Anesthesia dispensing robot

Abstract

The present invention provides an anesthesia dispensing robot, which relates to the field of robot technology and includes: a drug storage area, a drug taking component, a drug review component, a drug preparation component, a display component, an input component, and a processor arranged in the robot's main structure; the drug storage area includes multiple storage partitions, each storage partition is used to store a type of anesthetic drug; the drug taking component is used to take anesthetic drugs from the drug storage area according to the drug taking instruction of the processor; the drug review component is used to weigh the amount of anesthetic drugs taken; the drug preparation component is used to mix the taken anesthetic drugs and produce anesthetic medical devices when the amount of anesthetic drugs taken is consistent with the drug taking instruction, and the processor is used to generate the drug taking instruction. According to the present invention, based on the doctor's modification and confirmation of the recommended anesthetic drug dispensing information, the recommended anesthetic drug dispensing information can be optimized to improve the accuracy of the recommendation.

Description

Technical Field

[0001] The present invention relates to the technical field of robots, and in particular to an anesthesia dispensing robot. Background Art

[0002] In related technology, CN119427393A discloses an automatic medication dispensing robot and method. The robot includes: a main controller, a human-computer interaction system, a drug detection system, and a drug processing system. The human-computer interaction system is used to input and obtain user information and medication information, and confirms the user information and medication information before dispensing the medication, and feeds back the confirmed information to the main controller; the drug detection system is used to detect the status of the medication and feed back the medication status information to the main controller; the main controller is used to generate instructions based on the feedback information from the human-computer interaction system and the drug detection system; and the drug processing system is used to dispense the corresponding medication according to the instructions of the main controller and release the corresponding medication after the status is detected. This solution improves the efficiency and accuracy of medication management, enables continuous medication dispensing, and monitors the quality of medication in real time to ensure that the medication meets the requirements for administration. In addition, one robot can be used by multiple people and has the characteristics of a compact structure and high medication dispensing efficiency.

[0003] CN118386225A discloses an intelligent dispensing robot based on visual recognition technology, wherein the intelligent dispensing robot includes: a medicine box attribute information acquisition subsystem for acquiring the medicine box attribute information of a first target medicine box; an automatic medicine loading subsystem for automatically replenishing the first target medicine box based on the medicine box attribute information based on the visual recognition technology; a medicine dispensing demand acquisition subsystem for acquiring the medicine dispensing demand of a target patient; a medicine dispensing plan determination subsystem for determining the medicine dispensing plan based on the medicine dispensing demand and the medicine box attribute information; and an intelligent medicine dispensing subsystem for intelligently dispensing medicine based on the medicine dispensing plan. The intelligent dispensing robot based on visual recognition technology of this solution introduces visual recognition technology and automatically fills the first target medicine box based on the medicine box attribute information. In addition, the medicine dispensing plan is determined based on the medicine dispensing demand and the medicine box attribute information, and the medicine is dispensed based on the medicine dispensing plan. There is no need for manual participation in the medicine dispensing process, which avoids the situation of medicine dispensing errors and is more intelligent.

[0004] Therefore, in the related art, although robots can be used for automatic prescription, they are unable to recommend prescription information to doctors, and are even more unable to further optimize the recommended prescription information based on the doctor's modification and confirmation of the recommended prescription information. Summary of the Invention

[0005] The present invention provides an anesthesia dispensing robot that can solve the technical problem that related technologies are unable to recommend dispensing information to doctors, and are even more unable to further optimize the recommended dispensing information based on the doctor's modification and confirmation of the recommended dispensing information.

[0006] According to a first aspect of the present invention, there is provided an anesthesia dispensing robot, comprising:

[0007] Robot main structure and moving components;

[0008] The mobile component is used to carry and move the robot main structure, and the robot main structure includes a robot shell, and the robot shell is provided with a drug storage area, a drug taking component, a drug review component, a drug dispensing component, a display component, an input component and a processor;

[0009] The drug storage area includes a plurality of storage partitions, each storage partition is used to store a type of anesthetic drugs;

[0010] The drug taking component is used to take at least one type of anesthetic drug from the drug storage area according to the drug taking instruction of the processor;

[0011] The drug review component is used to weigh the amount of anesthetic drugs taken to determine whether the amount taken is consistent with the drug taking instruction;

[0012] The drug preparation component is used to mix the taken anesthetic drugs and prepare anesthetic drug medical devices when the amount of anesthetic drugs taken is consistent with the drug taking instruction;

[0013] The processor is configured to:

[0014] Generate anesthetic drug dispensing information based on patient information and doctor information collected by the input component;

[0015] displaying the anesthetic drug dispensing information via a display component;

[0016] After receiving the doctor's confirmation information on the anesthetic drug dispensing information through the input component, the drug taking instruction is generated.

