WO2018103131A1 - 一种生物组织模拟材料的配方比例确定方法及系统 - Google Patents

一种生物组织模拟材料的配方比例确定方法及系统 Download PDF

Info

Publication number
WO2018103131A1
WO2018103131A1 PCT/CN2016/110365 CN2016110365W WO2018103131A1 WO 2018103131 A1 WO2018103131 A1 WO 2018103131A1 CN 2016110365 W CN2016110365 W CN 2016110365W WO 2018103131 A1 WO2018103131 A1 WO 2018103131A1
Authority
WO
WIPO (PCT)
Prior art keywords
biological tissue
neural network
formula
component
fitting function
Prior art date
Application number
PCT/CN2016/110365
Other languages
English (en)
French (fr)
Inventor
李景振
聂泽东
刘宇航
Original Assignee
深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳先进技术研究院 filed Critical 深圳先进技术研究院
Publication of WO2018103131A1 publication Critical patent/WO2018103131A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
    • G16B5/00—ICT specially adapted for modelling or simulations in systems biology, e.g. gene-regulatory networks, protein interaction networks or metabolic networks

Definitions

  • the invention relates to a preparation technology of a biological tissue simulation material, in particular to a method and a system for determining a formulation ratio in a preparation process of a biological tissue simulation material.
  • the object of the present invention is to provide a method for determining the proportion of a biological tissue simulation material, which can make the preparation of the biological tissue simulation material have the advantages of low cost, less time required, high precision, convenience and convenience. .
  • Another object of the present invention is to provide a formula ratio determining system for a biological tissue simulating material, which can make the preparation of the biological tissue simulating material have the advantages of low cost, high precision, less time required, and convenience.
  • the technical solution adopted by the present invention is: a method for determining a formula ratio of a biological tissue simulation material, and the steps of the method include:
  • the obtained dielectric property fitting function is input to the BP neural network for data processing, thereby outputting the proportional value of each component in the recipe.
  • the step of inputting the obtained dielectric characteristic fitting function into the BP neural network for data processing to output the proportional value of each component in the recipe is preceded by the step of establishing a BP neural network.
  • step of establishing a BP neural network includes:
  • training data of the BP neural network wherein the training data includes training input data and training output data, the training input data includes a dielectric property fitting function corresponding to each component in the formula, and a biological tissue simulation to be prepared A dielectric property fitting function of the material, the training output data includes a ratio optimal value of each component in the recipe, and then using the obtained BP neural network initial weight, initial threshold, and training data to train the BP neural network Until the training end requirement is met;
  • the BP neural network obtained after the training is used as the BP neural network to be established.
  • the initial weight and the initial threshold of the BP neural network are obtained by using a fish swarm algorithm.
  • the step of inputting the obtained dielectric characteristic fitting function into the BP neural network for data processing to output the proportional value of each component in the recipe is followed by a determining processing step, and the determining processing step is specifically:
  • the determining processing step specifically includes:
  • the input factor includes a dielectric property fitting function corresponding to each component in the formula and a ratio value of each component in the output recipe;
  • the output factor comprises a dielectric property error between the biological tissue simulated material to be prepared and the real biological tissue
  • the membership value with the largest value is selected from the evaluation matrix, and then it is judged whether the selected membership degree is less than the preset value, and if so, it indicates that the proportion value of each component in the output formula meets the preset judgment.
  • the standard which ends at this time, on the contrary, indicates that the ratio value of each component in the output formula does not meet the preset evaluation criteria, and returns to the step of re-executing the establishment of the BP neural network.
  • the dielectric property fitting function corresponding to the component is a curve function fitted by the dielectric properties of the components at different frequencies.
  • the dielectric properties include a dielectric constant and/or an electrical conductivity.
  • a formula ratio determining system for a biological tissue simulation material comprising:
  • the obtaining unit is configured to obtain a dielectric property fitting function corresponding to each component in the formula and a dielectric property fitting function of the biological tissue simulation material to be prepared;
  • a processing unit configured to input the obtained dielectric property fitting function into the BP neural network for data processing, thereby outputting a proportional value of each component in the formula.
  • processing unit is preceded by an establishing unit for establishing a BP neural network.
  • the establishing unit includes:
  • a training module configured to acquire training data of a BP neural network, wherein the training data includes training input data and training output data, the training input data includes a dielectric characteristic fitting function corresponding to each component in the formula, and a required
  • the prepared biological tissue simulates a dielectric property fitting function of the material, the training output data includes a ratio optimal value of each component in the formula, and then uses the obtained initial weight, initial threshold, and training data of the BP neural network, thereby BP neural network is trained until the training end requirement is met;
  • the BP neural network acquisition module is used to use the BP neural network obtained after the training is completed as the BP neural network to be established.
  • the initial weight and the initial threshold of the BP neural network are obtained by using a fish swarm algorithm.
  • processing unit is further provided with a determination processing unit, wherein the determination processing unit is configured to determine whether the ratio value of each component in the output recipe meets a preset criterion, and if so, ends, otherwise, returns to re-execution establishment.
  • the determining processing unit specifically includes:
  • An input factor obtaining module configured to obtain an input factor, wherein the input factor includes a dielectric property fitting function corresponding to each component in the formula and a proportional value of each component in the output recipe;
  • An output factor acquisition module configured to obtain an output factor, wherein the output factor includes a dielectric property error between the biological tissue simulation material to be prepared and the real biological tissue;
  • a judging matrix establishing module configured to establish a judging matrix from an input factor to an output factor according to the obtained input factor and output factor
  • a judging processing module configured to select a membership value having the largest value from the evaluation matrix according to the maximum membership principle, and then determine whether the selected membership degree is less than a preset value, and if so, a ratio of each component in the output recipe The value meets the preset criterion and ends at this time. Otherwise, it indicates that the ratio value of each component in the output formula does not meet the preset criterion, and returns to the data processing flow corresponding to the re-execution unit.
  • the dielectric property fitting function corresponding to the component is a curve function fitted by the dielectric properties of the components at different frequencies.
  • the dielectric properties include a dielectric constant and/or an electrical conductivity.
  • the beneficial effects of the present invention are: by using the method of the invention, in the preparation process, only the dielectric property fitting function corresponding to each component in the formulation needs to be acquired and acquired, and the known biological tissue to be prepared is obtained. After simulating the dielectric property fitting function of the materials, they are input into the trained BP neural network for data processing, and the ratio values of the respective components can be obtained, so that the producer can obtain the ratio value of each component according to the output. Realize the production of biological tissue simulation materials. It can be seen that the method of the present invention does not need to involve excessive human operation, compared to the conventional formula ratio determination scheme. It also does not require multiple experiments, which makes the preparation of biological tissue simulation materials have the advantages of low cost, less time required, high precision, convenience and quickness.
  • Another beneficial effect of the present invention is that by using the system of the present invention, in the preparation process, only the acquisition unit is required to obtain the dielectric property fitting function corresponding to each component in the formulation, and to obtain the known desired preparation.
  • the biological tissue simulates the dielectric property fitting function of the material, and then the processing unit inputs the obtained dielectric property fitting function into the trained BP neural network for data processing, thereby outputting the ratio value of each component in the formula, so that the producer
  • the production of biological tissue simulation materials can be achieved based on the ratio of the components output. It can be seen that compared with the traditional formula ratio determination scheme, the system of the invention does not need to involve too many human operations, and does not need multiple experiments, so that the preparation of the biological tissue simulation material is low in cost, less time required, and accurate. High, convenient and fast.
  • FIG. 1 is a schematic flow chart showing the steps of a method for determining a formulation ratio of a biological tissue simulation material according to the present invention
  • FIG. 2 is a schematic flow chart showing the steps of a specific embodiment of a method for determining a formulation ratio of a biological tissue simulation material according to the present invention
  • FIG. 3 is a schematic flow chart showing the steps of a specific embodiment of the step of establishing a BP neural network in FIG. 2;
  • FIG. 4 is a schematic flow chart showing the steps of a specific embodiment of the determining processing step of FIG. 2;
  • FIG. 5 is a schematic structural block diagram of a formula ratio determining system of a biological tissue simulation material according to the present invention.
  • FIG. 6 is a structural block diagram of a specific embodiment of a formula ratio determining system for a biological tissue simulating material according to the present invention.
  • Figure 7 is a structural block diagram of a specific embodiment of the establishing unit of Figure 6;
  • FIG. 8 is a structural block diagram of a specific embodiment of the determination processing unit in FIG. 6.
  • a method for determining a formulation ratio of a biological tissue simulation material the steps of the method include:
