Central infection medication decision support system based on AI

Abstract

The invention relates to the technical field of health management and computers, in particular to an AI-based central infection medication decision support system, comprising a receiving unit; the construction unit is used for classifying the historical patients according to the historical patient characteristics, obtaining a medication prescription corresponding to each classification, constructing a real world database, and establishing a medication relation matrix and a prognosis relation matrix; the auxiliary medicine taking unit is used for receiving a group of new patient characteristics input by the user, inputting a medicine taking prescription corresponding to the category to the user, and when the new patient does not belong to any category, taking the medicine taking prescription; and outputting a theoretical prescription formed by medication selection, medication metering, associated medication and medication modes to a user according to the medication relation matrix. The method is used for providing medication reference for doctors, so that the accuracy of central nervous infection medication is ensured, and subsequent adverse reactions are reduced.

Description

Technical Field

[0001] The present invention relates to the fields of health management and computer technology, and in particular to an AI-based central nervous system infection medication decision support system. Background Art

[0002] Postoperative central nervous system infections (NCNSIs) are serious complications that can prolong hospitalization and increase the morbidity and mortality of patients. According to statistics, the infection rate of postoperative CNSIs in neurosurgery is 4.6% to 25%, and the mortality rate of postoperative meningitis and / or ventriculitis in neurosurgery is 3% to 33%. Even if CNSIs are cured, patients generally have varying degrees of neurological dysfunction.

[0003] Linezolid is an effective drug for postoperative central nervous system infection in neurosurgery. It can be rapidly distributed to well-perfused organs and tissues after intravenous or oral administration. During the use of linezolid, achieving therapeutic antimicrobial drug concentrations at the site of infection is a necessary condition for optimal in vivo efficacy. For pathogens in general ward patients, it is easy to achieve the goal of MICs of 1-2 mg / L according to the instructions (600 mg intravenous injection twice a day). However, in special patient populations (critically ill patients, patients with renal insufficiency, obese patients, pediatric patients, and patients receiving P-glycoprotein (P-gp) regulators (rifampicin)), their drug clearance (CL) and / or distribution volume (Vd) vary greatly in vivo and between individuals, and conventional doses may result in insufficient drug exposure or potential toxic side effects. However, inter-individual PK variability of linezolid in plasma and cerebrospinal fluid concentrations has also been observed in critically ill patients, who are known to have experienced pathophysiological changes (enhanced or impaired renal function, hyperdynamic circulatory state), which can significantly change the PK of antibiotics. Taking individual factors into consideration, some studies have used a higher dosing regimen than the standard 600 mg (twice a day) to optimize the therapeutic effect of linezolid and address the problem of treatment failure in critically ill neurosurgery patients. However, high-dose and long-term use of linezolid may increase the risk of adverse drug reactions, usually thrombocytopenia.

[0004] For special patient populations, administering conventional doses of linezolid to treat central nervous system infections increases the risk of treatment failure, but increasing the dosage may increase the risk of adverse drug reactions. Therefore, how to optimize the linezolid treatment regimen to ensure efficacy while reducing adverse reactions has become a major issue that needs to be addressed in the current treatment of central nervous system infections. Summary of the invention

[0005] To solve the above problems, the present invention provides an AI-based central nervous system infection medication decision support system, which is used to provide physicians with medication references, thereby ensuring the accuracy of central nervous system infection medications and reducing the occurrence of subsequent adverse reactions.

[0006] In order to achieve the above-mentioned purpose, the technical scheme of the present invention is as follows: a central infection medication decision support system based on AI, a central infection medication decision support system based on AI, comprising:

[0007] Receiving unit: The receiving unit is used to accept the medication prescription, follow-up results and a corresponding set of historical patient characteristics input by the user;

[0008] Construction unit: The construction unit is used to classify historical patients according to historical patient characteristics, obtain the corresponding medication prescriptions for each classification, and build a real-world database. The construction unit is also used to build the relationship between medication selection, medication dosage, related medication, and medication methods in medication prescriptions and the corresponding historical patient characteristics, and establish a medication relationship matrix. It is also used to build the relationship between follow-up results and medication selection, medication dosage, related medication, and medication methods in medication prescriptions, and build a prognosis relationship matrix;

[0009] Auxiliary medication unit: The auxiliary medication unit is used to receive a group of new patient characteristics input by the user, and determine the category to which the new patient belongs based on the new patient characteristics, and input the medication prescription corresponding to the category to the user. When the new patient does not belong to any category, the historical patient characteristics corresponding to the group of new patient characteristics are selected one by one, and the corresponding medication selection, medication dosage, related medication and medication method are output according to the medication relationship matrix, and the theoretical prescription consisting of medication selection, medication dosage, related medication and medication method is output to the user;

