Risk assessment system based on medical record, electronic equipment and storage medium

Abstract

The invention discloses a risk assessment system based on medical records, an electronic device and a storage medium, the system comprises an acquisition module used for acquiring a to-be-processed medical record and medical record text data, a first extraction module used for calling a trained text extraction large model, performing extraction processing on a text belonging to a target semantic type in the medical record text data, and a second extraction module used for calling a trained text extraction large model; the training module is used for training medical record text data to obtain target text data extracted from the medical record text data, the risk assessment module is used for calling a trained risk assessment model, performing assessment processing on the target text data and predicting risk type information corresponding to the target text data, and the result prediction module is used for calling a trained result prediction model and predicting risk type information corresponding to the target text data. And performing bleeding result prediction on the target text data and the risk type information to obtain a bleeding risk result, and taking the bleeding risk result as a risk assessment result of the to-be-processed medical record. According to the invention, the risk assessment result corresponding to the medical record can be accurately assessed, and medical auxiliary information is provided for clinicians.

Description

Technical Field

[0001] The present invention relates to the field of medical technology assistance, and in particular to a medical record-based risk assessment system, electronic equipment, and storage medium. Background Art

[0002] Gastrointestinal bleeding is a common clinical syndrome caused by a variety of diseases, including peptic ulcers, inflammatory bowel disease, diverticula, hemorrhoids, colon polyps, rectal prolapse, vascular abnormalities in the digestive tract, and malignant tumors. It can lead to a range of adverse outcomes, including blood transfusions and surgery, and in severe cases, can even be life-threatening. Currently, clinically used gastrointestinal bleeding scores are primarily used to screen patients at very low risk who do not require hospitalization before outpatient care, but have low predictive value for adverse outcomes in hospitalized patients.

[0003] Therefore, how to conveniently, quickly and accurately assess the risk of adverse gastrointestinal bleeding outcomes corresponding to medical records is an urgent problem that needs to be solved. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide a medical record-based risk assessment system, electronic device, and storage medium to conveniently, quickly, and accurately assess the risk of adverse gastrointestinal bleeding outcomes corresponding to the medical records.

[0005] In a first aspect, an embodiment of the present invention provides a medical record-based risk assessment system, comprising: an acquisition module, a first extraction module, a risk assessment module, and an outcome prediction module; The acquisition module is used to acquire the medical records to be processed and acquire the medical record text data corresponding to the medical records to be processed; The first extraction module is used to call the trained text extraction model to extract text belonging to the target semantic type in the medical record text data to obtain target text data extracted from the medical record text data; The risk assessment module is used to call the trained risk assessment model, perform assessment processing on the target text data, and predict the risk type information corresponding to the target text data; The result prediction module is used to call the trained result prediction model, predict the bleeding result of the target text data and the risk type information, obtain the bleeding risk result, and use the bleeding risk result as the risk assessment result of the medical record to be processed.

[0006] In some embodiments, the medical record-based risk assessment system further includes a first construction module and a first training module; The first construction module is used to construct a training data set, wherein the training data set includes training medical record text data of a plurality of different training medical records, and a plurality of annotated text data of the plurality of training medical record text data that have been annotated, each of the annotated text data being text data of a target semantic type; The first training module is used to take the training medical record text data in the training data set as input and the annotated text data as output, and fine-tune the large language model to be trained until convergence, thereby obtaining a text extraction large model for extracting text data belonging to the target semantic type in the text data.

[0007] In some embodiments, the construction module includes a collection unit, an optimization unit, a determination unit, and a labeling unit; The collection unit is used to collect initial medical records of patients with multiple diseases in multiple departments; The optimization unit is used to optimize the initial medical record to obtain an optimized medical record; The determining unit is configured to determine a corresponding medical record template according to the disease type of each disease, wherein the medical record template contains text data of a target semantic type; The labeling unit is used to label and timestamp the text data belonging to the target semantic type in the optimized medical records of the same disease type based on the medical record template, to obtain multiple training medical records, and to extract the training medical record text data of the training medical records from the multiple training medical records, as well as the labeled text data and the first timestamp information after labeling and timestamp processing.

[0008] In some embodiments, the medical record-based risk assessment system further includes a determination module, an analysis module, and a second construction module; The determination module is used to determine descriptive text data about each disease based on prior knowledge of multiple diseases; The analysis module is used to analyze the description text data to obtain disease description text data belonging to the target semantic type in the description text data; The second construction module is used to construct a medical record template for each disease type based on the disease description text data.

[0009] In some embodiments, the optimization unit includes an anonymization processing subunit, a format unification subunit, and a elimination processing subunit; The anonymization processing subunit is used to perform anonymization processing on the initial medical record to remove personal information and identity identification code from the initial medical record to obtain a first medical record; The format unification subunit is configured to perform format unification processing on the first medical record to convert the document format of the first medical record into a target document format to obtain a second medical record; The elimination processing sub-unit is used to perform low-quality elimination processing on the second medical record to eliminate sentences or paragraphs in the second medical record that contain grammatical errors, typos, repetitions, and missing cases, so as to obtain an optimized medical record after optimization processing.

[0010] In some embodiments, the medical record-based risk assessment system further includes a second extraction module, a second training module, and a third training module; The second extraction module is configured to extract disease outcome text data from the training medical record text data of a plurality of different training medical records, wherein the disease outcome text data includes a plurality of different types of risk type information, and the risk type information is correspondingly annotated with second timestamp information; The second training module is configured to take the annotated text data of the same disease in the training medical record text data as input, take the disease outcome text data of the same disease as output, train the deep learning model to be trained until convergence, and obtain a trained risk assessment model; The third training module is used to take the annotated text data and the corresponding first timestamp information as input data, and take the risk type information matching the input data and the corresponding second timestamp information as output data, and train the result prediction model to be trained until convergence to obtain the trained result prediction model.

