Teaching target feedback method and system based on artificial intelligence model

Abstract

The invention discloses a teaching target feedback method and system based on an artificial intelligence model, and relates to the technical field of virtual diagnosis and treatment teaching, and the method comprises the steps: creating a virtual patient for a virtual diagnosis and treatment window in a customized manner according to a current teaching task; the AI model is used for adaptively updating illness state data for the diagnosis and treatment behaviors and the deduction of natural time; monitoring diagnosis and treatment behaviors and illness state evolution data in real time; calculating a disease deterioration score of each diagnosis and treatment node; according to the calculation result of the deterioration score, allocating an evaluation weight to each diagnosis and treatment node; creating a fine-grained ability quantification formula based on the distributed evaluation weight, and calculating a quantification result of the diagnosis and treatment ability of the student; and generating a feedback report of the current teaching task based on the quantitative result of the diagnosis and treatment ability of each student. According to the invention, the deep analysis capability of clinical thinking is enhanced, and the authenticity of teaching data is enhanced; the technical limitation that a traditional virtual diagnosis and treatment system focuses on the process, thinking, result and process is broken through.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual diagnosis and treatment teaching, and in particular to a teaching target feedback method and system based on an artificial intelligence model. Background Art

[0002] With the innovation of medical education model, virtual diagnosis and treatment teaching system has gradually become an important means of clinical skills training. However, traditional clinical teaching is limited by the uneven distribution of physical medical resources and it is difficult to meet the needs of large-scale medical talent training. Although existing virtual patient systems (such as Kaplan, i-Human, etc.) can simulate basic consultation scenarios, they generally have problems such as insufficient intelligence and single feedback dimension. Their feedback mechanism is mostly based on matching evaluation of preset rule bases, which cannot dynamically evaluate the learner's clinical thinking process and has weak analysis ability for unstructured diagnosis and treatment behaviors. Therefore, it is urgent to build an intelligent feedback system with real-time multi-dimensional analysis capabilities, realize fine-grained evaluation of diagnosis and treatment behaviors through deep learning methods, and establish an adaptive learning mechanism to improve the precision level of medical education. Summary of the invention

[0003] The present invention provides a teaching target feedback method based on an artificial intelligence model, comprising: Step 1. Customize and create a virtual patient for the virtual diagnosis and treatment window according to the current teaching task; Step 2: Use the pre-trained AI model to adaptively update the virtual patient's condition data based on the trainee's diagnosis and treatment behaviors and the natural time delay; Step 3: Monitor the diagnosis and treatment behaviors of trainees in the virtual diagnosis and treatment window in real time, as well as the disease evolution data of virtual patients, and record the monitored data, and generate a complete clinical data chain after time series; Step 4: Calculate the disease progression score of each diagnosis and treatment node based on the patient's disease evolution data in the clinical data chain; Step 5: Assign evaluation weights to each diagnosis and treatment node based on the calculation results of the disease progression score; Step 6: Create a fine-grained ability quantification formula based on the assigned evaluation weights, and use the created quantification formula to calculate the quantitative results of the trainees' diagnosis and treatment abilities; Step 7. Generate a feedback report on the current teaching task based on the quantitative results of each student’s diagnosis and treatment ability.

[0004] As described above, a teaching goal feedback method based on an artificial intelligence model, in which a virtual patient is customized for a virtual diagnosis and treatment window according to the current teaching task, is specifically divided into the following sub-steps: Define the basic information and medical history of the virtual patient; Define fictitious patient complaints and secondary symptoms; The dialog engine loads the defined virtual patient into the virtual diagnosis and treatment window.

[0005] As described above, a teaching goal feedback method based on an artificial intelligence model is used, in which a pre-trained AI model is used to adaptively update the condition data of a virtual patient based on the diagnosis and treatment behavior of the trainee and the delay of natural time, which is specifically divided into the following sub-steps: Set the conversion ratio between system time and virtual diagnosis and treatment window time, and use the virtual diagnosis and treatment window time as the reference to regularly use the pre-trained AI model to update the current condition data of the virtual patient; The diagnosis and treatment behaviors performed by the trainees are extracted into a sequence of representation vectors through the dialogue engine; The representation vector sequence output by the dialogue engine is input into the pre-trained AI model in real time, and the current condition data of the virtual patient is updated according to the output results of the model.