[0010] Prognosis and outcome analysis unit: The prognosis and outcome analysis unit is used to receive the medication regimen input by the user, and output the predicted follow-up results using the prognosis relationship matrix according to the medication selection, medication dosage, associated medication and medication method contained in the medication regimen;

[0011] Adverse reaction warning unit: The adverse reaction warning unit is used to screen historical patient characteristics, obtain drug-induced adverse reaction information of historical patients, and output the drug-induced adverse reaction information corresponding to the historical patient when the new patient characteristics and corresponding medication regimen input by the user are the same as the historical patient characteristics and medication prescriptions of the historical patient.

[0012] Furthermore, it also includes a prediction unit, which is used to select each group of historical patient characteristics into two parts, predict the other part of historical patient characteristics several times according to any part of the historical patient characteristics, so as to form several groups of prediction characteristics, and predict the prediction prescription corresponding to each group of prediction characteristics to form a prediction database. The auxiliary medication unit is also used to compare the input new patient characteristics with the prediction characteristics in the prediction database after the query in the real-world database fails, select a group of prediction characteristics in the prediction database with the greatest similarity to the new patient characteristics, and output the prediction prescription corresponding to the group of prediction characteristics.

[0013] Furthermore, the receiving unit is also used to receive several medication prescriptions, follow-up results and a corresponding set of historical patient characteristics during the historical patient treatment process input by the user, and the construction unit is used to update the real-world database based on the medication prescriptions, follow-up results and a corresponding set of historical patient characteristics.

[0014] Furthermore, the prediction unit is also used to predict prediction features and prediction prescriptions based on a number of medication prescriptions, follow-up results and a corresponding set of historical patient characteristics in the historical patient treatment process input by the user received by the receiving unit, and to update the prediction database after each prediction feature and prediction prescription are obtained, and to use the obtained historical patient characteristics and medication prescriptions to correct the original prediction features and prediction data in the prediction database.

[0015] Furthermore, the receiving unit is also used to judge the effectiveness of the medication prescription according to the follow-up consultation results, and after the medication prescription is judged to be invalid, screen out the corresponding medication prescription, follow-up consultation results and a corresponding set of historical patient characteristics.

[0016] Furthermore, the receiving unit is also used to continuously receive medication regimens, follow-up results and a corresponding set of new patient characteristics during the treatment of new patients. The construction component is also used to update the real-world database after the new patient treatment process is completed based on the medication regimens, follow-up results and a corresponding set of new patient characteristics received each time.

[0017] Furthermore, it also includes a monitoring unit, which is used to collect voice information of new patients, judge the status of new patients based on the voice information, and output alarm information to the user when the new patient has adverse reactions corresponding to the medication prescription.

[0018] Furthermore, the monitoring unit is also used to receive allergy information of a new patient input by a user, and to adjust the theoretical prescription or the predicted prescription when the auxiliary medication unit outputs the theoretical prescription or the prediction unit outputs the predicted prescription.

[0019] Furthermore, the monitoring unit is also used to judge the drug-induced adverse reaction information of the new patient based on the new patient's follow-up results, and to query the new patient's adverse reaction information in the real-world database. When the same adverse reaction information appears in the real-world database, the medication prescription of the historical patient corresponding to the adverse reaction information is output.

[0020] Furthermore, the monitoring unit is also used to mark all medication prescriptions of the historical patient after outputting the medication prescription of the historical patient corresponding to the adverse reaction information.

[0021] The technical principles and beneficial effects of the above scheme are as follows:

[0022] 1. In this solution, by constructing a real-world database, medical staff can input the characteristics of new patients into the system, query the prescriptions used by patients with corresponding characteristics in the real-world database, and judge the feasibility of the medication plan formulated for the new patient based on the output medication plan. The medication plan formulated for the new patient can also be adjusted based on the output medication plan so that the medication plan formulated for the new patient can achieve the best expected effect, thereby achieving the purpose of assisting clinical medication.