[0017] By adopting the above technical solution, the present invention can achieve the following technical effects:

[0018] According to the present invention, recommended anesthetic drug dispensing information can be generated based on patient information and doctor information, and a drug dispensing instruction can be generated after the doctor confirms it. Therefore, based on the doctor's modification and confirmation of the recommended anesthetic drug dispensing information, the recommended anesthetic drug dispensing information can be further optimized to improve the accuracy of the recommendation. When dispensing based on the drug dispensing instruction, the drug review component can be used to review the amount of anesthetic drugs to improve the accuracy and safety of the dispensing. When determining the correlation coefficient, the similarity of medication between each doctor and the current doctor when treating similar patients can be calculated, thereby determining the correlation coefficient between the sample doctor information and the doctor information, which is used to accurately describe the similarity of medication habits between each doctor and the current doctor, as well as the reference value of each doctor's medication data for predicting the current doctor's medication habits. When training a narcotic drug dosage prediction model, a conditional function can be used to determine the dosage error when the narcotic drug dosage prediction model makes a correct or incorrect prediction. Based on the similarity between the physical condition of the sample patient and the current patient, and the correlation coefficient between the sample doctor information and the current doctor's information, the reference value of historical data for predicting the current doctor's medication habits is determined, thereby obtaining a loss function for the narcotic drug dosage prediction model. This allows for targeted training on historical data with high reference value, improving training efficiency and the accuracy of the narcotic drug dosage prediction model. When determining the judgment conditions, the consistency between the narcotic drug dosage and the drug administration instructions can be judged from three aspects: the standard deviation of the error between the actual weight of the narcotic drug and the used weight, the maximum error, and the number and type of narcotic drugs that were inaccurately taken. This improves the accuracy and comprehensiveness of the judgment. When determining the supplementary extraction flow rate, the relationship between the flow rate extracted by the pump extraction component and the actual weight of the anesthetic drug can be determined in the form of a linear function, and the supplementary extraction flow rate can be determined by a conditional function. In the case where the weight error of the anesthetic drug is large, the flow rate corresponding to the weight error of the anesthetic drug is determined by the flow weighing function, and the supplementary extraction flow rate is obtained to compensate for the weight error of the anesthetic drug and improve the extraction accuracy of the anesthetic drug. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic diagram of an anesthesia dispensing robot according to an embodiment of the present invention is exemplarily shown. DETAILED DESCRIPTION

[0020] Figure 1 A schematic diagram of an anesthesia dispensing robot according to an embodiment of the present invention is exemplarily shown, wherein the anesthesia dispensing robot includes:

[0021] Robot main structure and moving components;

[0022] The mobile component is used to carry and move the robot main structure, and the robot main structure includes a robot shell, and the robot shell is provided with a drug storage area, a drug taking component, a drug review component, a drug dispensing component, a display component, an input component and a processor;

[0023] The drug storage area includes a plurality of storage partitions, each storage partition is used to store a type of anesthetic drugs;

[0024] The drug taking component is used to take at least one type of anesthetic drug from the drug storage area according to the drug taking instruction of the processor;

[0025] The drug review component is used to weigh the amount of anesthetic drugs taken to determine whether the amount taken is consistent with the drug taking instruction;

[0026] The drug preparation component is used to mix the taken anesthetic drugs and prepare anesthetic drug medical devices when the amount of anesthetic drugs taken is consistent with the drug taking instruction;

[0027] The processor is configured to:

[0028] Generate anesthetic drug dispensing information based on patient information and doctor information collected by the input component;

[0029] displaying the anesthetic drug dispensing information via a display component;

[0030] After receiving the doctor's confirmation information on the anesthetic drug dispensing information through the input component, the drug taking instruction is generated.

[0031] According to an embodiment of the present invention, the anesthesia dispensing robot can generate recommended anesthetic drug dispensing information based on patient information and doctor information, and generate drug collection instructions after the doctor's confirmation. Therefore, based on the doctor's modification and confirmation of the recommended anesthetic drug dispensing information, the recommended anesthetic drug dispensing information can be further optimized to improve the accuracy of the recommendation. When dispensing based on the drug collection instructions, the drug review component can be used to review the amount of anesthetic drugs taken, thereby improving the accuracy and safety of the dispensing.

[0032] According to one embodiment of the present invention, the mobile component can be a wheeled mobile component or a tracked mobile component, the mobile component can carry the robot main structure, the robot main structure can be covered by the robot shell, and a drug storage area, a drug retrieval component, a drug review component, a drug preparation component, a display component, an input component and a processor can be set in the robot shell.

[0033] According to one embodiment of the present invention, the partition of the drug storage area can be a medicine bottle for storing a type of anesthetic drug, the drug extraction component can be a high-precision pump that can extract liquid drugs, and can extract anesthetic drugs according to the amount described in the drug extraction instruction, and output the extracted anesthetic drugs to the drug review component. Each drug storage area can correspond to a drug review component. The drug review component can be a precision weighing component that can weigh the weight of the extracted drugs and determine whether the weight is consistent with the drug extraction instruction. The drug preparation component can mix the extracted anesthetic drugs when the amount of anesthetic drugs extracted is consistent with the drug extraction instruction, and make the mixed drugs into anesthetic medical devices, for example, syringes that have extracted anesthetic drugs, etc. The present invention does not limit the specific type of anesthetic medical devices.