  • the obtained dielectric property fitting function is input to the BP neural network for data processing, thereby outputting the proportional value of each component in the recipe.
  • the formulation refers to a preparation formula of a biological tissue simulation material to be prepared.
  • a dielectric property fitting function corresponding to the component is a curve function fitted by a dielectric property of the component at different frequencies, that is, the The composition will correspond to different dielectric properties at different frequencies, and the curve function obtained by curve fitting the dielectric properties corresponding to the different frequencies will be the fitting of the dielectric properties corresponding to the components. function.
  • a dielectric property fitting function for the biological tissue mimic material to be prepared is a curve function fitted by the dielectric properties of the biological tissue mimic material to be prepared at different frequencies.
  • the dielectric property includes a dielectric constant and/or electrical conductivity, that is, a dielectric property fitting function corresponding to the component, which specifically includes a dielectric constant fitting. Function and / or conductivity fitting function. It can be seen from the above that for the dielectric constant fitting function corresponding to the component, the specific components correspond to different dielectric constants at different frequencies, and the dielectric constants corresponding to the different frequencies are performed. Curve function obtained after curve fitting; likewise, for the conductivity fitting function, specifically, the components correspond to different conductivity at different frequencies, and the conductance corresponding to the different frequencies The curve function obtained after the curve fitting.
  • a dielectric property fitting function for the biological tissue simulation material to be prepared specifically includes a dielectric constant fitting function and a conductivity fitting function of the biological tissue simulation material to be prepared.
  • the obtained dielectric characteristic fitting function is input into the BP neural network for data processing, thereby outputting a BP neural value before the step of outputting the proportional value of each component in the formula. Network this step.
  • the step of establishing the BP neural network and the step of fitting the dielectric property fitting function corresponding to each component in the formulation and the dielectric property fitting function of the biological tissue simulation material to be prepared, they are The sequential logical sequence between them may be adjusted according to actual needs.
  • the step of establishing a BP neural network is to set a dielectric property fitting function corresponding to each component in the acquisition formula and a biological tissue to be prepared. Dielectric of simulated material Feature fitting function before this step.
  • the step of inputting the dielectric characteristic fitting function into the BP neural network for data processing to output the proportional value of each component in the recipe is followed by a determining processing step.
  • the determining processing step is specifically:
  • Example 1 Acquisition of the fitting function of dielectric properties corresponding to each component in the formulation
  • the components contained are deionized water, gelatin, agar, sodium chloride, aluminum powder, edible oil, detergent, and a total of seven components.
  • the seven components in addition to participating in the adjustment of the dielectric properties of the biological tissue simulating material, they also have other functions, such as: 1, deionized water, which is used to dissolve gelatin, agar and sodium chloride; 2, gelatin, which acts as a gel; 3, agar, which is used to increase the melting point of biological tissue simulation materials; 4, sodium chloride and aluminum powder, which can significantly change the conductivity of biological tissue simulation materials; 5. Edible oil, which can significantly change the dielectric constant of biological tissue simulation materials; 6.
  • the dielectric property fitting function includes a dielectric constant fitting function and a conductivity fitting function.
  • the obtaining step preferably includes:
  • the first step is to collect the dielectric constant and conductivity of each component at different frequencies
  • dielectric properties ie dielectric constant and conductivity
  • impedance analyzers and coaxial probes For each component in the formulation, their dielectric properties, ie dielectric constant and conductivity, can be measured by impedance analyzers and coaxial probes. Specific measurement methods include: 1. For deionized water, cooking oil and detergent Agent, since all three components are liquid, the coaxial probe can be separately immersed in these three components, and then the dielectric constant and conductivity at different frequencies can be obtained by an impedance analyzer; Sodium and aluminum powders, because of their powder form, can directly press the coaxial probe on the powder to obtain the dielectric constant and conductivity at different frequencies; 3.
  • the dielectric constant and conductivity of each component at different frequencies should be repeatedly measured more than 20 times, and the number of data points acquired each time is 800;
  • the dielectric constant and the electrical conductivity of the obtained components at different frequencies are fitted to obtain a dielectric constant fitting function and a conductivity fitting function corresponding to each component;
  • the least squares method is preferably used to fit the dielectric constant and conductivity of each component at different frequencies, and the specific fitting steps are as follows:
  • Step 1 According to the data obtained by the impedance analyzer, draw a rough graphic scatter plot, and determine the number n of fitting polynomials according to the graphical scatter plot;
  • Step 3 According to the first formula above, It is a 0, a 1, a 2 ... a n polyvalent function, depending on the conditions polyhydric extremum function of a 0, a 1, a 2 ... a n for calculating the partial derivatives, of available
  • the second formula is as follows:
  • the third formula is a linear equation for a 0 , a 1 , a 2 ... a n , and by solving the system of equations, a 0 , a 1 can be obtained respectively. , the value of a 2 ... a n ;
  • Step 5 According to the values of a 0 , a 1 , a 2 ... a n , the dielectric constant fitting function and the conductivity fitting function corresponding to each component can be obtained, as shown in the fourth formula:
  • the dielectric constant fitting function and the conductivity fitting function of the biological tissue simulation material are all known parameters and can be directly obtained.
  • Example 2 Formulation ratio determination method of biological tissue simulation material
  • a method for determining the formulation ratio of a biological tissue simulation material has the following specific steps.
  • Step S1 Establish a BP neural network.
  • the step S1 includes:
  • the BP neural network includes an input layer, an implicit layer and an output layer, and the input signal is transmitted from the input layer node to each hidden layer node, and finally to the output layer output;
  • the number of nodes of the input layer is set to 16, respectively, corresponding to deionized water, gelatin, agar, sodium chloride, aluminum powder, edible oil, detergent, and the corresponding components of the seven components.
  • a dielectric constant fitting function and a conductivity fitting function, and a dielectric constant fitting function and a conductivity fitting function of the biological tissue simulation material to be prepared that is, the training input data of the BP neural network includes each of the formulas a dielectric constant fitting function and a conductivity fitting function corresponding to the composition, and a dielectric constant fitting function and a conductivity fitting function of the biological tissue simulation material to be prepared; setting the node of the output layer to 1,
  • the value is the optimal ratio of the formula, that is, the training output data of the BP neural network is the optimal ratio of each component in the formula; the number of layers of the hidden layer is set to 1, and the number of nodes is preferably determined by the following formula: :
  • I is the number of nodes of the input layer.
  • the value is 16, and J is the number of nodes of the hidden layer. Therefore, the size of J is determined to be 4;
  • the initial weight and the initial threshold of the BP neural network are preferably obtained by using a fish swarm algorithm, and the specific obtaining steps include:
  • S1022 Initializing parameters of the fish swarm algorithm, wherein the parameters to be initialized by the fish swarm algorithm include the fish size of the fish size, the perceived distance of the artificial fish fish_dis tan ge, the maximum step size of the artificial fish moving fish_step, the congestion factor ⁇ , the maximum The number of iterations max_gen and the target value G;
  • the state F i is re-randomly selected to determine whether the forward condition is satisfied; if the preset number of trials is repeated, if the condition is still not satisfied, the sixth formula below is executed, and the sixth formula is as follows:
  • x i,j (k+1) x i,j (k)+rand ⁇ step(x center,j -x i,j (k))
  • x i,j (k+1) x i,j (k)+rand ⁇ step(x max,j -x i,j (k))
  • S1026 Determine whether the selected food concentration value is greater than a food concentration value on the bulletin board. If yes, replace the food concentration value on the bulletin board with the selected food concentration value, and vice versa, the food concentration value on the bulletin board. constant;
  • S1027 Determine whether the suspension condition is met. Specifically, determine whether the number of iterations reaches the maximum number of iterations or determine whether the accuracy error of the solution has been met. If yes, the individual information corresponding to each artificial fish is used as the initial weight and initiality obtained. Threshold, otherwise, returning to step S1025 again until the abort condition is satisfied;
  • the BP neural network obtained after the training is completed is used as a BP neural network to be established.
  • Step S2 obtaining a dielectric constant fitting function and a conductivity fitting function for each component currently prepared when preparing the desired biological tissue simulation material, and obtaining a known biological tissue simulation material to be prepared. Corresponding dielectric constant fitting function and conductivity fitting function, and then inputting the obtained dielectric constant fitting function and conductivity fitting function into the above-mentioned established BP neural network for data processing and outputting each component Proportional value.
  • step S2 is followed by a determining process step S3, which is specifically:
  • step S102 Determining whether the ratio value of each component in the output formula meets the preset criterion, and if so, ending, the ratio of the formula of the output is the final determined formula ratio, and vice versa, returning to the step of re-executing the establishment of the BP neural network
  • the fish population algorithm is used to recalculate the new initial weight and the initial threshold, and then the BP neural network is retrained based on the new initial weight and the initial threshold until the training end requirement is met, and then the The dielectric property fitting function is input to the newly trained BP neural network for data processing;