[0023] 2. This solution, by establishing a medication relationship matrix, constructs the relationship between various characteristics of historical patients in the real-world database for medication prescriptions. When there is no set of characteristics that are completely identical to those of new patients in the real-world database, the medication relationship matrix is ​​used to find the medication prescriptions corresponding to each characteristic, thereby forming a theoretical prescription for user reference. Compared with the existing technology, this solution provides users with theoretical prescriptions when faced with scenarios such as the need to use linezolid and new patients are special populations. This can help users adjust the conventional linezolid medication regimen, avoid possible adverse reactions of conventional medications for special patients, and reduce the occurrence of adverse reactions in subsequent new patients or poor treatment effects due to insufficient medication dosage.

[0024] 3. This solution, through the design of the prognosis and outcome analysis unit and the adverse reaction warning unit, helps users judge the rationality and effectiveness of the medication plan by allowing users to input the corresponding medication plan, the possible conditions and adverse reactions of new users. Compared with the existing technology, this solution can predict the condition of new patients after using the medicine, help users adjust the medication plan, and thus reduce the probability of adverse reactions in new patients after taking the medicine.

[0025] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1It is a structural schematic diagram of an embodiment of an AI-based central nervous system infection medication decision support system of the present invention;

[0027] Figure 2 This is a prescription prediction flow chart of an embodiment of the AI-based central nervous system infection medication decision support system of the present invention;

[0028] Figure 3 This is a flow chart of the prediction of follow-up results of an embodiment of the AI-based central nervous system infection medication decision support system of the present invention;

[0029] Figure 4 This is a flowchart of adverse reaction prediction of an embodiment of the AI-based central nervous system infection medication decision support system of the present invention.

[0030] The figure marks in the drawings of the specification include: 1. receiving unit; 2. building unit; 3. prediction unit; 4. monitoring unit; 5. auxiliary medication unit; 6. adverse reaction warning unit; 7. prognosis and outcome analysis unit. DETAILED DESCRIPTION

[0031] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] The following is further described in detail through specific implementation methods:

[0033] Example:

[0034] As attached Figure 1 Shown: An AI-based central infection medication decision support system, including:

[0035] Receiving unit 1: used to accept historical user's medication prescriptions, follow-up results and a corresponding set of historical patient characteristics input by the user, wherein the follow-up results include the patient's biochemical test results after medication, and the corresponding set of historical patient characteristics include age, weight, blood drug concentration, adverse reactions and treatment effectiveness; it is also used to receive the medication prescriptions, follow-up results and the first set of historical patient characteristics corresponding to several nodes in the entire treatment process of the historical patient input by the user; the receiving unit 1 is used to judge the effectiveness of the medication according to the follow-up results. In the follow-up results, if the biochemical test results are not up to standard, the medication prescription is judged as invalid. At this time, the set of historical patient characteristics corresponding to the medication prescription is screened out to ensure the effectiveness of the system output structure; the receiving unit 1 is also used to continuously receive the medication plans, follow-up results and a corresponding set of new patient characteristics corresponding to each node in the treatment process of new patients.

[0036] Construction unit 2: used to classify historical patients according to historical patient characteristics, such as classifying them according to age, weight, blood drug concentration, adverse reactions and treatment effectiveness in turn, and obtaining the corresponding medication prescriptions for each classification to construct a real-world database. It is also used to construct the relationship between medication selection, medication dosage, related medication and medication methods in medication prescriptions and the corresponding historical patient characteristics, and establish a medication relationship matrix. It is also used to construct the relationship between follow-up results and medication selection, medication dosage, related medication and medication methods in medication prescriptions, and establish a prognosis relationship matrix; it is also used to update the real-world database according to the medication prescriptions, follow-up results and the first group of historical patient characteristics corresponding to the node at several nodes in the entire treatment cycle of the historical patient; it is also used to update the medication plan, follow-up results and a corresponding set of new patient characteristics at each node in the diagnosis and treatment process of the new patient after the treatment process of the new patient is completed to the real-world database, thereby enriching the real-world database and improving the accuracy of the output results.

[0037] For example: historical patient A, his set of historical patient characteristics are age 1, weight 2, blood drug concentration 3, adverse reactions 4 and therapeutic effectiveness 5. The medication prescription used by patient A is A. When entering data later, age 1, weight 2, blood drug concentration 3, adverse reactions 4 and therapeutic effectiveness 5 are classified as the same category as patient A, and the medication prescriptions of this category are combined with prescription A to obtain the medication prescriptions of this category of patients, and a real-world database is constructed based on historical patient characteristics and corresponding medication prescriptions; prescription A includes medication selection a, medication dosage b, related medication c and medication method d, and the associations between a, b, c and d and 1, 2, 3, 4 and 5 are constructed respectively, thereby constructing a medication relationship matrix, and at the same time, the associations between patient A's biochemical test results and 1, 2, 3, 4 and 5 are constructed, thereby constructing a prognosis relationship matrix.

