Deep learning-based critical value reporting method and system

Abstract

The invention relates to the technical field of medical data, and discloses a critical value reporting method and system based on deep learning, and the method comprises the steps: training a pre-constructed initial evaluation model based on a historical sample data set, obtaining a critical value evaluation model, obtaining the medical record data of a to-be-estimated patient, importing the medical record data of the to-be-estimated patient into the critical value evaluation model, and obtaining a critical value report result. The method comprises the following steps: obtaining a critical value evaluation report, sending the critical value evaluation report to a pre-constructed doctor's advice system, confirming the doctor's advice report based on the doctor's advice system, encrypting the doctor's advice report to obtain encrypted data, and sending the encrypted data to a pre-constructed report query end. And after confirming that each responsibility query end in the report query end generates a return report, completing a critical value report based on deep learning. According to the method, the critical degree of the patient data can be evaluated by using the critical value evaluation model to provide diagnosis and treatment reference for doctors, and meanwhile, the doctor's advice report is pushed to a plurality of persons in charge in an encrypted form, so that supervision and implementation of each person in charge are realized.

Description

Technical Field

[0001] The present invention relates to the field of medical data technology, and in particular to a critical value reporting method and system based on deep learning. Background Art

[0002] In the medical industry, great emphasis is placed on risk monitoring and control. When a crisis occurs, having a more valuable reminder to inform busy clinicians on the front line is a very important requirement and demand.

[0003] Therefore, in the medical testing industry, critical values ​​refer to abnormal results of a certain test or a certain type of test. When such abnormal test results appear, it indicates that the patient may be in a life-threatening state. Clinicians need to obtain test information in a timely manner and quickly give patients effective intervention measures or treatments to save their lives. Otherwise, serious consequences may occur and the best rescue opportunity will be lost. Due to its importance, critical value reporting is a very important task and requirement.

[0004] However, although the above method can provide clinicians with critical value reports as a reference, the effect is not as good as expected. Critical value reports are often based on the reference range of the test items. The risks are not uniform for clinical needs and specific circumstances. For example, the same antiphospholipid antibody results should not have the same risk cutoff values ​​in the rheumatology department and the reproductive department. Therefore, it is necessary to assess the patient's criticality based on the department and treatment needs, and strengthen the supervision of each person in charge or physician. Summary of the invention

[0005] The present invention provides a critical value reporting method based on deep learning. A computer can use a critical value assessment model to assess the criticality of patient data to provide a diagnosis and treatment reference for physicians. At the same time, the medical order report is pushed to multiple responsible persons in an encrypted form to achieve supervision and implementation of each responsible person.

[0006] To achieve the above object, the present invention provides a critical value reporting method based on deep learning, comprising:

[0007] Acquire a medical record data set, perform preprocessing operations on the medical record data set to obtain a historical sample data set, train a pre-built initial assessment model based on the historical sample data set, and obtain a critical value assessment model, wherein both the initial assessment model and the critical value assessment model are models composed of neural networks;

[0008] Obtain the medical record data of the patient to be evaluated, import the medical record data of the patient to be evaluated into the critical value evaluation model, and obtain the critical value evaluation report;

[0009] Sending the critical value assessment report to a pre-built medical order system, if receiving an automatic diagnosis instruction from a control terminal of the pre-built medical order system, generating a recommended diagnosis report based on the automatic diagnosis instruction, and confirming the recommended diagnosis report as an initial medical order report, otherwise, obtaining an initial medical order report based on the medical order system after receiving the critical value assessment report;

[0010] After confirming that the initial medical order report is a preset feasible report by using a pre-constructed report feasibility detection method, the medical order report is obtained, an encryption operation is performed on the medical order report to obtain encrypted data, and the encrypted data is sent to a pre-constructed report query end, wherein the report query end includes a plurality of responsible query ends;

[0011] After confirming that each responsible query end in the report query end has generated a report, the critical value report based on deep learning is completed.

[0012] Optionally, performing a preprocessing operation on the medical record data set to obtain a historical sample data set includes:

[0013] Extract medical record data from the medical record data set in sequence, and perform the following operations on the extracted medical record data:

[0014] Based on the extracted medical record data, a historical diagnosis item set is obtained, historical diagnosis items are sequentially extracted from the historical diagnosis item set, and a scoring criterion sequence corresponding to the extracted historical diagnosis items is obtained from a pre-built scoring criterion sequence library, and item scores corresponding to the medical record data in the historical diagnosis items are obtained based on the scoring criterion sequence, and item scores are identified using the historical diagnosis items to obtain unit diagnosis evaluation data;

[0015] Summarize the unit diagnosis and evaluation data to obtain the unit diagnosis and evaluation data set corresponding to the medical record data;

[0016] After the unit diagnosis and evaluation data set is sent to the pre-built physician evaluation system, the critical value evaluation of the medical record data is performed using the physician evaluation system after receiving the unit diagnosis and evaluation data set and the pre-built physician evaluation method to obtain the critical value evaluation score corresponding to the medical record data;

[0017] Extracting basic information sets based on medical record data, correlating the basic information sets, unit diagnosis assessment data sets, and critical value assessment scores to obtain unit diagnosis sample data;

[0018] Summarize the unit diagnosis sample data to obtain the historical sample data set corresponding to the medical record data set.

[0019] Optionally, before obtaining the scoring criterion sequence corresponding to the extracted historical diagnosis item from the pre-built scoring criterion sequence library, the method further includes:

[0020] Obtain a classification code sequence, extract classification codes from the classification code sequence in sequence, and perform the following operations on the extracted classification codes:

[0021] Extract the corresponding patient data set from the medical record data set based on the classification code, and perform the following operations on each patient data in the patient data set:

[0022] Based on the patient data, the project diagnosis data set under the classification code is extracted, and the following operations are performed on each project diagnosis data in the project diagnosis data set:

[0023] If the project diagnostic data is preset complete data, confirming the project diagnostic data as available data;

[0024] Otherwise, identify the missing data types in the project diagnostic data;

[0025] If the missing data type is a pre-constructed detection data type, the project diagnostic data is eliminated; if the missing data type is a pre-constructed standard data type, a project database corresponding to the project diagnostic data is obtained, and a data filling operation is performed on the project diagnostic data based on the project database to obtain filled project diagnostic data, and the filled project diagnostic data is confirmed as available data;

[0026] The available data are aggregated to obtain an available data set, and a scoring criterion sequence library is constructed based on the available data set.

[0027] Optionally, constructing a scoring criterion sequence library based on an available data set comprises:

[0028] Extracting available data from the available data sets in sequence, identifying and summarizing available diagnostic items corresponding to the extracted available data, and obtaining an available diagnostic item set;

[0029] Perform a deduplication operation on the available diagnostic item set to obtain a total diagnostic item set, extract target diagnostic items from the total diagnostic item set in sequence, extract corresponding available item historical data sets from the available data sets based on the target diagnostic items, and construct a scoring criterion sequence for the extracted target diagnostic items based on the available item historical data sets;

[0030] The scoring criterion sequences are aggregated to obtain a scoring criterion sequence library.

[0031] Optionally, obtaining the classification coding sequence includes:

[0032] Extract medical record data from the medical record data set in sequence, extract surgical codes from the extracted medical record data, summarize the surgical codes to obtain a surgical code set, extract surgical codes from the surgical code set in sequence, and identify the department identification code corresponding to the extracted surgical code based on a pre-built hospital department category table, splice the department identification code and the surgical code to obtain an initial classification code;

[0033] Based on the department identification codes, the initial classification codes are respectively summarized to obtain an initial classification code set, wherein the initial classification code set includes multiple department code sets, the department code set includes multiple department codes, and the department identification codes corresponding to the multiple department codes are the same;

[0034] Department code sets are extracted from the initial classification code set in sequence, a recursive operation is performed on each department code set in the extracted department code set, and a simplified coding operation is performed on the recursive department code set to obtain a classification code sequence, wherein the classification code sequence includes multiple classification codes, and the classification code is associated with the surgery code and the department identification code corresponding to the surgery code.

