A method and system for medical decision support based on multi-modal data

Abstract

The application discloses a kind of medical decision support method and system based on multi-modal data, it is related to medical decision field, it solves the problem that medical decision support technology exists and the effect of decision application is not good, including steps S1: the illness text of target patient is collected and is interacted with inquiry model, obtains illness interaction strategy text, medical index analysis is carried out to illness interaction strategy text, preliminary collection data of medical decision is obtained according to analysis result, step S2: common decision patient is obtained to illness treatment decision, and symptom analysis is carried out to common decision patient, and decision sample patient screening data is obtained according to analysis result, step S3: decision applicability analysis is carried out to illness treatment decision, and decision applicability analysis data is obtained according to analysis result, step S4: medical decision support is carried out to target patient according to decision applicability analysis data, the scientificity and operability of the present application can be improved decision support.

Description

Technical Field

[0001] This invention belongs to the field of medical decision-making and involves artificial intelligence technology. Specifically, it is a medical decision support method and system based on multimodal data. Background Technology

[0002] Existing AI-powered diagnostic technologies have the following shortcomings when providing medical decision support: 1. Existing AI consultation technology lacks the ability to accurately filter and reduce noise in the information collection and processing stage for medical scenarios. Patients may input a large amount of life details unrelated to their condition, vague subjective feelings, and repetitive symptom descriptions during the consultation process. AI models often incorporate this redundant information into their decision-making basis, and then output content containing repetitive suggestions and irrelevant correlation analysis, resulting in a lack of readability in medical decisions. 2. Existing AI-based diagnostic technologies struggle to accurately calculate complex clinical logic. Medical decision support often involves multi-dimensional and non-linear complex logic, including the interaction of multiple coexisting diseases, the synergistic and antagonistic effects of treatment plans, and the comprehensive consideration of individual differences. However, existing AI-based diagnostic technologies rely heavily on structured data matching, simple rule-based reasoning, or shallow machine learning models, which cannot systematically model and accurately calculate these complex logics. Their decision-making process is often limited to the correspondence between a single symptom and a disease, and cannot combine pathological mechanisms, disease stages, and individual patient characteristics for global consideration and in-depth reasoning like clinicians. As a result, the final decision recommendations are prone to being one-sided and may even contradict actual clinical needs.

[0003] To this end, we propose a medical decision support method and system based on multimodal data. Summary of the Invention

[0004] In view of the shortcomings of existing technologies, the purpose of this invention is to provide a medical decision support method and system based on multimodal data, which aims to improve the comprehensiveness and accuracy of medical decisions.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a medical decision support method based on multimodal data, comprising the following steps: Step S1: Collect the patient's medical condition text and interact with the consultation model to obtain the medical condition interaction strategy text. Analyze the medical condition interaction strategy text using medical indicators and obtain preliminary data for medical decision-making based on the analysis results. Step S2: Based on the preliminary data collected for medical decision-making, obtain patients with shared decision-making for disease treatment decisions, perform symptom analysis on patients with shared decision-making, and screen patients with symptom matching based on the analysis results to obtain patient screening data for decision-making samples; Step S3: Based on the patient screening data of the decision sample, conduct a decision applicability analysis on the disease treatment decision, and obtain decision applicability analysis data based on the analysis results; Step S4: Provide medical decision support to the target patients based on the decision applicability analysis data.

[0006] Furthermore, in step S1, the specific steps are as follows: Step S11: Acquire patients who need medical decision support, obtain multiple medical patients, and randomly select one target patient from the multiple acquired medical patients. Step S12: Conduct a consultation with the target patient using an AI consultation model, and obtain the consultation results to get the patient's medical condition text; Step S13: Input the patient's condition text into the AI ​​interaction model in the form of medical consultation, and obtain the text of the response strategy output by the AI ​​interaction model to obtain the condition interaction strategy text; Step S14: Obtain the medical decisions provided by the disease interaction strategy text, and randomly select one sample medical decision from the multiple medical decisions obtained. Perform disease correlation analysis on the sample medical decision, and divide the sample medical decision into disease treatment decisions and non-disease treatment decisions based on the analysis results. Step S15: Divide each medical decision into disease treatment decisions and non-disease treatment decisions to obtain preliminary medical decision data.

[0007] Furthermore, in step S14, the specific steps are as follows: Extract disease keywords from the interactive strategy text to obtain patient disease keywords, and use the patient disease keywords as search text to collect several historical patients with the same disease as the target patient. For each historical patient, the actual medical decisions corresponding to the treatment are obtained, resulting in multiple historical diagnostic decisions. If the multiple historical diagnostic decisions obtained do not include the sample medical decisions, the sample medical decisions are directly classified as non-disease treatment decisions. If the acquired multiple historical diagnostic decisions include sample medical decisions, then select the historical patients who have received patients with sample medical decisions to obtain multiple patients with the same decision. Clinical symptoms are collected from target patients to obtain a set of target patient symptoms. Clinical symptoms of patients who made the same decision are also collected to obtain a set of historical patient symptoms. If the target patient's symptom set is a subset of the historical patient's symptom set, then the sample medical decision is classified as a disease treatment decision. If the target patient's symptom set is not a subset of the historical patient's symptom set, then clinical symptoms that exist in the target patient's symptom set but are not included in the historical patient's symptom set are acquired to obtain the symptoms to be assessed. The contraindications for the corresponding medical decision are acquired. If the symptoms to be assessed contain contraindications, then the sample medical decision is classified as a non-disease treatment decision. If the symptoms to be assessed do not contain contraindications, then the sample medical decision is classified as a disease treatment decision.

[0008] Furthermore, in step S2, the specific steps are as follows: Step S21: Obtain preliminary medical decision collection data, obtain the treatment decisions for the target patient's condition based on the preliminary medical decision collection data, and take the intersection of the patients with the same decision for each treatment decision to obtain multiple patients with shared decisions; Step S22: Randomly select a characteristic decision patient from the multiple shared decision patients, perform a disease commonality analysis between the characteristic decision patient and the target patient, and obtain the symptom commonality deviation degree corresponding to the characteristic decision patient based on the analysis results; Step S23: Obtain the symptom commonality deviation degree corresponding to each patient with shared decision-making, and set a preset benchmark value for the symptom commonality deviation degree. If the symptom commonality deviation degree is less than or equal to the preset benchmark value, the patients with shared decision-making are screened as symptom-matching patients. If the symptom commonality deviation degree is greater than the preset benchmark value, the patients with shared decision-making are screened as symptom-difference patients, thus obtaining the patient screening data for the decision sample.

[0009] Furthermore, in step S22, the specific steps are as follows: Step S221: Collect clinical symptoms of the target patient during the course of the disease to obtain the target patient symptom set; collect clinical symptoms of the characteristic decision patient during the course of the disease to obtain the characteristic patient symptom set; obtain the common clinical symptoms of the target patient symptom set and the characteristic patient symptom set. Step S222: Calculate the ratio of the number of shared clinical symptoms to the number of symptoms present in the target patient's symptom set to obtain the target symptom shared ratio; Step S223: Randomly select one characteristic common symptom from the multiple common clinical symptoms obtained, perform characteristic common symptom onset analysis for the target patient, and draw the first common symptom onset polyline based on the analysis results; Step S224: Perform characteristic common symptom onset analysis for patients with characteristic decision-making, and draw a polyline of the second common symptom onset based on the analysis results; Step S225: Plot the first common symptom onset polyline and the second common symptom onset polyline together on the symptom analysis coordinate system, obtain the first characteristic symptom monitoring period, and obtain the coordinate x-axis region covered by the first characteristic symptom monitoring period to obtain the target comparison coordinate region, and create a symptom comparison traversal line perpendicular to the coordinate x-axis within the target comparison coordinate region. Step S226: Within the target comparison coordinate area, arbitrarily select a feature comparison marker point. When the symptom comparison traversal line is at the feature comparison marker point, obtain the intersection point of the symptom comparison traversal line and the first common symptom onset broken line to obtain the first onset feature intersection point. Obtain the intersection point of the symptom comparison traversal line and the second common symptom onset broken line to obtain the second onset feature intersection point. Calculate the difference in the vertical coordinates between the first and second onset feature intersection points. Calculate the ratio of the absolute value of the difference to the first onset feature intersection point to obtain the symptom index deviation degree corresponding to the feature comparison marker point. Step S227: Use the symptom comparison traversal line to traverse each coordinate point in the target comparison coordinate area, obtain the symptom index deviation degree corresponding to each coordinate point according to the traversal result, and calculate the average of the multiple symptom index deviation degrees to obtain the average symptom deviation. Step S228: Compare the obtained deviation values ​​of multiple symptom indicators, mark the symptom indicator deviation with the largest value as the peak symptom deviation, mark the symptom indicator deviation with the smallest value as the trough symptom deviation, calculate the difference between the average symptom deviation and the trough symptom deviation to obtain the first normalized feature value, calculate the difference between the peak symptom deviation and the trough symptom deviation to obtain the second normalized feature value, and calculate the ratio of the first normalized feature value to the second normalized feature value to obtain the symptom deviation normalization ratio corresponding to the common symptoms. Step S229: Obtain the symptom deviation normalization ratio corresponding to each common clinical symptom, and calculate the symptom commonality deviation degree corresponding to the feature decision patient by combining the symptom deviation normalization ratio corresponding to each common clinical symptom and the target symptom commonality ratio.

