A Knowledge Graph Efficacy-Assisted Supervision System and Method Based on Multimodal Driving

Abstract

The present invention discloses a knowledge graph-assisted efficacy supervision system and method based on multi-modal driving, belonging to the technical field of efficacy-assisted supervision; constructing a first entity set and a first relationship set between patients and electronic medical records, constructing a second entity set and a second relationship set between patients and B-ultrasound images to form a patient personalized knowledge graph; attaching a time series cycle label to the lesions, denoted as cycle lesions; calculating the efficacy prediction value of the patient; calculating the lesion change rate of the cycle lesions; calculating the comprehensive efficacy evaluation value of the patient; calculating the optimized efficacy prediction value between different patients; calculating the optimal treatment plan; by using image analysis technology to extract lesion information, the present invention constructs a dynamic association between the condition and the treatment plan through a knowledge graph, quantifies the lesion change and efficacy prediction; finally, optimizes the treatment plan through the analysis of the efficacy similarity between patients, recommends the optimal treatment plan for patients, thereby improving the accuracy of medical decision-making and the scientific nature of the treatment effect.

Description

Technical Field

[0001] The present invention relates to the technical field of auxiliary supervision of treatment effects, and specifically to a knowledge graph-based auxiliary supervision system and method for treatment effects driven by multi-modalities. Background Art

[0002] In recent years, with the rapid development of medical informatization technology, the integrated application of electronic medical record (EMR) systems and medical image data has provided new opportunities for patient health management and treatment effect evaluation; as an important tool in the field of artificial intelligence, knowledge graph technology effectively represents the complex relationships between entities by constructing a semantic network of multi-dimensional data and has been widely applied in the medical field; especially during the treatment process, the development of a patient's condition and data for evaluating treatment effects often come from various types of data sources, including EMR texts, imaging data such as B-ultrasound; through the fusion of these multi-modal data, doctors can comprehensively grasp the patient's treatment process and thus make more accurate treatment decisions; however, most current treatment effect evaluation methods still rely on single-modal data, lacking dynamic tracking and comprehensive analysis of changes in the condition, and it is difficult to comprehensively and timely evaluate the treatment effects of patients.

[0003] The deficiencies in the prior art are mainly manifested in several aspects; firstly, most treatment effect evaluation systems are limited to the analysis based on text data and it is difficult to effectively integrate medical image data, resulting in the spatial changes of the patient's lesions not being fully reflected in the evaluation results; secondly, the prior art mostly relies on static historical data, lacking dynamic tracking and analysis of the changes in the patient's condition over time, and it is difficult to effectively predict the future development of the condition and treatment effects; in addition, the individual differences between patients have not been fully considered, resulting in insufficient personalization of treatment plans and affecting the accuracy and scientific nature of clinical treatment. Summary of the Invention

[0004] The purpose of the present invention is to provide a knowledge graph-based auxiliary supervision system and method for treatment effects driven by multi-modalities to solve the problems raised in the above background art.

[0005] To solve the above technical problems, the present invention provides the following technical solutions:

[0006] A method for auxiliary supervision of the efficacy of a knowledge graph based on multi-modal driving. This method includes the following steps: establishing a patient database, constructing an electronic medical record data set and a B-ultrasound image data set of patients; extracting entities and relationships between patients and electronic medical records; constructing a first entity set and a first relationship set between patients and electronic medical records, and constructing an initial knowledge graph; using image analysis technology to extract lesion regions in B-ultrasound images, and constructing a second entity set and a second relationship set between patients and B-ultrasound images; adding the second entity set and the second relationship set to the initial knowledge graph to form a patient personalized knowledge graph; attaching a time series cycle label to the lesion, denoted as a cyclic lesion; calculating the efficacy prediction value of the patient; calculating the lesion change rate of the cyclic lesion; based on the efficacy prediction value and the lesion change rate, calculating the comprehensive efficacy evaluation value of the patient; based on the comprehensive efficacy evaluation value, constructing an efficacy evaluation set and updating it to the patient personalized knowledge graph; calculating the optimized efficacy prediction values between different patients; based on the optimized efficacy prediction values, calculating the optimal treatment plan.

[0007] As a preferred solution of the method for auxiliary supervision of the efficacy of a knowledge graph based on multi-modal driving described in the present invention, establish a patient database, denoted as HZ = {HZ a |a ∈ [1, A]}, where HZ a represents the a-th patient, and A represents the total number of patients; obtain the electronic medical record data and B-ultrasound image data of the patient HZ a , and respectively construct an electronic medical record data set DB = {DB a |a ∈ [1, A]} and a B-ultrasound image data set BC = {BC a |a ∈ [1, A]}, where DB a represents the electronic medical record data of the a-th patient, and BC a represents the B-ultrasound image data of the a-th patient; the electronic medical record data includes the disease symptoms, treatment plan and medications of the patient HZ a .

[0008] Based on the electronic medical record data set and the patient database, extract the entities and relationships between patients and electronic medical records. The entities include the patient HZ a , the disease symptom S i , the treatment plan T b and the medication D c , where S i represents the i-th disease symptom, T b represents the b-th treatment plan, and D c represents the c-th medication; the relationships include having, receiving and using; the having represents the having relationship between the patient HZ a and the disease symptom S i , and the receiving represents the receiving relationship between the patient HZ a and the treatment plan Tb The acceptance relationship between them, where the usage indicates that the patient HZ a and the drug D c The usage relationship between them.

[0009] As a preferred solution of the method for auxiliary supervision of the curative effect of the knowledge graph based on multi-modal driving described in the present invention, a first entity set and a first relationship set between the patient and the electronic medical record are constructed, which are respectively denoted as the first entity set FE = {patient HZ a , diseased symptom S i , treatment plan T b , drug D c}, and the first relationship set FR = {suffering from, accepting, using}. Based on the first entity set E and the first relationship set R, an initial knowledge graph is constructed, denoted as G = {FE, FR}.

[0010] Using image analysis technology to extract the lesion area L in the B-ultrasound image, and record the location and size of the lesion. The lesion area refers to the abnormal area of tissues and organs shown in the B-ultrasound image; construct a second entity set and a second relationship set between the patient and the B-ultrasound image, which are respectively denoted as the second entity set SE = {patient HZ a , lesion location, lesion size} and the second relationship set SR = {belonging to, located at, measured value}. The belonging means that the lesion belongs to the patient HZ a , the located at means that the lesion is located at the lesion location in the B-ultrasound image, and the measured value means the measured lesion size of the lesion.