  • the BP neural network training data is re-acquired, the fish population algorithm is used to recalculate the new initial weight and initial threshold, and then the BP neural network is retrained based on the new training data, initial weight and initial threshold. Until the training end requirement is met, the obtained dielectric property fitting function is then input into the newly trained BP neural network for data processing.
  • step S3 As a preferred implementation of the foregoing step S3, as shown in FIG. 4, it specifically includes:
  • the input factor includes a dielectric constant fitting function corresponding to each component in the formula, a conductivity fitting function, and a proportional value of each component in the output formula; specifically, the input factor Can be expressed as:
  • ⁇ u 1 , u 2 , u 3 ⁇ are respectively the dielectric constant fitting function, the conductivity fitting function and the proportion value of the deionized water
  • the dielectric constant fitting function, the conductivity fitting function and the proportion of the ratio corresponding to gelatin, respectively, ⁇ u 7 , u 8 , u 9 ⁇ are respectively the dielectric constant fitting function and conductance corresponding to agar.
  • the rate fitting function and its proportion value, ⁇ u 10 , u 11 , u 12 ⁇ are respectively the dielectric constant fitting function, the conductivity fitting function and the proportion value of sodium chloride.
  • ⁇ u 13 , u 14 , u 15 ⁇ are respectively the dielectric constant fitting function, the conductivity fitting function and the proportion of the ratio of the aluminum powder, ⁇ u 16 , u 17 , u 18 ⁇ respectively
  • the dielectric constant fitting function, the conductivity fitting function and the proportion of the oil corresponding to the oil, ⁇ u 19 , u 20 , u 21 ⁇ are respectively the dielectric constant fitting function and conductance corresponding to the detergent. Rate fitting function and its proportion value;
  • the output factor includes a dielectric characteristic error between the biological tissue simulation material to be prepared and the real biological tissue; specifically, the output factor may be expressed as:
  • V ⁇ v 1 , v 2 , v 3 , v 4 , v 5 , v 6 , v 7 , v 8 , v 9 , v 10 ⁇
  • v 1 represents an error of less than 10%
  • v 2 represents a representative error of less than 20%
  • v 3 represents an error of less than 30%, and so on;
  • step S304 it is determined whether the evaluation matrix needs to be normalized, if yes, the evaluation matrix is normalized and then step S305 is performed; otherwise, step S305 is directly executed;
  • the maximum membership degree is selected from the evaluation matrix according to the maximum membership principle, and then determining whether the selected membership degree is less than a preset value.
  • the preset value is 20%, and if so, It means that the ratio value of each component in the output formula meets the preset criterion, that is, the ratio value of each component in the formula output by the BP neural network is reasonable, and at this time, the formula ratio of the output is final. Determine the proportion of the obtained formula, the producer can prepare the required biological tissue simulation material according to the ratio of the formula, and vice versa, the ratio value of each component in the output formula does not meet the preset evaluation standard.
  • the fish population algorithm is used to recalculate the new initial weight and the initial threshold, and then the BP neural network is re-trained based on the new initial weight and the initial threshold until the training end requirement is met, and the new Train the BP neural network to re-process, or return to re-execute the step of establishing BP neural network S1, re-acquire BP God
  • the fish population algorithm is used to recalculate the new initial weight and initial threshold, and then BP neural network is retrained based on the new training data, initial weight and initial threshold until the training end requirement is met. So far, the newly trained BP neural network is used to re-process.
  • Example 3 Formulation ratio determination system for biological tissue simulation materials
  • a determination system of the biological tissue simulation material as shown in FIG. 5, which includes:
  • the obtaining unit 401 is configured to acquire a dielectric property fitting function corresponding to each component in the formula and a dielectric property fitting function of the biological tissue simulation material to be prepared;
  • the processing unit 402 is configured to input the obtained dielectric property fitting function into the BP neural network for data processing, thereby outputting a proportional value of each component in the formula.
  • the acquisition unit 401 and the processing unit 402 described they can implement their functions through a processor.
  • the dielectric property fitting function corresponding to the component is a curve function fitted by the dielectric properties of the components at different frequencies.
  • the dielectric properties include a dielectric constant and/or conductivity.
  • the processing unit 402 is previously provided with an establishing unit 400 for establishing a BP neural network.
  • the establishing unit 400 For the order relationship between the establishing unit 400 and the obtaining unit 401, their sequential logical order may be set according to actual conditions, and preferably, the establishing unit 400 is disposed before the obtaining unit 401.
  • the establishing unit 400 includes:
  • the training module 4001 is configured to acquire training data of the BP neural network, where the training data includes training input data and training output data, where the training input data includes a dielectric characteristic fitting function corresponding to each component in the formula and the a dielectric property fitting function of the biological tissue simulation material to be prepared, the training output data including a ratio optimal value of each component in the formula, and then using the obtained initial weight, initial threshold, and training data of the BP neural network, thereby Train the BP neural network until the training end requirement is met;
  • the BP neural network acquisition module 4002 is configured to use the BP neural network obtained after the training is completed as the BP neural network to be established.
  • the initial weight and the initial threshold of the BP neural network are obtained by using a fish swarm algorithm.
  • the processing unit 402 is further provided with a determination processing unit 403, which is configured to determine whether the ratio values of the components in the output recipe are consistent.
  • the preset criterion if yes, ends, otherwise, returns to the data processing flow corresponding to the re-execution unit 400.
  • the determining processing unit 403 specifically includes:
  • the input factor obtaining module 4031 is configured to obtain an input factor, where the input factor includes a dielectric characteristic fitting function corresponding to each component in the formula and a proportional value of each component in the output recipe;
  • An output factor acquisition module 4032 configured to obtain an output factor, wherein the output factor includes a dielectric characteristic error between the biological tissue simulation material to be prepared and the real biological tissue;
  • the evaluation matrix establishing module 4033 is configured to establish an evaluation matrix from the input factor to the output factor according to the obtained input factor and the output factor;
  • the determining processing module 4034 is configured to select, according to the maximum membership principle, the membership degree with the largest value from the evaluation matrix, and then determine whether the selected membership degree is less than a preset value, and if yes, indicate the components of the output formula.
  • the proportional value meets the preset criterion and ends at this time. Otherwise, it indicates that the ratio value of each component in the output formula does not meet the preset criterion, and returns to the data processing flow corresponding to the re-execution unit 400.
  • the determining processing module 4034 is specifically configured to determine whether the evaluation matrix needs to be normalized. If yes, the evaluation matrix is normalized and then the evaluation step is performed. Otherwise, the evaluation step is directly performed, and the evaluation is performed.
  • the step is specifically: according to the maximum membership principle, selecting the membership degree with the largest value from the evaluation matrix, and then determining whether the selected membership degree is less than a preset value, and if so, indicating that the ratio of each component in the output formula matches The preset criterion is ended at this time. Otherwise, it means that the ratio value of each component in the output formula does not meet the preset criterion, and the data processing flow corresponding to the re-execution unit 400 is returned, and the fish group algorithm is used for recalculation.
  • the new initial weight and the initial threshold are obtained, and then the BP neural network is retrained based on the new initial weight and the initial threshold until the training end requirement is met, or the training data of the BP neural network is reacquired, and the fish population is used.
  • the algorithm recalculates the new initial weight and initial threshold, and then based on the new training number
  • the initial weights and the initial threshold, re-BP neural network is trained to meet the training requirements so far until the end.
  • the creator only needs to fit the dielectric property fitting function corresponding to each component in the collected formula, and the desired preparation.
  • the biological tissue simulation material's dielectric property fitting function is input into the trained BP neural network for data processing, and then the ratio value of each component can be obtained, so that the producer can realize the biological according to the obtained formula ratio.
  • the present invention adopts a dielectric constant fitting function and a conductivity fitting function corresponding to each component, and a dielectric constant fitting function and a conductivity fitting function of the biological tissue simulation material to be prepared, as BP.
  • the training input data of the neural network can make the proportional value output by the BP neural network closer to the actual situation in the subsequent application, the application is more practical, and the dielectric constant fitting function and conductivity fitting corresponding to each component are fitted.
  • the acquisition step of the function is also relatively simple, and the operation is convenient to implement.
  • using the fish swarm algorithm to obtain the initial threshold and weight of the BP neural network can make the training process of the BP neural network faster; using the evaluation matrix It is reasonable to judge the proportional value of each component of the output, which is not only simple in steps, easy to implement, but also more accurate in judgment.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Physiology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biotechnology (AREA)
  • Evolutionary Biology (AREA)
  • Medical Informatics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Investigating Or Analyzing Materials By The Use Of Electric Means (AREA)