[0011] In some embodiments, the medical record-based risk assessment system further includes a sending module; The sending module is configured to send the risk assessment result to a corresponding clinician when the risk assessment result of the medical record to be processed is a target risk assessment result.

[0012] In some embodiments, the target semantic type includes personal basic information type, physical sign type, disease history type, medication history type, and examination and inspection type.

[0013] In a second aspect, an embodiment of the present invention provides an electronic device, the electronic device including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are implemented: Obtaining the medical records to be processed and obtaining the medical record text data corresponding to the medical records to be processed; Calling the trained text extraction model to extract text belonging to the target semantic type in the medical record text data to obtain target text data extracted from the medical record text data; Calling the trained risk assessment model to evaluate the target text data and predict the risk type information corresponding to the target text data; The trained result prediction model is called to predict the bleeding result on the target text data and the risk type information to obtain a bleeding risk result, and the bleeding risk result is used as the risk assessment result of the medical record to be processed.

[0014] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented: Obtaining the medical records to be processed and obtaining the medical record text data corresponding to the medical records to be processed; Calling the trained text extraction model to extract text belonging to the target semantic type in the medical record text data to obtain target text data extracted from the medical record text data; Calling the trained risk assessment model to evaluate the target text data and predict the risk type information corresponding to the target text data; The trained result prediction model is called to predict the bleeding result on the target text data and the risk type information to obtain a bleeding risk result, and the bleeding risk result is used as the risk assessment result of the medical record to be processed.

[0015] An embodiment of the present invention provides a medical record-based risk assessment system, electronic device and storage medium. The system calls the trained text extraction model and risk assessment model through a first extraction module and a risk assessment module, automatically extracts the required target text data from the medical record text data of the medical record to be processed, and automatically predicts the risk type information corresponding to the medical record to be processed, and inputs the extracted target text data and the predicted risk type information into the trained result prediction model to predict the bleeding risk result. The system can conveniently, quickly and accurately evaluate the risk assessment results corresponding to the medical record to be processed, and provide medical auxiliary information for clinicians. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of a medical record-based risk assessment method provided by an embodiment of the present invention; Figure 2 This is a flow chart of a method for training a large text extraction model provided by an embodiment of the present invention; Figure 3 This is a schematic diagram of a marking process provided by an embodiment of the present invention; Figure 4 This is a flow chart of a method for training a risk assessment model provided by an embodiment of the present invention; Figure 5 This is a schematic diagram of risk factor prediction based on time series provided by an embodiment of the present invention; Figure 6 This is a schematic structural diagram of a medical record-based risk assessment system provided by an embodiment of the present invention; Figure 7 This is a schematic structural diagram of an electronic device provided by an embodiment of the present invention; Figure 8 This is another structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0018] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0019] As used herein, the term "including" and its variations are open-ended, i.e., "including but not limited to." The term "based on" means "based, at least in part, on." The term "one embodiment" means "at least one embodiment," the term "another embodiment" means "at least one additional embodiment," and the term "some embodiments" means "at least some embodiments." Other terms are defined in the following description.

[0020] Gastrointestinal bleeding (GTB) is a common clinical syndrome caused by a variety of diseases, including peptic ulcers, inflammatory bowel disease, diverticula, hemorrhoids, colon polyps, rectal prolapse, vascular abnormalities in the digestive tract, and malignant tumors. It can lead to a series of adverse outcomes, such as blood transfusions and surgery, and in severe cases, can even be life-threatening. Currently, GTB scores used in clinical practice are mostly used to screen extremely low-risk patients who do not require hospitalization before outpatient treatment, but have low predictive value for adverse outcomes in hospitalized patients.

[0021] Therefore, how to conveniently, quickly and accurately assess the risk of adverse gastrointestinal bleeding outcomes corresponding to medical records is an urgent problem that needs to be solved.

[0022] In order to solve the technical problems existing in the related art, the embodiment of the present invention provides a risk assessment method based on medical records. Figure 1 , Figure 1 1 is a flow chart of a medical record-based risk assessment method provided by an embodiment of the present invention, the method comprising steps 101 to 104; Step 101: Obtain the medical record to be processed, and obtain the medical record text data corresponding to the medical record to be processed.

[0023] In this embodiment, the medical records to be processed may be medical records collected by clinicians for patients currently being treated, or they may be medical records provided by patients or other personnel. Specifically, the medical records to be processed may be paper medical records or electronic medical records. In the case of paper medical records, this embodiment can obtain the medical record text data in the paper medical records through text recognition; in the case of electronic medical records, this embodiment can directly export the medical record text data in the electronic medical records.

[0024] It should be noted that the medical records to be processed provided in this embodiment can be medical records of any disease type related to gastrointestinal bleeding, such as peptic ulcer, inflammatory bowel disease, diverticulum, hemorrhoids, colon polyps, rectal prolapse, vascular abnormalities of the digestive tract, and malignant tumors; they can also be medical records of any disease type related to other symptoms.

[0025] This embodiment is mainly described by taking the risk assessment of gastrointestinal bleeding as an example, but the embodiment of the present invention can also be applied to the risk assessment of any other disease, and is not specifically limited here.

[0026] Step 102: Call the trained text extraction model to extract the text belonging to the target semantic type in the medical record text data to obtain the target text data extracted from the medical record text data.

[0027] The target semantic types provided in this embodiment are information used to assess the risk of a condition. This information is pre-determined key indicators that are highly correlated with the condition. For example, for gastrointestinal bleeding, the target semantic types provided in this embodiment may include basic personal information, physical signs, medical history, medication history, and examination and testing.