[0011] As a preferred solution of the method for auxiliary supervision of the curative effect of the knowledge graph based on multi-modal driving described in the present invention, add the second entity set and the second relationship set to the initial knowledge graph G = {FE, FR} to form the final patient personalized knowledge graph, denoted as HG = {FE ∪ SE, FR ∪ SR}; construct a time series, denoted as SJ = {t|t ∈ [1, T]}, where t represents the t-th time point and T represents the total number of time points, and attach a time series cycle label to the lesion L, denoted as the periodic lesion L t .

[0012] Calculate the curative effect prediction value P(HZ a , S a , T i , D b ) of the patient HZ c ), and the calculation formula is as follows:

[0013] P(HZ a , S i , T b , D c ) = α1·f(S i ) + α2·f(Tb ) + α3·f(D c );

[0014]

[0015] f(D c ) = d c ·p c ·e c ;

[0016] where f(S i ) represents the characteristic function of the disease symptoms, f(T b ) represents the characteristic function of the treatment plan, f(D c ) represents the drug characteristic function, α1, α2, and α3 respectively represent the preset weights of the disease symptoms, treatment plan, and drug, s i represents the severity of the disease symptom S i , which is determined according to the doctor's rating standard, d i represents the duration of the symptom, f i represents the frequency of the symptom occurrence, I b represents the intensity of the treatment plan, E b represents the suitability of the treatment method, which is scored according to the matching degree between the treatment plan and the patient's indications, D b represents the duration of the treatment course, d c represents the dose of the drug D c , p c represents the frequency of drug use, e c represents the efficacy score of the drug set based on clinical data.

[0017] In the present invention, this formula is a weighted sum based on the weights of the patient's symptoms, treatment plan, and drug; by quantifying the severity, duration, frequency of the disease symptoms, the intensity, suitability, and treatment course of the treatment plan, and the dose, frequency, and clinical effect of the drug, a comprehensive efficacy prediction value is obtained; the core function of this formula is to quantify the influence of the patient's current symptoms, treatment plan, and drug on the treatment effect; by converting the text information (such as symptoms, treatment plan, drug information) in the electronic medical record into quantifiable features, and then combining the corresponding weights, the efficacy prediction value of each patient is calculated; this value can be used as the basis for evaluating the effectiveness of the treatment plan and provides support for subsequent efficacy evaluation and optimization.

[0018] Calculate the lesion change rate ΔL t of the periodic lesion L t at the time point t, and the calculation formula is as follows:

[0019]

[0020] where ΔLt Indicates the periodic lesion L t The lesion change rate, where L1 represents the initial lesion size before treatment, and t≠1.

[0021] In the present invention, this formula is used to measure the degree of shrinkage of the lesion during the treatment process and reflect the treatment effect. The ratio of the difference between the initial lesion size and the current lesion size to the initial lesion size represents the lesion change rate; its main function is to dynamically track the changes of the patient's lesion and quantify the degree of shrinkage or deterioration of the lesion during the treatment process; by combining the lesion size in the time series data, doctors can intuitively see the changes in the lesion before and after treatment; this indicator helps to evaluate the actual performance of the treatment effect, especially in the case where image data analysis needs to be combined, and plays a role in dynamically monitoring the progress of the disease.

[0022] Based on the efficacy prediction value P(HZ a ,S i ,T b ,D c ) and the lesion change rate ΔL t , calculate the comprehensive efficacy evaluation value R(HZ a ,S a ,T i ,D b ,t) of the patient HZ c , and the calculation formula is as follows:

[0023] R(HZ a ,S i ,T b ,D c ,t) = β·P(HZ a ,S i ,T b ,D c ) + γ·ΔL t ;

[0024] Among them, β and γ represent preset parameters, which are adjusted according to different diseases, symptoms and treatment stages.

[0025] In the present invention, the comprehensive efficacy evaluation value combines the efficacy prediction value with the lesion change rate to form an evaluation index for the overall efficacy; it fuses data from different sources (electronic medical record data and image data) to generate a more comprehensive efficacy evaluation; this index can reflect the overall health status of the patient and provide doctors with the efficacy analysis results based on multiple dimensions to support medical decision-making.

[0026] As a preferred solution of the method for auxiliary supervision of the efficacy of the knowledge graph based on multi-modal drive described in the present invention, based on the comprehensive efficacy evaluation value R(HZ a ,S i ,T b, D c , t), construct the efficacy evaluation set PG = {[S i , T b , D c , R(HZ a , S i , T b , D c , t)]}, and update the efficacy evaluation set PG to the patient's personalized knowledge graph HG. The update formula is: TG = HG ∪ PG, where TG represents the updated patient's personalized knowledge graph.

[0027] Based on the updated patient's personalized knowledge graph TG, pre - calculate the similarity between patient HZ a and patient HZ o , denoted as sim(HZ a , HZ o ), where HZ o represents the o - th patient, and o ≠ a; based on the similarity sim(HZ a and patient HZ o ), calculate the optimized efficacy prediction value of patient HZ a and patient HZ o . The calculation formula is as follows: a o o i

[0028]

[0029] where R(HZ o , S i , T b , D c , t) represents the comprehensive efficacy evaluation value of patient HZ o , and R opt (HZ a , S i ) represents the optimized efficacy prediction value of patient HZ a and patient HZ o .

[0030] In the present invention, this formula is calculated through the similarity of efficacy and the comprehensive efficacy evaluation value between different patients, providing more personalized efficacy prediction for patients; through the similarity analysis of cross - patient data, the accuracy of efficacy prediction is improved, providing a more personalized treatment evaluation for patients; by referring to the treatment effects of other similar patients, this formula can provide a more valuable reference analysis for the efficacy prediction of the current patient.

[0031] Based on the optimized efficacy prediction value R a of patient HZ o and patient HZ opt(HZ a ,S i ), calculate the optimal treatment plan, and the calculation formula is as follows:

[0032] B(T b ,D c ) = argmax Tb,Dc [R opt (HZ a ,S i )];

[0033] Among them, B(T b ,D c ) represents the optimal treatment plan.

