input devices such as sensors or visual information acquisition devices; output devices such as display screens; storage devices such as tapes, hard disks, etc.; and communication devices. The communication device can allow the computer device to communicate with other devices (such as edge computing devices) wirelessly or by wire to exchange data. Figure 5A computer device having various devices is shown, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0168] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processor, all or part of the steps of the medical archive management method of the embodiment of the present disclosure are executed.

[0169] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.

[0170] According to the computer-readable storage medium of the embodiment of the present disclosure, non-transitory computer-readable instructions are stored thereon. When the non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the medical record management method of each embodiment of the present disclosure are executed.

[0171] The above-mentioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or mobile hard disk), media with built-in rewritable non-volatile memory (e.g., memory card) and media with built-in ROM (e.g., ROM box).

[0172] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.

[0173] The basic principles of the present disclosure are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present disclosure are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present disclosure. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, and are not limitations. The above details do not limit the present disclosure to the necessity of adopting the above specific details to be implemented.

[0174] In the present disclosure, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. The block diagrams of the devices, devices, equipment, and systems involved in the present disclosure are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0175] Additionally, as used herein, "or" used in a list of items beginning with "at least one" indicates a disjunctive list, so that, for example, a list of "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C).

[0176] The word "exemplary" does not mean that the described example is preferred or better than other examples.

[0177] It should also be noted that in the system and method of the present disclosure, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present disclosure.

[0178] Various changes, substitutions, and modifications of the techniques described herein may be made without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of the present disclosure is not limited to the specific aspects of the processes, machines, manufactures, compositions of events, means, methods, and actions described above. Currently existing or later to be developed processes, machines, manufactures, compositions of events, means, methods, or actions that perform substantially the same functions or achieve substantially the same results as the corresponding aspects described herein may be utilized. Thus, the appended claims include such processes, machines, manufactures, compositions of events, means, methods, or actions within their scope.

[0179] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0180] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present disclosure to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A medical record management method, characterized in that: include: Collecting patient information and establishing a medical record for each patient based on the patient information; In response to a medical staff's request to enter new patient information, the medical staff's voice signal is collected and converted into text information; Matching the text information with the medical records to query a number of target fields; Convert the text information into a value of a target field and map it into the medical record; In response to the submission request of the medical staff, the medical file is self-checked, and when key information is missing or wrong, the medical staff is reminded to correct the medical file.

2. The medical record management method according to claim 1, characterized in that: The patient information includes medical images of the patient; Standardize medical images; According to the type of medical image and the detection part of the medical image, available processing technology is selected to pre-process the medical image; The processed medical image is input into the image processing model, and the diagnosis result output by the image processing model is associated with the medical image before processing and stored in the medical file.

3. The medical record management method according to claim 1, characterized in that: The collecting of voice signals of medical staff and converting them into text messages includes: Train the preset Gaussian mixture model to obtain the initial parameter set; Preprocessing and feature extraction of the speech signal to obtain speech features; Based on the parameter set, the speech feature and the Q function, predict the next word and obtain the current Q value, where Q represents the state action value; Using a temporal difference learning function to update the current Q value, and optimizing the Gaussian mixture model based on the update result; Use the Riemann stochastic gradient descent function to obtain a new set of parameters; Updating the Q function using a new set of parameters; The next word is converted into text information.

4. The medical record management method according to claim 3, characterized in that: The expression of the Q function is as follows: Where Q(s, a) is the current Q value; s is the speech feature; a is the next word predicted; K is the number of Gaussian components; w k is the mixing weight of the kth Gaussian component; μ k is the mean vector of the kth Gaussian component; ∑ k is the covariance matrix of the kth Gaussian component; N(μ k ,∑ k ) is the probability density function of the kth Gaussian component; The expression of the temporal difference learning function is as follows: d t =r+γ·max a′ Q(s′, a′)-Q(s, a); Among them, δ t is the temporal difference error; r is the immediate reward; γ is the discount factor; s′ is the next speech feature; a′ is the word selected under the next speech feature; max a′ Q(s′, a′) is the maximum Q value under the next speech feature; The expression of the Riemann stochastic gradient descent function is as follows: θ′=Retr(θ-η·grad R Q(θ); Among them, θ′ is the new parameter set; θ is the current parameter set; Retr(·) is the retraction operation; η is the Riemann learning rate; grad R is the Riemann gradient; Q(θ) is the updated current Q value.

5. The medical record management method according to claim 1, characterized in that: The text information is matched with the medical file to query several target fields, including: Using word segmentation technology to convert the text information into a word sequence; Performing vector conversion on each word in the word sequence to obtain multiple word vectors; Based on the self-attention mechanism, obtain the dynamic score of each word vector; Dynamically weight the dynamic scores to obtain multiple weighted word vectors; Clustering the multiple weighted word vectors into a plurality of groups, and combining the word vectors of each group into a new vector; The medical file includes multiple file pages, and all original fields on the current file page being used by the medical staff are extracted; A similarity value between the new vector and each original field is obtained, and an original field whose similarity value is greater than a preset threshold is selected as a target field that matches the new vector.

6. The medical record management method according to claim 5, characterized in that: The calculation formula of the dynamic score is as follows: Among them, E i is the dynamic score of the i-th word vector; D ii' is the self-attention score between the i-th word vector and the i'th word vector, D ii' =AttentionScore(v i ,v i' ), AttentionScore(·) is the attention score function, v i is the i-th word vector, v i' is the i'th word vector, i∈(1,I), i'∈(1,I), and I is the total number of word vectors.

7. The medical record management method according to claim 6, characterized in that: The calculation formula of the weighted word vector is as follows: in, is the i-th weighted word vector; τ i is the dynamic weight of the i-th word vector, E i” is the dynamic score of the “i”-th word vector.

8. A computer device, characterized in that: The computer device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the medical record management method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the medical record management method described in any one of claims 1-7.

10. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images