[0028] Exemplarily, text data of personal basic information type may include but is not limited to age, gender, height, weight, etc., physical sign type may include but is not limited to blood pressure, heart rate, abdominal tenderness, mental state, etc., medical history type may include but is not limited to cirrhosis, liver cancer, heart failure, digestive tract tumors, etc., medication history type may include but is not limited to aspirin, clopidogrel, warfarin, heparin, ticagrelor, etc., examination and test type may include but is not limited to hemoglobin, albumin, international normalized ratio, blood creatinine, urea nitrogen, etc.

[0029] With the rapid development of artificial intelligence technology, traditional deep learning has made profound progress in the fields of lesion image recognition and lesion feature extraction, and can assist doctors in completing tasks such as lesion recognition and examination report writing in the field of medical imaging. A large language model is a neural network named by billions of adjustable parameters in its structural framework. Its training data set contains large-scale graphic and text data, and its parameter volume is usually in the millions to billions, far exceeding that of ordinary deep learning models. Therefore, it can obtain better feature extraction and learning capabilities, and at the same time has a strong ability to extract information from unstructured text. Therefore, the embodiment of the present invention can use a trained text extraction large model to extract and process the text of the target semantic type in the medical record text data of the medical record to be processed, so as to extract the key indicators with high correlation with the symptoms in the medical record to be processed, that is, the target text data of the target semantic type, so that the risk assessment of the relevant symptoms in the medical record to be processed can be carried out based on the target text data.

[0030] In some embodiments, in order to obtain a trained text extraction model, before step 102, the medical record-based risk assessment method provided in this embodiment may also include a technical solution for a training method of a text extraction model. For details, see Figure 2 , Figure 2 This is a flow chart of a training method for a large text extraction model provided by an embodiment of the present invention. Figure 2 As shown, it includes steps 201 to 202; Step 201 is to construct a training data set, wherein the training data set includes training medical record text data of a plurality of different training medical records, and a plurality of annotated text data among the plurality of training medical record text data that have been annotated, and each of the annotated text data is text data of the target semantic type.

[0031] In this embodiment, the text data in the training medical record text data can be annotated according to a preset target semantic type to identify the annotated text data belonging to the target semantic type. In this way, the annotated data and the training medical record text data can be used as training pairs to fine-tune the pre-set large language model to be trained. It should be noted that the training medical records provided in this embodiment can be the historical medical records of any patient in the medical record system.

[0032] In some embodiments, this embodiment can also annotate the training medical record text data based on a pre-constructed medical record template related to the symptoms in the training medical record to obtain annotated text data. Specifically, the steps of constructing a training data set provided by this embodiment can specifically include: collecting initial medical records of patients with multiple diseases in multiple departments; optimizing the initial medical records to obtain optimized medical records after optimization; determining the corresponding medical record template according to the disease type of each disease, wherein the medical record template contains text data belonging to the target semantic type; based on the medical record template, annotating and timestamping the text data belonging to the target semantic type in the optimized medical records of the same disease type to obtain multiple training medical records, and extracting the training medical record text data of the training medical records from the multiple training medical records, as well as the annotated text data and the first timestamp information after annotating and timestamping.

[0033] Among them, in order to obtain a medical record template related to the disease, before the step of determining the corresponding medical record template according to the disease type of each of the diseases, the method provided in this embodiment may also include: determining the descriptive text data about each of the diseases based on prior knowledge of multiple diseases; analyzing the descriptive text data to obtain the disease description text data belonging to the target semantic type in the descriptive text data; and constructing a medical record template for each of the diseases based on the disease description text data.

[0034] It should be noted that in this embodiment, a team of multiple clinical experts can determine descriptive text data about each of the aforementioned diseases based on clinical guidelines, expert consensus, field reviews, textbooks, and other authoritative literature. These descriptive text data are then analyzed to obtain disease description text data belonging to the target semantic type within the descriptive text data, thereby selecting the disease description text data to construct a medical record template for the corresponding disease. In this way, this embodiment can, based on the pre-constructed medical record template, annotate the text data belonging to the target semantic type in the optimized medical records of the same disease type.

[0035] For example, taking gastrointestinal bleeding as an example, see Figure 3 , Figure 3 This is a schematic diagram of a marking process provided by an embodiment of the present invention. Figure 3 As shown, the text data in the left frame is text data in a medical record about gastrointestinal bleeding, and the text data in the right frame is annotated text data belonging to the target semantic type in the medical record text data on the left.

[0036] In this embodiment, in order to obtain high-quality training medical record text data to improve the text extraction accuracy of the text extraction model, the step of optimizing the initial medical record to obtain the optimized medical record after optimization provided in this embodiment can specifically include: anonymizing the medical record information of the initial medical record to remove the personal information and identity identification code in the initial medical record to obtain a first medical record; performing format unification processing on the first medical record to convert the document format of the first medical record into a target document format to obtain a second medical record; performing low-quality elimination processing on the second medical record to eliminate sentences or paragraphs in the second medical record that contain grammatical errors, typos, repetitions, and missing cases to obtain an optimized medical record after optimization.

[0037] Among them, anonymization processing is mainly used to remove personal information and patient personal information such as identification codes from original medical records; format unification processing is mainly used to unify medical records into a target document format that is convenient for extracting text data, such as converting medical records in PDF format and WORD format to TXT format; low-quality elimination processing is mainly used to eliminate medical records containing but not limited to sentences or paragraphs with grammatical errors, typos, repetitions, missing cases, etc. In this way, the optimization processing provided by the embodiment of the present invention can effectively improve the quality of training data, so that after the large language model to be trained is trained with high-quality training data, the text extraction accuracy of the trained text extraction large model can be improved, thereby improving the assessment accuracy of subsequent risk assessment.

[0038] In step 202, the training medical record text data in the training data set is used as input, and the annotated text data is used as output. The large language model to be trained is fine-tuned until convergence, thereby obtaining a large text extraction model for extracting text data belonging to the target semantic type in the text data.