[0034] If the optimized efficacy prediction value R a of patient HZ o is greater than the comprehensive efficacy evaluation value R(HZ opt (HZ a ,S i ) of patient HZ a , then recommend the optimal treatment plan B(T a ,S i ,T b ,D c ,t) to patient HZ a . b ,D c ) to patient HZ

[0035] In the present invention, this formula selects the plan that can maximize the efficacy by comparing the optimized efficacy prediction values of multiple treatment plans; according to the optimized efficacy prediction value, it automatically recommends the optimal treatment plan, which can help doctors select the best-effective plan from multiple treatment plans, improving the scientificity and accuracy of medical decisions; with the help of this formula, doctors can formulate personalized treatment plans for patients, thereby improving the treatment effect.

[0036] A multi-modal driven knowledge graph efficacy assisted supervision system, which includes: a patient data management module, a knowledge graph construction module, an efficacy evaluation module, and a personalized efficacy optimization module.

[0037] The patient data management module: establishes a patient database, constructs the electronic medical record data set and B-ultrasound image data set of the patient; extracts the entities and relationships between the patient and the electronic case.

[0038] The knowledge graph construction module: constructs the first entity set and the first relationship set between the patient and the electronic case, and constructs an initial knowledge graph; uses image analysis technology to extract the lesion area in the B-ultrasound image, and constructs the second entity set and the second relationship set between the patient and the B-ultrasound image.

[0039] The efficacy evaluation module: Add the second entity set and the second relationship set to the initial knowledge graph to form a patient personalized knowledge graph; attach a time series cycle label to the lesion, denoted as the cycle lesion; calculate the efficacy prediction value of the patient; calculate the lesion change rate of the cycle lesion; calculate the comprehensive efficacy evaluation value of the patient based on the efficacy prediction value and the lesion change rate.

[0040] The personalized efficacy optimization module: Based on the comprehensive efficacy evaluation value, construct an efficacy evaluation set and update it to the patient personalized knowledge graph; calculate the optimized efficacy prediction values between different patients; calculate the optimal treatment plan based on the optimized efficacy prediction values.

[0041] Furthermore, the patient data management module includes a database construction unit and an entity and relationship extraction unit.

[0042] The database construction unit: Establish a patient database, denoted as HZ = {HZ a |a ∈ [1, A]}, where HZ a represents the a-th patient and A represents the total number of patients; obtain the electronic medical record data and B-ultrasound image data of patient HZ a and respectively construct an electronic medical record data set DB = {DB a |a ∈ [1, A]} and a B-ultrasound image data set BC = {BC a |a ∈ [1, A]}, where DB a represents the electronic medical record data of the a-th patient and BC a represents the B-ultrasound image data of the a-th patient; the electronic medical record data includes the disease symptoms, treatment plan, and medications of patient HZ a .

[0043] The entity and relationship extraction unit: Based on the electronic medical record data set and the patient database, extract the entities and relationships between the patient and the electronic medical record. The entities include patient HZ a , disease symptom S i , treatment plan T b , and medication D c , where S i represents the i-th disease symptom, T b represents the b-th treatment plan, and D c represents the c-th medication; the relationships include having, receiving, and using; having represents the having relationship between patient HZ a and disease symptom S i , receiving represents the receiving relationship between patient HZ a and treatment plan T b , and using represents the using relationship between patient HZ a and medication D c .

[0044] Further, the knowledge graph construction module includes a first entity set and first relationship set construction unit and a second entity set and second relationship set construction unit.

[0045] The first entity set and first relationship set construction unit: constructs a first entity set and a first relationship set between patients and electronic medical records, denoted as the first entity set FE = {patient HZ a , disease symptoms S i , treatment plan T b , drugs D c}, and a first relationship set FR = {suffering from, receiving, using}, and constructs an initial knowledge graph based on the first entity set E and the first relationship set R, denoted as G = {FE, FR}.

[0046] The second entity set and second relationship set construction unit: uses image analysis technology to extract the lesion area L in the B-ultrasound image, and records the location and size of the lesion. The lesion area refers to the abnormal area of tissues and organs shown in the B-ultrasound image; constructs a second entity set and a second relationship set between patients and B-ultrasound images, denoted as the second entity set SE = {patient HZ a , lesion location, lesion size}, and a second relationship set SR = {belonging to, located at, measurement value}, where "belonging to" means that the lesion belongs to patient HZ a , "located at" means that the lesion is located at the lesion location in the B-ultrasound image, and "measurement value" means the measured lesion size of the lesion.

[0047] Further, the curative effect evaluation module includes a curative effect prediction and lesion change rate calculation unit and a comprehensive curative effect evaluation value calculation unit.

[0048] The curative effect prediction and lesion change rate calculation unit: adds the second entity set and the second relationship set to the initial knowledge graph G = {FE, FR} to form a final patient personalized knowledge graph, denoted as HG = {FE ∪ SE, FR ∪ SR}; constructs a time series, denoted as SJ = {t|t ∈ [1, T]}, where t represents the t-th time point and T represents the total number of time points, and attaches a time series period label to the lesion L, denoted as periodic lesion L t ; calculates the curative effect prediction value P(HZ a , S a , T i , D b , D c ).

[0049] The comprehensive curative effect evaluation value calculation unit: calculates the lesion change rate ΔL of the periodic lesion L t at time point t; based on the curative effect prediction value P(HZ t ​a , S i , T b , D c ), and the lesion change rate ΔL t , calculate the comprehensive efficacy evaluation value R(HZ a ) of the patient HZ a , S i , T b , D c , t).

[0050] Furthermore, the personalized efficacy optimization module includes a knowledge graph update unit and an optimal treatment plan recommendation unit.

[0051] The knowledge graph update unit: Based on the comprehensive efficacy evaluation value R(HZ a , S i , T b , D c , t), construct an efficacy evaluation set PG = {[S i , T b , D c , R(HZ a , S i , T b , D c , t)]}, update the efficacy evaluation set PG to the patient's personalized knowledge graph HG, and the update formula is: TG = HG ∪ PG, where TG represents the updated patient's personalized knowledge graph; based on the updated patient's personalized knowledge graph TG, pre-calculate the similarity between patient HZ a and patient HZ o , denoted as sim(HZ a , HZ o ), where HZ o represents the o-th patient, and o ≠ a; based on the similarity sim(HZ a and patient HZ o , calculate the optimized efficacy prediction value R a of patient HZ o and patient HZ a o opt (HZ a , S i ).