Abstract

一种生物组织模拟材料的配方比例确定方法及系统，系统包括获取单元(401)和处理单元(402)，方法包括：首先获取配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数，然后将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值(S2)。通过使用生物组织模拟材料的配方比例确定方法及系统，在制备生物组织模拟材料的过程中便无需涉及过多的人为操作，也无需多次实验，从而达到制备具有成本低廉、所需时间少、精准度高、方便快捷等效果，可广泛应用于生物组织模拟材料的制备领域中。

Description

一种生物组织模拟材料的配方比例确定方法及系统 技术领域
本发明涉及生物组织模拟材料的制备技术，尤其涉及一种在生物组织模拟材料的制备过程中的配方比例确定方法及系统。
背景技术
模拟生物组织介电特性(如介电常数、电导率)的模拟材料在生物电磁剂量学、磁共振成像、超声成像以及其它医学领域发展中起到了至关重要的作用。同时，随着科技的迅猛发展，生物组织模拟材料的需求量越来越大，因此，对生物组织模拟材料的制备效率要求越来越高。然而，目前对于制备生物组织模拟材料的方法，其主要是采用重复试验法，即通过制备多种样品后，通过多次实验才能确定生物组织模拟材料的制备配方中各个成分的比例，这样则会存在制备成本过高、实验时间长而导致制备效率低下、易受环境干扰而导致精度低下等缺陷，难以满足越来越大的需求量，并且影响了生物组织模拟材料的进一步发展和应用。
发明内容
为了解决上述的技术问题，本发明的目的是提供一种生物组织模拟材料的配方比例确定方法，使生物组织模拟材料的制备可具有成本低廉、所需时间少、精准度高、方便快捷等优点。
本发明的另一目的是提供一种生物组织模拟材料的配方比例确定系统，使生物组织模拟材料的制备可具有成本低廉、精度高、所需时间少、方便快捷等优点。
本发明所采取的技术方案是：一种生物组织模拟材料的配方比例确定方法，该方法的步骤包括有：
获取配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数；
将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值。
进一步，所述将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值这一步骤之前设有建立BP神经网络这一步骤。
进一步，所述建立BP神经网络这一步骤，其包括：
获取BP神经网络的训练数据，其中，所述训练数据包括训练输入数据和训练输出数据，所述训练输入数据包括配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数，所述训练输出数据包括配方中各个成分的比例最优值，然后利用获得的BP神经网络的初始权值、初始阈值以及训练数据，从而对BP神经网络进行训练，直到满足训练结束要求；
将训练结束后得到的BP神经网络作为所需建立的BP神经网络。
进一步，所述BP神经网络的初始权值和初始阈值是采用鱼群算法来获取的。
进一步，所述将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值这一步骤之后设有判断处理步骤，所述判断处理步骤具体为：
判断输出的配方中各个成分的比例值是否符合预设的评判标准，若是，则结束，反之，则返回重新执行建立BP神经网络这一步骤。
进一步，所述判断处理步骤具体包括：
获取输入因子，其中，所述输入因子包括配方中各个成分所对应的介电特性拟合函数以及输出的配方中各个成分的比例值；
获取输出因子，其中，所述输出因子包括所需制备的生物组织模拟材料与真实的生物组织之间的介电特性误差；
根据获取的输入因子和输出因子，建立从输入因子到输出因子的评判矩阵；
根据最大隶属原则，从评判矩阵中选出数值最大的隶属度，然后判断所述选出的隶属度是否小于预设值，若是，则表示输出的配方中各个成分的比例值符合预设的评判标准，此时结束，反之，则表示输出的配方中各个成分的比例值不符合预设的评判标准，返回重新执行建立BP神经网络这一步骤。
进一步，所述成分所对应的介电特性拟合函数是由成分在不同频率下的介电特性所拟合出的曲线函数。
进一步，所述介电特性包括介电常数和/或电导率。
本发明所采取的另一技术方案是：一种生物组织模拟材料的配方比例确定系统，该系统包括有：
获取单元，用于获取配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数；
处理单元，用于将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值。
进一步，所述处理单元之前设有用于建立BP神经网络的建立单元。
进一步，所述建立单元包括：
训练模块，用于获取BP神经网络的训练数据，其中，所述训练数据包括训练输入数据和训练输出数据，所述训练输入数据包括配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数，所述训练输出数据包括配方中各个成分的比例最优值，然后利用获得的BP神经网络的初始权值、初始阈值以及训练数据，从而对BP神经网络进行训练，直到满足训练结束要求；
BP神经网络获取模块，用于将训练结束后得到的BP神经网络作为所需建立的BP神经网络。
进一步，所述BP神经网络的初始权值和初始阈值是采用鱼群算法来获取的。
进一步，所述处理单元之后设有判断处理单元，所述判断处理单元用于判断输出的配方中各个成分的比例值是否符合预设的评判标准，若是，则结束，反之，则返回重新执行建立单元所对应的数据处理流程。
进一步，所述判断处理单元具体包括：
输入因子获取模块，用于获取输入因子，其中，所述输入因子包括配方中各个成分所对应的介电特性拟合函数以及输出的配方中各个成分的比例值；
输出因子获取模块，用于获取输出因子，其中，所述输出因子包括所需制备的生物组织模拟材料与真实的生物组织之间的介电特性误差；
评判矩阵建立模块，用于根据获取的输入因子和输出因子，建立从输入因子到输出因子的评判矩阵；
判断处理模块，用于根据最大隶属原则，从评判矩阵中选出数值最大的隶属度，然后判断所述选出的隶属度是否小于预设值，若是，则表示输出的配方中各个成分的比例值符合预设的评判标准，此时结束，反之，则表示输出的配方中各个成分的比例值不符合预设的评判标准，返回重新执行建立单元所对应的数据处理流程。
进一步，所述成分所对应的介电特性拟合函数是由成分在不同频率下的介电特性所拟合出的曲线函数。
进一步，所述介电特性包括介电常数和/或电导率。
本发明的有益效果是：通过使用本发明的方法，在制备过程中，仅需要对配方中各个成分所对应的介电特性拟合函数进行采集获取，以及获取已知的所需制备的生物组织模拟材料的介电特性拟合函数后,将它们输入至训练好的BP神经网络中进行数据处理，便能获得各个成分的比例值，这样制作者根据所输出的各个成分的比例值，便可实现生物组织模拟材料的制作。由此可见，相较于传统的配方比例确定方案，本发明的方法无需涉及过多的人为操作， 也无需多次实验，令生物组织模拟材料的制备具有成本低廉、所需时间少、精准度高、方便快捷等优点。