[0035] Optionally, after each responsible query end in the confirmation report query end generates a report, the following steps are included:

[0036] Extract the responsible query ends from the report query ends in sequence, and perform the following operations on the extracted responsible query ends:

[0037] Based on the preset query time, determine whether an access request instruction is received from the responsible query end; if an access request instruction is received from the responsible query end, perform an identity authentication operation on the responsible query end to obtain identity authentication data; if it is confirmed that the identity authentication data is consistent with the authentication data pre-built in the responsible query end, perform a decryption operation on the encrypted data to obtain decrypted data, and generate a report based on the decrypted data;

[0038] Otherwise, the delayed query time is obtained, and the delayed query time is used as the preset query time, and the step of determining whether an access request instruction is received from the responsible query end based on the preset query time is returned until a report is generated.

[0039] Optionally, constructing a scoring criterion sequence for the extracted target diagnostic items based on the available item history data set includes:

[0040] Extracting a corresponding target diagnostic data set from the available project historical data set based on the target diagnostic project, and constructing a frequency distribution graph based on the target diagnostic data set, wherein the target diagnostic data set includes a plurality of target diagnostic data, and the target diagnostic data includes: a diagnostic value, an empirical diagnostic upper limit value, and an empirical diagnostic lower limit value, and the horizontal axis of the frequency distribution graph is the diagnostic value, and the vertical axis is the number of occurrences of the diagnostic value in the target diagnostic data set;

[0041] Extracting diagnostic values ​​greater than a preset frequency from the frequency distribution diagram to obtain a reference value set to be evaluated, using the physician evaluation system and the physician evaluation method to obtain an evaluation score corresponding to each reference value to be evaluated in the reference value set to be evaluated, to obtain an evaluation score set, associating the evaluation scores in the evaluation score set with the diagnostic values, and arranging the evaluation score set in ascending order of the evaluation scores to obtain a scoring reference data set;

[0042] Extracting scoring reference data from the scoring reference data in sequence, and extracting the next scoring reference data adjacent to the scoring reference data to obtain adjacent scoring reference data, and constructing a scoring interval and a diagnostic value interval based on the scoring reference data and the adjacent scoring reference data;

[0043] Summarize the scoring intervals and the diagnostic value intervals respectively to obtain the scoring interval set and the diagnostic value interval set;

[0044] Extract target diagnostic data from the target diagnostic data set in sequence to obtain data to be normalized, identify the diagnostic value interval corresponding to the data to be normalized from the diagnostic value interval set, obtain the target interval, map the data to be normalized to the target interval, identify the scoring interval corresponding to the target interval from the scoring interval set, obtain the target scoring interval, perform proportional scoring calculation on the data to be normalized based on the target scoring interval, and obtain a normalized score;

[0045] The normalized scores and the evaluation scores are summarized in order from small to large to obtain a scoring criteria sequence.

[0046] Optionally, the physician evaluation system and the pre-built physician evaluation method after receiving the unit diagnosis evaluation data set are used to perform a critical value evaluation on the medical record data to obtain a critical value evaluation score corresponding to the medical record data, including:

[0047] After the unit diagnosis evaluation data set is sent to the physician evaluation system, a critical value evaluation table is obtained based on the physician evaluation system, wherein the critical value evaluation table includes: percentage evaluation criteria, the unit diagnosis evaluation data set and medical record data;

[0048] Based on a preset diagnosis and treatment physician review method, a plurality of diagnosis and treatment physician units are obtained, and a plurality of initial evaluation scores are obtained based on the plurality of diagnosis and treatment physician units, wherein there is one and only one account in the diagnosis and treatment physician unit, and the initial evaluation score corresponds to the diagnosis and treatment physician unit one by one;

[0049] The average of multiple initial assessment scores is calculated to obtain a critical value assessment score.

[0050] Optionally, the pre-built initial assessment model is trained based on the historical sample data set to obtain a critical value assessment model, including:

[0051] The historical sample data set is divided based on multiple department identification codes to obtain multiple department classification data sets, and the following operation is performed on each of the multiple department classification data sets:

[0052] The department classification data set is divided by using preset sample training values ​​to obtain a training sample set and a verification sample set, a disease information set and a disease critical value score set are extracted from the training sample set, and the disease information set and the disease critical value score set are used to train a pre-constructed classification initial evaluation model, wherein the classification initial evaluation model is a model constructed by a neural network, and the classification initial evaluation model includes: a disease information input layer, a pooling layer, a fully connected layer and a disease critical value score output layer, the pooling layer adopts a maximum pooling strategy, the pooling function size is 2×2, and the fully connected layer includes 1024 unit nodes. The loss function in the classification evaluation model is:

[0053]

[0054] Among them, LOST MSE represents the mean square error loss function, N represents the number of training samples in the training sample set, i represents the disease criticality score corresponding to the i-th training sample, Represents the disease criticality score predicted by the classification assessment model;

[0055] The validation sample set is used to perform reverse update calculation on the classification initial evaluation model to obtain the classification evaluation model, the classification evaluation model is summarized to obtain the initial evaluation model, and a retraining sample set is randomly extracted from the historical sample data set based on the sample training value. The initial evaluation model is iterated using the retraining sample set until the number of iteration steps of the iterative operation reaches the preset number of iteration steps to obtain the critical value evaluation model.

[0056] To achieve the above object, the present invention also provides a critical value reporting system based on deep learning, comprising:

[0057] A critical value report acquisition module is used to acquire a medical record data set, perform preprocessing operations on the medical record data set, obtain a historical sample data set, train a pre-built initial evaluation model based on the historical sample data set, and obtain a critical value evaluation model, wherein both the initial evaluation model and the critical value evaluation model are models composed of neural networks, acquire the medical record data of the patient to be evaluated, import the medical record data of the patient to be evaluated into the critical value evaluation model, and obtain a critical value evaluation report;

[0058] A doctor's order report acquisition module is used to send the critical value assessment report to a pre-built doctor's order system. If an automatic diagnosis instruction is received from the pre-built doctor's order system control terminal, a recommended diagnosis report is generated based on the automatic diagnosis instruction, and the recommended diagnosis report is confirmed as an initial doctor's order report. Otherwise, the initial doctor's order report is obtained based on the doctor's order system after receiving the critical value assessment report, and the doctor's order report is obtained after confirming that the initial doctor's order report is a preset feasible report using a pre-built report feasibility detection method;

[0059] The reporting report generation module is used to perform encryption operations on the medical order report, obtain encrypted data, and send the encrypted data to a pre-built report query terminal, wherein the report query terminal includes multiple responsible query terminals. After confirming that each responsible query terminal in the report query terminal has generated a reporting report, the critical value report based on deep learning is completed.

[0060] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0061] A memory storing at least one instruction;

[0062] The processor executes the instructions stored in the memory to implement the critical value reporting method based on deep learning described above.

[0063] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned deep learning-based critical value reporting method.

[0064] In order to solve the problems described in the background technology, the present invention first obtains a medical record data set, performs preprocessing operations on the medical record data set to obtain a historical sample data set, trains a pre-constructed initial evaluation model based on the historical sample data set to obtain a critical value evaluation model, obtains the medical record data of the patient to be evaluated, imports the medical record data of the patient to be evaluated into the critical value evaluation model, and obtains a critical value evaluation report. By evaluating the patient based on the medical record data and the critical value evaluation model, a critical value evaluation report is obtained. The physician can judge whether diagnosis and treatment are needed and the effect of diagnosis and treatment based on the obtained critical value evaluation report, which can provide a reliable basis for the diagnosis and treatment of clinicians, and assist physicians to better provide patients with safe, effective and timely diagnosis and treatment services. The critical value assessment report is sent to the pre-built medical order system. If an automatic diagnosis instruction is received from the pre-built medical order system control end, a recommended diagnosis report is generated based on the automatic diagnosis instruction, and the recommended diagnosis report is confirmed as the initial medical order report. Otherwise, the initial medical order report is obtained based on the medical order system after receiving the critical value assessment report. The present invention can automatically diagnose and generate the initial medical order report, and can also generate the initial medical order report according to the medical system through the doctor's diagnosis. After confirming that the initial medical order report is a preset feasible report using the pre-built report feasibility detection method, the medical order report is obtained. In the present invention, if multiple doctors believe that the initial medical order report is feasible, the initial medical order report is confirmed as a preset feasible report, ensuring the security of the medical order report. The medical order report is encrypted to obtain encrypted data, and the encrypted data is sent to the pre-built report query end, wherein the report query end includes multiple responsible query ends. After confirming that each responsible query end in the report query end generates a report, the critical value report based on deep learning is completed. The present invention also encrypts the medical order report and pushes it to multiple responsible persons. Only when it is confirmed that multiple responsible persons have received the medical report, the critical value report is confirmed to be completed, which is conducive to supervising each responsible person. Therefore, the present invention can use the critical value assessment model to assess the criticality of patient data to provide a reference for diagnosis and treatment for doctors, and at the same time, push the medical order report to multiple responsible persons in an encrypted form to achieve supervision and implementation of each responsible person. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A flowchart of a critical value reporting method based on deep learning provided by an embodiment of the present invention;

[0066] Figure 2 A functional module diagram of a critical value reporting system based on deep learning provided by one embodiment of the present invention;

[0067] Figure 3 A schematic diagram of the structure of an electronic device for implementing the critical value reporting method based on deep learning provided by an embodiment of the present invention;

[0068] Description of reference numerals:

[0069] 1. Electronic device; 10. Processor; 11. Memory; 12. Bus.

