Knowledge graph construction method, terminal and medium based on BNCT

Abstract

The present application provides a BNCT-based knowledge graph construction method, a BNCT treatment effect prediction method, a terminal, and a computer storage medium. The method extracts sub-data from each treatment data set corresponding to the target tumor, and constructs an initial BNCT knowledge graph based on the association relationship between each sub-data; and utilizes a sum-product network method to strengthen the causal relationship in the initial BNCT knowledge graph network to obtain a final BNCT knowledge graph. The method provided in the present application can not only quickly construct a BNCT knowledge graph based on causal relationships based on treatment plan data and patient data, but also, based on the BNCT knowledge graph, can conveniently obtain the treatment effect corresponding to the target tumor treatment plan, providing decision support for auxiliary determination of treatment plan formulation.

Description

Technical Field

[0001] The present application belongs to the technical field of boron concentration acquisition, and relates to a BNCT-based knowledge graph construction method, a BNCT treatment effect prediction method, a terminal, and a computer storage medium. Background Art

[0002] Boron Neutron Capture Therapy (BNCT), as an emerging targeted tumor therapy, achieves "cellular-level targeted blasting" of cancer cells through the precise capture reaction of boron-10 nuclides with thermal neutrons, while minimizing damage to surrounding healthy tissues. Its clinical decision-making relies heavily on the multimodal knowledge fusion of nuclear physics, radiobiology, and oncology. In existing technologies, the formulation of BNCT treatment plans is often based on manual experience, that is, based on the knowledge and experience of BNCT experts. However, this formulation method often fails to quickly, efficiently, and accurately integrate and construct interdisciplinary knowledge systems (such as neutron dose calculation models, boron drug metabolic kinetic parameters, tumor microenvironment characteristics, etc.), and cannot quickly process the complex relationship between massive clinical cases and dynamic medical evidence. As a result, the BNCT treatment knowledge architecture obtained by the traditional manual experience-based knowledge retrieval model not only has efficiency bottlenecks such as long decision-making cycles and lagging knowledge updates, but may even limit the homogeneity of treatment plans due to expert cognitive biases, thereby reducing the accuracy of decision-making and affecting treatment efficacy.

[0003] Therefore, how to quickly and accurately obtain the therapeutic effect corresponding to the BNCT treatment plan has become a technical problem that needs to be solved in this field. Summary of the Invention

[0004] In view of the shortcomings of the existing technology mentioned above, the purpose of this application is to provide a BNCT-based knowledge graph construction method, BNCT treatment effect prediction method, terminal and medium, which are used to solve the problems of long decision-making cycle and delayed knowledge update when obtaining BNCT treatment knowledge architecture in existing methods.

[0005] To achieve the above-mentioned and other related objectives, the present application provides, in a first aspect, a method for constructing a knowledge graph based on BNCT, comprising:

[0006] A treatment data set corresponding to BNCT of a target tumor is obtained; the treatment data set includes treatment plan data and treatment effect data in each historical treatment; based on the causal relationship between each data in the treatment data set in BNCT treatment, an initial BNCT knowledge graph is constructed, and a sample data set corresponding to the initial BNCT knowledge graph is constructed; wherein the causal relationship includes a weighted causal relationship and a joint causal relationship; among each node corresponding to the treatment plan data, a node that satisfies the weighted causal relationship is extracted as a sum node; and among each node corresponding to the treatment effect data, a node that satisfies the joint causal relationship is extracted as a product node; according to the sample data set, a sum-product network method is used to obtain the causal relationship probability corresponding to each sum node and the causal relationship probability corresponding to the product node to obtain a target BNCT knowledge graph.

[0007] In some embodiments of the first aspect of the present invention, constructing a sample dataset corresponding to the initial BNCT knowledge graph includes:

[0008] Based on the node information in the initial BNCT knowledge graph, feature variables corresponding to the node information are extracted from the treatment data set; each feature variable belonging to the same treatment case is constructed as a sample data strip; and the construction of each sample data strip is performed on the treatment data set to obtain a sample data set corresponding to the initial BNCT knowledge graph.

[0009] In some embodiments of the first aspect of the present invention, the implementation method of obtaining the causal relationship probability corresponding to each summation node and the causal relationship probability corresponding to the product node using the sum-product network method includes:

[0010] The frequency weight calculation method is used to obtain the causal relationship probability of the summation root node corresponding to the summation node; the association effect probability calculation method is used to obtain the causal relationship probability of the product root node corresponding to the product node.

[0011] In some embodiments of the first aspect of the present invention, the method of using the frequency weight calculation method to obtain the causal relationship probability of the summation root node corresponding to the summation node includes:

[0012] In the sample data set, extract the sample subset corresponding to the current summation node, and obtain the number of sample data items in the sample subset as the total number of samples of the current summation node; in the sample subset, extract the current number of samples corresponding to the current summation root node; based on the current number of samples and the total number of samples, calculate the proportion of the current sample data to the total number of samples as the causal relationship probability corresponding to the current summation root node.

[0013] In some embodiments of the first aspect of the present invention, the method for calculating the probability of an associated effect to obtain the probability of a causal relationship of a product root node corresponding to a product node includes:

[0014] In the initial BNCT knowledge graph, a node group corresponding to the current product node is determined; based on the causal relationship, a treatment variable, a confounding variable and an outcome variable are determined in the node group; based on the causal relationship, an association probability function is constructed between the outcome variable, the treatment variable and the confounding variable; based on the sample data in the sample data subset, a relationship probability calculation is performed according to the association probability function to obtain a causal probability value corresponding to the product node group.

[0015] In some embodiments of the first aspect of the present invention, the method for extracting the summing node includes:

[0016] For a single node among the nodes corresponding to the treatment plan data, determine whether it is a child node; when the node is a child node, determine whether there is a weighted causal relationship between the root nodes corresponding to the current child node and the current child node, and if so, mark the child node as a summation node; and the method for extracting the product node includes: for a single node among the nodes corresponding to the treatment effect data, determine whether the node is a child node; when the node is a child node, determine whether there is a joint causal relationship between the root nodes corresponding to the current child node and the current child node, and if so, mark the child node as a product node.