[0039] After obtaining a training dataset constructed from high-quality training data, the large language model to be trained is fine-tuned using the training dataset until convergence, thereby obtaining a high-precision large text extraction model.

[0040] It should be noted that when fine-tuning the large language model to be trained using a training dataset, the training data in the training dataset can be randomly allocated in a certain ratio, for example, an 8:2 ratio, to obtain a training set and a test set. This allows the convergence of the large language model to be tested using a test set that is different from the training set. In this embodiment, a low-rank adaptive (LoRA) fine-tuning method can be used to fine-tune the large language model.

[0041] After obtaining a high-precision text extraction model, the required target text data can be easily, quickly and accurately extracted from the medical records to be processed, so as to improve the evaluation accuracy of subsequent evaluation and processing of these target text data.

[0042] Step 103: Call the trained risk assessment model to perform assessment processing on the target text data and predict the risk type information corresponding to the target text data.

[0043] Step 104 , calling the trained result prediction model, performing bleeding result prediction on the target text data and the risk type information, obtaining a bleeding risk result, and using the risk type information as the risk assessment result of the medical record to be processed.

[0044] The bleeding risk results provided in this embodiment may be risk coefficients with numerical values ​​for characterizing different bleeding risks, with different numerical risk coefficients corresponding to different bleeding risks. For example, a risk coefficient of 0-15 may correspond to a very low risk of bleeding; a risk coefficient of 15-25 may correspond to a low risk of bleeding; a risk coefficient of 25-35 may correspond to a medium risk of bleeding; a risk coefficient of 35-45 may correspond to a high risk of bleeding; a risk coefficient of 45-55 may correspond to a high risk of bleeding; and a risk coefficient of 55 or above may correspond to a very high risk of bleeding.

[0045] In this embodiment, in order to obtain the trained risk assessment model and outcome prediction model, before step 103, the medical record-based risk assessment method provided in this embodiment may also include a technical solution for the training method of the risk assessment model and outcome prediction model. For details, see Figure 4 , Figure 4 FIG. 1 is a flow chart of a method for training a risk assessment model according to an embodiment of the present invention. Figure 4 As shown, it includes steps 401 to 403; Step 401: extract disease outcome text data from a plurality of different training medical records, wherein the disease outcome text data includes a plurality of different types of risk type information, and the risk type information is correspondingly marked with second timestamp information.

[0046] Among them, since the training medical records provided in this embodiment can be historical medical records in the medical record system, the training medical records contain disease outcome text data in which clinicians write disease conclusions under different timestamp information. Specifically, the disease outcome text data may include corresponding risk type information under different second timestamp information. Therefore, this embodiment can directly extract the disease outcome text data and the corresponding second timestamp information from the training medical record text data.

[0047] Specifically, the various types of risk type information contained in the disease outcome text data provided in this embodiment may include but is not limited to information such as conservative medical treatment, blood transfusion treatment, endoscopic treatment, surgical treatment, and death; the bleeding risk result information may include corresponding risk type information at different time stamps in the future.

[0048] In step 402, the annotated text data of the same disease in the training medical record text data is used as input, and the disease outcome text data of the same disease is used as output, and the deep learning model to be trained is trained until convergence to obtain a trained risk assessment model.

[0049] In this embodiment, the main purpose is to train the deep learning model to be trained by using the annotated text data and the disease outcome text data of the same disease as a training pair. The deep learning model provided in this embodiment can be any convolutional neural network model that can achieve prediction function.

[0050] In one case, the deep learning model to be trained can be trained until convergence by a preset number of training times, that is: the annotated text data of the same disease in the training medical record text data is input into the deep learning model to be trained for prediction processing to obtain the predicted risk assessment result; a preset loss function is used to calculate the loss value of the disease outcome text data corresponding to the input annotated text data relative to the risk assessment result, and the model parameters of the deep learning model to be trained are fine-tuned according to the loss value; the annotated text data is continuously used to train the deep learning model to be trained and the model parameters are fine-tuned until the calculated number of training times reaches a preset number of training times, for example, 50,000 times or 100,000 times, and then it can be determined that the deep learning model has converged.

[0051] In another case, the deep learning model to be trained can be trained until convergence so that the accuracy of the predicted risk assessment results reaches a preset accuracy threshold, that is: the annotated text data of the same disease in the training medical record text data is input into the deep learning model to be trained for prediction processing to obtain the predicted risk assessment result; a preset loss function is used to calculate the loss value of the disease outcome text data corresponding to the input annotated text data relative to the risk assessment result, and the model parameters of the deep learning model to be trained are fine-tuned according to the loss value; the annotated text data is continuously used to train the deep learning model to be trained and the model parameters are fine-tuned until the accuracy of the risk assessment results predicted by the deep learning model to be trained reaches a preset accuracy threshold, such as 95% or 98%, and then it can be determined that the deep learning model has converged.

[0052] The above only describes two methods of model convergence. Other methods for determining model convergence can also be used and are not specifically limited here.

[0053] Step 403: Take the annotated text data and the corresponding first timestamp information as input data, take the risk type information matching the input data and the corresponding second timestamp information as output data, train the result prediction model to be trained until convergence, and obtain the trained result prediction model.

[0054] In this embodiment, the annotated text data and the corresponding first timestamp information, as well as the risk type information and the corresponding second timestamp information, are used as entity tag-outcome-time sequence pairs with time series characteristics to train the outcome prediction model to be trained. The outcome prediction model to be trained provided in this embodiment can be a prediction model based on a Long Short Term Memory Network (LSTM).

[0055] It should be noted that the training convergence method of the large language model provided by the present invention and / or the training convergence method of the result prediction model to be trained can also refer to the above-mentioned method of training the deep learning model to be trained until convergence, which will not be repeated here.