[0052] The optimal treatment plan recommendation unit: Based on the optimized efficacy prediction value R a of patient HZ o and patient HZ opt (HZ a , S i ), calculate the optimal treatment plan B(T b ​​, D c ); If the patient HZ a and the patient HZ o 's optimized efficacy prediction value R opt (HZ a , S i ) is greater than the patient HZ a 's comprehensive efficacy evaluation value R(HZ a , S i , T b , D c , t), then the optimal treatment plan B(T a , D b ) is recommended for the patient HZ c .

[0053] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: In a multi-modal driven knowledge graph efficacy assisted supervision system and method provided by the present invention, by integrating electronic medical records, B-ultrasound images and time series data, a personalized knowledge graph is constructed to achieve a comprehensive evaluation of the patient's efficacy; First, collect and structure the patient's multi-modal data to lay a foundation for subsequent analysis; Secondly, use image analysis technology to extract lesion information, and construct a dynamic association between the condition and the treatment plan through the knowledge graph to quantify the lesion change and efficacy prediction; Finally, optimize the treatment plan through the analysis of the similarity of efficacy between patients, and recommend the optimal treatment plan for patients, thereby improving the accuracy of medical decision-making and the scientific nature of the treatment effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention.

[0055] Figure 1 is a schematic diagram of the steps of a multi-modal driven knowledge graph efficacy assisted supervision method of the present invention;

[0056] Figure 2 is a schematic diagram of the structure of a multi-modal driven knowledge graph efficacy assisted supervision system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0058] Please refer to Figure 1, in the first embodiment: A knowledge graph efficacy-assisted supervision method based on multi-modal driving is provided, and the method includes the following steps:

[0059] Step S1: Establish a patient database, construct an electronic medical record data set and a B-ultrasound image data set of patients; extract entities and relationships between patients and electronic medical records.

[0060] Specifically, establish a patient database, denoted as HZ = {HZ a |a ∈ [1, A]}, where HZ a represents the a-th patient, and A represents the total number of patients; obtain the electronic medical record data and B-ultrasound image data of patient HZ a , and respectively construct an electronic medical record data set DB = {DB a |a ∈ [1, A]} and a B-ultrasound image data set BC = {BC a |a ∈ [1, A]}, where DB a represents the electronic medical record data of the a-th patient, and BC a represents the B-ultrasound image data of the a-th patient; the electronic medical record data includes the disease symptoms, treatment plans, and medications of patient HZ a .

[0061] Furthermore, based on the electronic medical record data set and the patient database, extract entities and relationships between patients and electronic medical records. The entities include patient HZ a , disease symptom S i , treatment plan T b , and medication D c , where S i represents the i-th disease symptom, T b represents the b-th treatment plan, and D c represents the c-th medication; the relationships include having, receiving, and using; having means the having relationship between patient HZ a and disease symptom S i , receiving means the receiving relationship between patient HZ a and treatment plan T b , and using means the using relationship between patient HZ a and medication D c .

[0062] Step S2: Construct a first entity set and a first relationship set between patients and electronic medical records, and construct an initial knowledge graph; use image analysis technology to extract the lesion area in the B-ultrasound image, and construct a second entity set and a second relationship set between patients and B-ultrasound images.

[0063] Specifically, construct a first entity set and a first relationship set between patients and electronic medical records, denoted as the first entity set FE = {patient HZa , the disease symptom S i , the treatment plan T b , the drug D c}, and the first relationship set FR = {suffering from, receiving, using}. Based on the first entity set E and the first relationship set R, an initial knowledge graph is constructed, denoted as G = {FE, FR}.

[0064] Furthermore, image analysis technology is used to extract the lesion area L in the B-ultrasound image, and the position and size of the lesion are recorded. The lesion area refers to the abnormal area of tissues and organs shown in the B-ultrasound image; a second entity set and a second relationship set between the patient and the B-ultrasound image are constructed, denoted as the second entity set SE = {patient HZ a , lesion position, lesion size} and the second relationship set SR = {belonging to, located at, measurement value}, where "belonging to" means the lesion belongs to the patient HZ a , "located at" means the lesion is located at the lesion position in the B-ultrasound image, and "measurement value" means the measured lesion size of the lesion.

[0065] Step S3: Add the second entity set and the second relationship set to the initial knowledge graph to form a patient personalized knowledge graph; attach a time series cycle label to the lesion, denoted as the periodic lesion; calculate the efficacy prediction value of the patient; calculate the lesion change rate of the periodic lesion; based on the efficacy prediction value and the lesion change rate, calculate the comprehensive efficacy evaluation value of the patient.

[0066] Specifically, add the second entity set and the second relationship set to the initial knowledge graph G = {FE, FR} to form the final patient personalized knowledge graph, denoted as HG = {FE ∪ SE, FR ∪ SR}; construct a time series, denoted as SJ = {t|t ∈ [1, T]}, where t represents the t-th time point and T represents the total number of time points, and attach a time series cycle label to the lesion L, denoted as the periodic lesion L t .

[0067] Furthermore, calculate the efficacy prediction value P(HZ a , S a , T i , D b ) of the patient HZ c . The calculation formula is as follows:

[0068] P(HZ a , S i , T b , D c ) = α1·f(S i ) + α2·f(T b ) + α3·f(D c );

[0069]

[0070] f(D c ) = d c ·p c ·e c ;

[0071] Among them, f(S i ) represents the characteristic function of the disease symptoms, f(T b ) represents the characteristic function of the treatment plan, f(D c ) represents the drug characteristic function, α1, α2, and α3 respectively represent the preset weights of the disease symptoms, treatment plan, and drug, s i represents the severity of the disease symptoms S i , which is determined according to the doctor's rating standard, d i represents the duration of the symptoms, f i represents the occurrence frequency of the symptoms, I b represents the intensity of the treatment plan, E b represents the suitability of the treatment method, which is scored according to the matching degree between the treatment plan and the patient's indications, D b represents the duration of the treatment course, d c represents the dose of the drug D c , p c represents the usage frequency of the drug, e c represents the efficacy score of the drug set based on clinical data.