[0038] Auxiliary medication unit 5: is used to accept a set of new patient characteristics of a new patient input by the user, and compare the set of new patient characteristics with the historical patient physical signs in the real-world database. When a set of historical patient characteristics identical to the new patient characteristics appears, the medication prescription corresponding to the set of historical patient characteristics is output; when a set of historical patient characteristics identical to the new patient characteristics does not exist in the real-world database, the new patient characteristics in the set of new patient characteristics are mined one by one according to the medication relationship matrix to obtain the medication selection, medication dosage, related medication and medication method corresponding to each new patient characteristic, and the above contents are combined to obtain a theoretical prescription including medication selection, medication dosage, related medication and medication method.

[0039] For example: when the user inputs a new patient B, whose new patient characteristics are age 1.1, weight 2.1, blood drug concentration 3.1, adverse reactions 4.1, and treatment efficiency 5.1, if there is a set of historical patient characteristics of 1.1, 2.1, 3.1, 4.1, and 5.1 in the real-world database, the medication prescription corresponding to the set of historical patient characteristics is output; if the set of historical patient characteristics does not exist, the medication relationship matrix is ​​used to mine medication selection, medication measurement, related medication, and medication methods according to 1.1, 2.1, 3.1, 4.1, and 5.1 in turn, and a theoretical prescription of medication selection, medication measurement, related medication, and medication methods is formed, and the theoretical prescription is output to the user.

[0040] The prognosis and outcome analysis unit 7 is used to accept the medication plan input by the user, including medication selection, medication dosage, related medication and medication method, and use the prognosis relationship matrix to output the medication plan in sequence, including the follow-up results corresponding to the medication selection, medication dosage, related medication and medication method, and combine the above contents to form a predicted follow-up result.

[0041] The adverse reaction warning unit 6 is used to obtain drug-induced adverse reaction information based on a set of historical patient characteristics input by the user, and when the medication regimen of a new user input by the user and a set of new user characteristics exist in the real-world database, the adverse reaction information corresponding to the medication regimen and historical patient characteristics is output.

[0042] For example: Based on the historical patient characteristics 1, 2, 3, 4 and 5 of patient A input by the user, and obtaining the drug-induced adverse reaction information 4, when the user inputs the information of a new patient C, if a set of new patient characteristics of C is 1, 2, 3, 4 and 5, and the medication regimen is A, then C and A can correspond, and the adverse reaction information 4 of A is output.

[0043] The prediction unit 3 is used to select each group of historical patient characteristics into two parts, select any one part, use the prediction model to predict the other part for multiple times, thereby forming multiple groups of prediction features, and predict the prediction prescription corresponding to each group of predictions, and build a prediction database based on these prediction features and prediction prescriptions. When the new patient characteristics cannot correspond to the historical patient characteristics and medication prescriptions of historical patients in the real-world database, the new patient characteristics and medication regimen of the new patient are put into the prediction database for comparison, and a group of prediction features with the greatest similarity to the new patient characteristics in the prediction database are selected, and the prediction prescription corresponding to the group of prediction features is output for user reference. This scheme enriches the database through prediction based on the existing data, reduces the problem of lack of output results and inability to output effective results due to missing data in the early stage of the scheme application, and compared with the scheme using the closest data in the real-world database, this scheme can avoid the inaccuracy caused by the large difference between the reference data and the input data, expand the prediction range of the system, and improve the applicability of the system. During use, according to the number of times the output data is selected, the data that is selected more times is more consistent with the actual situation, then this part of the data is selected to train the prediction model to improve the accuracy of the prediction model; the prediction unit 3 can also be used to predict a set of historical patient characteristics of each node according to several medication prescriptions, follow-up results and a corresponding set of historical patient characteristics in the historical patient treatment process, obtain multiple sets of prediction features of the node, and obtain the prediction prescription corresponding to each set of prediction features, update the prediction database, and modify the original prediction features and prediction data in the prediction database according to the changing trends of historical patient characteristics and medication prescriptions, so as to improve the accuracy of the data in the prediction database and improve the reference value of the output plan.