[0070] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0071] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0072] The embodiment of the present application provides a critical value reporting method based on deep learning. The execution subject of the critical value reporting method based on deep learning includes but is not limited to at least one of the electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the critical value reporting method based on deep learning can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0073] Reference Figure 1 FIG. 1 is a flow chart of a critical value reporting method based on deep learning provided by an embodiment of the present invention. In this embodiment, the critical value reporting method based on deep learning includes:

[0074] S1. Obtain a medical record data set, perform preprocessing operations on the medical record data set, and obtain a historical sample data set.

[0075] It is understandable that the medical record data set is a collection of multiple medical record data, and the medical record data is extracted from pre-built electronic medical records. For example, after a patient visits a hospital, the doctor prescribes an electronic medical record for the patient, and the information in the electronic medical record is extracted and summarized. The sum of the obtained information is the medical record data. Optionally, the medical record data includes: name, gender, age, historical diagnosis items, surgical codes, etc. For specific content and explanations, please refer to subsequent embodiments.

[0076] It should be noted that the medical record data in the embodiments of the present invention are all data when the disease has been diagnosed and surgery is required or has been performed. However, it should be understood that the medical record data also includes diagnostic data corresponding to historical diagnostic items of the patient at different time periods. For example, a patient needs surgery due to high blood pressure. Prior to this, the patient's blood pressure was diagnosed multiple times. Among the multiple diagnoses, some of the diagnostic values ​​were evaluated by the doctor and surgery was not required. Some of the diagnostic values ​​were evaluated by the doctor and surgery was required. These records are all included in the historical diagnosis items.

[0077] Furthermore, the preprocessing operation is performed on the medical record data set to obtain the historical sample data set, including:

[0078] Extract medical record data from the medical record data set in sequence, and perform the following operations on the extracted medical record data:

[0079] Based on the extracted medical record data, a historical diagnosis item set is obtained, historical diagnosis items are sequentially extracted from the historical diagnosis item set, and a scoring criterion sequence corresponding to the extracted historical diagnosis items is obtained from a pre-built scoring criterion sequence library, and item scores corresponding to the medical record data in the historical diagnosis items are obtained based on the scoring criterion sequence, and item scores are identified using the historical diagnosis items to obtain unit diagnosis evaluation data;

[0080] Summarize the unit diagnosis and evaluation data to obtain the unit diagnosis and evaluation data set corresponding to the medical record data;

[0081] After the unit diagnosis and evaluation data set is sent to the pre-built physician evaluation system, the critical value evaluation of the medical record data is performed using the physician evaluation system after receiving the unit diagnosis and evaluation data set and the pre-built physician evaluation method to obtain the critical value evaluation score corresponding to the medical record data;

[0082] Extracting basic information sets based on medical record data, correlating the basic information sets, unit diagnosis assessment data sets, and critical value assessment scores to obtain unit diagnosis sample data;

[0083] Summarize the unit diagnosis sample data to obtain the historical sample data set corresponding to the medical record data set.

[0084] It is understandable that the historical diagnosis item set is a collection of multiple historical diagnosis items, and the historical diagnosis items are items detected by patients stored in the medical record data. Exemplarily, a patient underwent heart surgery, and before that, an electrocardiogram was performed. In the electrocardiogram test, multiple items that need to be tested are included, for example: the first item is atrial fibrillation with ventricular preexcitation, the shortest RR interval ≤ 250ms, the second item is frequent premature ventricular contractions, QT interval prolongation QTc ≥ 550ms, and the third item is 2nd degree type 2 atrioventricular block. The first item, the second item, and the third item are all historical diagnosis items, and if the first item, the second item, and the third item are all historical diagnosis items in the medical record data, then the historical diagnosis item set is a set consisting of the first item, the second item, and the third item.

[0085] Furthermore, the scoring criterion sequence library is a database of criteria constructed for scoring all historical diagnostic item sets. The scoring criterion sequences correspond one-to-one to historical diagnostic items. The scoring criterion sequence is a criterion sequence used to score each diagnostic value in the diagnostic value set in the historical diagnostic items. The value obtained by scoring the diagnostic value using the scoring criterion sequence is the item score. Please refer to the subsequent embodiments for specific examples.

[0086] It is understandable that the specific implementation method of using historical diagnostic items to identify item scores is: using an identifier to represent the historical diagnostic item. For example, if the historical diagnostic item is "atrial fibrillation with ventricular preexcitation, the shortest RR interval ≤250ms", and the identifier of the historical diagnostic item is set to RRleast250, when the item score is 80, the historical diagnostic item is used to identify the item score, and the obtained unit diagnostic assessment data is [RRleast250-80].

[0087] It should be noted that the physician evaluation system is an online system that multiple physicians can log in and issue medical advice and review and evaluate medical records. It can be logged in through terminals including but not limited to computers, mobile phones, and hospital self-service machines. There are many similar physician evaluation systems in the prior art, and the implementation of the present invention is not limited to this. The physician evaluation method is a method of asking multiple experienced physicians to conduct empirical evaluations on medical record data. It should be noted that even if the embodiment of the present invention obtains project scores in each historical diagnosis project, it is still necessary for the physician to determine whether the medical record data is dangerous. The critical value evaluation score is the score given to the critical value degree of the medical record data according to the physician evaluation system and the physician evaluation method. The basic information set is a set of personal basic physical data of the patient stored in the medical record data, such as gender, age, height, weight, etc.

[0088] It can be understood that the method of associating the basic information set, the unit diagnosis assessment data set and the critical value assessment score is: the present information set, the unit diagnosis assessment data set and the critical value assessment score are summarized, and the resulting data set is the unit diagnosis sample data, and the historical sample data set is the set of all unit diagnosis sample data corresponding to the medical record data set.

[0089] Furthermore, before obtaining the scoring criterion sequence corresponding to the extracted historical diagnosis item from the pre-built scoring criterion sequence library, the method further includes:

[0090] Obtain a classification code sequence, extract classification codes from the classification code sequence in sequence, and perform the following operations on the extracted classification codes:

[0091] Extract the corresponding patient data set from the medical record data set based on the classification code, and perform the following operations on each patient data in the patient data set:

[0092] Based on the patient data, the project diagnosis data set under the classification code is extracted, and the following operations are performed on each project diagnosis data in the project diagnosis data set:

[0093] If the project diagnostic data is preset complete data, confirming the project diagnostic data as available data;

[0094] Otherwise, identify the missing data types in the project diagnostic data;

[0095] If the missing data type is a pre-constructed detection data type, the project diagnostic data is eliminated; if the missing data type is a pre-constructed standard data type, a project database corresponding to the project diagnostic data is obtained, and a data filling operation is performed on the project diagnostic data based on the project database to obtain filled project diagnostic data, and the filled project diagnostic data is confirmed as available data;

[0096] The available data are aggregated to obtain an available data set, and a scoring criterion sequence library is constructed based on the available data set.

[0097] It is understandable that the surgical data is the data corresponding to the surgical items that the patient will be or have completed as recorded in the medical record data, including but not limited to: surgical operation name, main code, additional code, category and entry options. The contents included in the surgical data refer to the "National Clinical Version 3.0 of the Surgical Operation Classification Code", so the surgical operation name, main code, additional code, category and entry options are all existing professional terms and will not be repeated here. The surgical code is the code obtained by splicing the main code and additional code in the surgical data, which is used to uniquely identify the surgical data.