[0006] The present invention also provides a teaching goal feedback system based on an artificial intelligence model, comprising: a virtual diagnosis and treatment module, a disease dynamic simulation module, a clinical data recording module, a diagnosis and treatment capability quantification module, and a teaching feedback module; The virtual diagnosis and treatment module is used to customize and create virtual patients for the virtual diagnosis and treatment window according to the current teaching task; The dynamic disease simulation module is used to dynamically simulate the evolution of the disease of virtual patients using a pre-trained AI model; The clinical data recording module is used to record the diagnosis and treatment behaviors made by trainees in the virtual diagnosis and treatment window, as well as the real-time condition data of virtual patients; The diagnostic and treatment ability quantification module is used to deeply analyze the trainees’ clinical thinking based on the recorded data and quantify the trainees’ diagnostic and treatment abilities; The teaching feedback module is used to generate a feedback report on the current teaching task based on the quantitative results of each student's diagnosis and treatment ability.

[0007] The beneficial effects achieved by the present invention are as follows: the in-depth analytical ability of clinical thinking is enhanced, and the authenticity of teaching data is enhanced; the technical limitations of the traditional virtual diagnosis and treatment system of "focusing on process but not thinking, focusing on results but not process" are broken through. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0009] Figure 1This is a flow chart of a teaching objective feedback method based on an artificial intelligence model provided in Example 1 of the present application; Figure 2 This is a schematic diagram of a teaching objective feedback system based on an artificial intelligence model provided in Example 2 of the present application. DETAILED DESCRIPTION

[0010] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0011] Embodiment 1 like Figure 1 As shown, the first embodiment of the present application provides a teaching goal feedback method based on an artificial intelligence model, including: Step S10: Customize and create a virtual patient for the virtual diagnosis and treatment window according to the current teaching task; Virtual patients are the basis for achieving the current teaching task and are created by the teacher. The creation process is divided into the following sub-steps: Step S11: define the basic information and medical history of the virtual patient; Basic information is information directly displayed to students, including: the patient's age, gender, occupation, marital status, etc. The medical history requires students to ask the patient about it, including: the patient's past medical history, family medical history, allergy history, etc.

[0012] Step S12: defining the main complaint and secondary symptoms of the virtual patient; The chief complaint refers to the patient's most obvious symptom or discomfort, and the secondary symptom refers to other symptoms accompanying the chief complaint. At the same time, in this step, it is also necessary to initialize the virtual patient's medical data, that is, various physiological indicators related to the patient's condition, such as blood sugar, blood pressure, white blood cell count, etc.; here it is necessary to indicate the patient's exact and accurate diagnosis results as the final evaluation basis.

[0013] Step S13: loading the defined virtual patient into the dialogue engine of the virtual diagnosis and treatment window; The dialogue engine of the virtual diagnosis and treatment window can automatically play the role of a patient according to preset terms and complete basic dialogues during the consultation process. There are many existing dialogue engines, such as Rasa, Dialogflow, GPT-4, Google Bard, etc., which can be integrated into the existing virtual diagnosis and treatment window as needed.