[0017] In some embodiments of the first aspect of the present invention, when the product node is the tumor volume change rate and the product root node is the boron drug dosage, the node group includes: neutron beam irradiation time, neutron beam energy, boron drug dosage, tumor boron concentration, tumor volume, and tumor volume change rate; the boron drug dosage is set as the treatment variable, the neutron beam irradiation time and neutron beam energy are set as confounding variables, and the tumor volume concentration, tumor volume, and tumor volume change rate are set as outcome variables; and the association probability function between the outcome variable, the treatment variable, and the confounding variables is constructed as follows:

[0018]

[0019] Where, P is the probability of the associated effect; P(U i1 ,U i2 |T i ) is the probability of the association effect under the influence of the confounding variable on the treatment variable, P(S i |T i ,U i1 ,U i2) is the probability of the associated effect of the treatment variable on the outcome variable under the influence of multiple confounding variables; U1 is the neutron beam irradiation time, U2 is the neutron beam energy; S2 is the tumor volume change rate; the tumor volume change rate is represented by a weighted combination or functional relationship of tumor volume concentration and tumor volume, satisfying:

[0020] S2=α·S0·U2+β·U1·ln(S1)

[0021] Among them, α is the influence coefficient of tumor boron concentration, β is the influence coefficient of irradiation time; S0 is the tumor boron concentration; S1 is the tumor volume.

[0022] To achieve the above-mentioned and other related objectives, the present application further provides, in a second aspect, a method for predicting BNCT treatment effects based on a knowledge graph, comprising:

[0023] Obtain a pre-constructed target BNCT knowledge graph; use sample data to train the target BNCT knowledge graph to obtain a trained knowledge graph; based on the trained knowledge graph, obtain the tumor treatment effect of the patient under a given treatment plan; wherein, the target BNCT knowledge graph is obtained using any of the BNCT-based knowledge graph construction methods described above.

[0024] In a third aspect, the present application provides a terminal comprising: a processor and a memory, wherein the memory is communicatively connected to the processor; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes any of the BNCT-based knowledge graph construction methods described above, or executes the BNCT treatment effect prediction method based on the knowledge graph as described above.

[0025] In a fourth aspect, the present application provides a computer storage medium storing a computer program, which, when executed by a processor, implements any of the above-described BNCT-based knowledge graph construction methods, or executes the above-described BNCT treatment effect prediction method based on the knowledge graph.

[0026] As described above, the BNCT-based knowledge graph construction method, BNCT treatment effect prediction method, terminal and medium provided in this application, by constructing a BNCT knowledge graph based on causal relationships, realize the systematic integration of treatment plan data and patient data, not only realizing the conversion of fragmented knowledge into a structured relationship network, but also through knowledge reasoning, the treatment effect corresponding to the treatment plan can be obtained, thereby assisting doctors in predicting the treatment effects corresponding to different treatment plans; and, by constructing a dynamic knowledge update mechanism, the BNCT knowledge graph is updated based on the latest treatment sample data, thereby further improving the accuracy of the knowledge graph, thereby greatly improving the accuracy of personalized treatment plan design. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Shown is a flow chart of the BNCT-based knowledge graph construction method provided in an embodiment of the present application;

[0028] Figure 2 Shown is a flow chart of step S20 in one embodiment of the present application;

[0029] Figure 3 Shown is a flow chart of step S40 in one embodiment of the present application;

[0030] Figure 4 Shown is a flow chart of a method for obtaining the causal relationship probability corresponding to a single summation root node in one embodiment of the present application;

[0031] Figure 5 Shown is a flow chart of a method for obtaining the causal relationship probability corresponding to a single product root node in one embodiment of the present application;

[0032] Figure 6 A logic diagram showing the causal relationship between boron drug dosage and tumor volume change rate in one embodiment of the present application;

[0033] Figure 7 Shown is a flow chart of the BNCT treatment effect prediction method based on the knowledge graph provided in an embodiment of the present application;

[0034] Figure 8 Shown is a schematic diagram of the structure of the terminal described in the embodiment of the present application. DETAILED DESCRIPTION

[0035] The following describes the embodiments of the present application through specific examples. Those skilled in the art can easily understand the other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0036] It should be noted that in the following description, reference is made to the accompanying drawings, which describe several embodiments of the present application. It should be understood that other embodiments may also be used, and that mechanical, structural, electrical, and operational changes may be made without departing from the spirit and scope of the present application. The following detailed description should not be considered restrictive, and the scope of the embodiments of the present application is limited only by the claims of the published patents. The terms used herein are only for describing specific embodiments and are not intended to limit the present application. Spatially related terms, such as "upper", "lower", "left", "right", "below", "below", "lower", "above", "upper", etc., may be used in the text to facilitate the description of the relationship between one element or feature shown in the figure and another element or feature.

[0037] Furthermore, as used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context indicates otherwise. It should be further understood that the terms "comprise," "include," and "include" indicate the presence of the stated features, operations, elements, components, items, categories, and / or groups, but do not preclude the presence, occurrence, or addition of one or more other features, operations, elements, components, items, categories, and / or groups. The terms "or" and "and / or" used herein are to be interpreted as inclusive, or mean any one or any combination.

[0038] In order to solve the technical problems in the prior art, the present application provides a BNCT-based knowledge graph construction method in the first aspect, which is used to construct a target BNCT knowledge graph based on causal association for the target tumor type, so as to obtain the treatment effect prediction results corresponding to the treatment plan based on the target BNCT knowledge graph.

[0039] Among them, the target tumor type is the tumor type treated by the BNCT treatment plan, including but not limited to melanoma, and glioma.