[0072] For example, assume that the severity s1 is 3, the duration of the symptoms d1 is 2, the frequency f1 is 4, the intensity of the treatment plan I1 is 1.5, the suitability E1 is 0.9, the duration of the treatment course D1 is 10, the dose d1 is 2, the usage frequency of the drug p1 is 3, the efficacy score e1 is 0.8, and α1, α2, and α3 are 0.4, 0.3, and 0.3 respectively. Substituting into the formula for calculation, we get f(D1) = 2 * 3 * 0.8 = 4.8, and the efficacy prediction value P(HZ1, S i , T b , D c ) = 0.4 * 1.2 + 0.3 * 0.14 + 0.3 * 4.8 = 0.48 + 0.042 + 1.44 = 1.962.

[0073] When calculating at time point t, the lesion change rate ΔL t of the periodic lesion L t is calculated as follows:

[0074]

[0075] Among them, ΔLt Indicates the periodic lesion L t The lesion change rate, L1 represents the initial lesion size before treatment, and t≠1.

[0076] For example, assume that the initial lesion size L1 before treatment is 10, and the periodic lesion L t is 6. Substituting into the formula, the lesion change rate ΔL of the periodic lesion L t is calculated to be 0.4. t is 0.4.

[0077] Furthermore, based on the efficacy prediction value P(HZ a ,S i ,T b ,D c ) and the lesion change rate ΔL t , calculate the comprehensive efficacy evaluation value R(HZ a ,S a ,T i ,D b ,t) of the patient HZ c . The calculation formula is as follows:

[0078] R(HZ a ,S i ,T b ,D c ,t) = β·P(HZ a ,S i ,T b ,D c ) + γ·ΔL t ;

[0079] Among them, β and γ represent preset parameters, which are adjusted according to different conditions, symptoms and treatment stages.

[0080] For example, assume that β and γ are 0.7 and 0.3 respectively. Substituting into the formula, the comprehensive efficacy evaluation value R(HZ1,S i ,T b ,D c ,t) of the patient HZ1 is calculated to be 0.7*1.962 + 0.3*0.4 = 1.37 + 0.12 = 1.49.

[0081] Step S4: Based on the comprehensive efficacy evaluation value, construct an efficacy evaluation set and update it to the patient's personalized knowledge graph; calculate the optimized efficacy prediction value between different patients; based on the optimized efficacy prediction value, calculate the optimal treatment plan.

[0082] Specifically, based on the comprehensive efficacy evaluation value R(HZ a ,S i ,T b ,D c,t), construct the efficacy evaluation set PG = {[S i ,T b ,D c ,R(HZ a ,S i ,T b ,D c ,t)]}, and update the efficacy evaluation set PG to the patient's personalized knowledge graph HG. The update formula is: TG = HG ∪ PG, where TG represents the updated patient's personalized knowledge graph.

[0083] Furthermore, based on the updated patient's personalized knowledge graph TG, through the similarity between the preset patient HZ a and the patient HZ o , denoted as sim(HZ a ,HZ o ), where HZ o represents the o-th patient, and o ≠ a; based on the similarity sim(HZ a and the patient HZ o , calculate the optimized efficacy prediction value of the patient HZ a and the patient HZ o . The calculation formula is as follows: a o opt R

[0084] (HZ a ,S i ) = sim(HZ a ,HZ o )·R(HZ o ,S i ,T b ,D c ,t);

[0085]

[0086] where R(HZ o ,S i ,T b ,D c ,t) represents the comprehensive efficacy evaluation value of the patient HZ o , and R opt (HZ a ,S i ) represents the optimized efficacy prediction value of the patient HZ a and the patient HZ o .

[0086] For example, assume that the similarity sim(HZ1, HZ2) between patient HZ1 and patient HZ2 is 1.5, and R(HZ2, S i ,T b ,D c, t) is 1.6, substituting it into the formula to calculate the optimized curative effect prediction value R opt (HZ a , S i ) = 1.5 * 1.6 = 2.4.

[0087] Furthermore, based on the patient HZ a and the patient HZ o 's optimized curative effect prediction value R opt (HZ a , S i ), calculate the optimal treatment plan, and the calculation formula is as follows:

[0088]

[0089] Among them, B(T b , D c ) represents the optimal treatment plan.

[0090] If the optimized curative effect prediction value R a of the patient HZ o and the patient HZ opt (HZ a , S i ) is greater than the comprehensive curative effect evaluation value R(HZ a ) of the patient HZ a , S i , T b , D c , t), then recommend the optimal treatment plan B(T a , D b ) for the patient HZ c .

[0091] Please refer to Figure 2 , in the second embodiment: Provide a knowledge graph efficacy-assisted supervision system based on multi-modal driving, and the system includes: a patient data management module, a knowledge graph construction module, an efficacy evaluation module, and a personalized efficacy optimization module.

[0092] The patient data management module: Establish a patient database, construct an electronic medical record data set and a B-ultrasound image data set of the patient; Extract the entities and relationships between the patient and the electronic case.

[0093] The knowledge graph construction module: Construct a first entity set and a first relationship set between the patient and the electronic case, and construct an initial knowledge graph; Use image analysis technology to extract the lesion area in the B-ultrasound image, and construct a second entity set and a second relationship set between the patient and the B-ultrasound image.

[0094] The efficacy evaluation module: Add the second entity set and the second relationship set to the initial knowledge graph to form a patient personalized knowledge graph; attach a time series cycle label to the lesion, denoted as the cycle lesion; calculate the efficacy prediction value of the patient; calculate the lesion change rate of the cycle lesion; calculate the comprehensive efficacy evaluation value of the patient based on the efficacy prediction value and the lesion change rate.

[0095] The personalized efficacy optimization module: Construct an efficacy evaluation set based on the comprehensive efficacy evaluation value and update it to the patient personalized knowledge graph; calculate the optimized efficacy prediction values between different patients; calculate the optimal treatment plan based on the optimized efficacy prediction values.

[0096] Furthermore, the patient data management module includes a database construction unit and an entity and relationship extraction unit.