[0034] According to one embodiment of the present invention, when medication is required, the doctor can input patient information and doctor information through an input component (e.g., a keyboard or a touch screen, etc.), and the processor can generate recommended anesthetic drug dispensing information based on the patient information and the doctor information. For example, anesthetic drug dispensing information can be recommended based on the doctor's medication habits and the patient's physical condition, etc.

[0035] According to one embodiment of the present invention, anesthetic drug dispensing information is generated based on the patient information and doctor information collected by the input component, including: determining the patient's physiological indicator data based on the patient information; determining the doctor's identity information based on the doctor's information; training a anesthetic drug dosage prediction model based on the doctor's identity information and the patient's physiological indicator data to obtain a trained anesthetic drug dosage prediction model; inputting the patient's physiological indicator data and the doctor's identity information into the trained anesthetic drug dosage prediction model to obtain anesthetic drug dispensing information.

[0036] According to one embodiment of the present invention, a narcotic drug dosage prediction model can simulate a physician's medication habits, thereby generating recommended narcotic drug dispensing information based on the physician's medication habits and the patient's physiological indicators (e.g., height, weight, blood pressure, blood sugar, medical history, etc.). Furthermore, before each recommendation, the model can be trained based on historical data, the current physician's information, and the patient's physiological indicators. This results in a trained narcotic drug dosage prediction model that better matches the current physician's medication habits, and can then recommend narcotic drug dispensing information that better matches the current physician's medication habits.

[0037] According to one embodiment of the present invention, a narcotic drug dosage prediction model is trained based on the doctor's identity information and the patient's physiological indicator data to obtain a trained narcotic drug dosage prediction model, including: obtaining identity information corresponding to sample doctor information; obtaining a correlation coefficient between the sample doctor information and the doctor information; obtaining sample patient information corresponding to the sample doctor information; determining sample physiological indicator data of the sample patient based on the sample patient information; obtaining sample drug taking instructions for the sample patient based on the sample doctor information; inputting the sample doctor's identity information and the sample physiological indicator data of the sample patient into the narcotic drug dosage prediction model to obtain sample narcotic drug dispensing information; obtaining a loss function of the narcotic drug dosage prediction model based on the sample narcotic drug dispensing information, the sample drug taking instructions, the correlation coefficient, the sample physiological indicator data and the physiological indicator data; training the narcotic drug dosage prediction model based on the loss function of the narcotic drug dosage prediction model to obtain a trained narcotic drug dosage prediction model.

[0038] According to one embodiment of the present invention, a historical data database contains sample doctor information for multiple doctors, including the current doctor and other doctors. Identity information corresponding to each sample doctor information can be obtained. Furthermore, the correlation coefficient between each doctor and the current doctor can be obtained. For example, if a doctor and the current doctor have a teacher-student relationship and their medication habits are similar, using this doctor's historical data can be highly valuable in predicting the current doctor's medication habits.

[0039] According to one embodiment of the present invention, obtaining the correlation coefficient between the sample doctor information and the doctor information includes: determining the correlation coefficient between the i-th sample doctor information and the doctor information according to formula (1): , (1)

[0040] in, is the sample physiological index vector composed of the sample physiological index data of the jth sample patient corresponding to the i-th sample doctor information, for The transposed vector of is the historical physiological indicator vector composed of the physiological indicator data of the sth historical patient information corresponding to the doctor information, Based on The dosage vector consisting of the dosage data of multiple anesthetic drugs obtained from the sample drug taking instructions of the sample patient with the maximum value, for The transposed vector of Based on A dosage vector consisting of dosage data of multiple anesthetic drugs obtained from the drug taking instructions of historical patients with the maximum value, is the number of historical patient information corresponding to the doctor information, is the number of sample patients corresponding to the i-th sample doctor information, s≤ ,j≤ , and s, , j and All are positive integers.

[0041] According to one embodiment of the present invention, is the cosine similarity between the physiological indicator vector of the jth sample patient corresponding to the i-th sample doctor information and the historical physiological indicator vector of the s-th historical patient information corresponding to the current doctor's doctor information, It represents the maximum similarity between the historical physiological index vector of the patients treated by the current doctor and the physiological index vector of the patients treated by the i-th doctor, that is, the similarity between the most similar patients treated by the two doctors. This represents the cosine similarity of the dosage vectors of anesthetic drugs used by the current doctor and the i-th doctor when treating the most similar patients, that is, the similarity of medication use between the two when treating the most similar patients. By multiplying the above two items, the similarity of the medication habits of the i-th doctor and the current doctor when treating the most similar patients can be determined. When this similarity is high, it means that the medication habits of the i-th doctor and the current doctor are similar, and the correlation between the i-th doctor and the current doctor is high, which may be a teacher-student relationship, etc. The medication data of the i-th doctor has a greater reference value for predicting the medication habits of the current doctor. Therefore, this similarity can be determined as the correlation coefficient between the i-th sample doctor information and the doctor information. Similarly, the correlation coefficient between each sample doctor information and the doctor information can be calculated.