[0010] Furthermore, in step S223, the specific steps are as follows: The time point at which the target patient first develops the common characteristic symptoms is obtained to obtain the first target symptom time point. The time point corresponding to the current moment is obtained to obtain the second target symptom time point. The time period between the first target symptom time point and the second target symptom time point is set as the first characteristic symptom monitoring period. During the first characteristic symptom monitoring period, the time points when the target patients experienced the common characteristic symptoms were acquired and marked as Z1 symptom onset time points, Z2 symptom onset time points, ..., Za symptom onset time points according to the order of their occurrence. The duration of symptom attacks at the Z1 symptom onset time point of the target patient is obtained, thus obtaining the Z1 attack duration. The time interval between the Z1 and Z2 symptom onset time points is calculated, thus obtaining the Z1 consecutive attack interval. The ratio of the Z1 attack duration to the Z1 consecutive attack interval is calculated, thus obtaining the Z1 symptom attack duration ratio. The duration of symptom attacks at the Z2 symptom onset time point of the target patient is obtained, thus obtaining the Z2 attack duration. The time interval between the Z2 and Z3 symptom onset time points is calculated. The Z2 consecutive attack interval duration is obtained, and the ratio of the Z2 attack duration to the Z2 consecutive attack interval duration is calculated to obtain the Z2 symptom attack duration ratio. Similarly, the symptom attack duration of the target patient at the Za-1 symptom attack time point is obtained to obtain the Za-1 attack duration. The time interval between the Za-1 symptom attack time point and the Za symptom attack time point is calculated to obtain the Za-1 consecutive attack interval duration. The ratio of the Za-1 attack duration to the Za-1 consecutive attack interval duration is calculated to obtain the Za-1 symptom attack duration ratio. In the existing Cartesian coordinate system, the time value is created as the horizontal axis and the symptom onset time ratio is created as the vertical axis to obtain the symptom analysis coordinate system. In the symptom analysis coordinate system, the coordinate points with the symptom onset time point as the horizontal axis and the symptom onset time ratio as the vertical axis are marked as symptom manifestation coordinate points. The multiple symptom manifestation coordinate points are connected in sequence to obtain the first common symptom onset polyline.

[0011] Furthermore, in step S3, the specific steps are as follows: Step S31: Obtain preliminary medical decision collection data, and obtain treatment decisions for the target patient's condition based on the preliminary medical decision collection data, and mark the obtained treatment decisions as J1 treatment decision, J2 treatment decision, J3 treatment decision, ... Jc treatment decision; Step S32: Obtain decision sample patient screening data, and obtain patients matching the symptoms corresponding to the target patients based on the decision sample patient screening data; Step S33: Perform clinical treatment index analysis on J1 treatment decision for symptom-matched patients, and obtain the common time period and quantitative interval of J1 treatment decision based on the analysis results; Step S34: Obtain the common time period for treatment decisions from J2 to Jc, and the quantitative interval for decisions from J2 to Jc, respectively; Step S35: Define the common treatment decision period from J1 to Jc and the quantitative interval of decision from J1 to Jc as decision applicability analysis data.

[0012] Furthermore, in step S33, the specific steps are as follows: A J1 treatment timeline is created by using the time point corresponding to the first visit of symptom-matched patients as the starting time point, and the historical periods of clinical treatment of symptom-matched patients by taking J1 treatment decisions are marked in the J1 treatment timeline to obtain multiple J1 decision treatment periods. The time intersection of the multiple J1 decision-making treatment periods is taken to obtain the common time periods of J1 treatment decisions; For each symptom-matched patient, the union of abnormal physiological indicators that appeared during the common treatment decision period of J1 was taken to obtain multiple abnormal indicators for patients. The obtained abnormal indicators for patients were labeled as P1 decision-related indicators, P2 decision-related indicators, P3 decision-related indicators, ... Pd decision-related indicators. Patients who received J1 treatment decisions and underwent historical treatment were acquired, resulting in multiple J1 treatment sample patients. Patients with abnormal P1 decision-related indicators were also acquired, resulting in multiple abnormal sample patients. The ratio of the number of abnormal sample patients to the number of J1 treatment sample patients was calculated to obtain the P1 indicator analysis weight. Repeat the process of obtaining the analysis weights for the P1 indicator, and obtain the analysis weights for the P2 decision-related indicators to the Pd decision analysis indicators respectively, to obtain the analysis weights for the P2 indicators to the Pd indicators. Randomly select one sample of symptom-matched patients from the multiple symptom-matched patients, perform clinical indicator analysis on the sample of symptom-matched patients, and obtain the quantitative value of the J1 decision index corresponding to the sample of symptom-matched patients based on the analysis results. The historical periods in which J1 treatment decisions were made in clinical treatment were obtained to identify characteristic treatment periods. The characteristic treatment period is discretized into several indicator monitoring points. The indicator deviation of P1 decision-related indicators at each indicator monitoring point is obtained to obtain the time deviation of P1 indicators. The median of the normal indicator interval corresponding to the P1 decision-related indicators is collected to obtain the benchmark value of P1 indicators. The ratio of the time deviation of P1 indicators to the benchmark value of P1 indicators is calculated to obtain multiple P1 indicator deviation degrees.

[0013] Furthermore, in step S33, the specific steps are as follows: The median deviation of the multiple P1 indicators is calculated to obtain the median deviation of the P1 indicators. The difference between each P1 indicator deviation and the median deviation of the P1 indicators is calculated, and the absolute value of the obtained difference is taken. The results are then renamed from the first indicator deviation to the qth indicator deviation in ascending order of the absolute value. The standard deviation of the first indicator deviation and the second indicator deviation is calculated to obtain the standard deviation of the first indicator deviation. The standard deviation of the first indicator deviation to the third indicator deviation is calculated to obtain the standard deviation of the second indicator deviation. And so on, the standard deviation of the first indicator deviation to the qth indicator deviation is calculated to obtain the standard deviation of the (q-1)th indicator deviation. Compare the standard deviations of the first indicator deviation to the standard deviations of the (q-1)th indicator deviation, and set the standard deviation of the indicator deviation with the largest value as the peak value of the standard deviation of the indicator deviation. The average deviation of the indicators involved in the calculation of the peak standard deviation of the indicator deviation is calculated to obtain the comprehensive deviation value of the P1 indicator corresponding to the sample symptom-matched patients. Repeat the process of obtaining the comprehensive deviation value of the P1 indicator, and obtain the comprehensive deviation values ​​of the indicators corresponding to the decision-related indicators from P2 to Pd respectively, to obtain the comprehensive deviation degree of the P2 indicator to the comprehensive deviation degree of the Pd indicator. Calculate the product of the comprehensive deviation value of indicator P1 and the analysis weight of indicator P1 to obtain the weighted comprehensive deviation value of indicator P1. Calculate the product of the comprehensive deviation value of indicator P2 and the analysis weight of indicator P2 to obtain the weighted comprehensive deviation value of indicator P2. And so on, calculate the product of the comprehensive deviation value of indicator Pd and the analysis weight of indicator Pd to obtain the weighted comprehensive deviation value of indicator Pd. Calculate the average value of the weighted comprehensive deviation values ​​of indicators P1 to Pd to obtain the quantitative value of decision indicator J1. The J1 decision index quantification value corresponding to the symptom-matched patients in the repeated sample is obtained. The J1 decision index quantification value corresponding to each symptom-matched patient is obtained. The average value of the obtained J1 decision index quantification value is calculated to obtain the first indicator quantification feature value. The standard deviation value of the obtained J1 decision index quantification value is calculated to obtain the second indicator quantification feature value. The sum of the first indicator quantification feature value and the second indicator quantification feature value is used to create the upper limit of the J1 decision quantification interval. The difference between the first indicator quantification feature value and the second indicator quantification feature value is used to create the lower limit of the J1 decision quantification interval, thus obtaining the J1 decision quantification interval.