[0097] The database construction unit: Establish a patient database, denoted as HZ = {HZ a |a ∈ [1, A]}, where HZ a represents the a-th patient and A represents the total number of patients; obtain the electronic medical record data and B-ultrasound image data of the patient HZ a and respectively construct an electronic medical record data set DB = {DB a |a ∈ [1, A]} and a B-ultrasound image data set BC = {BC a |a ∈ [1, A]}, where DB a represents the electronic medical record data of the a-th patient and BC a represents the B-ultrasound image data of the a-th patient; the electronic medical record data includes the disease symptoms, treatment plan, and medications of the patient HZ a .

[0098] The entity and relationship extraction unit: Based on the electronic medical record data set and the patient database, extract the entities and relationships between the patient and the electronic medical record. The entities include the patient HZ a , the disease symptom S i , the treatment plan T b , and the medication D c , where S i represents the i-th disease symptom, T b represents the b-th treatment plan, and D c represents the c-th medication; the relationships include has, receives, and uses; the has relationship represents the relationship between the patient HZ a and the disease symptom S i , the receives relationship represents the relationship between the patient HZ a and the treatment plan T b , and the uses relationship represents the relationship between the patient HZ a and the medication D c .

[0099] Further, the knowledge graph construction module includes a first entity set and first relationship set construction unit and a second entity set and second relationship set construction unit.

[0100] The first entity set and first relationship set construction unit: constructs a first entity set and a first relationship set between patients and electronic medical records, denoted as the first entity set FE = {patient HZ a , disease symptoms S i , treatment plan T b , drugs D c}, and a first relationship set FR = {suffering from, receiving, using}. Based on the first entity set E and the first relationship set R, an initial knowledge graph is constructed, denoted as G = {FE, FR}.

[0101] The second entity set and second relationship set construction unit: uses image analysis technology to extract the lesion area L in the B-ultrasound image, and records the location and size of the lesion. The lesion area refers to the abnormal area of tissues and organs shown in the B-ultrasound image; constructs a second entity set and a second relationship set between patients and B-ultrasound images, denoted as the second entity set SE = {patient HZ a , lesion location, lesion size}, and a second relationship set SR = {belonging to, located at, measurement value}. The "belonging to" indicates that the lesion belongs to the patient HZ a , the "located at" indicates that the lesion is located at the lesion location in the B-ultrasound image, and the "measurement value" indicates the measured lesion size of the lesion.

[0102] Further, the treatment effect evaluation module includes a treatment effect prediction and lesion change rate calculation unit and a comprehensive treatment effect evaluation value calculation unit.

[0103] The treatment effect prediction and lesion change rate calculation unit: adds the second entity set and the second relationship set to the initial knowledge graph G = {FE, FR} to form the final patient personalized knowledge graph, denoted as HG = {FE ∪ SE, FR ∪ SR}; constructs a time series, denoted as SJ = {t|t ∈ [1, T]}, where t represents the t-th time point and T represents the total number of time points, and attaches a time series period label to the lesion L, denoted as the periodic lesion L t ; calculates the treatment effect prediction value P(HZ a , S a , T i , D b , D c ).

[0104] The comprehensive treatment effect evaluation value calculation unit: calculates the lesion change rate ΔL t of the periodic lesion L t at the time point t; based on the treatment effect prediction value P(HZa , S i , T b , D c ), and the lesion change rate ΔL t , calculate the comprehensive efficacy evaluation value R(HZ a ) of the patient HZ a , S i , T b , D c , t).

[0105] Further, the personalized efficacy optimization module includes a knowledge graph update unit and an optimal treatment plan recommendation unit.

[0106] The knowledge graph update unit: Based on the comprehensive efficacy evaluation value R(HZ a , S i , T b , D c , t), construct an efficacy evaluation set PG = {[S i , T b , D c , R(HZ a , S i , T b , D c , t)]}, update the efficacy evaluation set PG to the patient's personalized knowledge graph HG, and the update formula is: TG = HG ∪ PG, where TG represents the updated patient's personalized knowledge graph; Based on the updated patient's personalized knowledge graph TG, preset the similarity between patient HZ a and patient HZ o , denoted as sim(HZ a , HZ o ), where HZ o represents the o-th patient, and o ≠ a; Based on the similarity sim(HZ a and patient HZ o , calculate the optimized efficacy prediction value R a of patient HZ o and patient HZ a o opt (HZ a , S i ).

[0107] The optimal treatment plan recommendation unit: Based on the optimized efficacy prediction value R a of patient HZ o and patient HZ opt (HZ a , S i ), calculate the optimal treatment plan B(T b )​, D c ); If the patient HZ a and the patient HZ o 's optimized efficacy prediction value R opt (HZ a , S i ) is greater than the patient HZ a 's comprehensive efficacy evaluation value R(HZ a , S i , T b , D c , t), then the optimal treatment plan B(T a , D b ) is recommended for the patient HZ c .