[0098] For example, in the "National Clinical Version 3.0 of the Surgical Procedure Classification Code", the main code corresponding to therapeutic vascular ultrasound of the head is 00.0101, and the additional code is empty. Therefore, the surgical code is 00.0101. By searching the surgical code in the "National Clinical Version 3.0 of the Surgical Procedure Classification Code", the only corresponding surgical operation name can be found, which is therapeutic vascular ultrasound of the head.

[0099] Further, the classification coding sequence is a coding sequence that includes department information and surgical coding information. It is understandable that the scoring criteria sequence library is constructed according to different departments. This is because the same test data corresponds to different criticality levels for different departments. Exemplarily, the same result of antiphospholipid antibodies should not be the same in the risk cutoff values ​​of rheumatology and reproductive departments. For the cases and needs of the reproductive department, the diagnostic value in a certain historical diagnostic project may have been able to cause miscarriage, and the corresponding critical value assessment score and unit diagnostic assessment data set have met the needs of critical value reporting, while the rheumatology department may feel that a higher antiphospholipid antibody diagnostic value is required to trigger the disease standard of rheumatism. Therefore, the embodiment of the present invention only matches and trains each diagnostic value of the patient group that has been diagnosed, that is, the medical record data of known diseases is digitized for critical values, so that when new data to be diagnosed is obtained, critical value warnings can be performed according to the corresponding disease or surgery.

[0100] For example, a patient may need vasopressor infusion, or has already undergone vasopressor infusion. The patient is then evaluated based on medical record data and a critical value assessment model. The physician can determine whether the vasopressor infusion value is reached based on the critical value assessment report, or how the surgical effect is after vasopressor infusion. This can provide a reliable basis for the clinician's diagnosis and treatment, and assist the physician in better providing safe, effective, and timely diagnosis and treatment services to the patient.

[0101] It is understandable that, since the scoring criteria sequence library is constructed based on different departments, each patient data in the patient data set is data from the same diagnosis and treatment department. The project diagnosis data set is a collection of multiple project diagnosis data, and the project diagnosis data is the historical diagnosis data in the hospital department corresponding to the classification code. The difference between the medical record data set and the patient data set is that the patient data set is a collection of multiple medical record data extracted from the hospital department corresponding to the classification code.

[0102] Furthermore, if the project diagnostic data is the preset complete data, the project diagnostic data includes: a project diagnostic value, a project empirical diagnostic upper limit value, and a project empirical diagnostic lower limit value, wherein the project diagnostic value is a detection data type, the project empirical diagnostic upper limit value and the project empirical diagnostic lower limit value are both standard data types, the detection data type is data that can only be obtained when the patient is tested, and the standard data type is a standard value that can be consulted in the project database.

[0103] Exemplarily, the project diagnostic data is hypertension diagnostic data. Therefore, if the hypertension diagnostic data includes the patient's hypertension diagnostic value, the hypertension empirical diagnostic upper limit value and the hypertension empirical diagnostic lower limit value, the hypertension empirical diagnostic upper limit value is the upper limit of the hypertension diagnostic value medically considered. If the hypertension diagnostic value is greater than the hypertension empirical diagnostic upper limit value, it means that the patient's hypertension corresponding to the project diagnostic data has reached a critical state and medical intervention is necessary. The hypertension empirical diagnostic upper limit value and the hypertension empirical diagnostic lower limit value are both stored in the project database corresponding to the hypertension diagnostic data. If any of the blood pressure empirical diagnostic upper limit value or the hypertension empirical diagnostic lower limit value does not exist in the hypertension diagnostic data, the missing data can be called from the project database corresponding to the hypertension diagnostic data. If the hypertension diagnostic value does not exist in the hypertension diagnostic data, it means that the project diagnostic data cannot be completed only through data analysis, so it needs to be eliminated.

[0104] It can be understood that, according to the above example, the missing data type includes the detection data type and the standard data type.

[0105] Further, the obtaining of the classification code sequence includes:

[0106] Extract medical record data from the medical record data set in sequence, extract surgical codes from the extracted medical record data, summarize the surgical codes to obtain a surgical code set, extract surgical codes from the surgical code set in sequence, and identify the department identification code corresponding to the extracted surgical code based on a pre-built hospital department category table, splice the department identification code and the surgical code to obtain an initial classification code;

[0107] Based on the department identification codes, the initial classification codes are respectively summarized to obtain an initial classification code set, wherein the initial classification code set includes multiple department code sets, the department code set includes multiple department codes, and the department identification codes corresponding to the multiple department codes are the same;

[0108] Department code sets are extracted from the initial classification code set in sequence, a recursive operation is performed on each department code set in the extracted department code set, and a simplified coding operation is performed on the recursive department code set to obtain a classification code sequence, wherein the classification code sequence includes multiple classification codes, and the classification code is associated with the surgery code and the department identification code corresponding to the surgery code.

[0109] It is understandable that the surgical code set is a set obtained by summarizing surgical codes. The hospital department category table is a table composed of multiple hospital departments, and the multiple hospital departments in the hospital department category table have corresponding department identification codes. The department identification code is a code that can uniquely identify the hospital department corresponding to the department identification code in the hospital department category table. The initial classification code is spliced ​​by the department identification code and the surgical code, that is, the initial classification code associates the department identification code with the surgical code, and includes both medical department information and surgical information.

[0110] It should be noted that the purpose of obtaining the initial classification code set by respectively aggregating the initial classification codes based on the department identification code is to classify the initial classification codes according to the department identification code so that there are multiple department code sets in the classified initial classification code set, and the department code set includes multiple department codes, and the department identification codes corresponding to the multiple department codes are the same. The initial classification code after the initial classification codes are classified and aggregated into the set corresponding to the same department identifier is the department code.

[0111] It is understandable that since the embodiment of the present invention constructs the classification code sequence of different departments based on all the medical record data in the medical record data, there are repeated department codes in the department code set. For example, if there are two or more patients who need monoclonal antibody treatment, there will be a corresponding number of initial classification codes in the department code set. The recursive operation is to remove the repeated initial classification codes in the department code set, and make the multiple initial classification codes retained in the department code set different, so as to obtain the recursive department code set. The simplified coding operation is to simplify the department code into a code with a preset character length, so as to shorten the code length and reduce the amount of data calculation.

[0112] It is understandable that the association of the classification code with the surgery code and the department identification code corresponding to the surgery code means that the surgery code and the department identification code can be identified from the classification code.

[0113] Furthermore, constructing a scoring criteria sequence library based on the available data set includes:

[0114] Extracting available data from the available data sets in sequence, identifying and summarizing available diagnostic items corresponding to the extracted available data, and obtaining an available diagnostic item set;

[0115] Perform a deduplication operation on the available diagnostic item set to obtain a total diagnostic item set, extract target diagnostic items from the total diagnostic item set in sequence, extract corresponding available item historical data sets from the available data sets based on the target diagnostic items, and construct a scoring criterion sequence for the extracted target diagnostic items based on the available item historical data sets;

[0116] The scoring criterion sequences are aggregated to obtain a scoring criterion sequence library.

[0117] It is understandable that the available diagnostic items are items included in the available data, which are consistent with the meaning of the historical diagnostic items, except that the available diagnostic items are diagnostic items corresponding to the available data, that is, the data in the available diagnostic items are all complete data. One available data only includes one available diagnostic item.

[0118] Further, the available diagnostic item set is a set of multiple available diagnostic items. The deduplication operation is performed on the available diagnostic item set to obtain a total diagnostic item set, which is obtained by removing duplicate available diagnostic items in the available diagnostic item set and aggregating the retained available diagnostic items, and there are no identical available diagnostic items in the total diagnostic item set.

[0119] It should be noted that the target diagnostic item is an available diagnostic item in the total diagnostic item set, and the available item historical data set is a set consisting of all available data of the target diagnostic item in the available data set.

[0120] In an embodiment of the present invention, a total set of diagnostic items is first constructed based on the available data sets. There are no identical target diagnostic items in the total set of diagnostic items. That is to say, all diagnostic items in the available data sets are checked and summarized into a set. This set is the total set of diagnostic items. The total set of diagnostic items includes multiple target diagnostic items. Therefore, in order to construct a scoring standard sequence for each target diagnostic item, an available item historical data set is also extracted from the available data sets as a parameter for constructing the scoring standard sequence.

[0121] It can be understood that the scoring criteria sequences are summarized and the resulting set is a scoring criteria sequence library, that is, there are multiple scoring criteria sequences in the scoring criteria sequence library, and the scoring criteria sequences correspond one-to-one to the target diagnostic items. The significance of this is that when new medical record data is received, the items in the new medical record data that are the same as the target diagnostic items can be evaluated.