本发明的另一有益效果是：通过使用本发明的系统，在制备过程中，仅需要获取单元获取得到配方中各个成分所对应的介电特性拟合函数，以及获取已知的所需制备的生物组织模拟材料的介电特性拟合函数，然后处理单元将获得的介电特性拟合函数输入至训练好的BP神经网络中进行数据处理，从而输出配方中各个成分的比例值，这样制作者根据所输出的各个成分的比例值，便可实现生物组织模拟材料的制作。由此可见，相较于传统的配方比例确定方案，本发明的系统无需涉及过多的人为操作，也无需多次实验，令生物组织模拟材料的制备具有成本低廉、所需时间少、精准度高、方便快捷等优点。
附图说明
图1是本发明一种生物组织模拟材料的配方比例确定方法的步骤流程示意图；
图2是本发明一种生物组织模拟材料的配方比例确定方法的一具体实施例步骤流程示意图；
图3是图2中建立BP神经网络这一步骤的一具体实施例步骤流程示意图；
图4是图2中判断处理步骤的一具体实施例步骤流程示意图；
图5是本发明一种生物组织模拟材料的配方比例确定系统的结构框示意图；
图6是本发明一种生物组织模拟材料的配方比例确定系统的一具体实施例结构框示意图；
图7是图6中所述建立单元的一具体实施例结构框示意图；
图8是图6中所述判断处理单元的一具体实施例结构框示意图。
具体实施方式
如图1所示，一种生物组织模拟材料的配方比例确定方法，该方法的步骤包括有：
获取配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数；
将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值。其中，所述的配方是指所需制备的生物组织模拟材料的制备配方。
进一步作为本方法实施例的优选实施方式，对于所述的成分所对应的介电特性拟合函数，其是由成分在不同频率下的介电特性所拟合出的曲线函数，即所述的成分在不同频率下会对应不同的介电特性，对这些在不同频率下所对应的介电特性进行曲线拟合后所得出的曲线函数，其便为所述成分所对应的介电特性拟合函数。同样，对于所述所需制备的生物组织模拟材料的介电特性拟合函数，其是由所需制备的生物组织模拟材料在不同频率下的介电特性所拟合出的曲线函数。
进一步作为本方法实施例的优选实施方式，所述介电特性包括介电常数和/或电导率，即对于所述成分所对应的介电特性拟合函数，其具体包括有介电常数拟合函数和/或电导率拟合函数。由上述可知，对于所述成分所对应的介电常数拟合函数，其具体为，所述的成分在不同频率下对应不同的介电常数，对这些在不同频率下所对应的介电常数进行曲线拟合后所得出的曲线函数；同样，对于所述的电导率拟合函数，其具体为，所述的成分在不同频率下对应不同的电导率，对这些在不同频率下所对应的电导率进行曲线拟合后所得出的曲线函数。另外，对于所述所需制备的生物组织模拟材料的介电特性拟合函数，其具体包括所需制备的生物组织模拟材料的介电常数拟合函数和电导率拟合函数。
进一步作为本方法实施例的优选实施方式，所述将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值这一步骤之前设有建立BP神经网络这一步骤。而对于所述建立BP神经网络这一步骤，以及所述获取配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数这一步骤，它们之间的先后逻辑顺序可根据实际需求来调整，优选地，所述建立BP神经网络这一步骤设置在所述获取配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电 特性拟合函数这一步骤之前。
进一步作为本方法实施例的优选实施方式，所述将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值这一步骤之后设有判断处理步骤，所述判断处理步骤具体为：
判断输出的配方中各个成分的比例值是否符合预设的评判标准，若是，则结束，反之，则返回重新执行建立BP神经网络这一步骤。
实施例1、配方中各个成分所对应的介电特性拟合函数的获取
在本实施例中，对于制备生物组织模拟材料的配方，其优选确定所包含的成分有去离子水、明胶、琼脂、氯化钠、铝粉、食用油、洗洁剂，共7种成分。其中，对于所述的7种成分，它们除了参与调节生物组织模拟材料的介电特性外，还分别具有其它功能，例如：1、去离子水，其用于溶解明胶、琼脂和氯化钠；2、明胶，其起到凝胶剂的作用；3、琼脂，其用于提高生物组织模拟材料的熔点；4、氯化钠和铝粉，它们能显著地改变生物组织模拟材料的电导率；5、食用油，其能显著地改变生物组织模拟材料的介电常数；6、洗洁剂，其用于降低去离子水与食用油之间的表面张力，促进配方中各成分之间的融合。另外，在本实施例中，所述介电特性拟合函数包括介电常数拟合函数和电导率拟合函数。
对于上述配方中的7种成分，它们各自所对应的介电常数拟合函数和电导率拟合函数，其获取步骤优选包括有：
第一步、采集各个成分在不同频率下的介电常数和电导率；
对于配方中的各个成分，它们的介电特性，即介电常数和电导率，可通过阻抗分析仪和同轴探头来测量，具体测量方法包括：1、对于去离子水、食用油和洗洁剂，由于这三种成分均为液体，因此可分别将同轴探头浸入到这三种成分中，然后通过阻抗分析仪来获取其在不同频率下的介电常数和电导率；2、对于氯化钠和铝粉，由于其形态为粉末状，因此可直接将同轴探头按压在粉末上面，从而获得其在不同频率下的介电常数和电导率；3、对于明胶和琼脂，由于其为固体，且其表面粗糙，因此需将其碾碎为粉末状后按照上述对于氯化钠和铝粉的测量方法来进行测量，从而获得它们在不同频率下的介电常数和电导率；
优选地，在上述采集过程中，每种成分在不同频率下的介电常数和电导率应重复测量20次以上，并且每次所获取的数据点的数目为800个；
第二步、对获得的各个成分在不同频率下的介电常数和电导率进行拟合，从而获得各个成分所对应的介电常数拟合函数和电导率拟合函数；
其中，优选采用最小二乘法来对各个成分在不同频率下的介电常数和电导率进行拟合，具体拟合步骤如下所示：
步骤1.根据通过阻抗分析仪所获得的数据，画出粗略的图形散点图，根据图形散点图确定拟合多项式的次数n；
步骤2.根据需拟合的数据点(xi,yj)，其中i＝0,1,..m，m为采集的数据点的数目，如800，然后计算
Figure PCTCN2016110365-appb-000001
令
Figure PCTCN2016110365-appb-000002
即令第一公式
Figure PCTCN2016110365-appb-000003
为最小值；
步骤3.根据上述第一公式可知，
Figure PCTCN2016110365-appb-000004
为a0,a1,a2...an的多元函数，根据多元函 数求极值的条件，对a0,a1,a2...an进行求偏导运算，可得第二公式如下所示：
Figure PCTCN2016110365-appb-000005
即有第三公式如下所示：
Figure PCTCN2016110365-appb-000006
步骤4.根据上述第三公式可知，第三公式是关于a0,a1,a2...an的线性方程组，通过对该方程组进行求解，便可分别得到a0,a1,a2...an的值；
步骤5.根据a0,a1,a2...an的值，便可获得各成分所对应的介电常数拟合函数和电导率拟合函数，如第四公式所示：
Figure PCTCN2016110365-appb-000007