[0056] After obtaining the trained risk assessment model and result prediction model, the trained risk assessment model can be used to evaluate and process the target text data, thereby predicting the risk type information corresponding to the target text data, and using the trained result prediction model to predict the bleeding result of the target text data and risk type information to obtain the bleeding risk result, and finally the bleeding risk result is used as the risk assessment result of the medical record to be processed.

[0057] Optionally, this embodiment can connect the output end of the text extraction model to the input end of the risk assessment model, and connect the output end of the risk assessment model to the input end of the result policy model, so as to construct a multimodal large model risk assessment system, so that it can automatically extract relevant key indicators, i.e., target text data, based on the input medical record text data, and automatically predict the corresponding risk assessment results at different time stamps in the future, so as to achieve the purpose of conveniently, quickly and accurately evaluating the risk assessment results corresponding to the medical records, and thus provide medical auxiliary information for clinicians.

[0058] Specifically, the medical record-based risk assessment system provided by the embodiment of the present invention can predict the corresponding risk assessment results at different time stamps in the future. For example, see Figure 5 , Figure 5This is a schematic diagram of risk factor prediction based on time series provided by an embodiment of the present invention. Figure 5 As shown, Figure 5 The following figure shows the predicted risk coefficients of a patient's medical records, as determined by the medical record-based risk assessment system provided in this embodiment. For example, if the patient entered their medical records before January 11, 20xx, the medical record-based risk assessment system provided in this embodiment can output different risk coefficients for different time points on and after January 11, 20xx. This allows patients to clearly understand the risks associated with their medical records, and can also provide clinicians with auxiliary medical information, allowing them to prepare appropriate interventions based on the generated prediction results to prevent the patient's condition from worsening.

[0059] As an optional embodiment, this embodiment can also construct a comparison template between actual risk assessment results and intervention measures based on clinical guidelines and expert consensus, so that different intervention measures can be sent to clinical physicians based on the corresponding risk assessment results at different timestamps. Specifically, the medical record-based risk assessment method provided in this embodiment can be directly applied to local medical information query devices in various hospitals, allowing each patient to self-check the risk assessment results of relevant conditions in their own medical records.

[0060] Among them, the comparison template between the actual risk assessment results and the intervention measures can be constructed based on the specific risk coefficient value, clinical guidelines and expert consensus. For example, when the risk coefficient is 0-15, the corresponding intervention measure may be medical observation or discharge; when the risk coefficient is 15-25, the corresponding intervention measure may be conservative medical treatment; when the risk coefficient is 25-35, the corresponding intervention measure may be blood transfusion intervention; when the risk coefficient is 35-45, the corresponding intervention measure may be interventional or endoscopic hemostasis intervention; when the risk coefficient is 45-55, the corresponding intervention measure may be surgical intervention; when the risk coefficient is above 55, the corresponding intervention measure may be death risk, and referral to the ICU ward for care is recommended.

[0061] Thus, in this embodiment, the risk assessment system based on medical records provided in this embodiment can generate the following Figure 5 After the risk factor prediction results are shown, different intervention measures are prepared for the patients in advance according to the constructed comparison template of the actual risk assessment results and intervention measures.

[0062] Furthermore, in order to facilitate follow-up treatment for patients who are seriously ill but unaware of their condition, the medical record-based risk assessment method provided in this embodiment may also include: when the risk assessment result of the medical record to be processed is the target risk assessment result, the risk assessment result is sent to the corresponding clinician.

[0063] Among them, the target risk assessment result provided in this embodiment can be a bleeding risk result that indicates that the symptoms of the relevant disease in the medical record to be processed are serious and require timely treatment. For example, when the bleeding risk result is a higher risk, high risk or extremely high risk, the corresponding result requires timely treatment.

[0064] At this point, the medical record-based risk assessment method provided by the embodiment of the present invention is completed.

[0065] According to the method described in the above embodiment, this embodiment will be further described from the perspective of a medical record-based risk assessment system. The medical record-based risk assessment system can be implemented as an independent entity or integrated into an electronic device, such as a terminal. The terminal may include a mobile phone, a tablet computer, etc.

[0066] See Figure 6 , Figure 6 FIG. 1 is a schematic diagram of a structure of a medical record-based risk assessment system provided by an embodiment of the present invention. Figure 6 As shown, the medical record-based risk assessment system 500 provided by the embodiment of the present invention includes: an acquisition module 501, an extraction module 502, a risk assessment module 503 and a result prediction module 504; The acquisition module 501 is used to acquire the medical records to be processed and acquire the medical record text data corresponding to the medical records to be processed.

[0067] The extraction module 502 is used to call the trained text extraction model to extract the text belonging to the target semantic type in the medical record text data to obtain the target text data extracted from the medical record text data.

[0068] The risk assessment module 503 is used to call the trained risk assessment model, perform assessment processing on the target text data, and predict the risk type information corresponding to the target text data.

[0069] The result prediction module 504 is used to call the trained result prediction model, perform bleeding result prediction on the target text data and the risk type information, obtain a bleeding risk result, and use the bleeding risk result as the risk assessment result of the medical record to be processed.

[0070] In an embodiment of the present invention, the medical record-based risk assessment system 500 provided in this embodiment may further include a first construction module and a first training module; Among them, the first construction module is used to construct a training data set, which includes training medical record text data of multiple different training medical records, and multiple annotated text data in the multiple training medical record text data that have been annotated, and each of the annotated text data is text data of the target semantic type; the first training module is used to take the training medical record text data in the training data set as input and the annotated text data as output, and fine-tune the large language model to be trained until convergence, so as to obtain a text extraction large model for extracting text data belonging to the target semantic type in the text data.