[0122] Furthermore, constructing a scoring criterion sequence for the extracted target diagnostic items based on the available item history data set includes:

[0123] Extracting a corresponding target diagnostic data set from the available project historical data set based on the target diagnostic project, and constructing a frequency distribution graph based on the target diagnostic data set, wherein the target diagnostic data set includes a plurality of target diagnostic data, and the target diagnostic data includes: a diagnostic value, an empirical diagnostic upper limit value, and an empirical diagnostic lower limit value, and the horizontal axis of the frequency distribution graph is the diagnostic value, and the vertical axis is the number of occurrences of the diagnostic value in the target diagnostic data set;

[0124] Extracting diagnostic values ​​greater than a preset frequency from the frequency distribution diagram to obtain a reference value set to be evaluated, using the physician evaluation system and the physician evaluation method to obtain an evaluation score corresponding to each reference value to be evaluated in the reference value set to be evaluated, to obtain an evaluation score set, associating the evaluation scores in the evaluation score set with the diagnostic values, and arranging the evaluation score set in ascending order of the evaluation scores to obtain a scoring reference data set;

[0125] Extracting scoring reference data from the scoring reference data in sequence, and extracting the next scoring reference data adjacent to the scoring reference data to obtain adjacent scoring reference data, and constructing a scoring interval and a diagnostic value interval based on the scoring reference data and the adjacent scoring reference data;

[0126] Summarize the scoring intervals and the diagnostic value intervals respectively to obtain the scoring interval set and the diagnostic value interval set;

[0127] Extract target diagnostic data from the target diagnostic data set in sequence to obtain data to be normalized, identify the diagnostic value interval corresponding to the data to be normalized from the diagnostic value interval set, obtain the target interval, map the data to be normalized to the target interval, identify the scoring interval corresponding to the target interval from the scoring interval set, obtain the target scoring interval, perform proportional scoring calculation on the data to be normalized based on the target scoring interval, and obtain a normalized score;

[0128] The normalized scores and the evaluation scores are summarized in order from small to large to obtain a scoring criteria sequence.

[0129] It is understandable that the diagnostic value is the diagnostic parameter of the patient in the diagnostic item in the target diagnostic data. The empirical diagnostic upper limit and the empirical diagnostic lower limit correspond to the empirical diagnostic upper limit and the empirical diagnostic lower limit of hypertension described in the previous embodiment, and will not be repeated here.

[0130] It should be noted that the preset frequency is a set frequency number, which is used to extract diagnostic values ​​with a large number of occurrences. The reference value set to be estimated is a set of all diagnostic values ​​greater than the preset frequency. The reference value to be estimated is the diagnostic value in the reference value set to be estimated.

[0131] Exemplarily, the evaluation score corresponding to each reference value to be evaluated in the reference value set to be evaluated is obtained by using the physician evaluation system and the physician evaluation method: Assume that the target diagnostic item corresponding to the reference value set to be evaluated is: ventricular tachycardia (ventricular rate ≥ 150 times / min), and the reference value set to be evaluated is: 0, 23, 89 and 114, that is, in the frequency distribution diagram, the number of times the parameters 0, 23, 89 and 114 appear is greater than the preset frequency. The physician evaluation system and the physician evaluation method are used to score each reference value to be evaluated in 0, 23, 89 and 114, and the set of the obtained score is the evaluation score set. Assume that the evaluation score corresponding to 0 is 0, the evaluation score corresponding to 23 is 20, the evaluation score corresponding to 89 is 75, and the evaluation score corresponding to 114 is 100. Then, 0, 20, 75, and 100 are arranged in ascending order to obtain a scoring reference data set, which is: 0-0, 23-20, 89-75, 114-100. Extract the first scoring reference data 0-0 from the scoring reference data set, and extract the next scoring reference data 23-20 adjacent to the first scoring reference data. Then, the scoring interval is [0,20), the diagnostic value interval is [0,23), and so on. Examples are not given one by one here.

[0132] It is understandable that the score interval set and the diagnostic value interval set are sets consisting of all score intervals and all diagnostic value intervals, respectively. The data to be normalized is the diagnostic value in the target diagnostic data. The target interval is the diagnostic value interval corresponding to the normalized data, and the target score interval is the score interval corresponding to the target interval. Exemplarily, if the score interval is [0,20), the diagnostic value interval is [0,23), and the data to be normalized is 20, then the corresponding diagnostic value interval of the data to be normalized is [0,23), and the corresponding score interval is [0,20). Therefore, the following formula is used to calculate the proportional score of the normalized data:

[0133]

[0134] Wherein, X represents the normalized fraction, and according to the above formula, the normalized fraction is 17.3.

[0135] It can be understood that, for example, if the normalized score is 17.3 and the evaluation scores are 0 and 20, the scoring criteria sequence is: [0-0, 20-17.3, 23-20].

[0136] Furthermore, the physician evaluation system and the pre-built physician evaluation method after receiving the unit diagnosis evaluation data set are used to perform a critical value evaluation on the medical record data to obtain a critical value evaluation score corresponding to the medical record data, including:

[0137] After the unit diagnosis evaluation data set is sent to the physician evaluation system, a critical value evaluation table is obtained based on the physician evaluation system, wherein the critical value evaluation table includes: percentage evaluation criteria, the unit diagnosis evaluation data set and medical record data;

[0138] Based on a preset diagnosis and treatment physician review method, a plurality of diagnosis and treatment physician units are obtained, and a plurality of initial evaluation scores are obtained based on the plurality of diagnosis and treatment physician units, wherein there is one and only one account in the diagnosis and treatment physician unit, and the initial evaluation score corresponds to the diagnosis and treatment physician unit one by one;

[0139] The average of multiple initial assessment scores is calculated to obtain a critical value assessment score.

[0140] It is understandable that the percentage evaluation criteria is that the full score is 100 as the criterion for the total evaluation score. The initial evaluation score is the score given by the physicians in the treating physician unit to the medical record data. Only one physician in a treating physician unit can log in and use it. The treating physician review method is that the physician scores the medical record data based on his or her own experience and the unit's diagnosis and evaluation data set.

[0141] S2. Train a pre-built initial assessment model based on a historical sample data set to obtain a critical value assessment model, wherein both the initial assessment model and the critical value assessment model are models constructed by neural networks.

[0142] Furthermore, the pre-built initial assessment model is trained based on the historical sample data set to obtain a critical value assessment model, including:

[0143] The historical sample data set is divided based on multiple department identification codes to obtain multiple department classification data sets, and the following operation is performed on each of the multiple department classification data sets:

[0144] The department classification data set is divided by using preset sample training values ​​to obtain a training sample set and a verification sample set, a disease information set and a disease critical value score set are extracted from the training sample set, and the disease information set and the disease critical value score set are used to train a pre-constructed classification initial evaluation model, wherein the classification initial evaluation model is a model constructed by a neural network, and the classification initial evaluation model includes: a disease information input layer, a pooling layer, a fully connected layer and a disease critical value score output layer, the pooling layer adopts a maximum pooling strategy, the pooling function size is 2×2, and the fully connected layer includes 1024 unit nodes. The loss function in the classification evaluation model is:

[0145]

[0146] Among them, LOST MSE represents the mean square error loss function, N represents the number of training samples in the training sample set, i represents the disease criticality score corresponding to the i-th training sample, Represents the disease criticality score predicted by the classification assessment model;

[0147] The validation sample set is used to perform reverse update calculation on the classification initial evaluation model to obtain the classification evaluation model, the classification evaluation model is summarized to obtain the initial evaluation model, and a retraining sample set is randomly extracted from the historical sample data set based on the sample training value. The initial evaluation model is iterated using the retraining sample set until the number of iteration steps of the iterative operation reaches the preset number of iteration steps to obtain the critical value evaluation model.

[0148] It is understandable that the department classification data set is a set of multiple historical sample data classified according to medical departments. The sample training value is the ratio between the department classification data used for training in the department classification data set and the department classification data set. For example, if the sample training value is 80%, 80% of the department classification data in the department classification data set will be used to train the classification initial evaluation model. The disease information is the input data corresponding to the training sample. The disease information is extracted from the available data, so the disease information includes all the data in the corresponding available data. The disease critical value score is the critical value score corresponding to the training sample.