[0014] Step S20: using the pre-trained AI model to adaptively update the condition data of the virtual patient based on the diagnosis and treatment behavior performed by the trainee and the delay of natural time; The pre-trained AI model is used to adaptively update the virtual patient's condition data based on the diagnosis and treatment behaviors made by the trainees and the passage of natural time. The update process is specifically divided into the following sub-steps: Step S21: Set the conversion ratio between system time and virtual diagnosis and treatment window time, and use the virtual diagnosis and treatment window time as the standard to regularly use the pre-trained AI model to update the current condition data of the virtual patient; System time refers to the natural time output by the computer. The virtual diagnosis and treatment window time is a virtual time, which is used to speed up the evolution of the virtual patient's condition so that trainees can quickly and clearly observe the changes in the patient's condition. Its display format is also consistent with the natural time, but the speed of the watch is very fast. The conversion ratio set in this embodiment is 1:500, that is, one minute of natural time is equal to 500 minutes of virtual time. Then, based on the converted virtual time, the current condition data of the virtual patient is automatically updated every 800 minutes to simulate the evolution of the virtual patient's condition under the delay of natural time; the condition update data is output by the AI ​​model, which is trained with real clinical data and is used to enhance the authenticity of virtual diagnosis and treatment teaching. The input data consists of two parts. One part is the characterization vector sequence of diagnosis and treatment behavior. No diagnosis and treatment behavior is performed here, so this part of the input data is empty. The other part is the update cycle (in hours). The update cycle here is 800 minutes, which is converted to hours as 800 / 60≈13.33 hours. After the input data passes through the AI ​​model, the updated condition data can be output. The mathematical expression of the AI ​​model is: , where Y is the model output, is the vector of the patient’s k-th original condition data, is the self-evolution matrix of the k-th original disease data, is the response matrix of the k-th original disease data to the i-th diagnosis and treatment behavior, is the characterization vector of the input i-th diagnosis and treatment behavior, is the patient's tolerance coefficient to the i-th diagnosis and treatment behavior, i ranges from 1 to n, and n is the total number of input diagnosis and treatment behavior representation vectors. is the input update period, is the self-evolution cycle of the disease (defined during model training), C is the noise matrix of the patient's disease evolution, k takes values ​​from 1 to w, and w is the total number of the patient's original disease data.

[0015] Step S22: extracting the diagnosis and treatment behaviors performed by the trainees into a representation vector sequence through the dialogue engine; The encoder-decoder in the dialogue engine is used to realize the mutual conversion between word units and representation vectors, but the current dialogue engine cannot automatically identify which words said by the students represent medical treatment behaviors, so the dialogue engine needs to be fine-tuned to adapt it to the current application scenario: first, add an output head to the dialogue engine; then divide the medical treatment behavior-related sentences in the real medical records into word units and mark them as positive samples, and divide other irrelevant sentences into word units and mark them as negative samples; use the contrastive learning training method to align the latent variables in the newly added output head with the positive samples, so that it has the ability to identify medical treatment behaviors; finally, use the trained output head to output the identified medical treatment behaviors as a representation vector sequence.

[0016] Step S23: inputting the representation vector sequence output by the dialogue engine into the pre-trained AI model in real time, and updating the current condition data of the virtual patient according to the output result of the model; In this step, the AI ​​model input data consists of a complete characterization vector sequence and an update cycle. The characterization vector sequence is extracted in real time by the dialogue engine, and the update cycle is calculated by subtracting the virtual treatment window time when the disease data was last updated from the current virtual treatment window time. Note that the unit must be converted to hours.

[0017] Step S30: monitor the diagnosis and treatment behaviors performed by the trainees in the virtual diagnosis and treatment window in real time, as well as the disease evolution data of the virtual patient, and record the monitored data, and generate a complete clinical data chain after time series; Set up a listener to monitor the output action of the diagnosis and treatment behavior representation vector and the update action of the disease data in real time, and record the monitored data. After timing, a complete clinical data chain can be obtained. This data chain is the basic data for quantifying the students' diagnosis and treatment capabilities.

[0018] Step S40: Calculate the disease progression score of each diagnosis and treatment node according to the patient's disease progression data in the clinical data chain; The complete clinical data chain can be used to gradually analyze the clinical thinking of trainees to achieve fine-grained ability quantification. The diagnosis and treatment nodes are multiple data segments divided by the clinical data chain in time sequence. The patient's condition data is segmented and input into the condition deterioration scoring formula: Calculate the patient's condition deterioration score S in each time period, where is the value of the k-th disease data at the t-th time point, is the value of the kth disease data at the t-1th time point, t ranges from 2 to T, T is the total number of disease update time points in the current time period, is the deterioration mark of the kth condition data. If the condition data is lower, the healthier it is, the The value is 1. If the condition data is higher, the healthier it is, The value is -1. is the disease identification weight of the k-th disease data, is the time attenuation coefficient (here the value is 0.1~0.9), k ranges from 1 to w, and w is the total number of disease data.