[0042] In this way, the similarity of medication usage between each doctor and the current doctor when treating similar patients can be calculated, thereby determining the correlation coefficient between the sample doctor information and the doctor information, which is used to accurately describe the similarity of medication habits between each doctor and the current doctor, as well as the reference value of each doctor's medication data for predicting the current doctor's medication habits.

[0043] According to one embodiment of the present invention, sample doctor information for sample patients' sample medication instructions, i.e., medication dosage information after modification and confirmation by the doctor, can be obtained. Furthermore, the sample doctor's identity information and the sample patient's sample physiological indicator data can be input into a narcotic drug dosage prediction model to obtain sample narcotic drug dispensing information. The sample medication instructions are then used as labeled information, and the error between the sample narcotic drug dispensing information and the labeled information is determined, thereby determining the loss function of the narcotic drug dosage prediction model.

[0044] According to one embodiment of the present invention, the loss function of the anesthetic drug dosage prediction model is obtained based on the sample anesthetic drug dispensing information, the sample drug taking instruction, the correlation coefficient, the sample physiological indicator data and the physiological indicator data, including: obtaining the loss function of the anesthetic drug dosage prediction model according to formula (2): , (2)

[0045] in, is the correlation coefficient between the i-th sample doctor information and the doctor information, is the sample physiological index vector composed of the sample physiological index data of the jth sample patient corresponding to the i-th sample doctor information, for The transposed vector of , P is the physiological index vector composed of physiological index data, is the dosage data of the kth anesthetic drug obtained based on the i-th sample doctor's information and the sample drug taking instruction of the j-th sample patient, is the dosage data of the kth anesthetic drug obtained based on the sample anesthetic drug dispensing information of the jth sample patient based on the i-th sample doctor information, and are all preset coefficients greater than 1, and is the logical operator of "and", if is the conditional function, M is the number of sample doctor information, is the number of sample patients corresponding to the i-th sample doctor information, n is the number of types of anesthetic drugs, k≤n, j≤ , i≤M, and k, n, j, , i and M are all positive integers.

[0046] According to one embodiment of the present invention, in formula (2), the conditional function Indicates In the case of , otherwise 0. Indicates that the dosage data of the anesthetic drug in the sample drug withdrawal instruction is 0, while the dosage data of the anesthetic drug in the sample anesthetic drug dispensing information is not 0, that is, the actual dosage of the anesthetic drug used by the doctor is not 0, while the dosage of the anesthetic drug predicted by the anesthetic drug dosage prediction model is 0. It can be considered that the anesthetic drug dosage prediction model has made an error in its prediction. Therefore, the conditional function value can obtain a preset coefficient greater than 1. , otherwise 0. Conditional function Indicates In the case of , otherwise 0. Indicates that the dosage data of the anesthetic drug in the sample drug withdrawal instruction is not 0, while the dosage data of the anesthetic drug in the sample anesthetic drug dispensing information is 0. That is, the actual dosage of the anesthetic drug used by the doctor is 0, while the dosage of the anesthetic drug predicted by the anesthetic drug dosage prediction model is not 0. It can be considered that the anesthetic drug dosage prediction model has made an error in its prediction. Therefore, the conditional function value can obtain a preset coefficient greater than 1. , otherwise 0. Indicates In the case of , otherwise 0. It means that the dosage data of anesthetic drugs in the sample drug taking instruction is not 0, while the dosage data of anesthetic drugs in the sample anesthetic drug dispensing information is not 0, that is, the actual dosage of the anesthetic drug used by the doctor is not 0, while the dosage of the anesthetic drug predicted by the anesthetic drug dosage prediction model is not 0. It can be considered that the prediction of the anesthetic drug dosage prediction model is not wrong, but there may be a certain error. Therefore, the relative error between the two can be calculated. As the conditional function value. The above three conditional functions are added together to express the preset coefficient greater than 1 when the anesthetic drug dosage prediction model makes an error in the prediction. or , when the anesthetic drug dosage prediction model predicts no errors, the sum of the relative errors is less than 1 During training, the loss function value can be increased in the event of a prediction error, increasing the penalty and training intensity, improving training efficiency. Furthermore, when the prediction is correct, the relative error can be reduced, improving the prediction accuracy of the dosage of various anesthetic drugs. The sum of the three conditional function values ​​for each anesthetic drug can be calculated to obtain the total error in the dosage of each anesthetic drug for the jth sample patient based on the i-th sample doctor's information.