[0149] It can be understood that the disease information input layer is the input layer in the classification initial assessment model, and the disease critical value score output layer is the output layer in the classification initial assessment model. That is to say, in the classification initial assessment model, after the disease information is input, the output value obtained is the predicted disease critical value score.

[0150] It should be noted that the reverse update calculation is a prior art and will not be described in detail here.

[0151] It is understandable that the purpose of randomly extracting a retraining sample set from the historical sample data set based on the sample training value is to iteratively train the initial evaluation model. The technology of iterative training is a prior art and will not be described in detail here. Optionally, the preset number of iteration steps in the embodiment of the present invention is 1000.

[0152] S3. Obtain the medical record data of the patient to be evaluated, import the medical record data of the patient to be evaluated into the critical value evaluation model, and obtain a critical value evaluation report.

[0153] It is understandable that the patient medical record data to be evaluated is the patient medical record data that needs to be predicted. The critical value assessment report includes the disease critical value score of the patient medical record data to be evaluated predicted based on the critical value assessment model, and also includes the item scores corresponding to multiple test items in the patient medical record data to be evaluated. The acquisition of multiple item scores has been explained in detail in the embodiments of the present invention and will not be repeated here.

[0154] S4. Send the critical value assessment report to the pre-built medical order system. If an automatic diagnosis instruction is received from the pre-built medical order system control end, a recommended diagnosis report is generated based on the automatic diagnosis instruction, and the recommended diagnosis report is confirmed as the initial medical order report. Otherwise, the initial medical order report is obtained based on the medical order system after receiving the critical value assessment report.

[0155] It is understandable that the medical order system is a system for establishing medical orders, which is an existing technology. The medical order system control terminal is a control port for the physician to control the medical order system, which can be a computer, a mobile phone, etc. It is understandable that the medical order system has an embedded intelligent diagnosis system, which is a medical plan recommendation system based on artificial intelligence. There are many existing technologies that can be implemented, and no limitation is made here. The automatic diagnosis instruction is an instruction initiated by the physician through the medical order system control terminal to control the intelligent diagnosis system to generate a recommended diagnosis report. It is understandable that if the automatic diagnosis instruction is not received, it means that the physician expects to make a diagnosis by himself, so the physician can receive the critical value assessment report and establish an initial medical order report based on the critical value assessment report.

[0156] S5. After confirming that the initial medical order report is a preset feasible report using a pre-built report feasibility detection method, obtain the medical order report, perform encryption operations on the medical order report to obtain encrypted data, and send the encrypted data to a pre-built report query terminal, wherein the report query terminal includes multiple responsibility query terminals.

[0157] It can be understood that the report feasibility detection method is a method in which multiple doctors verify whether the initial medical order report is indeed feasible. The implementation method is to send the initial medical order report to multiple doctors, and multiple doctors independently judge whether the initial medical order report is feasible. If multiple doctors believe that the initial medical order report is feasible, the initial medical order report is confirmed as a preset feasible report, and the medical order report is obtained.

[0158] It should be noted that performing encryption operations on medical order reports to obtain encrypted data is an existing technology and will not be described in detail here. The responsibility query end is a port used by the person in charge who must receive and view the medical order report according to the current hospital rules. For example, multiple responsibility query ends include: the attending physician responsibility query end, the pharmacist responsibility query end, the patient responsibility query end, and if the patient is a minor, there may also be a guardian responsibility query end.

[0159] S6. After confirming that each responsible query end in the report query end has generated a report, the critical value report based on deep learning is completed.

[0160] Furthermore, after each responsible query end in the confirmation report query end generates a report, it includes:

[0161] Extract the responsible query ends from the report query ends in sequence, and perform the following operations on the extracted responsible query ends:

[0162] Based on the preset query time, determine whether an access request instruction is received from the responsible query end; if an access request instruction is received from the responsible query end, perform an identity authentication operation on the responsible query end to obtain identity authentication data; if it is confirmed that the identity authentication data is consistent with the authentication data pre-built in the responsible query end, perform a decryption operation on the encrypted data to obtain decrypted data, and generate a report based on the decrypted data;

[0163] Otherwise, the delayed query time is obtained, and the delayed query time is used as the preset query time, and the step of determining whether an access request instruction is received from the responsible query end based on the preset query time is returned until a report is generated.

[0164] It is understandable that the query time is the maximum time for receiving the report set manually. If no access request instruction is received from the responsible query end within the query period, it means that the responsible query end has not queried the medical order report. The access request instruction is an instruction initiated when the responsible query end needs to query the medical order report. The identity authentication operation is an operation to first confirm whether the user of the responsible query end is the preset responsible person when the responsible query end needs to query the medical order report. There are many existing technologies that can be implemented, for example, using dynamic codes, verifying identity cards, face recognition, etc., and the embodiments of the present invention do not limit this. The pre-constructed authentication data is the data pre-stored by the responsible query end according to the identity authentication operation, which is used to determine whether the user of the responsible query end is the preset responsible person.

[0165] It is understandable that performing a decryption operation on encrypted data to obtain decrypted data is a prior art and will not be described in detail here. The feedback report is a report used to report that the responsible query end has queried the medical order report, including the query time, query duration, IP address of the responsible query end, etc. If the responsible query end does not query the medical order report within the query time, the delayed query time will be obtained again to determine whether the responsible query end has queried the medical order report. Optionally, after the responsible query end has not queried the medical order report, a reminder can be made through manual telephone connection.

[0166] In order to solve the problems described in the background technology, the present invention first obtains a medical record data set, performs preprocessing operations on the medical record data set to obtain a historical sample data set, trains a pre-constructed initial evaluation model based on the historical sample data set to obtain a critical value evaluation model, obtains the medical record data of the patient to be evaluated, imports the medical record data of the patient to be evaluated into the critical value evaluation model, and obtains a critical value evaluation report. By evaluating the patient based on the medical record data and the critical value evaluation model, a critical value evaluation report is obtained. The physician can judge whether diagnosis and treatment are needed and the effect of diagnosis and treatment based on the obtained critical value evaluation report, which can provide a reliable basis for the diagnosis and treatment of clinicians, and assist physicians to better provide patients with safe, effective and timely diagnosis and treatment services. The critical value assessment report is sent to the pre-built medical order system. If an automatic diagnosis instruction is received from the pre-built medical order system control end, a recommended diagnosis report is generated based on the automatic diagnosis instruction, and the recommended diagnosis report is confirmed as the initial medical order report. Otherwise, the initial medical order report is obtained based on the medical order system after receiving the critical value assessment report. The present invention can automatically diagnose and generate the initial medical order report, and can also generate the initial medical order report according to the medical system through the doctor's diagnosis. After confirming that the initial medical order report is a preset feasible report using the pre-built report feasibility detection method, the medical order report is obtained. In the present invention, if multiple doctors believe that the initial medical order report is feasible, the initial medical order report is confirmed as a preset feasible report, ensuring the security of the medical order report. The medical order report is encrypted to obtain encrypted data, and the encrypted data is sent to the pre-built report query end, wherein the report query end includes multiple responsible query ends. After confirming that each responsible query end in the report query end generates a report, the critical value report based on deep learning is completed. The present invention also encrypts the medical order report and pushes it to multiple responsible persons. Only when it is confirmed that multiple responsible persons have received the medical report, the critical value report is confirmed to be completed, which is conducive to supervising each responsible person. Therefore, the present invention can use the critical value assessment model to assess the criticality of patient data to provide a reference for diagnosis and treatment for doctors, and at the same time, push the medical order report to multiple responsible persons in an encrypted form to achieve supervision and implementation of each responsible person.

[0167] like Figure 2 As shown, it is a functional module diagram of a critical value reporting system based on deep learning provided by one embodiment of the present invention.