可见，根据上述步骤1-5以及所获得的7种成分在不同频率下的介电常数和电导率，便能分别对去离子水、明胶、琼脂、氯化钠、铝粉、食用油、洗洁剂等七种成分的介电常数和电导率进行最小二乘法拟合，得到相应的拟合函数。
另外，对于所述所需制备的生物组织模拟材料的介电常数拟合函数和电导率拟合函数，它们均为已知参数，直接获取便可。
实施例2、生物组织模拟材料的配方比例确定方法
如图2所示，一种生物组织模拟材料的配方比例确定方法，其具体步骤如下所示。
步骤S1、建立BP神经网络。
如图3所示，所述步骤S1包括：
S101、BP神经网络的设定；
所述BP神经网络包括有输入层、隐含层和输出层，输入信号从输入层节点，依次传到各隐含层节点，最后传到输出层输出；
在本实施例中，设定输入层的节点数为16，分别一一对应去离子水、明胶、琼脂、氯化钠、铝粉、食用油、洗洁剂，这7种成分各自所对应的介电常数拟合函数和电导率拟合函数，以及所需制备的生物组织模拟材料的介电常数拟合函数和电导率拟合函数，即所述BP神经网络的训练输入数据包括配方中各个成分所对应的介电常数拟合函数和电导率拟合函数，以及所需制备的生物组织模拟材料的介电常数拟合函数和电导率拟合函数；设定输出层的节点为1，其值为配方比例最佳值，即所述BP神经网络的训练输出数据为配方中各个成分的比例最优值；设定隐含层的层数为1，其节点数目则优选采用以下公式来确定：
J＝log2(I)
其中，I为输入层的节点数目，本实施例中，其值为16，J为隐含层的节点数目，因此，J的大小确定为4；
S102、获取BP神经网络的初始权值和初始阈值，然后利用获得的BP神经网络的初始权值、初始阈值以及训练数据，从而对上述设定的BP神经网络进行训练，直到满足训练结束要求为止；
其中，对于所述BP神经网络的初始权值和初始阈值，其优选采用鱼群算法来获取，具体获取步骤包括有：
S1021、确定人工鱼的维度，所述人工鱼的维度包括BP神经网络的权值和阈值，即H＝H(v11,...vI1,u11,...,uk1,w1k,...,wpk,θk)；
S1022、对鱼群算法的参数进行初始化，其中，鱼群算法需初始化的参数包括有种群大小fish_size，人工鱼的感知距离fish_dis tan ge，人工鱼移动的最大步长fish_step，拥挤度因子δ，最大迭代次数max_gen和目标值G；
S1023、设置初始迭代次数k＝0，并在可行域内随机产生fish_size个人工鱼的个体，形成初始鱼群，且各个分量均为(-1,1)区间内的随机数；
S1024、计算初始鱼群中各人工鱼个体当前位置的食物浓度值Y，并比较大小，保留最大值进入公告板；
S1025、对各人工鱼分别模拟觅食行为、聚群行为和追尾行为，从而获得人工鱼分别模拟觅食行为、聚群行为和追尾行为后所得到的食物浓度值，然后选出数值最大的食物浓度值，并将该选出的食物浓度值所对应的行为作为实际执行；
其中，人工鱼的觅食行为、聚群行为和追尾行为的计算方法如下所示：
①、觅食行为的计算方法
设人工鱼当前的状态为Fi，在它的视野范围内(di,j≤fish_dis tan ge)随机选择一个状态Fj，如果Yi<Yj，则向该方向前进一步，执行以下第五公式，所述第五公式如下所示：
fi,j(k+1)＝fi,j(k)+rand·step(Fi,j-xi,j(k))
反之，则重新随机选择状态Fi，判断是否满足前进条件；重复预设尝试次数后，如果仍不满足条件，则执行式以下第六公式，所述第六公式如下所示：
xi,j(k+1)＝xi,j(k)+rand·step
②、聚群行为的计算方法
设人工鱼当前的状态为Fi，探索可见域内的伙伴数目为nf，形成集合Φi，若Φi≠φ，则在集合内探索中心位置Xcenter，计算该中心位置的食物浓度值Fcenter，如果满足
Figure PCTCN2016110365-appb-000008
则执行聚群行为公式，如以下第七公式所示：
xi,j(k+1)＝xi,j(k)+rand·step(xcenter,j-xi,j(k))
③、追尾行为的计算方法
设人工鱼当前的状态为Fi，探索可见域内的伙伴数目为nf，形成集合Φi，若Φi≠φ，则集合内探索食物浓度最大的伙伴Xmax，计算该处的食物浓度值Fmax，如果满足
Figure PCTCN2016110365-appb-000009
则执行追尾行为公式，如以下第八公式所示：
xi,j(k+1)＝xi,j(k)+rand·step(xmax,j-xi,j(k))
S1026、判断上述选出的食物浓度值是否大于公告板上的食物浓度值，若是，则将上述选出的食物浓度值取代公告板上的食物浓度值，反之，公告板上的食物浓度值则不变；
S1027、判断是否满足中止条件，具体地，判断迭代次数是否达到最大迭代次数或者判断是否已经满足求解的精度误差，若是，则将当前各人工鱼所对应的个体信息作为获得的初始权值和初始阈值，反之，则返回重新执行步骤S1025，直至满足中止条件；
S103、将训练结束后得到的BP神经网络作为所需建立的BP神经网络。
步骤S2、在制备所需的生物组织模拟材料时，对当前准备好的各个成分进行介电常数拟合函数和电导率拟合函数的获取，并且获取已知的所需制备的生物组织模拟材料所对应的介电常数拟合函数和电导率拟合函数，然后将获得的介电常数拟合函数和电导率拟合函数输入至上述建好的BP神经网络中进行数据处理后输出各个成分的比例值。
另外，由于BP神经网络所输出的配方中各成分的比例值，即配方比例，可能会存有不合理性的情况，因此为了使最终得到的各个成分的比例值更加精确，更加贴近实际，在所述步骤S2之后设有判断处理步骤S3，所述步骤S3具体为：
判断输出的配方中各个成分的比例值是否符合预设的评判标准，若是，则结束，所述输出的配方比例为最终确定得到的配方比例，反之，则返回重新执行建立BP神经网络这一步骤中的步骤S102，采用鱼群算法重新计算得出新的初始权值和初始阈值，然后基于新的初始权值和初始阈值重新对BP神经网络进行训练，直到满足训练结束要求为止，接着将获得的介电特性拟合函数输入至新训练好的BP神经网络中进行数据处理；
或者，判断输出的配方中各个成分的比例值是否符合预设的评判标准，若是，则结束，所述输出的配方比例为最终确定得到的配方比例，反之，则返回重新执行建立BP神经网络这一步骤，重新获取BP神经网络的训练数据，采用鱼群算法重新计算得出新的初始权值和初始阈值，然后基于新的训练数据、初始权值和初始阈值，重新对BP神经网络进行训练，直到满足训练结束要求为止，接着将获得的介电特性拟合函数输入至新训练好的BP神经网络中进行数据处理。
作为上述步骤S3的优选实施方式，如图4所示，其具体包括：
S301、获取输入因子，其中，所述输入因子包括配方中各个成分所对应的介电常数拟合函数、电导率拟合函数及输出的配方中各个成分的比例值；具体地，所述输入因子可表示为：
U＝{u1,u2,u3,u4,u5,u6,u7,u8,u9,u10,u11,u12,u13,u14,u15,u16,u17,u18,u19,u20,u21}
其中，{u1,u2,u3}分别为去离子水所对应的介电常数拟合函数、电导率拟合函数及其所占的比例值，{u4,u5,u6}分别为明胶所对应的介电常数拟合函数、电导率拟合函数及其所占的比例值，{u7,u8,u9}分别为琼脂所对应的介电常数拟合函数、电导率拟合函数及其所占的比例值，{u10,u11,u12}分别为氯化钠所对应的介电常数拟合函数、电导率拟合函数及其所占的比例值，{u13,u14,u15}分别为铝粉所对应的介电常数拟合函数、电导率拟合函数及其所占的比例值，{u16,u17,u18}分别为食用油所对应的介电常数拟合函数、电导率拟合函数及其所占的比例值，{u19,u20,u21}分别为洗洁剂所对应的介电常数拟合函数、电导率拟合函数及其所占的比例值；