[0040] See also Figure 1 , which is a flow chart of the method for constructing a knowledge graph based on BNCT provided in an embodiment of the present application; Figure 1 As shown, the method includes the following steps:

[0041] S10, obtain the treatment dataset of the target tumor during historical BNCT treatment;

[0042] The treatment data set is various types of data contained in the historical treatment process, including treatment plan data and treatment effect data;

[0043] The treatment plan data are various treatment parameters set or generated during the BNCT treatment process, including neutron source type, neutron source dose, irradiation angle, irradiation time, boron drug dosage and boron drug dose absorption rate;

[0044] The treatment effect data are data on tumor control and surrounding tissue effects after BNCT treatment, including information such as tumor radiation absorption rate, tumor volume change rate, and the probability of normal tissue canceration.

[0045] S20, performing data preprocessing on the treatment data set using a data mining method to obtain a processed treatment data set;

[0046] Specifically, when this step is executed, Figure 2 As shown, it includes the following sub-steps:

[0047] S21, performing data denoising processing on each data in the treatment data set;

[0048] Specifically, noise, errors, missing values, etc. in each data in the treatment data set are removed.

[0049] S22, performing standardization processing on each data in the treatment data set.

[0050] Specifically, for a single type of data, operations such as normalization, standardization or discretization are performed on the data under this type to convert data of different physical dimensions into preset physical dimensions, so that data of the same type have the same physical dimension.

[0051] In the present application, the above steps are performed on each of the treatment datasets to obtain each pre-processed treatment dataset.

[0052] In some optional embodiments, when the treatment dataset includes text data, step S22 further includes:

[0053] Data encoding is performed on text data so that the encoded data can be processed by calculation.

[0054] S30, constructing an initial BNCT knowledge graph based on the causal relationship between each data in the treatment dataset, and obtaining a sample dataset corresponding to the initial BNCT knowledge graph;

[0055] Among them, the causal relationship is the mathematical correlation between various data in BNCT treatment; the causal relationship refers to the mathematical dependence reflected by various types of data in BNCT treatment. The mathematical dependence is the statistical dependence and influence mechanism between different variables (such as drug dosage, treatment method, etc.), which is used to characterize the causal action path between various variables and provide a scientific basis for BNCT treatment effect evaluation and decision support.

[0056] In this application, the causal relationship is calculated using Bayesian causal probability to efficiently characterize the causal influence and mechanism of action between different variables during BNCT treatment.

[0057] It should be noted that, in some other embodiments, the causal relationship may also be expressed using other correlation probabilities such as the Pearson correlation coefficient, the Spearman rank correlation coefficient, etc., to reveal the correlation or causal relationship between the characteristic variables from different statistical risks.

[0058] Specifically, each data in the treatment data set is converted into a node in the initial BNCT knowledge graph; the causal relationship between each node is extracted, and the causal relationship is converted into a directed edge in the initial BNCT knowledge graph; the causal relationship between nodes is graphically represented using a directed graph method to construct an initial BNCT knowledge graph; wherein the nodes in the directed graph represent each data (entity) in the graph, and the causal relationship is represented as a directed edge in the graph.

[0059] After constructing the initial BNCT knowledge graph, based on the node information in the initial BNCT knowledge graph, feature variables corresponding to each node information are extracted from the pre-extracted treatment data set; each feature variable belonging to the same treatment case is used as a sample data strip; based on this method, sample data strips are extracted from the treatment data set to construct a sample data set corresponding to the initial BNCT knowledge graph; that is, the sample data set contains each sample data strip.

[0060] In an optional embodiment, the method for extracting the causal relationship between the nodes includes:

[0061] A treatment effect data in the treatment data set is used as the current child node,

[0062] Using a semantic analysis method, extracting treatment plan data having a causal relationship with the treatment effect data from the treatment data set as the current root node corresponding to the current child node;

[0063] The causal relationship between the current root node and the current child node is constructed as a directed edge between the current root node and the current child node.

[0064] Exemplarily, the treatment plan node in the initial BNCT knowledge graph includes:

[0065] 1) Boron dosage: such as "BPA 80ppm";

[0066] 2) Neutron source dose: such as "dose rate 2.5 Gy-eq";

[0067] 3) Mode of administration: such as “intravenous injection”;

[0068] 4) Initial concentration of boron drug: such as 60 ppm;

[0069] 5) Irradiation angle: such as 30°;

[0070] 6) Irradiation time: e.g. 15 minutes.

[0071] The treatment outcome nodes in the initial BNCT knowledge graph include:

[0072] 1) Tumor volume change rate: such as "volume reduced by 50%."

[0073] 2) Side effects: such as "mild nausea".

[0074] 3) Survival time: such as “survived 24 months”.

[0075] 4) Boron drug absorption dose: such as "12ppm".

[0076] Based on the causal relationship, a directed edge is constructed between nodes with the causal relationship; illustratively, the directed edge includes:

[0077] “Boron drug dosage” → “Boron drug absorption dose”

[0078] “Boron drug dose” → “tumor volume change rate”;

[0079] “Neutron source dose” → “side effects”.

[0080] For example, "tumor volume change rate" is a child node, and "boron drug dosage" is a node under the child node.

[0081] Optionally, after constructing the initial BNCT knowledge graph, the initial BNCT knowledge graph can be visualized using Neo4j tools, so as to more efficiently and conveniently obtain the causal relationship between the nodes in the graph based on the visualized initial BNCT knowledge graph.

[0082] S40, extracting nodes that satisfy weighted causal relationships from each node corresponding to the treatment plan data as summation nodes; and extracting nodes that satisfy joint causal relationships from each node corresponding to the treatment effect data as product nodes; based on the sample data set, using the sum-product network method, obtaining the causal relationship probability corresponding to each summation node and the causal relationship probability corresponding to the product node to obtain the target BNCT knowledge graph.

[0083] The sum-product network is a directed acyclic graph (DAG) structure including sum nodes and product nodes;

[0084] The summation node performs weighted summation on the outputs of the subnodes to represent the mixture of latent variables or the hierarchical clustering of data distribution or the probability weighting under different conditions;

[0085] The product nodes perform product operations on the outputs of the child nodes to represent the independence assumptions between variables or the joint distribution of local features.

[0086] For example, when considering the effect of different boron drug doses on the tumor volume change rate, the tumor volume change rate is used as the summation node (which is also a child node), and the different boron drug doses are used as the root node. The summation node is used to represent the probability of the effect of different boron drug doses on the tumor volume change rate (that is, the probability of causality).