[0108] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0109] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A multi-modal driven knowledge graph therapeutic effect auxiliary supervision method, characterized in that: The method comprises the following steps: Step S1: Establish a patient database, construct the patient's electronic medical record data set and B-ultrasound image data set; extract the entities and relationships between the patient and the electronic medical record; Step S2: construct a first entity set and a first relationship set between the patient and the electronic medical record, and construct an initial knowledge graph; use image analysis technology to extract the lesion area in the B-ultrasound image, and construct a second entity set and a second relationship set between the patient and the B-ultrasound image; Step S3: adding the second entity set and the second relationship set to the initial knowledge graph to form a personalized knowledge graph for the patient; attaching a time series period label to the lesion, recording it as a periodic lesion; calculating the patient's efficacy prediction value; calculating the lesion change rate of the periodic lesion; and calculating the patient's comprehensive efficacy evaluation value based on the efficacy prediction value and the lesion change rate; Step S4: Based on the comprehensive efficacy evaluation value, construct an efficacy evaluation set and update it to the patient's personalized knowledge graph; calculate the optimized efficacy prediction value between different patients; based on the optimized efficacy prediction value, calculate the optimal treatment plan; The specific implementation process of step S1 includes the following steps: Establish a patient database, denoted as HZ = {HZ a |a∈[1,A]}, where HZ a represents the ath patient, A represents the total number of patients; obtain the patient HZ a The electronic medical record data and B-ultrasound image data are used to construct the electronic medical record data set DB = {DB a |a∈[1,A]} and B-ultrasound image dataset BC={BC a |a∈[1,A]}, where DB a represents the electronic medical record data of the ath patient, BC a represents the B-ultrasound image data of the ath patient; the electronic medical record data includes patient HZ a symptoms, treatment options and medications; Based on the electronic medical record dataset and the patient database, entities and relationships between patients and electronic medical records are extracted, and the entities include patient HZ a 、Symptoms i Treatment plan b and drug D c , where S i represents the ith symptom, T b represents the bth treatment plan, D c represents the cth drug; the relationship includes suffering from, receiving and using; the suffering from represents the patient HZ a Symptoms i The relationship between the patient and the patient, the acceptance indicates that the patient HZ a Treatment plan T b The acceptance relationship between the patient HZ a With drug D c The usage relationship between them; The specific implementation process of step S2 includes the following steps: Construct the first entity set and the first relationship set between the patient and the electronic medical record, respectively denoted as the first entity set FE = {patient HZ a 、Symptoms i Treatment plan b , Drug D c } and the first relationship set FR = {suffering from, accepting, using}, based on the first entity set E and the first relationship set R, construct an initial knowledge graph, denoted as G = {FE, FR}; The image analysis technology is used to extract the lesion area L in the B-ultrasound image, and the position and size of the lesion are recorded. The lesion area refers to the abnormal area of ​​the tissue and organ displayed on the B-ultrasound image; the second entity set and the second relationship set between the patient and the B-ultrasound image are constructed, which are respectively recorded as the second entity set SE = {patient HZ a , lesion location, lesion size} and a second relationship set SR = {belongs to, is located at, measurement value}, where the belonging indicates that the lesion belongs to the patient HZ a , the located indicates that the lesion is located at the lesion position in the B-ultrasound image, and the measured value indicates the lesion size measured by the lesion; The specific implementation process of step S3 includes the following steps: The second entity set and the second relationship set are added to the initial knowledge graph G = {FE, FR} to form the final patient personalized knowledge graph, denoted as HG = {FE∪SE, FR∪SR}; a time series is constructed, denoted as SJ = {t|t∈[1,T]}, where t represents the tth time point and T represents the total number of time points. A time series period label is added to the lesion L, denoted as periodic lesion L t ; Calculate patient HZ a The predictive value of the efficacy P(HZ a ,S i ,T b ,D c ), the calculation formula is as follows: P(HZ a ,S i ,T b ,D c )=α1·f(S i )+α2·f(T b )+α3·f(D c ); f(D c )=d c ·p c ·e c ; Among them, f(S i ) represents the characteristic function of the disease symptoms, f(T b ) represents the characteristic function of the treatment plan, f(D c ) represents the drug characteristic function, α1, α2 and α3 represent the weights of the preset symptoms, treatment plans and drugs respectively, and s i Indicates symptoms of illness i The severity of i Indicates the duration of symptoms, f i Indicates the frequency of symptoms, I b Indicates the intensity of the treatment plan, E b Indicates the suitability of the treatment method, D b Indicates the duration of the treatment course, d c Indicates drug D c The dose, p c represents the frequency of drug use, e c It represents the efficacy score of a drug based on clinical data; Calculate the periodic lesion L at time point t t The lesion change rate ΔL t , the calculation formula is as follows: Where, ΔL t Indicates periodic lesion L t The lesion change rate, L1 represents the initial lesion size before treatment, and t≠1; Based on the efficacy prediction value P(HZ a ,S i ,T b ,D c ) and the lesion change rate ΔL t , calculate the patient's HZ a The comprehensive efficacy evaluation value R(HZ a ,S i ,T b ,D c ,t), the calculation formula is as follows: R(HZ a ,S i ,T b ,D c ,t)=β·P(HZ a ,S i ,T b ,D c )+γ·ΔL t ; Wherein, β and γ represent preset parameters.

2. According to claim 1, a multi-modal driven knowledge graph therapeutic effect auxiliary supervision method is characterized in that: The specific implementation process of step S4 includes the following steps: Based on the comprehensive efficacy evaluation value R(HZ a ,S i ,T b ,D c ,t), construct the efficacy evaluation set PG={[S i ,T b ,D c ,R(HZ a ,S i ,T b ,D c ,t)]}, update the efficacy evaluation set PG to the patient personalized knowledge graph HG, the update formula is: TG = HG ∪ PG, where TG represents the updated patient personalized knowledge graph; Based on the updated patient personalized knowledge graph TG, the patient HZ is preset a With patient HZ o The similarity between them is denoted as sim(HZ a ,HZ o ), where HZ o represents the oth patient, and o≠a; based on patient HZ a With patient HZ o The similarity between sim(HZ a ,HZ o ), calculate the patient's HZ a With patient HZ o The optimized efficacy prediction value is calculated as follows: R opt (HZ a ,S i )=sim(HZ a ,HZ o )·R(HZ o ,S i ,T b ,D c ,t); Among them, R(HZ o ,S i ,T b ,D c ,t) represents the patient HZ o The comprehensive efficacy evaluation value, R opt (HZ a ,S i ) indicates patient HZ a With patient HZ o The optimized efficacy prediction value of Based on the patient's HZ a With patient HZ o The optimized efficacy prediction value R opt (HZ a ,S i ), calculate the optimal treatment plan, the calculation formula is as follows: B(T b ,D c )=argmax Tb,Dc [R opt (HZ a ,S i )]; Among them, B(T b ,D c ) represents the optimal treatment plan; If the patient HZ a With patient HZ o The optimized efficacy prediction value R opt (HZ a ,S i ) is greater than the patient's HZ a The comprehensive efficacy evaluation value R(HZ a ,S i ,T b ,D c ,t), then it is the patient HZ a Recommend the best treatment plan B(T b ,D c ).