[0071] In some embodiments, the construction module provided in this embodiment includes a collection unit, an optimization unit, a determination unit, and a labeling unit; Among them, the collection unit is used to collect the initial medical records of patients with multiple diseases in multiple departments; the optimization unit is used to optimize the initial medical records to obtain optimized medical records after optimization; the determination unit is used to determine the corresponding medical record template according to the disease type of each disease, and the medical record template contains text data belonging to the target semantic type; the labeling unit is used to label and timestamp the text data belonging to the target semantic type in the optimized medical records of the same disease type based on the medical record template to obtain multiple training medical records, and extract the training medical record text data of the training medical records from the multiple training medical records, as well as the labeled text data and the first timestamp information after labeling and timestamp processing.

[0072] In some embodiments, the medical record-based risk assessment system 500 provided in this embodiment further includes a determination module, an analysis module, and a second construction module; Among them, the determination module is used to determine the descriptive text data about each of the diseases based on prior knowledge of multiple diseases; the analysis module is used to analyze the descriptive text data to obtain the disease description text data belonging to the target semantic type in the descriptive text data; the second construction module is used to construct the medical record template of each of the diseases based on the disease description text data.

[0073] In some other embodiments, the optimization unit provided in this embodiment includes an anonymization processing subunit, a format unification subunit, and a elimination processing subunit; Among them, the anonymous processing sub-unit is used to perform anonymization processing on the medical record information of the initial medical record to remove the personal information and identity identification code in the initial medical record to obtain a first medical record; the format unification sub-unit is used to perform format unification processing on the first medical record to convert the document format of the first medical record into a target document format to obtain a second medical record; the elimination processing sub-unit is used to perform low-quality elimination processing on the second medical record to eliminate sentences or paragraphs in the second medical record that contain grammatical errors, typos, repetitions, and missing cases to obtain an optimized medical record after optimization processing.

[0074] In one embodiment of the present invention, the medical record-based risk assessment system 500 provided in this embodiment further includes a second extraction module, a second training module, and a third training module; Among them, the second extraction module is used to extract disease outcome text data from the training medical record text data of multiple different training medical records, and the disease outcome text data includes multiple different types of risk type information, and the risk type information is correspondingly annotated with second timestamp information; the second training module is used to take the annotated text data of the same disease in the training medical record text data as input, and the disease outcome text data of the same disease as output, and train the deep learning model to be trained until convergence to obtain a trained risk assessment model; the third training module is used to take the annotated text data and the corresponding first timestamp information as input data, and the risk type information matching the input data and the corresponding second timestamp information as output data, and train the result prediction model to be trained until convergence to obtain a trained result prediction model.

[0075] In another embodiment of the present invention, the medical record-based risk assessment system 500 provided in this embodiment may further include a sending module, which is used to send the risk assessment result to the corresponding clinician when the risk assessment result of the medical record to be processed is the target risk assessment result.

[0076] Specifically, the target semantic types provided in this embodiment may include personal basic information type, physical sign type, disease history type, medication history type, and examination and testing type.

[0077] During specific implementation, the above modules and / or units can be implemented as independent entities, or can be arbitrarily combined to be implemented as the same or several entities. The specific implementation of the above modules and / or units can refer to the previous method embodiments. The specific beneficial effects that can be achieved can also be found in the beneficial effects in the previous method embodiments, which will not be repeated here.

[0078] Also, see Figure 7 , Figure 7 This is a structural diagram of an electronic device provided by an embodiment of the present invention. The electronic device may be a mobile terminal such as a smart phone, a tablet computer, or the like. Figure 7 As shown, the electronic device 600 includes a processor 601 and a memory 602. The processor 601 is electrically connected to the memory 602.

[0079] The processor 601 is the control center of the electronic device 600. It uses various interfaces and lines to connect various parts of the entire electronic device. By running or loading applications stored in the memory 602 and calling data stored in the memory 602, it executes various functions of the electronic device 600 and processes data, thereby monitoring the electronic device 600 as a whole.

[0080] In this embodiment, the processor 601 in the electronic device 600 will load the instructions corresponding to the processes of one or more applications into the memory 602 according to the following steps, and the processor 601 will run the applications stored in the memory 602, thereby implementing any step in the medical record-based risk assessment method provided in the above embodiment.

[0081] The electronic device 600 can implement the steps of any embodiment of the medical record-based risk assessment method provided in the embodiments of the present invention. Therefore, it can achieve the beneficial effects that can be achieved by any medical record-based risk assessment method provided in the embodiments of the present invention. Please refer to the previous embodiments for details and will not be repeated here.

[0082] See Figure 8 , Figure 8 is another structural diagram of an electronic device provided by an embodiment of the present invention, such as Figure 8 As shown, Figure 8 The electronic device 700 is a mobile terminal such as a smart phone or a laptop computer.

[0083] RF circuit 710 is used to receive and transmit electromagnetic waves, converting them into electrical signals, thereby enabling communication with a communications network or other devices. RF circuit 710 may include various existing circuit components for performing these functions, such as an antenna, a radio frequency transceiver, a digital signal processor, an encryption / decryption chip, a subscriber identity module (SIM) card, memory, and the like. RF circuit 710 can communicate with various networks, such as the Internet, an intranet, or a wireless network, or with other devices via a wireless network. These wireless networks may include cellular telephone networks, wireless local area networks, or metropolitan area networks. The wireless networks may utilize various communication standards, protocols, and technologies, including but not limited to Global System for Mobile Communication (GSM), Enhanced Data GSM Environment (EDGE), Wideband Code Division Multiple Access (WCDMA), Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Wireless Fidelity (Wi-Fi) (such as Institute of Electrical and Electronics Engineers standards IEEE 802.11a, IEEE 802.11b, IEEE802.11g, and / or IEEE802.11n), Voice over Internet Protocol (VoIP), Worldwide Interoperability for Microwave Access (Wi-Max), other protocols for email, instant messaging, and short messaging, and any other suitable communication protocols, including those currently undeveloped.