[0087] Also, when considering the effects of "boron dose" and "neutron irradiation dose" on "tumor volume change rate", the "tumor volume change rate" is used as the product node (which is also a child node), and the "boron dose" and "neutron irradiation dose" are used as root nodes respectively. Then, the product node can be used to characterize the probability of the influence of different boron doses on the tumor volume change rate under the influence of neutron irradiation dose, or can be used to characterize the probability of the influence of different neutron irradiation doses on the tumor volume change rate under the influence of boron dose.

[0088] Specifically, when step S40 is executed, Figure 3 As shown, the following steps are included:

[0089] S41, dividing the nodes of the initial BNCT knowledge graph into sum nodes and product nodes according to the causal relationship between the nodes;

[0090] Among them, the causal relationship forms include at least weighted causal relationship and joint causal relationship; the weighted causal relationship is a nonlinear causal relationship, corresponding to the summation node, and the joint causal relationship is a causal relationship of multiple factors, corresponding to the product node.

[0091] In this application, the nodes of the initial BNCT knowledge graph include nodes corresponding to treatment plan data and nodes corresponding to treatment effect data;

[0092] By determining the form of causal association between nodes, nodes that satisfy weighted causal relationships are extracted from each node corresponding to the treatment plan data as summation nodes; and by determining the form of causal association between nodes, nodes that satisfy joint causal relationships are extracted from each node corresponding to the treatment effect data as product nodes, so as to conveniently and efficiently determine the node type to which each node belongs, providing data preparation for the execution of subsequent steps.

[0093] Specifically, for a single node among the nodes corresponding to the treatment plan data, determine whether the node is a child node; when the node is a child node, obtain the root nodes corresponding to the current child node; determine whether there is a weighted causal relationship between each root node and the current child node, and if so, mark the child node as a summation node; execute this process for each node corresponding to the treatment plan data, thereby extracting each summation node.

[0094] Similarly, for each node corresponding to the treatment effect data, determine whether the node is a child node; when the node is a child node, obtain the root nodes corresponding to the current child node; determine whether there is a joint causal relationship between each root node and the current child node. If so, mark the child node as a product node; execute this process for each node corresponding to the treatment plan data, thereby extracting each product node.

[0095] In a specific embodiment, the correlation / independence between nodes in the initial BNCT knowledge graph network is determined based on expert experience or using a preset causal relationship model; based on the correlation between the nodes, each node in each initial BNCT knowledge graph network is divided into a sum node and a product node; illustratively, the causal relationship model provided by the BNCT treatment planning software (TPS) is used, and the input parameters of the TPS are used as variable nodes, and the output parameters of the TPS are used as sum nodes or product nodes;

[0096] By utilizing expert experience or a preset causal relationship model, the existing causal relationship between nodes can be fully utilized, so that the summation node and the product node can be determined more quickly and conveniently.

[0097] S42, for each summing node, using a frequency weight calculation method to obtain the causal relationship probability of each summing root node corresponding to the summing node;

[0098] The frequency weight is the frequency of occurrence of the sample data strip corresponding to the root node in the total sample data strip corresponding to the summation node;

[0099] For a single summation root node under a summation node, the corresponding causal relationship probability is obtained as follows: Figure 4 Shown, including:

[0100] S421, extracting a sample subset corresponding to the current summation node from the sample data set, and obtaining the number of sample data pieces in the sample subset as the total number of samples of the current summation node;

[0101] Specifically, according to the numerical range corresponding to the current summation node, in the sample data set, each sample data strip whose numerical value of the characteristic variable of the summation node is within the numerical range is extracted; the set of each sample data strip within the numerical range is used as the sample subset corresponding to the current summation node, and the number of sample data strips in the sample subset is counted as the total number of samples of the current summation node;

[0102] The characteristic variable of the summation node is the characteristic variable corresponding to / contained by the summation node; for example, when the summation node is "boron drug dosage is greater than 50", the characteristic variable corresponding to / contained by the summation node is "boron drug dosage".

[0103] For example, when the current summation node is "boron dosage is greater than 50", from the sample data set, sample data strips with values ​​greater than 50 for the summation node characteristic variable, i.e., "boron dosage", are extracted, totaling 100 strips; the 100 extracted sample data strips are used as the sample data subset of the current summation node; accordingly, the total number of samples of the current summation node is 100.

[0104] S422, extracting the current number of samples corresponding to the current summation root node from the sample subset;

[0105] Specifically, determining the value range corresponding to the current summation root node; based on the value range, counting the number of sample data pieces in the sample subset whose values ​​of the characteristic variables of the current summation root node are within the value range, and using the number as the current number of samples corresponding to the current summation root node;

[0106] Among them, the characteristic variable of the summation root node is the characteristic variable corresponding to / contained by the summation root node; for example, when the summation root node is "boron drug dosage 10-15", the characteristic variable corresponding to / contained by the summation root node is "boron drug dosage".

[0107] In this application, the value range corresponding to a single summation root node is a pre-defined numerical range, and satisfies: the total value range formed by the value range intervals corresponding to each summation root node (for the same characteristic variable) is the total value range of the characteristic variable corresponding to the summation root node in the sample data set.

[0108] Exemplarily, when the summation node is the boron dosage, its corresponding summation root node is the "boron dosage" of different value ranges, including: the "boron dosage" of the first interval (10-15), the "boron dosage" of the second interval (15-20), the "boron dosage" of the third interval (20-25), and the "boron dosage" of the fourth interval (25-30), which correspond to the first summation root node, the second root node, the third summation root node, and the fourth summation root node respectively; for the first summation root node, in the sample subset corresponding to the summation node (a total of 100 sample data strips), the number of sample data strips of "boron dosage" located in the first interval is counted, which is 30, and then the 30 is used as the current sample number corresponding to the current summation root node.