[0014] A medical decision support system based on multimodal data, the medical decision support system comprising: Data acquisition module: Collects textual information about the patient's condition and interacts with the consultation model to obtain textual information about the patient's condition interaction strategy. Analyzes the textual information about the patient's condition interaction strategy using medical indicators and obtains preliminary data for medical decision-making based on the analysis results. Data Analysis Module: Based on the preliminary data collected for medical decision-making, the module identifies patients with shared decision-making criteria for disease treatment, performs symptom analysis on these patients, and then uses the analysis results to screen patients with shared decision-making criteria by matching their symptoms, thus obtaining the patient screening data for the decision sample. Applicability Analysis Module: Based on the patient screening data of the decision sample, this module performs decision applicability analysis on the treatment decisions and obtains decision applicability analysis data based on the analysis results. Decision support module: Based on decision applicability analysis data, extract the common time periods and quantitative intervals corresponding to treatment decisions for different conditions, and provide medical decision support for target patients according to the common time periods and quantitative intervals.

[0015] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention collects and analyzes the patient's medical condition text to obtain initial interactive text. It then performs medical indicator analysis on the initial interactive text, and based on the analysis results, conducts preliminary screening of the medical decisions involved in the medical interactive text to obtain treatment decisions. For these treatment decisions, it identifies patients with shared decisions and analyzes their symptoms. Based on the analysis results, it screens patients with shared decisions for symptom matching and uses these symptom-matching patients as decision support samples for decision analysis. This effectively filters redundant information, ensuring that decision analysis is based on real cases that highly match the symptoms of the target patient, thus improving the accuracy and relevance of the initial screening for treatment decisions.

[0016] 2. This invention analyzes historical medical indicators for symptom-matched patients to determine their treatment decisions. Based on the analysis results, it obtains the common time periods and quantitative intervals for each treatment decision. Using these common time periods and quantitative intervals, it provides medical decision support for target patients, offering clear time-based guidance and quantitative references. This overcomes the limitations of existing technologies in accurately calculating complex clinical logic, shifting medical decision-making from experience-based judgment to data-driven approaches, and improving the scientific rigor and operability of decision support. Attached Figure Description