[0047] According to one embodiment of the present invention, It is the cosine similarity between the sample physiological indicator vector of the jth sample patient corresponding to the i-th sample doctor information and the physiological indicator item vector of the current patient, that is, the similarity between the physical condition of the jth sample patient corresponding to the i-th sample doctor information and the current patient. The higher the similarity, the greater the reference value of the dosage data of various anesthetic drugs of the j-th sample patient corresponding to the i-th sample doctor information for predicting the dosage data of various anesthetic drugs of the current patient. Therefore, the cosine similarity can be used as the weight of the total error of the dosage of various anesthetic drugs of the j-th sample patient for the i-th sample doctor information, so as to perform weighted summation to obtain the medication prediction error for the i-th doctor.

[0048] According to one embodiment of the present invention, is the correlation coefficient between the ith doctor and the current doctor. The higher the correlation coefficient, the higher the reference value of the ith doctor's medication habits for predicting the current doctor's medication habits. Therefore, The weight of the medication prediction error for the i-th doctor is used to perform a weighted summation to obtain the prediction error for each doctor, which can be used as the loss function for the anesthetic dosage prediction model. Furthermore, backpropagation can be performed based on the loss function to adjust the parameters of the anesthetic dosage prediction model. After multiple training cycles, a trained anesthetic dosage prediction model is obtained.

[0049] In this way, the dosage error when the anesthetic drug dosage prediction model predicts correctly or incorrectly can be determined through the conditional function, and the reference value of historical data for predicting the current doctor's medication habits can be determined based on the similarity between the physical condition of the sample patient and the current patient, and the correlation coefficient between the sample doctor information and the doctor information of the current doctor, thereby obtaining the loss function of the anesthetic drug dosage prediction model, which can be used to train historical data with high reference value in a targeted manner, thereby improving the training efficiency and the accuracy of the anesthetic drug dosage prediction model.

[0050] According to one embodiment of the present invention, after obtaining a trained anesthetic drug dosage prediction model, the current physician's information and the current patient's physiological indicator data can be input into the trained anesthetic drug dosage prediction model to obtain anesthetic drug dispensing information, i.e., the dosages of multiple anesthetic drugs recommended by the trained anesthetic drug dosage prediction model. The anesthetic drug dispensing information can be displayed on a display component. After viewing the anesthetic drug dispensing information, the physician can modify and confirm it through an input component. Upon receiving the physician's confirmation of the anesthetic drug dispensing information (or the modified anesthetic drug dispensing information), a drug dispensing instruction is generated, indicating the dosages of multiple anesthetic drugs deemed accurate by the physician. The drug dispensing component can then dispense at least one anesthetic drug according to the drug dispensing instruction.

[0051] According to one embodiment of the present invention, after anesthetic drugs are taken, the amount of each anesthetic drug taken can be reviewed by a drug review component. The drug review component is configured to: weigh the weights of multiple anesthetic drugs and send the weights to a processor; the processor is configured to: determine the weights of the multiple anesthetic drugs used based on the drug taking instruction; set multiple judgment conditions based on the weights used and the weights of the multiple anesthetic drugs; and determine that the amount of the anesthetic drugs taken is inconsistent with the drug taking instruction if at least one of the multiple judgment conditions is met.

[0052] According to one embodiment of the present invention, a plurality of judgment conditions are set according to the usage weight and the weight of the plurality of anesthetic drugs, including: obtaining judgment conditions C1, C2 and C3 according to formula (3), (3)

[0053] in, for The standard deviation of is the weight of the kth anesthetic drug used, is the weight of the kth anesthetic drug, max is the maximum value function, if is the conditional function, 、 、 、 is the preset coefficient, n is the number of types of narcotic drugs, k≤n, and both k and n are positive integers.

[0054] According to one embodiment of the present invention, in formula (3), the judgment condition C1 indicates that the standard deviation of the relative deviation between the used weight and the actual weight of the anesthetic drug is large, that is, greater than or equal to the preset coefficient In this case, the errors in the weights of various anesthetic drugs are large and the error levels vary, making it difficult to obtain accurate dosages of anesthetic drugs. Judgment condition C2 indicates that the maximum value of the relative deviation between the used weight and the actual weight is large, that is, greater than or equal to the preset coefficient. , indicating that there is a large error between the actual weight and the used weight of at least one anesthetic drug, and the amount of anesthetic drug taken is inaccurate. In the judgment condition C3, the conditional function Indicates , the conditional function value is 1, otherwise it is 0, so The relative deviation between the used weight and the actual weight is large (i.e., greater than or equal to the preset coefficient ) types and number of narcotic drugs, The percentage of anesthetic drugs with a large relative deviation between the used weight and the actual weight is greater than or equal to the preset coefficient. If at least one of the above three judgment conditions is met, it is determined that the amount of anesthetic drugs taken is inconsistent with the drug taking instruction and the amount of anesthetic drugs taken needs to be adjusted.