[0109] It should be noted that when constructing the initial BNCT knowledge graph, for each feature variable, its total value range in the sample data set, that is, the total value range, is first counted; then, according to the preset number of intervals n, the total value range is divided into n intervals, and a corresponding node is established for each interval to construct nodes corresponding to different value range intervals for the same feature variable, thereby realizing structured modeling of different value intervals of the feature variable, providing a basis for subsequent causal relationship modeling and reasoning.

[0110] Exemplarily, the characteristic variable of the summation node is the boron dose, and the total value range of the boron dose in the sample data set is 30-70; the number of intervals is predetermined to be 5; according to the number of intervals, the total numerical range of the boron dose is evenly divided into 5 value range intervals, corresponding to: 30-40, 40-50, 50-60 and 60-70 respectively; based on different value range intervals, corresponding summation nodes are constructed, namely the first summation node (boron dose 30-40), the second summation node (boron dose 40-50), the third summation node (boron dose 50-60) and the fourth summation node (boron dose 60-70).

[0111] Also exemplarily, the characteristic variable of the summation root node is the boron drug dosage, and the total value range of the boron drug dosage in the sample data set is 10-30; the number of intervals is predetermined to be 4; according to the number of intervals, the total numerical range of the boron drug dosage is evenly divided into 4 value range intervals, corresponding to: 10-15, 15-20, 20-25 and 25-30 respectively; based on different value range intervals, corresponding summation root nodes are constructed, namely the first summation root node (boron drug dosage 10-15), the second summation root node (boron drug dosage 15-20), the third summation root node (boron drug dosage 20-25) and the fourth summation root node (boron drug dosage 25-30).

[0112] S423, based on the current number of samples and the total number of samples, calculate the proportion of the current sample data to the total number of samples as the causal relationship probability corresponding to the current summation root node.

[0113] Specifically, after obtaining the current number of samples corresponding to the current summation root node and the total number of samples at the current summation node, the ratio of the current number of samples to the total number of samples is calculated and used as the causal relationship probability corresponding to the current summation root node. This quantifies the probability of each dose interval contributing to the total dose interval, providing a basis for subsequent causal inference and knowledge graph parameter optimization.

[0114] For example, as shown in Table 1 below, the summation node is a boron dosage of 50-60, and the corresponding total number of samples is 50; the summation root nodes corresponding to the summation node include: a first summation root node ("boron dosage" is 10-15), a second root node ("boron dosage" is 15-20), a third summation root node ("boron dosage" is 20-25), and a fourth summation root node ("boron dosage" is 25-30); wherein, the current number of samples corresponding to the first summation root node is 30, and the corresponding causal relationship probability is 30 / 50=0.6; the current number of samples corresponding to the second summation root node is 15, and the corresponding causal relationship probability is 10 / 50=0.2, and so on, the causal relationship probabilities corresponding to the third summation root node and the fourth summation root node are both 0.1.

[0115] Table 1. Probability of causal relationship corresponding to each summation root node under the summation node (boron dosage is 50-60)

[0116]

[0117] It can be seen from Table 1 above that for the summation node (boron drug dosage is 50-60), when the boron drug dosage is 10-15, the probability of affecting the summation node is the largest, which is 0.6.

[0118] S43, for each product node, using a calculation method for the association effect probability, to obtain the causal relationship probability of each product root node corresponding to the product node;

[0119] The probability of the association effect is the sum of the marginal probabilities of the outcome variable S distributed over each confounding variable U under the condition of a given treatment variable T;

[0120] In this application, the calculation method of the association effect probability includes:

[0121] 1) For a given processing variable T, obtain the values ​​of all possible confounding variables U under the processing variable T;

[0122] 2) For each confounding variable U, calculate its corresponding marginal probability, that is, the probability P(S|T,U) of the outcome variable S given the treatment variable T and the current confounding variable U, multiplied by the probability P(U|T) of the current confounding variable U under the same given T;

[0123] 3) Accumulate the marginal probabilities corresponding to each confounding variable U to obtain the probability of the associated effect.

[0124] For a single product root node under a product node, the corresponding causal relationship probability is obtained as follows: Figure 5 Shown, including:

[0125] S431, in the initial BNCT knowledge graph, determining a node group corresponding to the current product node, and determining a treatment variable (T) and a confounding variable (U) and an outcome variable (S) associated with the treatment variable in the node group based on a causal relationship;

[0126] In the initial BNCT knowledge graph, a node group corresponding to the current product node is determined based on a causal relationship; the node group is a combination of several nodes that have a causal relationship with the current product node.

[0127] In the node group, first determine the treatment variable (T), that is, the root node of the current product; then, identify all confounding variables (U) that interact with the treatment variable (T) and can affect the result variable (S); finally, determine the result variable (S), that is, the characteristic variable formed under the joint action of T and U, which contains the current product node; by clarifying the causal relationship between each variable in the node group, a structural basis is provided for subsequent causal inference and knowledge graph update.

[0128] In a specific embodiment, the initial BNCT knowledge graph includes the following nodes:

[0129] Dosage interval, tissue density, neutron beam irradiation time, neutron beam energy, boron drug dosage, tumor boron concentration, tumor volume, tumor volume change rate, and tumor volume change value;

[0130] When the tumor volume change rate is a product node and the boron drug dosage is a product root node, based on the causal relationship between the boron drug dosage and the tumor volume change rate, that is, in the causal logic process of the boron drug dosage (product root node) and the tumor volume change rate (product node), characteristic variables such as neutron beam irradiation time, neutron beam energy, tumor boron concentration, and tumor volume will affect the causal logic process; based on this, the neutron beam irradiation time, neutron beam energy, boron drug dosage, tumor boron concentration, tumor volume, and tumor volume change rate in the initial BNCT knowledge graph are divided into node groups; and the boron drug dosage is set as the treatment variable (T), the neutron beam irradiation time and neutron beam energy are set as the first confounding variable (U1) and the first confounding variable (U2), respectively, and the tumor volume concentration, tumor volume, and tumor volume change rate are set as the first result variable (S0), the second result variable (S1), and the third result variable (S2), respectively;

[0131] For ease of understanding, the characteristic variables contained in the above node group are defined as shown in Table 2 below:

[0132] Table 2 Definition of characteristic variables in node groups

[0133]

[0134] It should be noted that in different causal relationship chains, the same characteristic variable can serve as either a summation node or a product node.