[0168] The critical value reporting system 100 based on deep learning of the present invention can be installed in an electronic device. According to the functions implemented, the critical value reporting system 100 based on deep learning can include a critical value report acquisition module 101, a medical order report acquisition module 102 and a report generation module 103. The module of the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0169] The critical value report acquisition module 101 is used to acquire a medical record data set, perform a preprocessing operation on the medical record data set to obtain a historical sample data set, train a pre-constructed initial evaluation model based on the historical sample data set to obtain a critical value evaluation model, wherein both the initial evaluation model and the critical value evaluation model are models composed of neural networks, acquire the medical record data of the patient to be evaluated, import the medical record data of the patient to be evaluated into the critical value evaluation model, and obtain a critical value evaluation report;

[0170] The doctor's order report acquisition module 102 is used to send the critical value assessment report to the pre-built doctor's order system. If an automatic diagnosis instruction is received from the pre-built doctor's order system control terminal, a recommended diagnosis report is generated based on the automatic diagnosis instruction, and the recommended diagnosis report is confirmed as the initial doctor's order report. Otherwise, the initial doctor's order report is obtained based on the doctor's order system after receiving the critical value assessment report, and the doctor's order report is obtained after confirming that the initial doctor's order report is a preset feasible report using a pre-built report feasibility detection method;

[0171] The report generation module 103 is used to perform encryption operations on the medical order report to obtain encrypted data, and send the encrypted data to a pre-built report query terminal, wherein the report query terminal includes multiple responsibility query terminals, and after confirming that each responsibility query terminal in the report query terminal has generated a report, the critical value report based on deep learning is completed.

[0172] In detail, each module in the critical value reporting system 100 based on deep learning in the embodiment of the present invention is used in the same manner as described above. Figure 1 The same technical means as the critical value reporting method based on deep learning described in the text and can produce the same technical effects will not be repeated here.

[0173] like Figure 3 , is a schematic diagram of the structure of an electronic device for implementing a critical value reporting method based on deep learning provided by an embodiment of the present invention.

[0174] The electronic device 1 may include a processor 10, a memory 11 and a bus 12, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a critical value reporting method program based on deep learning.

[0175] Among them, the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. The memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as a mobile hard disk of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 also includes an internal storage unit of the electronic device 1 and an external storage device. The memory 11 can not only be used to store application software and various types of data installed in the electronic device 1, such as the code of the critical value reporting method program based on deep learning, but also can be used to temporarily store data that has been output or is to be output.

[0176] The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules (such as a critical value reporting method program based on deep learning, etc.) stored in the memory 11, and calls data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

[0177] The bus 12 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus 12 may be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0178] Figure 3 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 3The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0179] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management system, so that the power management system can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0180] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0181] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0182] The critical value reporting method program based on deep learning stored in the memory 11 of the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0183] Acquire a medical record data set, perform preprocessing operations on the medical record data set to obtain a historical sample data set, train a pre-built initial assessment model based on the historical sample data set, and obtain a critical value assessment model, wherein both the initial assessment model and the critical value assessment model are models composed of neural networks;

[0184] Obtain the medical record data of the patient to be evaluated, import the medical record data of the patient to be evaluated into the critical value evaluation model, and obtain the critical value evaluation report;

[0185] Sending the critical value assessment report to a pre-built medical order system, if receiving an automatic diagnosis instruction from a control terminal of the pre-built medical order system, generating a recommended diagnosis report based on the automatic diagnosis instruction, and confirming the recommended diagnosis report as an initial medical order report, otherwise, obtaining an initial medical order report based on the medical order system after receiving the critical value assessment report;

[0186] After confirming that the initial medical order report is a preset feasible report by using a pre-constructed report feasibility detection method, the medical order report is obtained, an encryption operation is performed on the medical order report to obtain encrypted data, and the encrypted data is sent to a pre-constructed report query end, wherein the report query end includes a plurality of responsible query ends;

[0187] After confirming that each responsible query end in the report query end has generated a report, the critical value report based on deep learning is completed.

[0188] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figures 1 to 3 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0189] Furthermore, if the module / unit integrated in the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or system that can carry the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0190] The present invention further 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 of an electronic device, the computer program can implement:

[0191] Acquire a medical record data set, perform preprocessing operations on the medical record data set to obtain a historical sample data set, train a pre-built initial assessment model based on the historical sample data set, and obtain a critical value assessment model, wherein both the initial assessment model and the critical value assessment model are models composed of neural networks;

[0192] Obtain the medical record data of the patient to be evaluated, import the medical record data of the patient to be evaluated into the critical value evaluation model, and obtain the critical value evaluation report;

[0193] Sending the critical value assessment report to a pre-built medical order system, if receiving an automatic diagnosis instruction from a control terminal of the pre-built medical order system, generating a recommended diagnosis report based on the automatic diagnosis instruction, and confirming the recommended diagnosis report as an initial medical order report, otherwise, obtaining an initial medical order report based on the medical order system after receiving the critical value assessment report;

[0194] After confirming that the initial medical order report is a preset feasible report by using a pre-constructed report feasibility detection method, the medical order report is obtained, an encryption operation is performed on the medical order report to obtain encrypted data, and the encrypted data is sent to a pre-constructed report query end, wherein the report query end includes a plurality of responsible query ends;

[0195] After confirming that each responsible query end in the report query end has generated a report, the critical value report based on deep learning is completed.

[0196] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative, and actual implementation may have other division methods.

[0197] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0198] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0199] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0200] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A critical value reporting method based on deep learning, characterized in that: The method comprises: Acquire a medical record data set, perform preprocessing operations on the medical record data set to obtain a historical sample data set, train a pre-built initial assessment model based on the historical sample data set, and obtain a critical value assessment model, wherein both the initial assessment model and the critical value assessment model are models composed of neural networks; Obtain the medical record data of the patient to be evaluated, import the medical record data of the patient to be evaluated into the critical value evaluation model, and obtain the critical value evaluation report; Sending the critical value assessment report to a pre-built medical order system, if receiving an automatic diagnosis instruction from a control terminal of the pre-built medical order system, generating a recommended diagnosis report based on the automatic diagnosis instruction, and confirming the recommended diagnosis report as an initial medical order report, otherwise, obtaining an initial medical order report based on the medical order system after receiving the critical value assessment report; After confirming that the initial medical order report is a preset feasible report by using a pre-constructed report feasibility detection method, the medical order report is obtained, an encryption operation is performed on the medical order report to obtain encrypted data, and the encrypted data is sent to a pre-constructed report query end, wherein the report query end includes a plurality of responsible query ends; After confirming that each responsible query end in the report query end has generated a report, the critical value report based on deep learning is completed.

2. The critical value reporting method based on deep learning according to claim 1, characterized in that: The preprocessing operation is performed on the medical record data set to obtain a historical sample data set, including: Extract medical record data from the medical record data set in sequence, and perform the following operations on the extracted medical record data: Based on the extracted medical record data, a historical diagnosis item set is obtained, historical diagnosis items are sequentially extracted from the historical diagnosis item set, and a scoring criterion sequence corresponding to the extracted historical diagnosis item is obtained from a pre-built scoring criterion sequence library, and item scores corresponding to the medical record data in the historical diagnosis item are obtained based on the scoring criterion sequence, and item scores are identified using the historical diagnosis item to obtain unit diagnosis evaluation data; Summarize the unit diagnosis and evaluation data to obtain the unit diagnosis and evaluation data set corresponding to the medical record data; After the unit diagnosis and evaluation data set is sent to the pre-built physician evaluation system, the critical value evaluation of the medical record data is performed using the physician evaluation system after receiving the unit diagnosis and evaluation data set and the pre-built physician evaluation method to obtain the critical value evaluation score corresponding to the medical record data; Extracting a basic information set based on medical record data, correlating the basic information set, unit diagnosis assessment data set, and critical value assessment score to obtain unit diagnosis sample data; Summarize the unit diagnosis sample data to obtain the historical sample data set corresponding to the medical record data set.

3. The critical value reporting method based on deep learning according to claim 2, characterized in that: Before obtaining the scoring criterion sequence corresponding to the extracted historical diagnosis item from the pre-built scoring criterion sequence library, the method further includes: Obtain a classification code sequence, extract classification codes from the classification code sequence in sequence, and perform the following operations on the extracted classification codes: Extract the corresponding patient data set from the medical record data set based on the classification code, and perform the following operations on each patient data in the patient data set: Based on the patient data, the project diagnosis data set under the classification code is extracted, and the following operations are performed on each project diagnosis data in the project diagnosis data set: If the project diagnostic data is preset complete data, confirming the project diagnostic data as available data; Otherwise, identify the missing data types in the project diagnostic data; If the missing data type is a pre-constructed detection data type, the project diagnostic data is eliminated; if the missing data type is a pre-constructed standard data type, a project database corresponding to the project diagnostic data is obtained, and a data filling operation is performed on the project diagnostic data based on the project database to obtain filled project diagnostic data, and the filled project diagnostic data is confirmed as available data; The available data are aggregated to obtain an available data set, and a scoring criterion sequence library is constructed based on the available data set.