[0017] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0018] Figure 1 This is a diagram illustrating the implementation steps of the present invention; Figure 2 This is an overall system block diagram of the present invention; Figure 3 This is a schematic diagram of the target comparison coordinate region of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1 Please see Figure 1 This invention provides a technical solution: a medical decision support method based on multimodal data, comprising the following steps: Step S1: Collect the patient's medical condition text and interact with the consultation model to obtain the medical condition interaction strategy text. Analyze the medical condition interaction strategy text using medical indicators and obtain preliminary data for medical decision-making based on the analysis results. The specific steps in step S1 are as follows: Step S11: Acquire patients who need medical decision support, obtain multiple medical patients, and randomly select one target patient from the multiple acquired medical patients. Step S12: Conduct a consultation with the target patient using an AI consultation model, and obtain the consultation results to get the patient's medical condition text; Step S13: Input the patient's condition text into the AI ​​interaction model in the form of medical consultation, and obtain the text of the response strategy output by the AI ​​interaction model to obtain the condition interaction strategy text; Step S14: Obtain the medical decisions provided by the disease interaction strategy text, and randomly select one sample medical decision from the multiple medical decisions obtained. Perform disease correlation analysis on the sample medical decision, and divide the sample medical decision into disease treatment decisions and non-disease treatment decisions based on the analysis results. In step S14, the specific steps are as follows: Extract disease keywords from the interactive strategy text to obtain patient disease keywords, and use the patient disease keywords as search text to collect several historical patients with the same disease as the target patient. For each historical patient, the actual medical decisions corresponding to the treatment are obtained, resulting in multiple historical diagnostic decisions. If the multiple historical diagnostic decisions obtained do not include the sample medical decisions, the sample medical decisions are directly classified as non-disease treatment decisions. If the acquired multiple historical diagnostic decisions include sample medical decisions, then select the historical patients who have received patients with sample medical decisions to obtain multiple patients with the same decision. Clinical symptoms are collected from target patients to obtain a set of target patient symptoms. Clinical symptoms of patients who made the same decision are also collected to obtain a set of historical patient symptoms. If the target patient's symptom set is a subset of the historical patient's symptom set, then the sample medical decision is classified as a disease treatment decision. If the target patient's symptom set is not a subset of the historical patient's symptom set, then clinical symptoms that exist in the target patient's symptom set but are not included in the historical patient's symptom set are acquired to obtain the symptoms to be evaluated. The decision-contraindication symptoms corresponding to the sample medical decision are acquired. If the symptoms to be evaluated have decision-contraindication symptoms, then the sample medical decision is classified as a non-disease treatment decision. If the symptoms to be evaluated do not have decision-contraindication symptoms, then the sample medical decision is classified as a disease treatment decision. Step S15: Divide each medical decision into disease-related treatment decisions and non-disease-related treatment decisions to obtain preliminary medical decision data; Step S2: Based on the preliminary data collected for medical decision-making, obtain patients with shared decision-making for disease treatment decisions, perform symptom analysis on patients with shared decision-making, and screen patients with symptom matching based on the analysis results to obtain patient screening data for decision-making samples; In step S2, the specific steps are as follows: Step S21: Obtain preliminary medical decision collection data, obtain the treatment decisions for the target patient's condition based on the preliminary medical decision collection data, and take the intersection of the patients with the same decision for each treatment decision to obtain multiple patients with shared decisions; Step S22: Randomly select a characteristic decision patient from the multiple shared decision patients, perform a disease commonality analysis between the characteristic decision patient and the target patient, and obtain the symptom commonality deviation degree corresponding to the characteristic decision patient based on the analysis results; In step S22, the specific steps are as follows: Step S221: Collect clinical symptoms of the target patient during the course of the disease to obtain the target patient symptom set; collect clinical symptoms of the characteristic decision patient during the course of the disease to obtain the characteristic patient symptom set; obtain the common clinical symptoms of the target patient symptom set and the characteristic patient symptom set. Step S222: Calculate the ratio of the number of shared clinical symptoms to the number of symptoms present in the target patient's symptom set to obtain the target symptom shared ratio; Step S223: Randomly select one characteristic common symptom from the multiple common clinical symptoms obtained, perform characteristic common symptom onset analysis for the target patient, and draw the first common symptom onset polyline based on the analysis results; In step S223, the specific steps are as follows: The time point at which the target patient first develops the common characteristic symptoms is obtained to obtain the first target symptom time point. The time point corresponding to the current moment is obtained to obtain the second target symptom time point. The time period between the first target symptom time point and the second target symptom time point is set as the first characteristic symptom monitoring period. During the first characteristic symptom monitoring period, the time points when the target patients experienced the common characteristic symptoms were acquired and marked as Z1 symptom onset time points, Z2 symptom onset time points, ..., Za symptom onset time points according to the order of their occurrence. The duration of symptom attacks at the Z1 symptom onset time point of the target patient is obtained, thus obtaining the Z1 attack duration. The time interval between the Z1 and Z2 symptom onset time points is calculated, thus obtaining the Z1 consecutive attack interval. The ratio of the Z1 attack duration to the Z1 consecutive attack interval is calculated, thus obtaining the Z1 symptom attack duration ratio. The duration of symptom attacks at the Z2 symptom onset time point of the target patient is obtained, thus obtaining the Z2 attack duration. The time interval between the Z2 and Z3 symptom onset time points is calculated. The Z2 consecutive attack interval duration is obtained, and the ratio of the Z2 attack duration to the Z2 consecutive attack interval duration is calculated to obtain the Z2 symptom attack duration ratio. Similarly, the symptom attack duration of the target patient at the Za-1 symptom attack time point is obtained to obtain the Za-1 attack duration. The time interval between the Za-1 symptom attack time point and the Za symptom attack time point is calculated to obtain the Za-1 consecutive attack interval duration. The ratio of the Za-1 attack duration to the Za-1 consecutive attack interval duration is calculated to obtain the Za-1 symptom attack duration ratio. In the existing Cartesian coordinate system, the time value is created as the horizontal axis and the symptom onset time ratio is created as the vertical axis to obtain the symptom analysis coordinate system. In the symptom analysis coordinate system, the coordinate points with the symptom onset time point as the horizontal axis and the symptom onset time ratio as the vertical axis are marked as symptom manifestation coordinate points. The multiple symptom manifestation coordinate points are connected in sequence to obtain the first common symptom onset polyline. Step S224: Perform characteristic common symptom onset analysis for patients with characteristic decision-making, and draw a polyline of the second common symptom onset based on the analysis results; Step S225: Plot the first common symptom onset polyline and the second common symptom onset polyline together on the symptom analysis coordinate system, obtain the first characteristic symptom monitoring period, and obtain the coordinate x-axis region covered by the first characteristic symptom monitoring period to obtain the target comparison coordinate region, and create a symptom comparison traversal line perpendicular to the coordinate x-axis within the target comparison coordinate region. Step S226: Within the target comparison coordinate area, arbitrarily select a feature comparison marker point. When the symptom comparison traversal line is at the feature comparison marker point, obtain the intersection point of the symptom comparison traversal line and the first common symptom onset broken line to obtain the first onset feature intersection point. Obtain the intersection point of the symptom comparison traversal line and the second common symptom onset broken line to obtain the second onset feature intersection point. Calculate the difference in the vertical coordinates between the first and second onset feature intersection points. Calculate the ratio of the absolute value of the difference to the first onset feature intersection point to obtain the symptom index deviation degree corresponding to the feature comparison marker point. Step S227: Use the symptom comparison traversal line to traverse each coordinate point in the target comparison coordinate area, obtain the symptom index deviation degree corresponding to each coordinate point according to the traversal result, and calculate the average of the multiple symptom index deviation degrees to obtain the average symptom deviation. Step S228: Compare the obtained deviation values ​​of multiple symptom indicators, mark the symptom indicator deviation with the largest value as the peak symptom deviation, mark the symptom indicator deviation with the smallest value as the trough symptom deviation, calculate the difference between the average symptom deviation and the trough symptom deviation to obtain the first normalized feature value, calculate the difference between the peak symptom deviation and the trough symptom deviation to obtain the second normalized feature value, and calculate the ratio of the first normalized feature value to the second normalized feature value to obtain the symptom deviation normalization ratio corresponding to the common symptoms. Step S229: Obtain the symptom deviation normalization ratio corresponding to each common clinical symptom, and calculate the symptom commonality deviation degree corresponding to the feature decision patient by combining the symptom deviation normalization ratio corresponding to each common clinical symptom and the target symptom commonality ratio. Step S23: Obtain the symptom commonality deviation degree corresponding to each patient with shared decision-making, set a preset benchmark value for symptom commonality deviation degree. If the symptom commonality deviation degree is less than or equal to the preset benchmark value, the patients with shared decision-making are screened as symptom-matching patients. If the symptom commonality deviation degree is greater than the preset benchmark value, the patients with shared decision-making are screened as symptom-difference patients, thus obtaining the patient screening data for the decision sample. Step S3: Based on the patient screening data of the decision sample, conduct a decision applicability analysis on the disease treatment decision, and obtain decision applicability analysis data based on the analysis results; In step S3, the specific steps are as follows: Step S31: Obtain preliminary medical decision collection data, and obtain treatment decisions for the target patient's condition based on the preliminary medical decision collection data, and mark the obtained treatment decisions as J1 treatment decision, J2 treatment decision, J3 treatment decision, ... Jc treatment decision; Step S32: Obtain decision sample patient screening data, and obtain patients matching the symptoms corresponding to the target patients based on the decision sample patient screening data; Step S33: Perform clinical treatment index analysis on J1 treatment decision for symptom-matched patients, and obtain the common time period and quantitative interval of J1 treatment decision based on the analysis results; In step S33, the specific steps are as follows: A J1 treatment timeline is created by using the time point corresponding to the first visit of symptom-matched patients as the starting time point, and the historical periods of clinical treatment of symptom-matched patients by taking J1 treatment decisions are marked in the J1 treatment timeline to obtain multiple J1 decision treatment periods. The time intersection of the multiple J1 decision-making treatment periods is taken to obtain the common time periods of J1 treatment decisions; For each symptom-matched patient, the union of abnormal physiological indicators that appeared during the common treatment decision period of J1 was taken to obtain multiple abnormal indicators for patients. The obtained abnormal indicators for patients were labeled as P1 decision-related indicators, P2 decision-related indicators, P3 decision-related indicators, ... Pd decision-related indicators. Patients who received J1 treatment decisions and underwent historical treatment were acquired, resulting in multiple J1 treatment sample patients. Patients with abnormal P1 decision-related indicators were also acquired, resulting in multiple abnormal sample patients. The ratio of the number of abnormal sample patients to the number of J1 treatment sample patients was calculated to obtain the P1 indicator analysis weight. Repeat the process of obtaining the analysis weights for the P1 indicator, and obtain the analysis weights for the P2 decision-related indicators to the Pd decision analysis indicators respectively, to obtain the analysis weights for the P2 indicators to the Pd indicators. Randomly select one sample of symptom-matched patients from the multiple symptom-matched patients, perform clinical indicator analysis on the sample of symptom-matched patients, and obtain the quantitative value of the J1 decision index corresponding to the sample of symptom-matched patients based on the analysis results. The historical periods in which J1 treatment decisions were made in clinical treatment were obtained to identify characteristic treatment periods. The characteristic treatment period is discretized into several indicator monitoring points. The indicator deviation of P1 decision-related indicators at each indicator monitoring point is obtained to obtain the time deviation of P1 indicators. The median of the normal indicator interval corresponding to the P1 decision-related indicators is collected to obtain the benchmark value of P1 indicators. The ratio of the time deviation of P1 indicators to the benchmark value of P1 indicators is calculated to obtain multiple P1 indicator deviation degrees. The median deviation of the multiple P1 indicators is calculated to obtain the median deviation of the P1 indicators. The difference between each P1 indicator deviation and the median deviation of the P1 indicators is calculated, and the absolute value of the obtained difference is taken. The results are then renamed from the first indicator deviation to the qth indicator deviation in ascending order of the absolute value. The standard deviation of the first indicator deviation and the second indicator deviation is calculated to obtain the standard deviation of the first indicator deviation. The standard deviation of the first indicator deviation to the third indicator deviation is calculated to obtain the standard deviation of the second indicator deviation. And so on, the standard deviation of the first indicator deviation to the qth indicator deviation is calculated to obtain the standard deviation of the (q-1)th indicator deviation. Compare the standard deviations of the first indicator deviation to the standard deviations of the (q-1)th indicator deviation, and set the standard deviation of the indicator deviation with the largest value as the peak value of the standard deviation of the indicator deviation. The average deviation of the indicators involved in the calculation of the peak standard deviation of the indicator deviation is calculated to obtain the comprehensive deviation value of the P1 indicator corresponding to the sample symptom-matched patients. Repeat the process of obtaining the comprehensive deviation value of the P1 indicator, and obtain the comprehensive deviation values ​​of the indicators corresponding to the decision-related indicators from P2 to Pd respectively, to obtain the comprehensive deviation degree of the P2 indicator to the comprehensive deviation degree of the Pd indicator. Calculate the product of the comprehensive deviation value of indicator P1 and the analysis weight of indicator P1 to obtain the weighted comprehensive deviation value of indicator P1. Calculate the product of the comprehensive deviation value of indicator P2 and the analysis weight of indicator P2 to obtain the weighted comprehensive deviation value of indicator P2. And so on, calculate the product of the comprehensive deviation value of indicator Pd and the analysis weight of indicator Pd to obtain the weighted comprehensive deviation value of indicator Pd. Calculate the average value of the weighted comprehensive deviation values ​​of indicators P1 to Pd to obtain the quantitative value of decision indicator J1. The J1 decision index quantification value corresponding to the symptom-matched patients in the repeated sample is obtained. The J1 decision index quantification value corresponding to each symptom-matched patient is obtained. The average value of the obtained J1 decision index quantification value is calculated to obtain the first indicator quantification feature value. The standard deviation value of the obtained J1 decision index quantification value is calculated to obtain the second indicator quantification feature value. The sum of the first indicator quantification feature value and the second indicator quantification feature value is used to create the upper limit of the J1 decision quantification interval. The difference between the first indicator quantification feature value and the second indicator quantification feature value is used to create the lower limit of the J1 decision quantification interval, thus obtaining the J1 decision quantification interval. Step S34: Obtain the common time period for treatment decisions from J2 to Jc, and the quantitative interval for decisions from J2 to Jc, respectively; Step S35: Define the common treatment decision period from J1 to Jc and the quantitative decision interval from J1 to Jc as decision applicability analysis data; Step S4: Provide medical decision support to the target patients based on the decision applicability analysis data; In step S4, the specific steps are as follows: Create a target treatment timeline by taking the time point corresponding to the first visit of the target patient as the starting time point, and mark the current time on the target treatment timeline to obtain the real-time decision time point; Obtain decision applicability analysis data, and based on the decision applicability analysis data, obtain the common time period for J1 treatment decisions to the common time period for Jc treatment decisions, as well as the quantitative range for J1 decisions to the quantitative range for Jc decisions; If the real-time decision-making time point falls within the common period of septic treatment decision-making, then the septic decision-making index quantification value is collected for the target patient. If the septic decision-making index quantification value is within the septic decision quantification range, then the septic treatment decision is pushed to the target patient through the AI ​​terminal.