[0019] Step S50: assigning an evaluation weight to each diagnosis and treatment node according to the calculation result of the disease progression score; Evaluation weights are assigned to each diagnosis and treatment node according to the disease worsening score. Diagnosis and treatment nodes with high disease worsening scores test students' clinical thinking ability more, so higher evaluation weights need to be assigned. Conversely, lower evaluation weights need to be assigned. In other words, the evaluation weight needs to be inversely proportional to the disease worsening score.

[0020] Step S60: creating a fine-grained capability quantification formula based on the assigned evaluation weights, and using the created quantification formula to calculate the quantification result of the trainee's diagnosis and treatment capability; First, the best treatment behavior for each treatment node is queried from the real medical records, that is, the treatment method with the best stage treatment effect that is closest to the current virtual patient's condition. Then, the difference between the patient's condition data before and after the implementation of the best treatment behavior and the condition data of the virtual patient before and after the trainee's treatment behavior is used to reflect the trainee's real-time judgment ability. Then, the best treatment path for the initial virtual patient is queried, that is, the treatment plan with the best overall treatment effect that is closest to the initial condition of the virtual patient. The treatment behaviors are defined as the treatment path after time series, and the difference between the trainee's treatment path and the best treatment path is compared to reflect the trainee's disease reasoning ability. Finally, the difference between the trainee's treatment time and the treatment time of the best treatment plan is used to reflect the trainee's treatment efficiency. Combining the above descriptions, the fine-grained ability quantification formula created is expressed as: , where G is the quantitative result of the trainees’ diagnosis and treatment ability, , , They are the assessment weights of real-time judgment ability, symptom reasoning ability, and diagnosis and treatment efficiency. It represents the absolute difference of the kth condition data before and after the student made the diagnosis and treatment behavior at the jth diagnosis and treatment node. It represents the absolute difference between the kth condition data before and after the implementation of the best treatment behavior for the jth treatment node. is the disease identification weight of the kth disease data, k ranges from 1 to w, w is the total number of disease data, is the evaluation weight of the jth diagnosis and treatment node, j ranges from 1 to m, and m is the total number of diagnosis and treatment nodes. Provide students with a diagnosis and treatment pathway. For the best diagnosis and treatment path, express and The graph edit distance between express Length, express Length, It takes time to diagnose and treat students. Diagnosis and treatment for the best treatment plan takes time.

[0021] The currently recorded data and the queried data are input into the quantitative formula to calculate the quantitative results of the trainees' diagnosis and treatment capabilities.

[0022] Step S70: generating a feedback report of the current teaching task based on the quantitative results of each trainee's diagnosis and treatment ability; Based on the generated feedback report, teaching staff can adjust teaching resources and courses in real time to suit the students' ability levels and improve teaching effectiveness. The feedback report includes but is not limited to the students' individual ability values, individual comprehensive ability values, overall average ability values, pass rates and excellent rates. The individual individual ability values ​​can reflect the individual's shortcomings and strengths, the individual comprehensive ability values ​​can reflect the individual's acceptance of the current teaching tasks, and the overall average ability values, pass rates and excellent rates can reflect the overall acceptance of the current teaching tasks by the students.

[0023] Embodiment 2 like Figure 2 As shown, the second embodiment of the present application provides a teaching goal feedback system based on an artificial intelligence model, including: a virtual diagnosis and treatment module 21, a disease dynamic simulation module 22, a clinical data recording module 23, a diagnosis and treatment ability quantification module 24, and a teaching feedback module 25; (1) A virtual diagnosis and treatment module 21, which is used to customize and create a virtual patient for the virtual diagnosis and treatment window according to the current teaching task; specifically, it includes: a virtual patient definition submodule and a virtual patient loading submodule; 1. Virtual patient definition submodule, used to define the basic information, medical history, main complaints and secondary symptoms of the virtual patient; Basic information is the information directly displayed to the trainees, including: the patient's age, gender, occupation, marital status, etc. The medical history requires the trainees to ask the patient about it, including: the patient's past medical history, family medical history, allergy history, etc.; the chief complaint refers to the patient's most obvious symptom or discomfort, and the secondary symptom refers to other symptoms accompanying the chief complaint. At the same time, in this step, it is also necessary to initialize the virtual patient's condition data, that is, various physiological indicators related to the patient's condition, such as blood sugar, blood pressure, white blood cell count, etc.; the patient's exact and accurate diagnosis results need to be noted here as the final evaluation basis.