4. The critical value reporting method based on deep learning as claimed in claim 3, characterized in that: The method of constructing a scoring criteria sequence library based on an available data set comprises: Extracting available data from the available data sets in sequence, identifying and summarizing available diagnostic items corresponding to the extracted available data, and obtaining an available diagnostic item set; Perform a deduplication operation on the available diagnostic item set to obtain a total diagnostic item set, extract target diagnostic items from the total diagnostic item set in sequence, extract corresponding available item historical data sets from the available data sets based on the target diagnostic items, and construct a scoring criterion sequence for the extracted target diagnostic items based on the available item historical data sets; The scoring criterion sequences are aggregated to obtain a scoring criterion sequence library.

5. The critical value reporting method based on deep learning according to claim 4, characterized in that: The obtaining of the classification coding sequence comprises: Extract medical record data from the medical record data set in sequence, extract surgical codes from the extracted medical record data, summarize the surgical codes to obtain a surgical code set, extract surgical codes from the surgical code set in sequence, and identify the department identification code corresponding to the extracted surgical code based on a pre-built hospital department category table, splice the department identification code and the surgical code to obtain an initial classification code; Based on the department identification codes, the initial classification codes are respectively summarized to obtain an initial classification code set, wherein the initial classification code set includes multiple department code sets, the department code set includes multiple department codes, and the department identification codes corresponding to the multiple department codes are the same; Department code sets are extracted from the initial classification code set in sequence, a recursive operation is performed on each department code set in the extracted department code set, and a simplified coding operation is performed on the recursive department code set to obtain a classification code sequence, wherein the classification code sequence includes multiple classification codes, and the classification code is associated with the surgery code and the department identification code corresponding to the surgery code.

6. The critical value reporting method based on deep learning according to claim 5, characterized in that: After each responsible query end in the confirmation report query end generates a report, it includes: Extract the responsible query ends from the report query ends in sequence, and perform the following operations on the extracted responsible query ends: Based on the preset query time, determine whether an access request instruction is received from the responsible query end; if an access request instruction is received from the responsible query end, perform an identity authentication operation on the responsible query end to obtain identity authentication data; if it is confirmed that the identity authentication data is consistent with the authentication data pre-built in the responsible query end, perform a decryption operation on the encrypted data to obtain decrypted data, and generate a report based on the decrypted data; Otherwise, the delayed query time is obtained, and the delayed query time is used as the preset query time, and the step of determining whether an access request instruction is received from the responsible query end based on the preset query time is returned until a report is generated.

7. The critical value reporting method based on deep learning according to claim 6, characterized in that: The step of constructing a scoring criterion sequence for the extracted target diagnostic items based on the available item history data set includes: Extracting a corresponding target diagnostic data set from the available project historical data set based on the target diagnostic project, and constructing a frequency distribution graph based on the target diagnostic data set, wherein the target diagnostic data set includes a plurality of target diagnostic data, and the target diagnostic data includes: a diagnostic value, an empirical diagnostic upper limit value, and an empirical diagnostic lower limit value, and the horizontal axis of the frequency distribution graph is the diagnostic value, and the vertical axis is the number of occurrences of the diagnostic value in the target diagnostic data set; Extracting diagnostic values ​​greater than a preset frequency from the frequency distribution diagram to obtain a reference value set to be evaluated, using the physician evaluation system and the physician evaluation method to obtain an evaluation score corresponding to each reference value to be evaluated in the reference value set to be evaluated, to obtain an evaluation score set, associating the evaluation scores in the evaluation score set with the diagnostic values, and arranging the evaluation score set in ascending order of the evaluation scores to obtain a scoring reference data set; Extracting scoring reference data from the scoring reference data in sequence, and extracting the next scoring reference data adjacent to the scoring reference data to obtain adjacent scoring reference data, and constructing a scoring interval and a diagnostic value interval based on the scoring reference data and the adjacent scoring reference data; Summarize the scoring intervals and the diagnostic value intervals respectively to obtain the scoring interval set and the diagnostic value interval set; Extract target diagnostic data from the target diagnostic data set in sequence to obtain data to be normalized, identify the diagnostic value interval corresponding to the data to be normalized from the diagnostic value interval set, obtain the target interval, map the data to be normalized to the target interval, identify the scoring interval corresponding to the target interval from the scoring interval set, obtain the target scoring interval, perform proportional scoring calculation on the data to be normalized based on the target scoring interval, and obtain a normalized score; The normalized scores and the evaluation scores are summarized in order from small to large to obtain a scoring criteria sequence.

8. The critical value reporting method based on deep learning according to claim 7, characterized in that: The physician evaluation system and the pre-built physician evaluation method after receiving the unit diagnosis evaluation data set are used to evaluate the critical value of the medical record data to obtain the critical value evaluation score corresponding to the medical record data, including: After the unit diagnosis evaluation data set is sent to the physician evaluation system, a critical value evaluation table is obtained based on the physician evaluation system, wherein the critical value evaluation table includes: percentage evaluation criteria, the unit diagnosis evaluation data set and medical record data; Based on a preset diagnosis and treatment physician review method, a plurality of diagnosis and treatment physician units are obtained, and a plurality of initial evaluation scores are obtained based on the plurality of diagnosis and treatment physician units, wherein there is one and only one account in the diagnosis and treatment physician unit, and the initial evaluation score corresponds to the diagnosis and treatment physician unit one by one; The average of multiple initial assessment scores is calculated to obtain a critical value assessment score.

9. The critical value reporting method based on deep learning according to claim 8, characterized in that: The pre-built initial assessment model is trained based on the historical sample data set to obtain a critical value assessment model, including: The historical sample data set is divided based on multiple department identification codes to obtain multiple department classification data sets, and the following operation is performed on each of the multiple department classification data sets: The department classification data set is divided by using preset sample training values ​​to obtain a training sample set and a verification sample set, a disease information set and a disease critical value score set are extracted from the training sample set, and the disease information set and the disease critical value score set are used to train a pre-constructed classification initial evaluation model, wherein the classification initial evaluation model is a model constructed by a neural network, and the classification initial evaluation model includes: a disease information input layer, a pooling layer, a fully connected layer and a disease critical value score output layer, the pooling layer adopts a maximum pooling strategy, the pooling function size is 2×2, and the fully connected layer includes 1024 unit nodes. The loss function in the classification evaluation model is: Among them, LOST MSE represents the mean square error loss function, N represents the number of training samples in the training sample set, i represents the disease criticality score corresponding to the i-th training sample, Represents the disease criticality score predicted by the classification assessment model; The validation sample set is used to perform reverse update calculation on the classification initial evaluation model to obtain the classification evaluation model, the classification evaluation model is summarized to obtain the initial evaluation model, and a retraining sample set is randomly extracted from the historical sample data set based on the sample training value. The initial evaluation model is iterated using the retraining sample set until the number of iteration steps of the iterative operation reaches the preset number of iteration steps to obtain the critical value evaluation model.

10. A critical value reporting system based on deep learning, characterized in that: The system comprises: A critical value report acquisition module is used to acquire a medical record data set, perform preprocessing operations on the medical record data set, obtain a historical sample data set, train a pre-built initial evaluation model based on the historical sample data set, and obtain a critical value evaluation model, wherein both the initial evaluation model and the critical value evaluation model are models composed of neural networks, acquire the medical record data of the patient to be evaluated, import the medical record data of the patient to be evaluated into the critical value evaluation model, and obtain a critical value evaluation report; A doctor's order report acquisition module is used to send the critical value assessment report to a pre-built doctor's order system. If an automatic diagnosis instruction is received from the pre-built doctor's order system control terminal, a recommended diagnosis report is generated based on the automatic diagnosis instruction, and the recommended diagnosis report is confirmed as an initial doctor's order report. Otherwise, the initial doctor's order report is obtained based on the doctor's order system after receiving the critical value assessment report, and the doctor's order report is obtained after confirming that the initial doctor's order report is a preset feasible report using a pre-built report feasibility detection method; The reporting report generation module is used to perform encryption operations on the medical order report, obtain encrypted data, and send the encrypted data to a pre-built report query terminal, wherein the report query terminal includes multiple responsible query terminals. After confirming that each responsible query terminal in the report query terminal has generated a reporting report, the critical value report based on deep learning is completed.

Images