3. A multi-modal driven knowledge graph therapeutic effect auxiliary supervision system, executing a multi-modal driven knowledge graph therapeutic effect auxiliary supervision method as claimed in claim 1, characterized in that: The system includes: a patient data management module, a knowledge graph construction module, an efficacy evaluation module and a personalized efficacy optimization module; The patient data management module: establishes a patient database, constructs a patient electronic medical record data set and a B-ultrasound image data set; extracts entities and relationships between patients and electronic medical records; The knowledge graph construction module: constructs a first entity set and a first relationship set between the patient and the electronic medical record, and constructs an initial knowledge graph; uses image analysis technology to extract the lesion area in the B-ultrasound image, and constructs a second entity set and a second relationship set between the patient and the B-ultrasound image; The efficacy evaluation module: adds the second entity set and the second relationship set to the initial knowledge graph to form a patient personalized knowledge graph; adds a time series period label to the lesion, recording it as a periodic lesion; calculates the patient's efficacy prediction value; calculates the lesion change rate of the periodic lesion; and calculates the patient's comprehensive efficacy evaluation value based on the efficacy prediction value and the lesion change rate; The personalized efficacy optimization module: constructs an efficacy evaluation set based on the comprehensive efficacy evaluation value, and updates it to the patient's personalized knowledge graph; calculates the optimized efficacy prediction value between different patients; and calculates the optimal treatment plan based on the optimized efficacy prediction value.

4. According to claim 3, a multi-modal driven knowledge graph therapeutic effect auxiliary supervision system is characterized by: The patient data management module includes a database construction unit and an entity and relationship extraction unit; The database construction unit: establishes a patient database, denoted as HZ = {HZ a |a∈[1,A]}, where HZ a represents the ath patient, A represents the total number of patients; obtain the patient HZ a The electronic medical record data and B-ultrasound image data are used to construct the electronic medical record data set DB = {DB a |a∈[1,A]} and B-ultrasound image dataset BC={BC a |a∈[1,A]}, where DB a represents the electronic medical record data of the ath patient, BC a represents the B-ultrasound image data of the ath patient; the electronic medical record data includes patient HZ a symptoms, treatment options and medications; The entity and relationship extraction unit: based on the electronic medical record data set and the patient database, extracts the entities and relationships between the patient and the electronic medical records, the entities include the patient HZ a 、Symptoms i Treatment plan b and drug D c , where S i represents the ith symptom, T b represents the bth treatment plan, D c represents the cth drug; the relationship includes suffering from, receiving and using; the suffering from represents the patient HZ a Symptoms i The relationship between the patient and the patient, the acceptance indicates that the patient HZ a Treatment plan T b The acceptance relationship between the patient HZ a With drug D c The usage relationship between them.

5. According to claim 4, a multi-modal driven knowledge graph therapeutic effect auxiliary supervision system is characterized by: The knowledge graph construction module includes a first entity set and a first relationship set construction unit and a second entity set and a second relationship set construction unit; The first entity set and the first relationship set construction unit: constructs the first entity set and the first relationship set between the patient and the electronic medical record, respectively denoted as the first entity set FE = {patient HZ a 、Symptoms i Treatment plan b , Drug D c } and the first relationship set FR = {suffering from, accepting, using}, based on the first entity set E and the first relationship set R, construct an initial knowledge graph, denoted as G = {FE, FR}; The second entity set and the second relationship set construction unit: use image analysis technology to extract the lesion area L in the B-ultrasound image, and record the position and size of the lesion, wherein the lesion area refers to the abnormal area of ​​the tissue and organ displayed on the B-ultrasound image; construct the second entity set and the second relationship set between the patient and the B-ultrasound image, respectively recorded as the second entity set SE = {patient HZ a , lesion location, lesion size} and a second relationship set SR = {belongs to, is located at, measurement value}, where the belonging indicates that the lesion belongs to the patient HZ a , the located indicates that the lesion is located at the lesion position in the B-ultrasound image, and the measured value indicates the lesion size measured by the lesion.

6. According to claim 5, a multi-modal driven knowledge graph therapeutic effect auxiliary supervision system is characterized by: The efficacy evaluation module includes an efficacy prediction and lesion change rate calculation unit and a comprehensive efficacy evaluation value calculation unit; The efficacy prediction and lesion change rate calculation unit: adds the second entity set and the second relationship set to the initial knowledge graph G = {FE, FR} to form the final patient personalized knowledge graph, denoted as HG = {FE∪SE, FR∪SR}; constructs a time series, denoted as SJ = {t|t∈[1,T]}, where t represents the tth time point and T represents the total number of time points, and adds a time series period label to the lesion L, denoted as periodic lesion L t ; Calculate the patient's HZ a The predictive value of the efficacy P(HZ a ,S i ,T b ,D c ); The comprehensive efficacy evaluation value calculation unit calculates the periodic lesion L at time point t. t The lesion change rate ΔL t Based on the predicted value of efficacy P(HZ a ,S i ,T b ,D c ) and the lesion change rate ΔL t , calculate the patient's HZ a The comprehensive efficacy evaluation value R(HZ a ,S i ,T b ,D c ,t).

7. The multi-modal driven knowledge graph therapeutic effect auxiliary monitoring system according to claim 6 is characterized by: The personalized efficacy optimization module includes a knowledge graph updating unit and an optimal treatment plan recommendation unit; The knowledge graph updating unit: based on the comprehensive efficacy evaluation value R(HZ a ,S i ,T b ,D c ,t), construct the efficacy evaluation set PG={[S i ,T b ,D c ,R(HZ a ,S i ,T b ,D c ,t)]}, update the efficacy evaluation set PG to the patient personalized knowledge graph HG, the update formula is: TG = HG ∪ PG, where TG represents the updated patient personalized knowledge graph; based on the updated patient personalized knowledge graph TG, preset patient HZ a With patient HZ o The similarity between them is denoted as sim(HZ a ,HZ o ), where HZ o represents the oth patient, and o≠a; based on patient HZ a With patient HZ o The similarity between sim(HZ a ,HZ o ), calculate the patient's HZ a With patient HZ o The optimized efficacy prediction value R opt (HZ a ,S i ); The optimal treatment plan recommendation unit: based on the patient's HZ a With patient HZ o The optimized efficacy prediction value R opt (HZ a ,S i ), calculate the optimal treatment plan B(T b ,D c ); if the patient HZ a With patient HZ o The optimized efficacy prediction value R opt (HZ a ,S i ) is greater than the patient's HZ a The comprehensive efficacy evaluation value R(HZ a ,S i ,T b ,D c ,t), then it is the patient HZ a Recommend the best treatment plan B(T b ,D c ).

Images