[0055] In this way, the standard deviation of the error between the actual weight of narcotic drugs and the used weight, the maximum error, and the number of types and quantities of inaccurately taken narcotic drugs can be used to judge whether the amount of narcotic drugs taken is consistent with the drug taking instructions, thereby improving the accuracy and comprehensiveness of the judgment.

[0056] According to one embodiment of the present invention, the drug dispensing component includes a pump-type extraction component, and the drug verification component includes a weighing component; the processor is further configured to: when the amount of anesthetic drugs dispensed is inconsistent with the drug dispensing instruction, obtain a flow weighing function based on the weights of the multiple anesthetic drugs obtained by the weighing component and the flow rates of the multiple anesthetic drugs extracted by the pump-type extraction component; obtain a supplementary extraction flow rate of the multiple anesthetic drugs based on the flow weighing function, the weights of the anesthetic drugs, and the drug dispensing instruction; and extract the multiple anesthetic drugs according to the supplementary extraction flow rate through the pump-type extraction component. That is, the relationship between the flow rate of the anesthetic drugs extracted by the pump-type extraction component and the actual weight of the anesthetic drugs is determined, and based on this relationship, determine how to supplement the flow rate of the anesthetic drugs so that the weight of the supplemented anesthetic drugs reaches or approaches the used weight of the anesthetic drugs.

[0057] According to one embodiment of the present invention, when the amount of anesthetic drugs taken is inconsistent with the drug taking instruction, a flow weighing function is obtained based on the weights of the multiple anesthetic drugs obtained by the weighing component and the flow rates of the multiple anesthetic drugs extracted by the pump-type extraction component, including: obtaining an undetermined coefficient equation of the flow weighing function according to formula (4), (4)

[0058] in, is the flow rate of the kth anesthetic drug, and is an undetermined coefficient; according to the flow rate and weight of multiple anesthetic drugs, the undetermined coefficient is solved to obtain the solution value of the undetermined coefficient; according to the solution value of the undetermined coefficient and the undetermined coefficient equation, the flow weighing function is obtained.

[0059] According to one embodiment of the present invention, in formula (4), to simplify calculations and reduce computing resource consumption, the undetermined coefficient equation of the flow rate weighing function can be set in the form of a linear equation to express the relationship between the flow rate and the actual weight of the anesthetic drug. The undetermined coefficient in the undetermined coefficient equation can be solved using the flow rates and actual weights of multiple anesthetic drugs to obtain the solution value of the undetermined coefficient, which can then be substituted into the undetermined coefficient equation to obtain the flow rate weighing function to express the relationship between the flow rate and the actual weight of the anesthetic drug.

[0060] According to one embodiment of the present invention, the supplementary extraction flow rate of multiple anesthetic drugs is obtained according to the flow weighing function, the weight of the anesthetic drug and the drug extraction instruction, including: obtaining the supplementary extraction flow rate of the kth anesthetic drug according to formula (5) ,

[0061] (5)

[0062] in, for The solution value of for The solution value of .

[0063] According to one embodiment of the present invention, in formula (5), the conditional function Indicates In the case of , otherwise it is 0. That is, when the relative deviation between the used weight and the actual weight of the kth anesthetic drug is greater than or equal to the preset coefficient (ie, when the error between the used weight and the actual weight is large), the supplementary extraction flow rate is If the relative deviation between the used weight and the actual weight is less than the preset coefficient, no supplementary extraction is required, that is, the supplementary extraction flow rate is 0. In order to use the error between the weight and the actual weight, the error is substituted into the flow weighing function to obtain the flow corresponding to the error. To compensate for the error, the flow corresponding to the error can be supplemented and extracted, that is, the supplementary extraction flow.

[0064] In this way, the relationship between the flow rate extracted by the pump extraction component and the actual weight of the anesthetic drug can be determined in the form of a linear function, and the supplementary extraction flow rate can be determined through a conditional function. In the case that the weight error of the anesthetic drug is large, the flow rate corresponding to the weight error of the anesthetic drug is determined through the flow weighing function, and the supplementary extraction flow rate is obtained to compensate for the weight error of the anesthetic drug and improve the extraction accuracy of the anesthetic drug.

[0065] According to one embodiment of the present invention, once the weights of multiple anesthetic drugs meet the required specifications, the drug dispensing component can be used to mix the drugs and produce anesthetic medical devices such as syringes. Furthermore, the doctor and patient information, as well as drug dispensing instructions, can be entered into a historical database. This facilitates subsequent training of the anesthetic dosage prediction model when other doctors use the anesthesia dispensing robot, thereby more accurately predicting other doctors' medication habits.