[0135] S432, constructing an association probability function between the outcome variable (S), the treatment variable (T), and the confounding variable (U) based on the causal relationship contained in the node group;

[0136] The relationship probability function is a function used to characterize the effect of the treatment variable (T) on the outcome variable (S) under the influence of each confounding variable (U).

[0137] According to the causal relationship, the causal relationship between the treatment variable (T), confounding variable (U) and outcome variable (S) is sorted out, and the association probability function between each variable is constructed.

[0138] In one embodiment, based on the causal relationship between boron drug dose and tumor volume change rate, the transmission process of the causal relationship is sorted out as follows: Figure 6 The logic diagram shown in Figure 6 As shown, the boron dose affects the dose received by the tumor by affecting the boron concentration in the tumor, which ultimately affects the rate of change of tumor volume. In addition, the causal logic transmission process from boron dose to tumor boron concentration is also affected by the time interval between drug administration and tissue density.

[0139] According to the calculation process of TPS, the associated probability function of each treatment variable under the confounding variable for the outcome variable is constructed as follows:

[0140]

[0141] Where, P is the probability of the associated effect; P(U i1 ,U i2 |T i ) is the probability of the association effect under the influence of the confounding variable on the treatment variable, P(S i |T i ,U i1 ,U i2 ) is the probability of the associated effect of the treatment variable on the outcome variable under the influence of multiple confounding variables.

[0142] Where S2 is the tumor volume change rate, which satisfies:

[0143] S2=α·S0·U2+β·U1·ln(S1)

[0144] Wherein, α is the influence coefficient of tumor boron concentration, and β is the influence coefficient of irradiation time; both are pre-set parameters; for example, α = -0.08, β = -0.05, where the negative sign indicates that the tumor volume is reduced.

[0145] S433 , performing relationship probability calculation based on the sample data in the sample data subset according to the association probability function to obtain a causal probability value corresponding to the product node group.

[0146] Specifically, in the sample data set, sample data strips containing the result variable (S), the treatment variable (T) and the confounding variable (U) are extracted to construct a sample data subset of the node group corresponding to the current product node;

[0147] Extracting sample data corresponding to the outcome variable (S), the treatment variable (T), and the confounding variable (U) in each of the sample data strips from the sample data subset;

[0148] Based on the sample data corresponding to the result variable (S), the treatment variable (T) and the confounding variable (U) in the sample data subset, performing a relationship probability calculation according to the association probability function to obtain a relationship probability value corresponding to each sample data item;

[0149] The relationship probability values ​​corresponding to each sample data strip are integrated to obtain a comprehensive value of the relationship probability, which is used as the causal probability value corresponding to the current product node group.

[0150] In a more specific embodiment, the sample data subset of the node group corresponding to the current product node is shown in Table 3 below;

[0151] Table 3 Sample data subsets and causal probability values

[0152]

[0153] Based on the above sample data subset, the causal probability value corresponding to the current product node is calculated as:

[0154]

[0155] Right now:

[0156] P(S2≤-20%|T=15)=(0.95×0.25)+(0.90×0.15)+(0.80×0.30)+(0.99×0.30)

[0157] =0.2375+0.135+0.24+0.297=0.9095

[0158] Conclusion: The above causal probability value P(S2≤-20%|T=15) is used to represent that when the boron dose is 15 mg / kg, after considering all confounding variables, the probability of tumor volume reduction exceeding 20% ​​is 90.95%.

[0159] It should be noted that in the above formula, P(S2≤-20%|T=15,U0,U1,U2,U3)·represents the probability value of the probability-weighted sum of all possible values ​​of all confounding variables that the tumor volume is reduced by more than 20% (S2≤-20%), which is used to eliminate the confounding effect of the confounding variable U to obtain P(S|T).

[0160] S44, combining the causal relationship probabilities of the sum root nodes and the causal relationship probabilities of the product root nodes to obtain a target BNCT knowledge graph.

[0161] Specifically, after obtaining the causal relationship probability of each sum root node and the causal relationship probability of each product root node, each causal relationship probability is added to the initial BNCT knowledge graph to obtain the target BNCT knowledge graph.

[0162] It should be noted that in some optional embodiments, the influence of confounding factors can be adjusted through hierarchical analysis or regression models, for example, estimating P(Treatment Outcome|Boron Dose, Neutron Flux, Tumor Type, Patient Age); and using causal inference methods (such as do-calculus) to calculate the intervention effect, for example, P(Treatment Outcome|do(Boron Dose=High)).

[0163] Since the accuracy of the constructed knowledge graph is positively correlated with the number of extracted treatment data sets, that is, the more treatment data sets extracted, the higher the accuracy of the constructed knowledge graph; based on this, in order to increase the number of obtained treatment data sets, in a specific embodiment, the method of obtaining the treatment data sets includes:

[0164] Extract treatment datasets related to BNCT treatment of target tumors from historical treatment records or research data;

[0165] Among them, the historical treatment record is the case record corresponding to each historical BNCT treatment performed on the target tumor type; that is, a single historical treatment has a corresponding single case record, and the case record contains various types of data generated during the corresponding historical treatment process.

[0166] The research data includes clinical trial reports, basic research papers, review articles, and other information. Knowledge texts on the mechanism of boron drug uptake by target tumors, the mechanism of cell damage caused by the interaction between boron drugs and neutrons, and the dose distribution patterns of different neutron sources in tumor tissues are extracted from each data;

[0167] Specifically, clinical data of patients receiving treatment are collected from the hospital medical record database or various research data are extracted from the medical database, and diagnostic information of BNCT is extracted from each clinical data or research data, including: basic information of patients, tumor characteristics, various parameters during the treatment process (such as boron drug concentration, neutron irradiation time, etc.) and treatment results (such as tumor recurrence, survival time, etc.).

[0168] Based on the same inventive concept, the present application also provides a BNCT treatment effect prediction method based on a knowledge graph, which is used to predict the tumor treatment effect of a given BNCT treatment plan and obtain the predicted results of the treatment effect.