[0021] Example 2 Please see Figure 2 Based on another concept of the same invention, a medical decision support system based on multimodal data is proposed. The specific working process of each module is as follows: The data acquisition module collects the patient's medical condition text and interacts with the consultation model to obtain the medical condition interaction strategy text. It then performs medical indicator analysis on the medical condition interaction strategy text and obtains preliminary data for medical decision-making based on the analysis results. Specifically as follows: Acquire patients who require medical decision support, obtain multiple medical patients, and arbitrarily select one target patient from the multiple acquired medical patients; The AI-powered consultation model is used to conduct consultations with target patients and obtain the consultation results to generate a textual description of the patient's condition. It should be noted here that: In this application, the patient's medical history may be described as follows: "He has had coronary heart disease for six years. After taking azithromycin for a week, he developed symptoms of tachycardia. Sometimes when he gets up from a stool, he feels his heart is beating very fast, as if his heart is about to jump out of his chest. This happens frequently and his heart rate increases during the initial period from rest to activity, accompanied by chest tightness and chest pain."

[0022] The patient's medical condition text is input into the AI ​​interaction model in the form of medical consultation, and the text of the response strategy output by the AI ​​interaction model is obtained to obtain the medical condition interaction strategy text. It should be noted here that: In this application, the AI ​​interaction model referred to herein is specifically deepseek.

[0023] The medical decisions provided by the disease interaction strategy text are obtained, and a sample medical decision is randomly selected from the multiple medical decisions obtained. The disease correlation analysis is performed on the sample medical decision, and the sample medical decision is divided into disease treatment decisions and non-disease treatment decisions based on the analysis results. It should be noted here that: In this application, the medical decisions involved include, but are not limited to, discontinuing azithromycin, performing electrocardiograms, and controlling tachycardia.

[0024] Specifically as follows: Extract disease keywords from the interactive strategy text to obtain patient disease keywords, and use the patient disease keywords as search text to collect several historical patients with the same disease as the target patient. It should be noted here that: In this application, if the patient's medical history is described as: "I have had coronary heart disease for six years. After taking azithromycin for a week, I developed symptoms of tachycardia. Sometimes when I get up from a stool, I feel my heart beating very fast, as if it is about to jump out of my chest. This happens frequently, and my heart rate increases shortly after transitioning from a resting state to an active state, accompanied by chest tightness and chest pain," then the relevant keywords for the medical history could be "coronary heart disease," "six years," "azithromycin," "rapid heart rate," and "chest tightness and chest pain." The historical patients referred to here must have each disease keyword present in the corresponding patient medical record text.

[0025] For each historical patient, the actual medical decisions corresponding to the treatment are obtained, resulting in multiple historical diagnostic decisions. If the multiple historical diagnostic decisions obtained do not include the sample medical decisions, the sample medical decisions are directly classified as non-disease treatment decisions. If the acquired multiple historical diagnostic decisions include sample medical decisions, then select the historical patients who have received patients with sample medical decisions to obtain multiple patients with the same decision. Clinical symptoms are collected from target patients to obtain a set of target patient symptoms. Clinical symptoms of patients who made the same decision are also collected to obtain a set of historical patient symptoms. If the target patient's symptom set is a subset of the historical patient's symptom set, then the sample medical decision is classified as a disease treatment decision. If the target patient's symptom set is not a subset of the historical patient's symptom set, then clinical symptoms that exist in the target patient's symptom set but are not included in the historical patient's symptom set are acquired to obtain the symptoms to be evaluated. The decision-contraindication symptoms corresponding to the sample medical decision are acquired. If the symptoms to be evaluated have decision-contraindication symptoms, then the sample medical decision is classified as a non-disease treatment decision. If the symptoms to be evaluated do not have decision-contraindication symptoms, then the sample medical decision is classified as a disease treatment decision. It should be noted here that: In this application, the decision-making contraindications referred to herein are specifically absolute contraindications. Absolute contraindications refer to symptoms or conditions in which a certain medical measure (medication, surgery, examination, etc.) absolutely cannot be used when a patient has a certain symptom or condition, otherwise it will cause uncontrollable extreme risks such as endangering life or irreversible damage, and there is no room for compromise.

[0026] The process of dividing the sample medical decisions into disease treatment decisions and non-disease treatment decisions was repeated, and each medical decision was divided into disease treatment decisions and non-disease treatment decisions to obtain preliminary medical decision data. The data analysis module acquires patients with shared decisions on disease treatment decisions based on the preliminary data collected for medical decision-making, performs symptom analysis on patients with shared decisions, and screens patients with shared decisions based on the analysis results to obtain patient screening data for decision sample. Specifically as follows: Acquire preliminary medical decision-making data, obtain treatment decisions for the target patient's condition based on the preliminary medical decision-making data, and find the intersection of patients with the same decision for each treatment decision to obtain multiple patients with shared decisions; It should be noted here that: In this application, the term "shared decision patients" specifically refers to patients who have undergone treatment based on every single treatment decision for their condition. Randomly select a characteristic decision patient from among the multiple patients with shared decision-making information, perform a disease commonality analysis between the characteristic decision patient and the target patient, and obtain the symptom commonality deviation degree corresponding to the characteristic decision patient based on the analysis results; Specifically as follows: The clinical symptoms of the target patients during the course of the disease are collected to obtain the target patient symptom set. The clinical symptoms of the characteristic decision patients during the course of the disease are collected to obtain the characteristic patient symptom set. The common clinical symptoms of the target patient symptom set and the characteristic patient symptom set are obtained to obtain the common clinical symptoms. The ratio of the number of shared clinical symptoms to the number of symptoms present in the symptom set of the target patient is calculated to obtain the target symptom shared ratio. From the multiple common clinical symptoms obtained, one characteristic common symptom is randomly selected, and the onset of the characteristic common symptom is analyzed for the target patient. Based on the analysis results, a polyline of the first common symptom is plotted. Specifically as follows: The time point at which the target patient first develops the common characteristic symptoms is obtained to obtain the first target symptom time point. The time point corresponding to the current moment is obtained to obtain the second target symptom time point. The time period between the first target symptom time point and the second target symptom time point is set as the first characteristic symptom monitoring period. During the first characteristic symptom monitoring period, the time points when the target patients experienced the common characteristic symptoms were acquired and marked as Z1 symptom onset time points, Z2 symptom onset time points, ..., Za symptom onset time points according to the order of their occurrence. It should be noted here that: In this application, Z1 symptom onset time point, Z2 symptom onset time point, ..., Za symptom onset time point are the marker symbols corresponding to the symptom onset time points, and a is an integer greater than 0; In this application, the onset time of Z1 symptoms coincides with the time of the first target symptom.