[0044] For example: a group of historical patient features of historical patient A is 1, 2, 3, 4 and 5. The five historical patient features are divided into two parts, one is 1 and 2, and the other is 3, 4 and 5. Multiple predictions are made based on 1 and 2 to obtain three groups of prediction features, namely (1, 2, 3.2, 4.2, 5.2), (1, 2, 3.5, 4.5, 5.5), and (1, 2, 3.7, 4.7, 5.7). The corresponding prediction prescriptions of these three groups of prediction features are B, C and D respectively. When the new patient D is input with the features (1, 2, 3.4, 4.4, 5.4) at this time, this group of features does not exist in the real-world database. Then this group of features is transferred to the prediction database for comparison. Among them, (1, 2, 3.2, 4.2 , 5.2) and (1, 2, 3.5, 4.5, 5.5) are similar to the characteristics of patient D, and (1, 2, 3.5, 4.5, 5.5) has the highest similarity. The predicted prescription C corresponding to (1, 2, 3.5, 4.5, 5.5) is selected as the reference for the patient's medication. During the entire diagnosis and treatment process of historical patient A, the historical patient characteristics of two nodes are (1, 2, 3.3, 4.1, 5.4) and (1, 2, 3.3, 4.1, 5.5), and in the corresponding medication prescriptions of the two nodes, the medication dosage of the previous node is greater than that of the latter node. At this time, the medication dosage is negatively correlated with the effectiveness, and the data in the prediction database that is positively correlated with the medication dosage is adjusted.

[0045] The monitoring unit 4 includes a microphone, which is used to collect voice information of the new patient during the treatment of the new patient. When the new patient's voice contains sentences expressing pain or other discomfort, such as words such as "sore throat", "stomach ache" and "palpitation", or the audio amplitude of the voice information is greater than a set value, the new patient may be screaming. At this time, the monitoring unit 4 outputs an alarm message to the user, and outputs an alarm message such as "Patient ×× is suspected of having an adverse reaction" to the user, prompting the user to check to avoid further deterioration of the new patient's condition. At the same time, it is also convenient for the user to promptly judge the effectiveness of the medication regimen corresponding to the new patient, thereby avoiding invalid information from entering the system, interfering with the structural reliability of the system output, and increasing the system's computing power. It is also used to receive the allergy information of new patients input by users, that is, drug allergy information. When the associated medication part of the theoretical prescription or predicted prescription requires the combined use of linezolid and cefoperazone sodium and sulbactam sodium, and the new patient is allergic to cefoperazone sodium and sulbactam sodium, the theoretical prescription or predicted prescription is adjusted at this time, such as replacing cefoperazone sodium and sulbactam sodium with other drugs with similar efficacy, etc., to avoid the appearance of drugs that the new patient is allergic to in the final output theoretical prescription or predicted prescription, thereby avoiding the user's use of the prescription causing physical damage to the patient and medical disputes. It is also used to determine whether the new patient has drug-induced adverse reactions based on the follow-up results of the new patient. When the new patient has adverse reactions, the real-world database is queried based on the adverse reactions. When the same adverse reaction information appears in the real-world database, the medication prescription of the historical patient corresponding to the adverse reaction information is output. After outputting the medication prescription of the historical patient corresponding to the adverse reaction information, all medication prescriptions of the historical patient are marked.

[0046] For example, when the follow-up result of the new patient D indicates that he has an adverse reaction F, F is put into the real-world database for query, and it is found that both patients E and patients F have adverse reactions F. At this time, the medication prescriptions used by patients E and patients F are entered. The user can compare the medication prescriptions of patients E and patients F with the new patient D to determine the drugs that may cause the adverse reaction and adjust the prescription. Subsequently, all the data entered into the real-world database by patients E and patients F are marked. When the information of patients E and patients F is subsequently output, the user is prompted that the prescriptions corresponding to patients E and patients B may cause adverse reactions, and the user needs to identify them, thereby further reducing the probability of medical accidents due to medication in the future.