[0169] Among them, the BNCT treatment plan includes treatment parameters corresponding to the treatment plan; for example, neutron source type, neutron source intensity, boron drug dosage, irradiation angle and irradiation time, etc.; the tumor treatment effect includes the degree of tumor reduction and the probability of canceration of normal tissue, etc.

[0170] See also Figure 7 , which is a flow chart of the BNCT treatment effect prediction method based on the knowledge graph provided in the embodiment of the present application; Figure 7 As shown, the method includes the following steps:

[0171] S100, obtaining a pre-built target BNCT knowledge graph;

[0172] In the embodiment, the target BNCT knowledge graph is constructed in the same manner as in the above embodiment and will not be described in detail here.

[0173] S200, using the sample data, performing a graph update on the target BNCT knowledge graph to obtain an updated knowledge graph;

[0174] Get the latest treatment dataset, based on the latest treatment dataset;

[0175] Using the sample data, the pre-built target BNCT knowledge graph is updated and the parameters of each node (such as weight and probability distribution) are optimized; during the optimization process, the EM algorithm or gradient descent method is used to achieve the convergence effect of the optimization process.

[0176] Specifically, the method of using each sample data to update the pre-built target BNCT knowledge graph and optimize each node parameter (such as weight and probability distribution) includes:

[0177] The number of historical sample data items in the historical treatment dataset is obtained as M, and the number of latest sample data items in the latest treatment dataset is obtained as N; the historical treatment dataset and the latest treatment dataset are merged to obtain an updated treatment dataset; in the updated treatment dataset, the proportion of historical sample data items is obtained as the first weight, and the proportion of latest sample data items is obtained as the second weight; based on the first weight and the second weight, the causal relationship probability between each node in the target BNCT knowledge graph is updated to obtain an updated causal relationship probability, which is

[0178] W c =AW a +BW b

[0179] Where W c is the updated causal relationship probability; W a is the causal relationship probability corresponding to the historical treatment data set; W b is the causal relationship probability corresponding to the latest treatment data set; A is the first weight; B is the second weight.

[0180] Among them, the causal relationship probability corresponding to the latest treatment data set is the causal relationship probability obtained based on the latest treatment data set using the BNCT-based knowledge graph construction method provided in the above embodiment.

[0181] S300, based on the updated knowledge graph, obtains the tumor treatment effect of the patient under a given treatment plan.

[0182] Wherein, the given treatment plan is a pre-established treatment plan for the target tumor;

[0183] Specifically, based on the given treatment plan, the treatment plan data contained in the treatment plan is extracted; based on the treatment plan data, combined with the patient data, a given data group is constructed; the given data group is input into the pre-trained knowledge graph, and based on the knowledge graph, the tumor treatment effect corresponding to the given data group is obtained through a search based on causal association relationships.

[0184] For example, in a trained knowledge graph, the question "Given the tumor type and boron drug properties, what is the probability of achieving complete remission under different neutron source parameters?" is input. Using the SPN method, the probability of tumor treatment effect is obtained by calculating the probabilities between related variable nodes, summation nodes, and product nodes, thereby realizing reasoning about the treatment effect.

[0185] This reasoning ability can help doctors more accurately assess the causal impact of different combinations of factors on treatment outcomes when formulating treatment plans.

[0186] See also Figure 5 , a schematic diagram of an optional hardware structure of a terminal provided in an embodiment of the present invention, which can be a mobile phone, computer, tablet device, personal digital assistant, factory backend processing equipment, etc. The terminal includes: at least one processor 61, memory 62, at least one network interface 64, and a user interface 63. The various components in the device are coupled together via a bus system 65. It will be understood that bus system 65 is used to enable connectivity and communication between these components. In addition to the data bus, the bus system also includes a power bus, a control bus, and a status signal bus.

[0187] The user interface 63 may include a display, a keyboard, a mouse, a trackball, a click gun, keys, buttons, a touch pad or a touch screen.

[0188] It will be appreciated that the memory 62 may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. The non-volatile memory may be a read-only memory (ROM) or a programmable read-only memory (PROM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memories represented by the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memories.

[0189] The memory 62 in the embodiment of the present invention is used to store various types of data to support the operation of the terminal. Examples of such data include: any executable program for operating on the terminal 60, such as an operating system 621 and an application 622; the operating system 621 includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application 622 can include various applications, such as a media player (MediaPlayer), a browser (Browser), etc., for implementing various application services. The implementation of the BNCT-based knowledge graph construction method or the BNCT treatment effect prediction method based on the knowledge graph provided in the embodiment of the present invention can be included in the application 622.

[0190] The method disclosed in the above embodiment of the present invention can be applied to the processor 61 or implemented by the processor 61. The processor 61 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by the hardware integrated logic circuit in the processor 61 or by instructions in the form of software. The above processor may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 61 can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiment of the present invention. The general-purpose processor 61 can be a microprocessor or any conventional processor, etc. The steps of the accessory optimization method provided in the embodiment of the present invention can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium, which is located in a memory. The processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0191] In an exemplary embodiment, the terminal 60 may be configured to execute the aforementioned method using one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs).

[0192] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when called by a processor, implements the BNCT-based knowledge graph construction method or the BNCT treatment effect prediction method based on the knowledge graph provided by the present invention.

[0193] Among them, a computer-readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. The computer-readable storage medium can be, for example, (but not limited to) an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, and a mechanical encoding device.

[0194] The computer-readable program characterized herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing / processing device.

[0195] In summary, the BNCT-based knowledge graph construction method, BNCT treatment effect prediction method, terminal and medium provided in this application, by constructing a target BNCT knowledge graph based on causal relationships, realizes the systematic integration of treatment plan data and patient data, not only realizing the conversion of fragmented knowledge into a structured relationship network, but also through knowledge reasoning, the treatment effect corresponding to the treatment plan can be obtained, thereby assisting doctors in predicting the treatment effects corresponding to different treatment plans; and, by constructing a dynamic knowledge update mechanism, the target BNCT knowledge graph is updated based on the latest treatment sample data, thereby further improving the accuracy of the knowledge graph, thereby greatly improving the accuracy of personalized treatment plan design.