[0027] The duration of symptom attacks at the Z1 symptom onset time point of the target patient is obtained, thus obtaining the Z1 attack duration. The time interval between the Z1 and Z2 symptom onset time points is calculated, thus obtaining the Z1 consecutive attack interval. The ratio of the Z1 attack duration to the Z1 consecutive attack interval is calculated, thus obtaining the Z1 symptom attack duration ratio. The duration of symptom attacks at the Z2 symptom onset time point of the target patient is obtained, thus obtaining the Z2 attack duration. The time interval between the Z2 and Z3 symptom onset time points is calculated. The Z2 consecutive attack interval duration is obtained, and the ratio of the Z2 attack duration to the Z2 consecutive attack interval duration is calculated to obtain the Z2 symptom attack duration ratio. Similarly, the symptom attack duration of the target patient at the Za-1 symptom attack time point is obtained to obtain the Za-1 attack duration. The time interval between the Za-1 symptom attack time point and the Za symptom attack time point is calculated to obtain the Za-1 consecutive attack interval duration. The ratio of the Za-1 attack duration to the Za-1 consecutive attack interval duration is calculated to obtain the Za-1 symptom attack duration ratio. In the existing Cartesian coordinate system, the time value is created as the horizontal axis and the symptom onset time ratio is created as the vertical axis to obtain the symptom analysis coordinate system. In the symptom analysis coordinate system, the coordinate points with the symptom onset time point as the horizontal axis and the symptom onset time ratio as the vertical axis are marked as symptom manifestation coordinate points. The multiple symptom manifestation coordinate points are connected in sequence to obtain the first common symptom onset polyline. For patients with characteristic decision-making, a characteristic common symptom onset analysis was performed, and a polyline plot of the second common symptom onset was drawn based on the analysis results; The first common symptom onset polyline and the second common symptom onset polyline are plotted together on the symptom analysis coordinate system to obtain the first characteristic symptom monitoring period. The coordinate x-axis region covered by the first characteristic symptom monitoring period is obtained to obtain the target comparison coordinate region. A symptom comparison traversal line perpendicular to the coordinate x-axis is created within the target comparison coordinate region. Please see Figure 3 Within the target comparison coordinate area, arbitrarily select a feature comparison marker point. When the symptom comparison traversal line is at the feature comparison marker point, obtain the intersection point of the symptom comparison traversal line and the first common symptom onset broken line to obtain the first onset feature intersection point. Obtain the intersection point of the symptom comparison traversal line and the second common symptom onset broken line to obtain the second onset feature intersection point. Calculate the difference in the vertical coordinates of the first and second onset feature intersection points. Calculate the ratio of the absolute value of the difference to the first onset feature intersection point to obtain the symptom index deviation degree corresponding to the feature comparison marker point. The symptom comparison traversal line is used to traverse each coordinate point in the target comparison coordinate area. Based on the traversal results, the symptom index deviation degree corresponding to each coordinate point is obtained, and the average of the obtained multiple symptom index deviation degrees is calculated to obtain the average symptom deviation. The obtained deviation scores of multiple symptom indicators are compared numerically. The symptom indicator deviation score with the largest value is marked as the peak symptom deviation score, and the symptom indicator deviation score with the smallest value is marked as the valley symptom deviation score. The difference between the average symptom deviation and the trough symptom deviation is calculated to obtain the first normalized characteristic value. The difference between the peak symptom deviation and the trough symptom deviation is calculated to obtain the second normalized characteristic value. The ratio of the first normalized characteristic value to the second normalized characteristic value is calculated to obtain the symptom deviation normalization ratio corresponding to the common symptoms. Repeat the process of obtaining the symptom deviation normalization ratio corresponding to the common symptoms, and obtain the symptom deviation normalization ratio corresponding to each common clinical symptom. The symptom commonality deviation degree corresponding to the characteristic decision patients is obtained by calculating the symptom deviation normalization ratio corresponding to each common clinical symptom and the target symptom commonality ratio. The common symptom bias of patients corresponding to feature-based decision-making is calculated using the following formula: ; Wherein, Gpd is the symptom commonality deviation degree corresponding to the patients in the feature decision, Pbzi is the symptom deviation normalization ratio, Gyb is the target symptom commonality ratio, and n is the corresponding quantity value of the common clinical symptoms. It should be noted here that: In this application, Pbzi can be the symptom deviation normalization ratio corresponding to any common clinical symptom.

[0028] Repeat the process of obtaining the symptom commonality deviation degree corresponding to patients with characteristic decisions, and obtain the symptom commonality deviation degree corresponding to each patient with common decisions; Set a preset baseline value for the common deviation of symptoms. If the common deviation of symptoms is less than or equal to the preset baseline value, patients with common decisions are screened as symptom-matching patients. If the common deviation of symptoms is greater than the preset baseline value, patients with common decisions are screened as symptom-difference patients, thus obtaining the decision sample patient screening data. It should be noted here that: The system acquires patients who have completed medical decision support in the past, and acquires patients whose symptoms match the patients in the past. This yields multiple patients whose symptoms match the past. The system also acquires the common deviation of symptoms for each patient whose symptoms match the past. The system compares the values ​​of the multiple common deviations of symptoms and marks the common deviation of symptoms as the preset baseline value. The applicability analysis module performs a decision applicability analysis on the disease treatment decision based on the patient screening data of the decision sample, and obtains decision applicability analysis data based on the analysis results; Specifically as follows: Acquire preliminary medical decision-making data, and based on this data, obtain treatment decisions for the target patient's condition. These decisions are then labeled as J1 treatment decision, J2 treatment decision, J3 treatment decision, ..., Jc treatment decision. It should be noted here that: In this application, J1, J2, J3, ..., Jc are the marking symbols corresponding to the treatment decisions for the disease, respectively.

[0029] Obtain patient screening data for the decision sample, and then obtain patients matching the symptoms of the target patients based on the patient screening data for the decision sample. Clinical treatment indicators were analyzed for J1 treatment decisions in symptom-matched patients. Based on the analysis results, common time periods and quantitative intervals of J1 treatment decisions were obtained. Specifically as follows: A J1 treatment timeline is created by using the time point corresponding to the first visit of symptom-matched patients as the starting time point, and the historical periods of clinical treatment of symptom-matched patients by taking J1 treatment decisions are marked in the J1 treatment timeline to obtain multiple J1 decision treatment periods. The time intersection of the multiple J1 decision-making treatment periods is taken to obtain the common time periods of J1 treatment decisions; For each symptom-matched patient, the union of abnormal physiological indicators that appeared during the common treatment decision period of J1 was taken to obtain multiple abnormal indicators for patients. The obtained abnormal indicators for patients were labeled as P1 decision-related indicators, P2 decision-related indicators, P3 decision-related indicators, ... Pd decision-related indicators. It should be noted here that: In this application, P1, P2, P3, ..., Pd are the marker symbols corresponding to the abnormal indicators of patients.

[0030] It should be noted here that: In this application, the abnormal physiological indicators referred to herein are specifically quantitative parameters reflecting the physiological functional state of the human body that exceed the normal reference range for medically recognized healthy individuals, or indicators that, although within the range, show a significant abnormal trend in the short term.

[0031] Patients who received J1 treatment decisions and underwent historical treatment were acquired, resulting in multiple J1 treatment sample patients. Patients with abnormal P1 decision-related indicators were also acquired, resulting in multiple abnormal sample patients. The ratio of the number of abnormal sample patients to the number of J1 treatment sample patients was calculated to obtain the P1 indicator analysis weight. It should be noted here that: In this application, the J1 treatment sample patients involved are not limited to those with the same condition as the target; the J1 treatment sample patients involved are simply those who have received J1 treatment decisions.

[0032] Repeat the process of obtaining the analysis weights for the P1 indicator, and obtain the analysis weights for the P2 decision-related indicators to the Pd decision analysis indicators respectively, to obtain the analysis weights for the P2 indicators to the Pd indicators. Randomly select one sample of symptom-matched patients from the multiple symptom-matched patients, perform clinical indicator analysis on the sample of symptom-matched patients, and obtain the quantitative value of the J1 decision index corresponding to the sample of symptom-matched patients based on the analysis results. The historical periods in which J1 treatment decisions were made in clinical treatment were obtained to identify characteristic treatment periods. The characteristic treatment period is discretized into several indicator monitoring points. The indicator deviation of P1 decision-related indicators at each indicator monitoring point is obtained to obtain the time deviation of P1 indicators. The median of the normal indicator interval corresponding to the P1 decision-related indicators is collected to obtain the benchmark value of P1 indicators. The ratio of the time deviation of P1 indicators to the benchmark value of P1 indicators is calculated to obtain multiple P1 indicator deviation degrees. It should be noted here that: In this application, the P1 indicator benchmark value referred to herein is specifically the average of the upper limit and the lower limit of the normal indicator range corresponding to the P1 decision-related indicator.

[0033] The median deviation of the multiple P1 indicators is calculated to obtain the median deviation of the P1 indicators. The difference between each P1 indicator deviation and the median deviation of the P1 indicators is calculated, and the absolute value of the obtained difference is taken. The results are then renamed from the first indicator deviation to the qth indicator deviation in ascending order of the absolute value. It should be noted here that: In this application, q refers to the quantitative value corresponding to the deviation of the P1 index, and q is an integer greater than 0.