[0066] According to an embodiment of the present invention, the anesthesia dispensing robot can generate recommended anesthetic drug dispensing information based on patient information and doctor information, and generate drug collection instructions after the doctor confirms it. Therefore, based on the doctor's modification and confirmation of the recommended anesthetic drug dispensing information, the recommended anesthetic drug dispensing information can be further optimized to improve the accuracy of the recommendation. When dispensing based on the drug collection instructions, the drug review component can be used to review the amount of anesthetic drugs to improve the accuracy and safety of the dispensing. When determining the correlation coefficient, the similarity of medication between each doctor and the current doctor when treating similar patients can be calculated, thereby determining the correlation coefficient between the sample doctor information and the doctor information, so as to accurately describe the similarity of medication habits between each doctor and the current doctor, as well as the reference value of each doctor's medication data for predicting the current doctor's medication habits. When training a narcotic drug dosage prediction model, a conditional function can be used to determine the dosage error when the narcotic drug dosage prediction model makes a correct or incorrect prediction. Based on the similarity between the physical condition of the sample patient and the current patient, and the correlation coefficient between the sample doctor information and the current doctor's information, the reference value of historical data for predicting the current doctor's medication habits is determined, thereby obtaining a loss function for the narcotic drug dosage prediction model. This allows for targeted training on historical data with high reference value, improving training efficiency and the accuracy of the narcotic drug dosage prediction model. When determining the judgment conditions, the consistency between the narcotic drug dosage and the drug administration instructions can be judged from three aspects: the standard deviation of the error between the actual weight of the narcotic drug and the used weight, the maximum error, and the number and type of narcotic drugs that were inaccurately taken. This improves the accuracy and comprehensiveness of the judgment. When determining the supplementary extraction flow rate, the relationship between the flow rate extracted by the pump extraction component and the actual weight of the anesthetic drug can be determined in the form of a linear function, and the supplementary extraction flow rate can be determined by a conditional function. In the case where the weight error of the anesthetic drug is large, the flow rate corresponding to the weight error of the anesthetic drug is determined by the flow weighing function, and the supplementary extraction flow rate is obtained to compensate for the weight error of the anesthetic drug and improve the extraction accuracy of the anesthetic drug.

Claims

1. An anesthesia dispensing robot, characterized in that: include: Robot main structure and moving components; The mobile component is used to carry and move the robot main structure, and the robot main structure includes a robot shell, and the robot shell is provided with a drug storage area, a drug taking component, a drug review component, a drug dispensing component, a display component, an input component and a processor; The drug storage area includes a plurality of storage partitions, each storage partition is used to store a type of anesthetic drugs; The drug taking component is used to take at least one type of anesthetic drug from the drug storage area according to the drug taking instruction of the processor; The drug review component is used to weigh the amount of anesthetic drugs taken to determine whether the amount taken is consistent with the drug taking instruction; The drug preparation component is used to mix the taken anesthetic drugs and prepare anesthetic drug medical devices when the amount of anesthetic drugs taken is consistent with the drug taking instruction; The processor is configured to: Generate anesthetic drug dispensing information based on patient information and doctor information collected by the input component; displaying the anesthetic drug dispensing information via a display component; After receiving the doctor's confirmation information on the anesthetic drug dispensing information through the input component, generating the drug taking instruction; Generate anesthetic drug dispensing information based on the patient and doctor information collected by the input component, including: Determining the patient's physiological indicator data based on the patient information; Determining the doctor's identity information based on the doctor's information; Training a narcotic drug dosage prediction model based on the doctor's identity information and the patient's physiological indicator data to obtain a trained narcotic drug dosage prediction model; Inputting the patient's physiological indicator data and the doctor's identity information into the trained anesthetic drug dosage prediction model to obtain anesthetic drug dispensing information; The anesthetic drug dosage prediction model is trained based on the doctor's identity information and the patient's physiological indicator data to obtain a trained anesthetic drug dosage prediction model, including: Obtain the identity information corresponding to the sample doctor information; Obtaining a correlation coefficient between the sample doctor information and the doctor information; Obtain sample patient information corresponding to sample doctor information; Determine sample physiological index data of the sample patient according to the sample patient information; Obtain sample doctor information and sample medication instructions for sample patients; Inputting the identity information of the sample doctors and the sample physiological index data of the sample patients into the anesthetic drug dosage prediction model to obtain the sample anesthetic drug dispensing information; Obtaining a loss function of a narcotic drug dosage prediction model based on the sample narcotic drug dispensing information, the sample drug taking instruction, the correlation coefficient, the sample physiological indicator data, and the physiological indicator data; The anesthetic drug dosage prediction model is trained according to the loss function of the anesthetic drug dosage prediction model to obtain a trained anesthetic drug dosage prediction model.