[0024] 2. Virtual patient loading submodule, used to load the defined virtual patient into the dialogue engine of the virtual diagnosis and treatment window; The dialogue engine of the virtual diagnosis and treatment window can automatically play the role of a patient according to preset terms and complete basic dialogues during the consultation process. There are many existing dialogue engines, such as Rasa, Dialogflow, GPT-4, Google Bard, etc., which can be integrated into the existing virtual diagnosis and treatment window as needed.

[0025] (2) a disease state dynamic simulation module 22, which is used to dynamically simulate the disease state evolution process of a virtual patient using a pre-trained AI model; specifically, it includes: a disease state natural evolution submodule, a diagnosis and treatment behavior extraction submodule, and a diagnosis and treatment behavior response submodule; 1. The natural evolution submodule of the condition is used to set the conversion ratio between the system time and the virtual diagnosis and treatment window time, and regularly use the pre-trained AI model to update the current condition data of the virtual patient based on the virtual diagnosis and treatment window time; System time refers to the natural time output by the computer. The virtual diagnosis and treatment window time is a virtual time, which is used to speed up the evolution of the virtual patient's condition so that trainees can quickly and clearly observe the changes in the patient's condition. Its display format is also consistent with the natural time, but the speed of the watch is very fast. The conversion ratio set in this embodiment is 1:500, that is, one minute of natural time is equal to 500 minutes of virtual time. Then, based on the converted virtual time, the current condition data of the virtual patient is automatically updated every 800 minutes to simulate the evolution of the virtual patient's condition under the delay of natural time; the condition update data is output by the AI ​​model, which is trained with real clinical data and is used to enhance the authenticity of virtual diagnosis and treatment teaching. The input data consists of two parts. One part is the characterization vector sequence of diagnosis and treatment behavior. No diagnosis and treatment behavior is performed here, so this part of the input data is empty. The other part is the update cycle (in hours). The update cycle here is 800 minutes, which is converted to hours as 800 / 60≈13.33 hours. After the input data passes through the AI ​​model, the updated condition data can be output. The mathematical expression of the AI ​​model is: , where Y is the model output, is the vector of the patient’s k-th original condition data, is the self-evolution matrix of the k-th original disease data, is the response matrix of the k-th original disease data to the i-th diagnosis and treatment behavior, is the characterization vector of the input i-th diagnosis and treatment behavior, is the patient's tolerance coefficient to the i-th diagnosis and treatment behavior, i ranges from 1 to n, and n is the total number of input diagnosis and treatment behavior representation vectors. is the input update period, is the self-evolution cycle of the disease (defined during model training), C is the noise matrix of the patient's disease evolution, k takes values ​​from 1 to w, and w is the total number of the patient's original disease data.

[0026] 2. The diagnosis and treatment behavior extraction submodule is used to extract the diagnosis and treatment behaviors made by the trainees into a representation vector sequence through the dialogue engine; The encoder-decoder in the dialogue engine is used to realize the mutual conversion between word units and representation vectors, but the current dialogue engine cannot automatically identify which words said by the students represent medical treatment behaviors, so the dialogue engine needs to be fine-tuned to adapt it to the current application scenario: first, add an output head to the dialogue engine; then divide the medical treatment behavior-related sentences in the real medical records into word units and mark them as positive samples, and divide other irrelevant sentences into word units and mark them as negative samples; use the contrastive learning method to align the latent variables in the newly added output head with the positive samples, so that it has the ability to identify medical treatment behaviors; finally, use the trained output head to output the identified medical treatment behaviors as a representation vector sequence.