[0084] The memory 720 can be used to store software programs and modules, such as the program instructions / modules corresponding to the medical record-based risk assessment method in the above-mentioned embodiment. The processor 780 executes various functional applications and medical record-based risk assessment by running the software programs and modules stored in the memory 720.

[0085] The memory 720 may include a high-speed random access memory (RAM) and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 720 may further include a memory remotely located relative to the processor 780, and such remote memory may be connected to the electronic device 700 via a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0086] The input unit 730 can be used to receive digital or character input and generate keyboard, mouse, joystick, optical, or trackball signal input related to user settings and function control. Specifically, the input unit 730 may include a touch-sensitive surface 731 and other input devices 732. The touch-sensitive surface 731, also known as a touch display or touchpad, can detect user touch operations on or near it (for example, operations performed on or near the touch-sensitive surface 731 using a finger, stylus, or any other suitable object or accessory) and drive corresponding connected devices according to pre-set programs. Optionally, the touch-sensitive surface 731 may include a touch detection device and a touch controller. The touch detection device detects the user's touch position and detects signals generated by the touch operation, transmitting the signals to the touch controller. The touch controller receives the touch information from the touch detection device, converts it into touch point coordinates, and then transmits it to the processor 780. The touch controller can also receive and execute commands from the processor 780. Furthermore, the touch-sensitive surface 731 can be implemented using various types of touch sensors, including resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch-sensitive surface 731, the input unit 730 may further include other input devices 732. Specifically, the other input devices 732 may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, power keys, etc.), a trackball, a mouse, and a joystick.

[0087] The display unit 740 can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces of the electronic device 700. These graphical user interfaces can be composed of graphics, text, icons, videos, or any combination thereof. The display unit 740 may include a display panel 741. Optionally, the display panel 741 can be configured in the form of an LCD (Liquid Crystal Display), an OLED (Organic Light-Emitting Diode), or the like. Furthermore, the touch-sensitive surface 731 can cover the display panel 741. When the touch-sensitive surface 731 detects a touch operation on or near it, it transmits the information to the processor 780 to determine the type of touch event. The processor 780 then provides a corresponding visual output on the display panel 741 based on the type of touch event. Although the touch-sensitive surface 731 and the display panel 741 are shown in the figure as two independent components to implement input and output functions, in some embodiments, the touch-sensitive surface 731 and the display panel 741 can be integrated to implement input and output functions.

[0088] The electronic device 700 may also include at least one sensor 750, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor. The ambient light sensor may adjust the brightness of the display panel 741 according to the brightness of the ambient light, and the proximity sensor may generate an interrupt when the flip cover is closed or closed. As a type of motion sensor, the gravity acceleration sensor can detect the magnitude of acceleration in all directions (generally three axes), and can detect the magnitude and direction of gravity when stationary. It can be used for applications that recognize the posture of the mobile phone (such as switching between horizontal and vertical screens, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tapping), etc.; as for other sensors that the electronic device 700 may also be configured with, such as a gyroscope, barometer, hygrometer, thermometer, infrared sensor, etc., they will not be described in detail here.

[0089] Audio circuit 760, speaker 761, and microphone 762 provide an audio interface between the user and electronic device 700. Audio circuit 760 can convert received audio data into electrical signals and transmit them to speaker 761, which then converts them into sound signals for output. Microphone 762, on the other hand, converts collected sound signals into electrical signals, which are then received by audio circuit 760 and converted into audio data. The audio data is then processed by processor 780 and transmitted via RF circuit 710 to, for example, another terminal. Alternatively, the audio data can be output to memory 720 for further processing. Audio circuit 760 may also include an earphone jack to allow communication between external headphones and electronic device 700.

[0090] Electronic device 700, through a transmission module 770 (e.g., a Wi-Fi module), can help users receive requests, send information, and so on, providing users with wireless broadband Internet access. Although transmission module 770 is shown in the figure, it is understood that it is not a required component of electronic device 700 and can be omitted as needed without changing the essence of the invention.

[0091] Processor 780 is the control center of electronic device 700. It connects all components of the phone using various interfaces and circuits. By running or executing software programs and / or modules stored in memory 720 and accessing data stored in memory 720, it executes various functions of electronic device 700 and processes data, thereby providing overall monitoring of the electronic device. Optionally, processor 780 may include one or more processing cores. In some embodiments, processor 780 may integrate an application processor and a modem processor. The application processor primarily handles the operating system, user interface, and application programs, while the modem processor primarily handles wireless communications. It is understood that the modem processor may not be integrated into processor 780.

[0092] Electronic device 700 also includes a power supply 790 (e.g., a battery) that supplies power to various components. In some embodiments, the power supply can be logically connected to processor 780 via a power management system, thereby enabling the power management system to manage charging, discharging, and power consumption. Power supply 790 can also include one or more DC or AC power supplies, a recharging system, a power failure detection circuit, a power converter or inverter, a power status indicator, and other arbitrary components.

[0093] Although not shown, the electronic device 700 also includes a camera (e.g., a front-facing camera, a rear-facing camera), a Bluetooth module, etc., which will not be described in detail here. Specifically, in this embodiment, the display unit of the electronic device is a touch screen display, and the mobile terminal also includes a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured such that one or more processors execute the one or more programs to implement any step of the medical record-based risk assessment method provided in the above embodiment.

[0094] During specific implementation, the above modules can be implemented as independent entities, or can be arbitrarily combined and implemented as the same or several entities. The specific implementation of the above modules can be found in the previous method embodiments and will not be repeated here.

[0095] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above-described embodiments can be accomplished through instructions, or by controlling related hardware through instructions. The instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. To this end, an embodiment of the present invention provides a storage medium storing a plurality of instructions that, when executed by a processor, can implement any step of the medical record-based risk assessment method provided in the above-described embodiments.