2. The anesthesia dispensing robot according to claim 1, characterized in that: Obtaining the correlation coefficient between the sample doctor information and the doctor information includes: According to the formula Determine the correlation coefficient AS between the i-th sample doctor information and the doctor information i , where P an,i,j is the sample physiological index vector composed of the sample physiological index data of the jth sample patient corresponding to the i-th sample doctor information, (P an,i,j ) T P an,i,j The transposed vector, P his,s UV is the historical physiological indicator vector composed of the physiological indicator data of the sth historical patient information corresponding to the doctor information. in,i,max Based on The dosage vector consisting of the dosage data of multiple anesthetic drugs obtained from the sample drug taking instructions of the sample patient with the maximum value, (UV in,i,max ) T UV in,i,max The transposed vector, UV his,max Based on The dosage vector consisting of the dosage data of multiple anesthetic drugs obtained from the historical drug usage instructions of the patient with the maximum value, m his is the number of historical patient information corresponding to the doctor information, m i is the number of sample patients corresponding to the i-th sample doctor information, s≤m his , j≤m i , and s, m his , j and m i All are positive integers.

3. The anesthesia dispensing robot according to claim 1, characterized in that: Obtaining a loss function of a narcotic drug dosage prediction model based on the sample narcotic drug dispensing information, the sample drug taking instruction, the correlation coefficient, the sample physiological indicator data, and the physiological indicator data, including: According to the formula Obtain the loss function LOSS of the anesthetic drug dosage prediction model, where AS i is the correlation coefficient between the i-th sample doctor information and the doctor information, P an,i,j is the sample physiological index vector composed of the sample physiological index data of the jth sample patient corresponding to the i-th sample doctor information, (P an,i,j ) T P an,i,j The transposed vector of P is the physiological index vector composed of physiological index data, U in,i,j,k is the dosage data of the kth anesthetic drug obtained based on the i-th sample doctor information and the sample drug taking instruction of the j-th sample patient, U di,i,j,k is the kth anesthetic drug dosage data obtained based on the i-th sample doctor information and the j-th sample patient's sample anesthetic drug dispensing information. α1 and α2 are both preset coefficients greater than 1, and is the "and" logical operator, if is the conditional function, M is the number of sample doctor information, and m i is the number of sample patients corresponding to the i-th sample doctor information, n is the number of types of anesthetic drugs, k≤n, j≤m i , i≤M, and k, n, j, m i , i and M are all positive integers.

4. The anesthesia dispensing robot according to claim 1, characterized in that: The drug review component is used to: Weighs various narcotic drugs and sends them to the processor; The processor is configured to: Determining the weight of multiple narcotic drugs to be used according to the drug dispensing instructions; setting a plurality of judgment conditions according to the usage weight and the weights of the plurality of anesthetic drugs; When at least one of the plurality of judgment conditions is satisfied, it is determined that the amount of anesthetic drug taken is inconsistent with the drug taking instruction.

5. The anesthesia dispensing robot according to claim 4, characterized in that: According to the usage weight and the weight of the plurality of anesthetic drugs, multiple judgment conditions are set, including: According to the formula Obtain judgment conditions C1, C2, and C3, where: for The standard deviation of W in,k is the weight of the kth anesthetic drug, W k is the weight of the kth anesthetic drug, max is the maximum value function, if is the conditional function, σ1, σ2, σ3 and σ4 are preset coefficients, n is the number of types of anesthetic drugs, k≤n, and both k and n are positive integers.

6. The anesthesia dispensing robot according to claim 5, characterized in that: The medicine taking component includes a pump extraction component, and the medicine review component includes a weighing component; The processor is further configured to: When the amount of anesthetic drugs taken is inconsistent with the drug taking instruction, a flow weighing function is obtained based on the weights of the multiple anesthetic drugs obtained by the weighing component and the flow rates of the multiple anesthetic drugs extracted by the pump extraction component; Obtaining a supplementary extraction flow rate of a plurality of anesthetic drugs according to the flow weighing function, the weight of the anesthetic drugs, and the drug taking instruction; A plurality of anesthetic drugs are extracted by the pump-type extraction component according to the supplementary extraction flow rate.

7. The anesthesia dispensing robot according to claim 6, characterized in that: In the case where the amount of anesthetic drugs taken is inconsistent with the drug taking instruction, a flow weighing function is obtained based on the weights of the multiple anesthetic drugs obtained by the weighing component and the flow rates of the multiple anesthetic drugs extracted by the pump-type extraction component, including: According to the formula F k =θ1W k +θ2 Obtain the undetermined coefficient equation of the flow weighing function, where F k is the flow rate of the kth anesthetic drug, θ1 and θ2 are unknown coefficients; Solving the undetermined coefficients according to the flow rates and weights of the multiple anesthetic drugs to obtain solution values ​​of the undetermined coefficients; A flow weighing function is obtained according to the solved value of the undetermined coefficient and the undetermined coefficient equation.

8. The anesthesia dispensing robot according to claim 7, characterized in that: According to the flow weighing function, the weight of the anesthetic drug and the drug taking instruction, a supplementary extraction flow rate of multiple anesthetic drugs is obtained, including: According to the formula Get the supplementary extraction flow F of the kth anesthetic drug k,su , where θ 1,A is the solution value of θ1, θ 2,A is the solution value of θ2.

Images