[0027] 3. The diagnosis and treatment behavior response submodule is used to input the representation vector sequence output by the dialogue engine into the pre-trained AI model in real time, and update the current condition data of the virtual patient according to the output results of the model; In this step, the AI ​​model input data consists of a complete characterization vector sequence and an update cycle. The characterization vector sequence is extracted in real time by the dialogue engine, and the update cycle is calculated by subtracting the virtual treatment window time when the disease data was last updated from the current virtual treatment window time. Note that the unit must be converted to hours.

[0028] (3) a clinical data recording module 23, which is used to record the diagnosis and treatment behaviors performed by the trainees in the virtual diagnosis and treatment window, as well as the real-time condition data of the virtual patient; Set up a listener to monitor the output action of the diagnosis and treatment behavior representation vector and the update action of the disease data in real time, and record the monitored data. After timing, a complete clinical data chain can be obtained. This data chain is the basic data for quantifying the students' diagnosis and treatment capabilities.

[0029] (4) Diagnosis and treatment ability quantification module 24, which is used to deeply analyze the clinical thinking of trainees based on the recorded data and quantify the diagnosis and treatment ability of trainees; specifically, it includes: disease progression scoring submodule, evaluation weight allocation submodule, and quantitative result calculation submodule; 1. The disease progression scoring submodule is used to calculate the disease progression score of each diagnosis and treatment node according to the patient's disease evolution process; The diagnosis and treatment nodes are multiple data segments divided by the clinical data chain in time sequence to achieve the effect of gradual analysis; the patient's condition data is segmented and input into the condition deterioration scoring formula: Calculate the patient's condition deterioration score S in each time period, where is the value of the k-th disease data at the t-th time point, is the value of the kth disease data at the t-1th time point, t ranges from 2 to T, T is the total number of disease update time points in the current time period, is the deterioration mark of the kth condition data. If the condition data is lower, the healthier it is, the The value is 1. If the condition data is higher, the healthier it is, The value is -1. is the disease identification weight of the k-th disease data, is the time attenuation coefficient (here the value is 0.1~0.9), k ranges from 1 to w, and w is the total number of disease data.

[0030] 2. Evaluation weight allocation submodule, which is used to allocate evaluation weights to each diagnosis and treatment node according to the disease progression score; Evaluation weights are assigned to each diagnosis and treatment node according to the disease worsening score. Diagnosis and treatment nodes with high disease worsening scores test students' clinical thinking ability more, so higher evaluation weights need to be assigned. Conversely, lower evaluation weights need to be assigned. In other words, the evaluation weight needs to be inversely proportional to the disease worsening score.

[0031] 3. The quantitative result calculation submodule is used to create a fine-grained capability quantification formula based on the assigned evaluation weights, and use the created quantification formula to calculate the quantitative results of the trainees' diagnosis and treatment capabilities.

[0032] First, the best treatment behavior for each treatment node is queried from the real medical records, that is, the treatment method with the best stage treatment effect that is closest to the current virtual patient's condition. Then, the difference between the patient's condition data before and after the implementation of the best treatment behavior and the condition data of the virtual patient before and after the trainee's treatment behavior is used to reflect the trainee's real-time judgment ability. Then, the best treatment path for the initial virtual patient is queried, that is, the treatment plan with the best overall treatment effect that is closest to the initial condition of the virtual patient. The treatment behaviors are defined as the treatment path after time series, and the difference between the trainee's treatment path and the best treatment path is compared to reflect the trainee's disease reasoning ability. Finally, the difference between the trainee's treatment time and the treatment time of the best treatment plan is used to reflect the trainee's treatment efficiency. Combining the above descriptions, the fine-grained ability quantification formula created is expressed as: , where G is the quantitative result of the trainees’ diagnosis and treatment ability, , , They are the assessment weights of real-time judgment ability, symptom reasoning ability, and diagnosis and treatment efficiency. It represents the absolute difference of the kth condition data before and after the student made the diagnosis and treatment behavior at the jth diagnosis and treatment node. It represents the absolute difference between the kth condition data before and after the implementation of the best treatment behavior for the jth treatment node. is the disease identification weight of the kth disease data, k ranges from 1 to w, w is the total number of disease data, is the evaluation weight of the jth diagnosis and treatment node, j ranges from 1 to m, and m is the total number of diagnosis and treatment nodes. Provide students with a diagnosis and treatment pathway. For the best diagnosis and treatment path, express and The graph edit distance between express Length, express Length, It takes time to diagnose and treat students. Diagnosis and treatment for the best treatment plan takes time.