[0034] The standard deviation of the first indicator deviation and the second indicator deviation is calculated to obtain the standard deviation of the first indicator deviation. The standard deviation of the first indicator deviation to the third indicator deviation is calculated to obtain the standard deviation of the second indicator deviation. And so on, the standard deviation of the first indicator deviation to the qth indicator deviation is calculated to obtain the standard deviation of the (q-1)th indicator deviation. Compare the standard deviations of the first indicator deviation to the standard deviations of the (q-1)th indicator deviation, and set the standard deviation of the indicator deviation with the largest value as the peak value of the standard deviation of the indicator deviation. The average deviation of the indicators involved in the calculation of the peak standard deviation of the indicator deviation is calculated to obtain the comprehensive deviation value of the P1 indicator corresponding to the sample symptom-matched patients. It should be noted here that: If the index deviation degree used to calculate the peak value of the standard deviation of the index deviation is from the first index deviation degree to the jth index deviation, then the average value of the first index deviation degree to the jth index deviation is the comprehensive deviation value of index P1. Repeat the process of obtaining the comprehensive deviation value of the P1 indicator, and obtain the comprehensive deviation values ​​of the indicators corresponding to the decision-related indicators from P2 to Pd respectively, to obtain the comprehensive deviation degree of the P2 indicator to the comprehensive deviation degree of the Pd indicator. Calculate the product of the comprehensive deviation value of indicator P1 and the analysis weight of indicator P1 to obtain the weighted comprehensive deviation value of indicator P1. Calculate the product of the comprehensive deviation value of indicator P2 and the analysis weight of indicator P2 to obtain the weighted comprehensive deviation value of indicator P2. And so on, calculate the product of the comprehensive deviation value of indicator Pd and the analysis weight of indicator Pd to obtain the weighted comprehensive deviation value of indicator Pd. Calculate the average value of the weighted comprehensive deviation values ​​of indicators P1 to Pd to obtain the quantitative value of decision indicator J1. The J1 decision index quantification value corresponding to the symptom-matched patients in the repeated sample is obtained. The J1 decision index quantification value corresponding to each symptom-matched patient is obtained. The average value of the obtained J1 decision index quantification value is calculated to obtain the first indicator quantification feature value. The standard deviation value of the obtained J1 decision index quantification value is calculated to obtain the second indicator quantification feature value. The sum of the first indicator quantification feature value and the second indicator quantification feature value is used to create the upper limit of the J1 decision quantification interval. The difference between the first indicator quantification feature value and the second indicator quantification feature value is used to create the lower limit of the J1 decision quantification interval, thus obtaining the J1 decision quantification interval. Repeat the process of obtaining the common time period for J1 treatment decisions and the quantitative interval of J1 decisions, and obtain the common time period for J2 treatment decisions to the common time period for Jc treatment decisions and the quantitative interval of J2 decisions to the quantitative interval of Jc decisions, respectively. The common time period for treatment decisions from J1 to Jc, and the quantitative interval for decisions from J1 to Jc, are defined as the data for decision applicability analysis. The decision support module provides medical decision support to target patients based on decision applicability analysis data. Specifically as follows: Create a target treatment timeline by taking the time point corresponding to the first visit of the target patient as the starting time point, and mark the current time on the target treatment timeline to obtain the real-time decision time point; Obtain decision applicability analysis data, and based on the decision applicability analysis data, obtain the common time period for J1 treatment decisions to the common time period for Jc treatment decisions, as well as the quantitative range for J1 decisions to the quantitative range for Jc decisions; If the real-time decision-making time point falls within the common period of septic treatment decision-making, then the septic decision-making index quantification value is collected for the target patient. If the septic decision-making index quantification value is within the septic decision quantification range, then the septic treatment decision is pushed to the target patient through the AI ​​terminal.

[0035] Compared to the problems described in the background art, this invention obtains initial interactive text by collecting and analyzing the patient's medical condition text, performs medical indicator analysis on the initial interactive text, conducts preliminary screening of medical decisions involved in the medical interactive text based on the analysis results, obtains treatment decisions, identifies patients with shared decisions for treatment decisions, analyzes the symptoms of patients with shared decisions, screens patients with symptom matching based on the analysis results, and uses symptom-matching patients as decision support analysis samples. This effectively filters redundant information, ensures that decision analysis is based on real cases that highly match the symptoms of the target patient, and improves the accuracy and targeting of the preliminary screening of treatment decisions. Furthermore, this invention analyzes historical medical indicators for symptom-matched patients regarding their treatment decisions. Based on the analysis results, it obtains the common time periods and quantitative intervals for each treatment decision, and provides medical decision support to target patients based on these common time periods and quantitative intervals. This provides clear time-based guidance and quantitative reference for medical decisions for target patients, overcoming the limitations of existing technologies in accurately calculating complex clinical logic. It shifts medical decision-making from experience-based judgment to data-driven approaches, improving the scientific rigor and operability of decision support.

[0036] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A medical decision support method based on multimodal data, characterized in that, include: Step S1: Collect the target patient's medical condition text and interact with the consultation model to obtain the medical condition interaction strategy text. Extract the medical condition keywords and search for matching patients in the history of the medical condition and their historical diagnosis decisions. Use this to determine whether the sample medical decision is a medical condition treatment decision. If the historical diagnosis decision does not contain the sample medical decision, it is directly determined to be a non-medical condition treatment decision. If it does contain it, collect the target patient's symptom set and compare it with the historical patient's symptom set. Combine this with the decision contraindication symptoms for further analysis to obtain the preliminary medical decision data. Step S2: Obtain the treatment decision corresponding to the target patient's condition. Take the intersection of patients with the same decision for each decision to obtain multiple patients with shared decisions. Select patients with characteristic decisions from these patients. Collect the symptom sets of the target patient and the patients with characteristic decisions. Determine the common clinical symptoms and calculate the common ratio of the target symptoms. Set the monitoring period for the common clinical symptoms. Calculate the symptom onset duration ratio and draw the symptom onset curves for both. Iterate through the coordinate system to calculate the symptom index deviation at each point and then obtain the symptom deviation normalization ratio. Combine the common ratio of the target symptoms to calculate the symptom commonality deviation. Compare the deviation of all patients with shared decisions with the preset benchmark value to screen out patients with matching symptoms and patients with discrepancies, and obtain the decision sample patient screening data. In step S2, the specific steps are as follows: Step S21: Obtain preliminary medical decision data, acquire the treatment decisions corresponding to the target patient's condition, and take the intersection of patients with the same decision for each treatment decision to obtain multiple patients with shared decisions; Step S22: Randomly select a characteristic decision patient from the multiple shared decision patients, perform a disease commonality analysis between the characteristic decision patient and the target patient, and obtain the symptom commonality deviation degree corresponding to the characteristic decision patient; Step S23: Obtain the symptom commonality deviation degree corresponding to each patient with shared decision, set a preset benchmark value for symptom commonality deviation degree. If the symptom commonality deviation degree is less than or equal to the preset benchmark value for symptom commonality deviation degree, then the patients with shared decision are screened as symptom matching patients. If it is greater than the preset benchmark value, then the patients with shared decision are screened as symptom difference patients, thus obtaining the decision sample patient screening data. In step S22, the specific steps are as follows: Step S221: Collect clinical symptoms of the target patient during the course of the disease to obtain the target patient symptom set; collect clinical symptoms of the characteristic decision patient during the course of the disease to obtain the characteristic patient symptom set; obtain the common clinical symptoms of the target patient symptom set and the characteristic patient symptom set. Step S222: Calculate the ratio of the number of shared clinical symptoms to the number of symptoms present in the target patient's symptom set to obtain the target symptom shared ratio; Step S223: Randomly select one characteristic common symptom from the multiple common clinical symptoms obtained, perform characteristic common symptom onset analysis for the target patient, and draw the first common symptom onset polyline based on the analysis results; Step S224: Perform characteristic common symptom onset analysis for patients with characteristic decision-making, and draw a polyline of the second common symptom onset based on the analysis results; Step S225: Perform coordinate interaction analysis on the first common symptom onset line and the second common symptom onset line, and obtain the symptom deviation normalization ratio corresponding to the characteristic common symptoms based on the analysis results; Step S226: Obtain the symptom deviation normalization ratio corresponding to each common clinical symptom, and calculate the symptom commonality deviation degree corresponding to the feature decision patient by combining the symptom deviation normalization ratio and the target symptom commonality ratio. Step S3: Based on the patient screening data of the decision sample, conduct a decision applicability analysis on the disease treatment decision, and obtain decision applicability analysis data based on the analysis results; Step S4: Based on the decision applicability analysis data, extract the common time periods and quantitative intervals corresponding to treatment decisions for different conditions, and provide medical decision support for target patients according to the common time periods and quantitative intervals.