S302、获取输出因子，其中，所述输出因子包括所需制备的生物组织模拟材料与真实的生物组织之间的介电特性误差；具体地，所述输出因子可表示为：
V＝{v1,v2,v3,v4,v5,v6,v7,v8,v9,v10}
其中，v1代表误差小于10％，v2代表代表误差小于20％，v3代表误差小于30％，依此类推；
S303、根据获取的输入因子和输出因子，建立从输入因子到输出因子的评判矩阵，该评判矩阵如下所示：
Figure PCTCN2016110365-appb-000010
S304、判断是否需要对评判矩阵进行归一化处理，若是，则对评判矩阵进行归一化处理后执行步骤S305，反之，则直接执行步骤S305；
S305、根据最大隶属原则，从评判矩阵中选出数值最大的隶属度，然后判断所述选出的隶属度是否小于预设值，在本实施例中，该预设值为20％，若是，则表示输出的配方中各个成分的比例值符合预设的评判标准，即所述BP神经网络所输出的配方中各个成分的比例值是合理的，此时结束，所述输出的配方比例为最终确定得到的配方比例，制作者按照这一配方比例来制备所需的生物组织模拟材料便可，反之，则表示输出的配方中各个成分的比例值不符合预设的评判标准，此时，则返回重新执行步骤S102，采用鱼群算法重新计算得出新的初始权值和初始阈值，然后基于新的初始权值和初始阈值重新对BP神经网络进行训练，直到满足训练结束要求为止，利用新训练好的BP神经网络来重新进行处理，或者，则返回重新执行建立BP神经网络这一步骤S1，重新获取BP神经网络的训练数据，采用鱼群算法重新计算得出新的初始权值和初始阈值，然后基于新的训练数据、初始权值和初始阈值，重新对BP神经网络进行训练，直到满足训练结束要求为止，利用新训练好的BP神经网络来重新进行处理。
上述方法实施例中所描述的技术特征均适用于以下的系统实施例中。
实施例3、生物组织模拟材料的配方比例确定系统
基于上述生物组织模拟材料的配方比例确定方法，建立与其相对应的系统，即一种生物组织模拟材料的确定系统，如图5所示，其包括有：
获取单元401，用于获取配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数；
处理单元402，用于将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值。对于所述的获取单元401和处理单元402，它们可通过处理器来实现其功能。
进一步作为本系统实施例的优选实施方式，所述成分所对应的介电特性拟合函数是由成分在不同频率下的介电特性所拟合出的曲线函数。
进一步作为本系统实施例的优选实施方式，所述介电特性包括介电常数和/或电导率。
进一步作为本系统实施例的优选实施方式，如图6所示，所述处理单元402之前设有用于建立BP神经网络的建立单元400。对于所述建立单元400和获取单元401之间的顺序关系，它们的先后逻辑顺序可根据实际情况来设定，而优选地，所述建立单元400设在获取单元401之前。
进一步作为本系统实施例的优选实施方式，如图7所示，所述建立单元400包括：
训练模块4001，用于获取BP神经网络的训练数据，其中，所述训练数据包括训练输入数据和训练输出数据，所述训练输入数据包括配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数，所述训练输出数据包括配方中各个成分的比例最优值，然后利用获得的BP神经网络的初始权值、初始阈值以及训练数据，从而对BP神经网络进行训练，直到满足训练结束要求；
BP神经网络获取模块4002，用于将训练结束后得到的BP神经网络作为所需建立的BP神经网络。
进一步作为本系统实施例的优选实施方式，所述BP神经网络的初始权值和初始阈值是采用鱼群算法来获取的。
进一步作为本系统实施例的优选实施方式，如图6所示，所述处理单元402之后设有判断处理单元403，所述判断处理单元403用于判断输出的配方中各个成分的比例值是否符合预设的评判标准，若是，则结束，反之，则返回重新执行建立单元400所对应的数据处理流程。
进一步作为本系统实施例的优选实施方式，如图8所示，所述判断处理单元403具体包括：
输入因子获取模块4031，用于获取输入因子，其中，所述输入因子包括配方中各个成分所对应的介电特性拟合函数以及输出的配方中各个成分的比例值；
输出因子获取模块4032，用于获取输出因子，其中，所述输出因子包括所需制备的生物组织模拟材料与真实的生物组织之间的介电特性误差；
评判矩阵建立模块4033，用于根据获取的输入因子和输出因子，建立从输入因子到输出因子的评判矩阵；
判断处理模块4034，用于根据最大隶属原则，从评判矩阵中选出数值最大的隶属度，然后判断所述选出的隶属度是否小于预设值，若是，则表示输出的配方中各个成分的比例值符合预设的评判标准，此时结束，反之，则表示输出的配方中各个成分的比例值不符合预设的评判标准，返回重新执行建立单元400所对应的数据处理流程。
对于上述判断处理模块4034，其具体用于判断是否需要对评判矩阵进行归一化处理，若是，则对评判矩阵进行归一化处理后执行评判步骤，反之，则直接执行评判步骤，所述评判步骤具体为，根据最大隶属原则，从评判矩阵中选出数值最大的隶属度，然后判断所述选出的隶属度是否小于预设值，若是，则表示输出的配方中各个成分的比例值符合预设的评判标准，此时结束，反之，则表示输出的配方中各个成分的比例值不符合预设的评判标准，返回重新执行建立单元400所对应的数据处理流程，采用鱼群算法重新计算得出新的初始权值和初始阈值，然后基于新的初始权值和初始阈值重新对BP神经网络进行训练，直到满足训练结束要求为止，或者，重新获取BP神经网络的训练数据，采用鱼群算法重新计算得出新的初始权值和初始阈值，然后基于新的训练数据、初始权值和初始阈值，重新对BP神经网络进行训练，直到满足训练结束要求为止。
由上述可得，通过使用上述本发明的方法和系统，在生物组织模拟材料的过程中，制作者仅需要将采集到的配方中各个成分所对应的介电特性拟合函数，以及所需制备的生物组织模拟材料的介电特性拟合函数，输入至训练好的BP神经网络中进行数据处理后，便能得到各个成分的比例值，这样制作者根据所获得的配方比例，便可实现生物组织模拟材料的制作。另外，本发明采用了各个成分所对应的介电常数拟合函数和电导率拟合函数，以及所需制备的生物组织模拟材料的介电常数拟合函数和电导率拟合函数，来作为BP神经网络的训练输入数据，这样可令后续应用时，BP神经网络所输出的比例值更贴近实际情况，应用实用性更高，而且各个成分所对应的介电常数拟合函数和电导率拟合函数的获取步骤也比较简单，操作实施便利性高；还有，采用鱼群算法来获取BP神经网络的初始阈值和权值，能令BP神经网络的训练过程更快速；采用评判矩阵这一方式来对输出的各成分的比例值进行合理性判断，其不仅步骤简单、易于实现，而且判断的准确度更高。
以上是对本发明的较佳实施进行了具体说明，但对本发明创造并不限于所述实施例，熟悉本领域的技术人员在不违背本发明精神的前提下还可做作出种种的等同变形或替换，这些等同的变形或替换均包含在本申请权利要求所限定的范围内。