[0196] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by one of ordinary skill in the art without departing from the spirit and technical concepts disclosed in this application shall be covered by the claims of this application.

Claims

1. A knowledge graph construction method based on BNCT, comprising: Obtain a treatment dataset of the target tumor during historical BNCT treatment; The treatment data set includes treatment plan data and treatment effect data of each historical treatment; Based on the causal relationship between each data in the treatment dataset in BNCT treatment, an initial BNCT knowledge graph is constructed, and a sample dataset corresponding to the initial BNCT knowledge graph is constructed; wherein the causal relationship includes a weighted causal relationship and a joint causal relationship; Extracting nodes that satisfy weighted causal relationships from each node corresponding to the treatment plan data as summation nodes; and extracting nodes that satisfy joint causal relationships from each node corresponding to the treatment effect data as product nodes; According to the sample data set, the sum-product network method is used to obtain the causal relationship probability corresponding to each sum node and the causal relationship probability corresponding to the product node to obtain the target BNCT knowledge graph.

2. The method for constructing a knowledge graph based on BNCT according to claim 1, characterized in that: The construction of a sample dataset corresponding to the initial BNCT knowledge graph includes: Based on the node information in the initial BNCT knowledge graph, extracting feature variables corresponding to the node information in the treatment dataset; Each characteristic variable belonging to the same treatment case is constructed as a sample data strip; The construction of each sample data strip is performed on the treatment dataset to obtain a sample dataset corresponding to the initial BNCT knowledge graph.

3. The method for constructing a knowledge graph based on BNCT according to claim 2, characterized in that: The implementation method of using the sum-product network method to obtain the causal relationship probability corresponding to each summation node and the causal relationship probability corresponding to the product node includes: The frequency weight calculation method is used to obtain the causal relationship probability of the summation root node corresponding to the summation node; The calculation method of the association effect probability is used to obtain the causal relationship probability of the product root node corresponding to the product node.

4. The method for constructing a knowledge graph based on BNCT according to claim 3, characterized in that: The method of obtaining the causal relationship probability of the summing root node corresponding to the summing node using the frequency weight calculation method includes: Extracting a sample subset corresponding to the current summing node from the sample data set, and obtaining the number of sample data pieces in the sample subset as the total number of samples of the current summing node; In the sample subset, extract the current number of samples corresponding to the current summation root node; Based on the current number of samples and the total number of samples, the ratio of the current sample data to the total number of samples is calculated as the causal relationship probability corresponding to the current summation root node.

5. The method for constructing a knowledge graph based on BNCT according to claim 3, characterized in that: The method for calculating the probability of the associated effect to obtain the probability of the causal relationship of the product root node corresponding to the product node includes: In the initial BNCT knowledge graph, a node group corresponding to the current product node is determined; based on the causal relationship, a treatment variable, a confounding variable, and an outcome variable are determined in the node group; Based on the causal relationship, constructing an association probability function between the outcome variable, the treatment variable and the confounding variable; Based on the sample data in the sample data subset, a relationship probability calculation is performed according to the association probability function to obtain a causal probability value corresponding to the product node group.

6. The method for constructing a knowledge graph based on BNCT according to claim 1, characterized in that: The method of extracting the summation node includes: For each node corresponding to the treatment plan data, determine whether it is a child node; If the node is a child node, determine whether there is a weighted causal relationship between the root nodes corresponding to the current child node and the current child node. If so, mark the child node as a summation node. And, the method of extracting the product node includes: For each node corresponding to the treatment effect data, determine whether the node is a child node; when If the node is a child node, determine whether there is a joint causal relationship between the root nodes corresponding to the current child node and the current child node. If so, mark the child node as a product node.

7. The method for constructing a knowledge graph based on BNCT according to claim 1, characterized in that: When the product node is the tumor volume change rate and the product root node is the boron drug dosage, the node group includes: neutron beam irradiation time, neutron beam energy, boron drug dosage, tumor boron concentration, tumor volume and tumor volume change rate; The boron drug dosage was set as the treatment variable, the neutron beam irradiation time and neutron beam energy were set as the confounding variables, and the tumor volume concentration, tumor volume and tumor volume change rate were set as the outcome variables; The association probability function between the outcome variable, the treatment variable and the confounding variable is constructed as follows: Where, P is the probability of the associated effect; P(U i1 ,U i2 |T i ) is the probability of the association effect under the influence of the confounding variable on the treatment variable, P(S i |T i ,U i1 ,U i2 ) is the probability of the associated effect of the treatment variable on the outcome variable under the influence of multiple confounding variables; U1 is the neutron beam irradiation time, U2 is the neutron beam energy; S2 is the tumor volume change rate; the tumor volume change rate is represented by a weighted combination or functional relationship of tumor volume concentration and tumor volume, satisfying: S2=α·S0·U2+β·U1·ln(S1) Among them, α is the influence coefficient of tumor boron concentration, β is the influence coefficient of irradiation time; S0 is the tumor boron concentration; S1 is the tumor volume.

8. A method for predicting BNCT treatment effects based on a knowledge graph, comprising: Obtain the pre-built target BNCT knowledge graph; Using the sample data, the target BNCT knowledge graph is trained to obtain a trained knowledge graph; Based on the trained knowledge graph, the tumor treatment effect of the patient under a given treatment plan is obtained; The target BNCT knowledge graph is obtained by using the BNCT-based knowledge graph construction method according to any one of claims 1 to 7.

9. A terminal, characterized in that: include: a processor and a memory, wherein the memory is communicatively connected to the processor; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes the BNCT-based knowledge graph construction method as described in any one of claims 1 to 7, or executes the BNCT treatment effect prediction method based on the knowledge graph as described in claim 8.

10. A computer storage medium storing a computer program, wherein: When the computer program is executed by a processor, it implements the BNCT-based knowledge graph construction method according to any one of claims 1 to 7, or implements the BNCT treatment effect prediction method based on the knowledge graph according to claim 8.

Images