2. The medical decision support method based on multimodal data according to claim 1, characterized in that, In step S1, the specific steps are as follows: Step S11: Acquire patients who require medical decision support and select target patients from them; Step S12: Conduct a consultation with the target patient using an AI consultation model, and obtain the consultation results to get the patient's medical condition text; Step S13: Input the patient's condition text into the AI ​​interaction model in the form of medical consultation, and obtain the text of the response strategy output by the AI ​​interaction model to obtain the condition interaction strategy text; Step S14: Obtain the medical decisions provided by the disease interaction strategy text, randomly select one sample medical decision from the multiple medical decisions obtained, perform disease correlation analysis on the sample medical decision, and divide the sample medical decision into disease treatment decisions and non-disease treatment decisions based on the analysis results. Step S15: Divide each medical decision into disease treatment decisions and non-disease treatment decisions to obtain preliminary medical decision data.

3. The medical decision support method based on multimodal data according to claim 2, characterized in that, In step S14, the specific steps are as follows: Extract disease keywords from the interactive strategy text to obtain patient disease keywords, and use the patient disease keywords as search text to collect several historical patients with the same disease as the target patient. The actual medical decisions corresponding to patients in the past are obtained to obtain historical diagnostic decisions. If the obtained historical diagnostic decisions do not include sample medical decisions, the sample medical decisions are directly classified as non-disease treatment decisions. If the obtained historical diagnostic decisions include sample medical decisions, then select the historical patients who have received patients with sample medical decisions to obtain multiple patients with the same decision. Clinical symptoms are collected from target patients to obtain a set of target patient symptoms. Clinical symptoms of patients who made the same decision are also collected to obtain a set of historical patient symptoms. If the target patient's symptom set is a subset of the historical patient's symptom set, then the sample medical decision is classified as a disease treatment decision. If the target patient's symptom set is not a subset of the historical patient's symptom set, then clinical symptoms that exist in the target patient's symptom set but are not included in the historical patient's symptom set are acquired to obtain the symptoms to be assessed. The decision-contraindication symptoms corresponding to the sample medical decision are acquired. If the symptoms to be assessed contain decision-contraindication symptoms, then the sample medical decision is classified as a non-disease treatment decision. If they do not contain any, then the sample medical decision is classified as a disease treatment decision.

4. The medical decision support method based on multimodal data according to claim 1, characterized in that, In step S223, the specific steps are as follows: A first characteristic symptom monitoring period is established for the target patients. Within the first characteristic symptom monitoring period, the time points when the target patients develop characteristic common symptoms are obtained, resulting in Za symptom onset time points. The duration of symptom onset at the Z1 symptom onset time point of the target patient is obtained to obtain the Z1 symptom duration. The time interval between the Z1 symptom onset time point and the Z2 symptom onset time point is calculated to obtain the Z1 consecutive onset interval. The ratio of the Z1 symptom duration to the Z1 consecutive onset interval is calculated to obtain the Z1 symptom duration ratio. This process is repeated to obtain the Za-1 symptom duration ratio. The coordinate points with the symptom onset time on the horizontal axis and the symptom onset time ratio on the vertical axis are marked as symptom dominance coordinate points. The obtained multiple symptom dominance coordinate points are connected in sequence to obtain the first common symptom onset polyline.

5. A medical decision support method based on multimodal data according to claim 1, characterized in that, The specific steps in step S3 are as follows: Step S31: Obtain preliminary medical decision collection data, obtain the treatment decision corresponding to the target patient's condition based on the preliminary medical decision collection data, and mark the obtained treatment decisions as J1 treatment decision to Jc treatment decision respectively. Step S32: Obtain decision sample patient screening data, and obtain patients matching the symptoms corresponding to the target patients based on the decision sample patient screening data; Step S33: Perform clinical treatment index analysis on J1 treatment decision for symptom-matched patients, and obtain the common time period and quantitative interval of J1 treatment decision based on the analysis results; Step S34: Obtain the common time period for J2 treatment decisions to the common time period for Jc treatment decisions, as well as the quantitative range for J2 decisions to the quantitative range for Jc decisions; Step S35: Define the common treatment decision period from J1 to Jc and the quantitative interval of decision from J1 to Jc as decision applicability analysis data.

6. A medical decision support method based on multimodal data according to claim 5, characterized in that, In step S33, the specific steps are as follows: A J1 treatment timeline is created by taking the time point corresponding to the first visit of symptom-matched patients as the starting time point. The historical periods of clinical treatment of symptom-matched patients by J1 treatment decisions are marked in the J1 treatment timeline to obtain multiple J1 decision treatment periods. The time intersection of the multiple J1 decision-making treatment periods is used to obtain the common time periods of J1 treatment decisions; For each symptom-matched patient, the union of abnormal physiological indicators that appeared during the common period of treatment decision-making in J1 was taken to obtain the decision-related indicators from P1 to Pd. Patients who received J1 treatment decisions and underwent historical treatment were acquired, resulting in multiple J1 treatment sample patients. Patients with abnormal P1 decision-related indicators were also acquired, resulting in multiple abnormal sample patients. The ratio of the number of abnormal sample patients to the number of J1 treatment sample patients was calculated to obtain the P1 indicator analysis weight. The weights of the P2 decision-related indicators to the Pd decision-making analysis indicators are obtained by analyzing the weights of the P2 indicators to the Pd indicators. Select sample symptom-matching patients from the multiple symptom-matching patients obtained, perform clinical indicator analysis on the sample symptom-matching patients, and obtain the quantitative value of the J1 decision index corresponding to the sample symptom-matching patients. Obtain the J1 decision index quantification value corresponding to each symptom-matched patient, and create a J1 decision quantification interval based on the J1 decision index quantification value.

7. A medical decision support method based on multimodal data according to claim 6, characterized in that, In step S33, the specific steps are as follows: The historical time periods during which J1 treatment decisions were made for clinical treatment of patients matched for sample symptoms were obtained to obtain characteristic treatment periods; The characteristic treatment period is discretized into several indicator monitoring points. The indicator deviation of P1 decision-related indicators at each indicator monitoring point is obtained to obtain the time deviation of P1 indicators. The median of the normal indicator interval corresponding to the P1 decision-related indicators is collected to obtain the benchmark value of P1 indicators. The ratio of the time deviation of P1 indicators to the benchmark value of P1 indicators is calculated to obtain multiple P1 indicator deviation degrees. Numerical analysis of the deviation of the P1 index was performed to obtain the comprehensive deviation value of the P1 index. Obtain the comprehensive deviation value of the indicators corresponding to the decision-related indicators of P2 to Pd, and obtain the comprehensive deviation degree of the indicators of P2 to Pd. Calculate the product of the comprehensive deviation value of indicator P1 and the analysis weight of indicator P1 to obtain the weighted comprehensive deviation value of indicator P1. Similarly, calculate the product of the comprehensive deviation value of indicator Pd and the analysis weight of indicator Pd to obtain the weighted comprehensive deviation value of indicator Pd. Calculate the average value of the weighted comprehensive deviation values ​​of indicators P1 to Pd to obtain the quantitative value of indicator J1.

8. A medical decision support system based on multimodal data, applicable to the medical decision support method based on multimodal data as described in any one of claims 1-7, characterized in that, The medical decision support system includes: Data acquisition module: Collects textual information about the patient's condition and interacts with the consultation model to obtain textual information about the patient's condition interaction strategy. Analyzes the textual information about the patient's condition interaction strategy using medical indicators and obtains preliminary data for medical decision-making based on the analysis results. Data Analysis Module: Based on the preliminary data collected for medical decision-making, the module identifies patients with shared decision-making criteria for disease treatment, performs symptom analysis on these patients, and then uses the analysis results to screen patients with shared decision-making criteria by matching their symptoms, thus obtaining patient screening data for the decision sample. Applicability Analysis Module: Based on the patient screening data of the decision sample, this module performs decision applicability analysis on the treatment decisions and obtains decision applicability analysis data based on the analysis results. Decision support module: Based on decision applicability analysis data, it extracts common time periods and quantitative intervals corresponding to treatment decisions for different conditions, and provides medical decision support for target patients based on the common time periods and quantitative intervals.

Images