[0047] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. An AI-based central nervous system infection medication decision support system, characterized in that: include: Receiving unit (1): The receiving unit is used to accept a user input of a historical patient's medication prescription, follow-up examination results, and a corresponding set of historical patient characteristics; Construction unit (2): Construction unit (2) is used to classify historical patients according to historical patient characteristics, and obtain the medication prescription corresponding to each classification, and construct a real-world database. Construction unit (2) is also used to construct the relationship between medication selection, medication dosage, related medication and medication method in the medication prescription and the corresponding historical patient characteristics, and establish a medication relationship matrix. It is also used to construct the relationship between the follow-up results and the medication selection, medication dosage, related medication and medication method in the medication prescription, and construct a prognosis relationship matrix. Auxiliary medication unit (5): The auxiliary medication unit (5) is used to receive a group of new patient characteristics input by the user, and determine the category to which the new patient belongs based on the new patient characteristics, and input a medication prescription corresponding to the category to the user. When the new patient does not belong to any category, the historical patient characteristics corresponding to the group of new patient characteristics are selected one by one, and the corresponding medication selection, medication dosage, related medication and medication method are output according to the medication relationship matrix, and a theoretical prescription consisting of medication selection, medication dosage, related medication and medication method is output to the user; Prognosis and outcome analysis unit (7): the prognosis and outcome analysis unit (7) is used to receive the medication regimen input by the user, and output the predicted follow-up result using the prognosis relationship matrix according to the medication selection, medication dosage, associated medication and medication method contained in the medication regimen; Adverse reaction warning unit (6): The adverse reaction warning unit (6) is used to screen historical patient characteristics, obtain drug-induced adverse reaction information of historical patients, and output drug-induced adverse reaction information corresponding to the historical patient when the new patient characteristics and corresponding medication regimen input by the user are the same as the historical patient characteristics and medication prescription of the historical patient.

2. The AI-based central nervous system infection medication decision support system according to claim 1, characterized in that: The system also includes a prediction unit (3), which is used to select each group of historical patient characteristics and divide them into two parts, predict the other part of historical patient characteristics several times according to any part of the historical patient characteristics, thereby forming several groups of prediction characteristics, and predict the prediction prescription corresponding to each group of prediction characteristics to form a prediction database. The auxiliary medication unit (5) is also used to compare the input new patient characteristics with the prediction characteristics in the prediction database after the query of the real-world database fails, select a group of prediction characteristics in the prediction database with the greatest similarity to the new patient characteristics, and output the prediction prescription corresponding to the group of prediction characteristics.

3. The AI-based central nervous system infection medication decision support system according to claim 2 is characterized in that: The receiving unit (1) is also used to receive a number of medication prescriptions, follow-up results and a corresponding set of historical patient characteristics during the historical patient treatment process input by the user, and the construction unit (2) is used to update the real-world database based on the medication prescriptions, follow-up results and a corresponding set of historical patient characteristics.

4. The AI-based central nervous system infection medication decision support system according to claim 3, characterized in that: The prediction unit (3) is also used to predict the prediction features and prediction prescriptions based on a number of medication prescriptions, follow-up results and a corresponding set of historical patient characteristics in the historical patient treatment process input by the user received by the receiving unit (1), and to update the prediction database after each prediction feature and prediction prescription are obtained, and to modify the original prediction features and prediction data in the prediction database using the obtained historical patient characteristics and medication prescriptions.

5. The AI-based central nervous system infection medication decision support system according to claim 4, characterized in that: The receiving unit (1) is also used to judge the effectiveness of the medication prescription according to the follow-up consultation result, and after the medication prescription is judged to be invalid, screen out the corresponding medication prescription, follow-up consultation result and a corresponding set of historical patient characteristics.

6. The AI-based central nervous system infection medication decision support system according to claim 5, characterized in that: The receiving unit (1) is also used to continuously receive medication regimens, follow-up results, and a corresponding set of new patient characteristics during the treatment of new patients. The construction component is also used to update the real-world database based on the medication regimens, follow-up results, and a corresponding set of new patient characteristics received each time after the new patient treatment process is completed.

7. The AI-based central nervous system infection medication decision support system according to claim 6, characterized in that: It also includes a monitoring unit (4), which is used to collect voice information of new patients, judge the status of new patients based on the voice information, and output alarm information to the user when the new patient has adverse reactions corresponding to the medication prescription.

8. The AI-based central nervous system infection medication decision support system according to claim 7, characterized in that: The monitoring unit (4) is also used to receive allergy information of a new patient input by a user, and to adjust the theoretical prescription or the predicted prescription when the auxiliary medication unit (5) outputs a theoretical prescription or the prediction unit (3) outputs a predicted prescription.

9. The AI-based central nervous system infection medication decision support system according to claim 8, characterized in that: The monitoring unit (4) is also used to judge the drug-induced adverse reaction information of the new patient based on the new patient's follow-up results, and to query the new patient's adverse reaction information in the real-world database. When the same adverse reaction information appears in the real-world database, the medication prescription of the historical patient corresponding to the adverse reaction information is output.

10. The AI-based central nervous system infection medication decision support system according to claim 9, characterized in that: The monitoring unit (4) is also used to mark all medication prescriptions of the historical patient after outputting the medication prescription of the historical patient corresponding to the adverse reaction information.

Images