[0096] The storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0097] Since the instructions stored in the storage medium can execute the steps in any embodiment of the medical record-based risk assessment method provided in the embodiments of the present invention, the beneficial effects that can be achieved by any medical record-based risk assessment method provided in the embodiments of the present invention can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

[0098] The above is a detailed introduction to a medical record-based risk assessment system, electronic device, and storage medium provided in the embodiments of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core ideas of the present application. At the same time, for those skilled in the art, according to the ideas of the present application, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as limiting the present application. Moreover, for those of ordinary skill in the art, without departing from the principles of the present invention, several improvements and modifications can be made, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A risk assessment system based on medical records, characterized in that: include: an acquisition module, a first extraction module, a risk assessment module, and a result prediction module; The acquisition module is used to acquire the medical records to be processed and acquire the medical record text data corresponding to the medical records to be processed; The first extraction module is used to call the trained text extraction model to extract text belonging to the target semantic type in the medical record text data to obtain target text data extracted from the medical record text data; The risk assessment module is used to call the trained risk assessment model, perform assessment processing on the target text data, and predict the risk type information corresponding to the target text data; The result prediction module is used to call the trained result prediction model, predict the bleeding result of the target text data and the risk type information, obtain the bleeding risk result, and use the bleeding risk result as the risk assessment result of the medical record to be processed.

2. The system according to claim 1, wherein The medical record-based risk assessment system further includes a first construction module and a first training module; The first construction module is used to construct a training data set, wherein the training data set includes training medical record text data of a plurality of different training medical records, and a plurality of annotated text data of the plurality of training medical record text data that have been annotated, each of the annotated text data being text data of a target semantic type; The first training module is used to take the training medical record text data in the training data set as input and the annotated text data as output, and fine-tune the large language model to be trained until convergence, thereby obtaining a text extraction large model for extracting text data belonging to the target semantic type in the text data.

3. The system according to claim 2, wherein: The construction module includes a collection unit, an optimization unit, a determination unit and a labeling unit; The collection unit is used to collect initial medical records of patients with multiple diseases in multiple departments; The optimization unit is used to optimize the initial medical record to obtain an optimized medical record; The determining unit is configured to determine a corresponding medical record template according to the disease type of each disease, wherein the medical record template contains text data of a target semantic type; The labeling unit is used to label and timestamp the text data belonging to the target semantic type in the optimized medical records of the same disease type based on the medical record template, to obtain multiple training medical records, and to extract the training medical record text data of the training medical records from the multiple training medical records, as well as the labeled text data and the first timestamp information after labeling and timestamp processing.

4. The system according to claim 3, wherein: The medical record-based risk assessment system further includes a determination module, an analysis module, and a second construction module; The determination module is used to determine descriptive text data about each disease based on prior knowledge of multiple diseases; The analysis module is used to analyze the description text data to obtain disease description text data belonging to the target semantic type in the description text data; The second construction module is used to construct a medical record template for each disease type based on the disease description text data.

5. The system according to claim 3, wherein: The optimization unit includes an anonymous processing subunit, a format unification subunit and a elimination processing subunit; The anonymization processing subunit is used to perform anonymization processing on the initial medical record to remove personal information and identity identification code from the initial medical record to obtain a first medical record; The format unification subunit is configured to perform format unification processing on the first medical record to convert the document format of the first medical record into a target document format to obtain a second medical record; The elimination processing sub-unit is used to perform low-quality elimination processing on the second medical record to eliminate sentences or paragraphs in the second medical record that contain grammatical errors, typos, repetitions, and missing cases, so as to obtain an optimized medical record after optimization processing.

6. The system according to claim 3, wherein: The medical record-based risk assessment system further includes a second extraction module, a second training module, and a third training module; The second extraction module is configured to extract disease outcome text data from the training medical record text data of a plurality of different training medical records, wherein the disease outcome text data includes a plurality of different types of risk type information, and the risk type information is correspondingly annotated with second timestamp information; The second training module is configured to take the annotated text data of the same disease in the training medical record text data as input, take the disease outcome text data of the same disease as output, train the deep learning model to be trained until convergence, and obtain a trained risk assessment model; The third training module is used to take the annotated text data and the corresponding first timestamp information as input data, and take the risk type information matching the input data and the corresponding second timestamp information as output data, and train the result prediction model to be trained until convergence to obtain the trained result prediction model.

7. The system according to claim 1, wherein: The medical record-based risk assessment system further includes a sending module; The sending module is configured to send the risk assessment result to a corresponding clinician when the risk assessment result of the medical record to be processed is a target risk assessment result.

8. The system according to any one of claims 1 to 7, wherein: The target semantic types include personal basic information type, physical sign type, disease history type, medication history type, and examination and testing type.

9. An electronic device, characterized in that: The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program: Obtaining the medical records to be processed and obtaining the medical record text data corresponding to the medical records to be processed; Calling the trained text extraction model to extract text belonging to the target semantic type in the medical record text data to obtain target text data extracted from the medical record text data; Calling the trained risk assessment model to evaluate the target text data and predict the risk type information corresponding to the target text data; The trained result prediction model is called to predict the bleeding result on the target text data and the risk type information to obtain a bleeding risk result, and the bleeding risk result is used as the risk assessment result of the medical record to be processed.

10. A computer-readable storage medium, characterized in that The computer readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented: Obtaining the medical records to be processed and obtaining the medical record text data corresponding to the medical records to be processed; Calling the trained text extraction model to extract text belonging to the target semantic type in the medical record text data to obtain target text data extracted from the medical record text data; Calling the trained risk assessment model to evaluate the target text data and predict the risk type information corresponding to the target text data; The trained result prediction model is called to predict the bleeding result on the target text data and the risk type information to obtain a bleeding risk result, and the bleeding risk result is used as the risk assessment result of the medical record to be processed.

Images