[0033] The currently recorded data and the queried data are input into the quantitative formula to calculate the quantitative results of the trainees' diagnosis and treatment capabilities.

[0034] (5) a teaching feedback module 25, used to generate a feedback report of the current teaching task based on the quantitative results of each trainee's diagnosis and treatment ability; Based on the generated feedback report, teaching staff can adjust teaching resources and courses in real time to suit the students' ability levels and improve teaching effectiveness. The feedback report includes but is not limited to the students' individual ability values, individual comprehensive ability values, overall average ability values, pass rates and excellent rates. The individual individual ability values ​​can reflect the individual's shortcomings and strengths, the individual comprehensive ability values ​​can reflect the individual's acceptance of the current teaching tasks, and the overall average ability values, pass rates and excellent rates can reflect the overall acceptance of the current teaching tasks by the students.

[0035] Corresponding to the above embodiment, an embodiment of the present invention provides a computer-readable storage medium, including: at least one memory and at least one processor; The memory is used to store one or more program instructions; A processor is used to run one or more program instructions to execute a teaching target feedback method based on an artificial intelligence model.

[0036] Corresponding to the above-mentioned embodiments, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer storage medium contains one or more program instructions, and the one or more program instructions are used by a processor to execute a teaching objective feedback method based on an artificial intelligence model.

[0037] The embodiment disclosed in the present invention provides a computer-readable storage medium, in which computer program instructions are stored. When the computer program instructions are executed on a computer, the computer executes the above-mentioned teaching objective feedback method based on an artificial intelligence model.

[0038] In the embodiment of the present invention, the processor may be an integrated circuit chip having the ability to process signals. The processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0039] The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor for execution, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The processor reads the information in the storage medium and completes the steps of the above method in combination with its hardware.

[0040] The storage medium may be a memory, which may be, for example, a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memory.

[0041] Among them, the non-volatile memory can be a read-only memory (ROM), a programmable ROM (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory.

[0042] The volatile memory may be a random access memory (RAM) which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDRSDRAM), enhanced synchronous DRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus DRAM (DRRAM).

[0043] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0044] Those skilled in the art will appreciate that in one or more of the above examples, the functions described in the present invention can be implemented using a combination of hardware and software. When software is used, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any media that facilitates the transmission of computer programs from one place to another. Storage media can be any available media that can be accessed by general or special-purpose computers.

[0045] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solution of the present invention should be included in the scope of protection of the present invention.

Claims

1. A teaching goal feedback method based on an artificial intelligence model, characterized in that: include: Step 1. Customize and create a virtual patient for the virtual diagnosis and treatment window according to the current teaching task; Step 2: Use the pre-trained AI model to adaptively update the virtual patient's condition data based on the trainee's diagnosis and treatment behaviors and the natural time delay; Step 3: Monitor the diagnosis and treatment behaviors of trainees in the virtual diagnosis and treatment window in real time, as well as the disease evolution data of virtual patients, and record the monitored data, and generate a complete clinical data chain after time series; Step 4: Calculate the disease progression score of each diagnosis and treatment node based on the patient's disease evolution data in the clinical data chain; Step 5: Assign evaluation weights to each diagnosis and treatment node based on the calculation results of the disease progression score; Step 6: Create a fine-grained ability quantification formula based on the assigned evaluation weights, and use the created ability quantification formula to calculate the quantitative results of the trainees' diagnosis and treatment abilities; Step 7. Generate a feedback report on the current teaching task based on the quantitative results of each student’s diagnosis and treatment ability.

2. The teaching objective feedback method based on an artificial intelligence model according to claim 1, characterized in that: Customize and create a virtual patient for the virtual diagnosis and treatment window according to the current teaching task, which is divided into the following sub-steps: Define the basic information and medical history of the virtual patient; Define fictitious patient complaints and secondary symptoms; The dialog engine loads the defined virtual patient into the virtual diagnosis and treatment window.

3. The teaching objective feedback method based on an artificial intelligence model according to claim 1, characterized in that: Use the pre-trained AI model to adaptively update the virtual patient's condition data based on the student's diagnosis and treatment behavior and the natural time delay. The specific steps are as follows: Set the conversion ratio between system time and virtual diagnosis and treatment window time, and use the virtual diagnosis and treatment window time as the reference to regularly use the pre-trained AI model to update the current condition data of the virtual patient; The diagnosis and treatment behaviors performed by the trainees are extracted into a sequence of representation vectors through the dialogue engine; The representation vector sequence output by the dialogue engine is input into the pre-trained AI model in real time, and the current condition data of the virtual patient is updated according to the output results of the model.

4. The teaching objective feedback method based on an artificial intelligence model according to claim 3 is characterized in that: The dialogue engine extracts the diagnosis and treatment behaviors performed by the trainees into a representation vector sequence, which is specifically divided into the following sub-steps: The dialogue engine adds a new output header; The diagnosis and treatment behavior sentences in the real medical records are divided into word units and marked as positive samples, and other sentences are divided into word units and marked as negative samples; The contrastive learning training method is used to align the latent variables in the newly added output head with the positive samples, so that it has the ability to identify diagnosis and treatment behaviors; The trained output head is used to output the identified diagnosis and treatment behaviors as a sequence of representation vectors.

5. A teaching objective feedback system based on an artificial intelligence model, used to execute a teaching objective feedback method based on an artificial intelligence model as claimed in any one of claims 1 to 4, characterized in that: include: Virtual diagnosis and treatment module, dynamic disease simulation module, clinical data recording module, diagnosis and treatment ability quantification module, teaching feedback module; The virtual diagnosis and treatment module is used to customize and create virtual patients for the virtual diagnosis and treatment window according to the current teaching task; The dynamic disease simulation module is used to dynamically simulate the evolution of the disease of virtual patients using a pre-trained AI model; The clinical data recording module is used to record the diagnosis and treatment behaviors made by trainees in the virtual diagnosis and treatment window, as well as the real-time condition data of virtual patients; The diagnostic and treatment ability quantification module is used to deeply analyze the trainees’ clinical thinking based on the recorded data and quantify the trainees’ diagnostic and treatment abilities; The teaching feedback module is used to generate a feedback report on the current teaching task based on the quantitative results of each student's diagnosis and treatment ability.

6. The teaching objective feedback system based on the artificial intelligence model according to claim 5, characterized in that: The virtual diagnosis and treatment module specifically includes: a virtual patient definition submodule and a virtual patient loading submodule; A virtual patient definition submodule is used to define the basic information, medical history, main complaints and secondary symptoms of the virtual patient; The virtual patient loading submodule is used to load the defined virtual patient into the dialogue engine of the virtual diagnosis and treatment window.

7. The teaching objective feedback system based on an artificial intelligence model according to claim 5, characterized in that: The disease dynamic simulation module specifically includes: disease natural evolution submodule, diagnosis and treatment behavior extraction submodule, diagnosis and treatment behavior response submodule; The natural evolution submodule is used to set the conversion ratio between system time and virtual diagnosis and treatment window time, and regularly use the pre-trained AI model to update the current condition data of the virtual patient based on the virtual diagnosis and treatment window time; The diagnosis and treatment behavior extraction submodule is used to extract the diagnosis and treatment behaviors performed by the trainees into a representation vector sequence through the dialogue engine; The diagnosis and treatment behavior response submodule is used to input the representation vector sequence output by the dialogue engine into the pre-trained AI model in real time, and update the current condition data of the virtual patient according to the output results of the model.

Images