Claims (16)

  1. 一种生物组织模拟材料的配方比例确定方法，其特征在于：该方法的步骤包括有：
    获取配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数；
    将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值。
  2. 根据权利要求1所述一种生物组织模拟材料的配方比例确定方法，其特征在于：所述将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值这一步骤之前设有建立BP神经网络这一步骤。
  3. 根据权利要求2所述一种生物组织模拟材料的配方比例确定方法，其特征在于：所述建立BP神经网络这一步骤，其包括：
    获取BP神经网络的训练数据，其中，所述训练数据包括训练输入数据和训练输出数据，所述训练输入数据包括配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数，所述训练输出数据包括配方中各个成分的比例最优值，然后利用获得的BP神经网络的初始权值、初始阈值以及训练数据，从而对BP神经网络进行训练；
    将训练结束后得到的BP神经网络作为所需建立的BP神经网络。
  4. 根据权利要求3所述一种生物组织模拟材料的配方比例确定方法，其特征在于：所述BP神经网络的初始权值和初始阈值是采用鱼群算法来获取的。
  5. 根据权利要求2-4任一项所述一种生物组织模拟材料的配方比例确定方法，其特征在于：所述将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值这一步骤之后设有判断处理步骤，所述判断处理步骤具体为：
    判断输出的配方中各个成分的比例值是否符合预设的评判标准，若是，则结束，反之，则返回重新执行建立BP神经网络这一步骤。
  6. 根据权利要求5所述一种生物组织模拟材料的配方比例确定方法，其特征在于：所述判断处理步骤具体包括：
    获取输入因子，其中，所述输入因子包括配方中各个成分所对应的介电特性拟合函数以及输出的配方中各个成分的比例值；
    获取输出因子，其中，所述输出因子包括所需制备的生物组织模拟材料与真实的生物组织之间的介电特性误差；
    根据获取的输入因子和输出因子，建立从输入因子到输出因子的评判矩阵；
    根据最大隶属原则，从评判矩阵中选出数值最大的隶属度，然后判断所述选出的隶属度是否小于预设值，若是，则表示输出的配方中各个成分的比例值符合预设的评判标准，此时结束，反之，则表示输出的配方中各个成分的比例值不符合预设的评判标准，返回重新执行建立BP神经网络这一步骤。
  7. 根据权利要求1-4任一项所述一种生物组织模拟材料的配方比例确定方法，其特征在于：所述成分所对应的介电特性拟合函数是由成分在不同频率下的介电特性所拟合出的曲线函数。
  8. 根据权利要求7所述一种生物组织模拟材料的配方比例确定方法，其特征在于：所述介电特性包括介电常数和/或电导率。
  9. 一种生物组织模拟材料的配方比例确定系统，其特征在于：该系统包括有：
    获取单元，用于获取配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数；
    处理单元，用于将获得的介电特性拟合函数输入至BP神经网络中进行数据处理，从而输出配方中各个成分的比例值。
  10. 根据权利要求9所述一种生物组织模拟材料的配方比例确定系统，其特征在于：所述处理单元之前设有用于建立BP神经网络的建立单元。
  11. 根据权利要求10所述一种生物组织模拟材料的配方比例确定系统，其特征在于：所述建立单元包括：
    训练模块，用于获取BP神经网络的训练数据，其中，所述训练数据包括训练输入数据和训练输出数据，所述训练输入数据包括配方中各个成分所对应的介电特性拟合函数以及所需制备的生物组织模拟材料的介电特性拟合函数，所述训练输出数据包括配方中各个成分的比例最优值，然后利用获得的BP神经网络的初始权值、初始阈值以及训练数据，从而对BP神经网络进行训练；
    BP神经网络获取模块，用于将训练结束后得到的BP神经网络作为所需建立的BP神经网络。
  12. 根据权利要求11所述一种生物组织模拟材料的配方比例确定系统，其特征在于：所述BP神经网络的初始权值和初始阈值是采用鱼群算法来获取的。
  13. 根据权利要求10-12任一项所述一种生物组织模拟材料的配方比例确定系统，其特征在于：所述处理单元之后设有判断处理单元，所述判断处理单元用于判断输出的配方中各个成分的比例值是否符合预设的评判标准，若是，则结束，反之，则返回重新执行建立单元所对应的数据处理流程。
  14. 根据权利要求13所述一种生物组织模拟材料的配方比例确定系统，其特征在于：所述判断处理单元具体包括：
    输入因子获取模块，用于获取输入因子，其中，所述输入因子包括配方中各个成分所对应的介电特性拟合函数以及输出的配方中各个成分的比例值；
    输出因子获取模块，用于获取输出因子，其中，所述输出因子包括所需制备的生物组织模拟材料与真实的生物组织之间的介电特性误差；
    评判矩阵建立模块，用于根据获取的输入因子和输出因子，建立从输入因子到输出因子的评判矩阵；
    判断处理模块，用于根据最大隶属原则，从评判矩阵中选出数值最大的隶属度，然后判断所述选出的隶属度是否小于预设值，若是，则表示输出的配方中各个成分的比例值符合预设的评判标准，此时结束，反之，则表示输出的配方中各个成分的比例值不符合预设的评判标准，返回重新执行建立单元所对应的数据处理流程。
  15. 根据权利要求9-12任一项所述一种生物组织模拟材料的配方比例确定系统，其特征在于：所述成分所对应的介电特性拟合函数是由成分在不同频率下的介电特性所拟合出的曲线函数。
  16. 根据权利要求15所述一种生物组织模拟材料的配方比例确定系统，其特征在于：所述介电特性包括介电常数和/或电导率。
PCT/CN2016/110365 2016-12-08 2016-12-16 一种生物组织模拟材料的配方比例确定方法及系统 WO2018103131A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201611123843.6 2016-12-08
CN201611123843.6A CN106682451B (zh) 2016-12-08 2016-12-08 一种生物组织模拟材料的配方比例确定方法及系统

Publications (1)

Publication Number Publication Date
WO2018103131A1 true WO2018103131A1 (zh) 2018-06-14

Family

ID=58868255

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/110365 WO2018103131A1 (zh) 2016-12-08 2016-12-16 一种生物组织模拟材料的配方比例确定方法及系统

Country Status (2)

Country Link
CN (1) CN106682451B (zh)
WO (1) WO2018103131A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114527567A (zh) * 2021-12-13 2022-05-24 中国北方车辆研究所 一种大视场镜头保护玻璃设计方法

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109558664B (zh) * 2018-11-22 2023-04-18 广东工业大学 一种注塑成型制造的复合物料配方制定方法
CN117373582B (zh) * 2023-12-06 2024-04-26 小米汽车科技有限公司 基于仿真实验模型的材料成分预测方法、装置及存储介质

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1876172A (zh) * 2005-06-02 2006-12-13 北京中医药大学 Bp神经网络模型方法制备药物组合物
CN101620249A (zh) * 2009-07-27 2010-01-06 张莉 人工电磁材料电磁参数测量的神经网络方法
US20110301856A1 (en) * 2005-11-14 2011-12-08 Mclean Hospital Corporation Methods and systems for drug screening and computational modeling based on biologically realistic neurons
CN102508042A (zh) * 2011-10-19 2012-06-20 中国人民解放军第四军医大学 测量生物组织介电谱特性的终端开路同轴探头及方法
CN104330642A (zh) * 2014-11-12 2015-02-04 中国人民解放军国防科学技术大学 测量生物组织各向异性介电谱特性的探头及其测算方法
CN104573273A (zh) * 2015-01-27 2015-04-29 江西理工大学 改性生物吸附剂构效关系模型构建方法及应用
CN104897449A (zh) * 2015-06-09 2015-09-09 南方医科大学 一种人体组织介电特性模拟材料的配置方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102258960A (zh) * 2011-05-26 2011-11-30 浙江迦南科技股份有限公司 一种基于神经网络系统的流化床配方自动生成方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1876172A (zh) * 2005-06-02 2006-12-13 北京中医药大学 Bp神经网络模型方法制备药物组合物
US20110301856A1 (en) * 2005-11-14 2011-12-08 Mclean Hospital Corporation Methods and systems for drug screening and computational modeling based on biologically realistic neurons
CN101620249A (zh) * 2009-07-27 2010-01-06 张莉 人工电磁材料电磁参数测量的神经网络方法
CN102508042A (zh) * 2011-10-19 2012-06-20 中国人民解放军第四军医大学 测量生物组织介电谱特性的终端开路同轴探头及方法
CN104330642A (zh) * 2014-11-12 2015-02-04 中国人民解放军国防科学技术大学 测量生物组织各向异性介电谱特性的探头及其测算方法
CN104573273A (zh) * 2015-01-27 2015-04-29 江西理工大学 改性生物吸附剂构效关系模型构建方法及应用
CN104897449A (zh) * 2015-06-09 2015-09-09 南方医科大学 一种人体组织介电特性模拟材料的配置方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114527567A (zh) * 2021-12-13 2022-05-24 中国北方车辆研究所 一种大视场镜头保护玻璃设计方法
CN114527567B (zh) * 2021-12-13 2023-10-27 中国北方车辆研究所 一种大视场镜头保护玻璃设计方法

Also Published As

Publication number Publication date
CN106682451A (zh) 2017-05-17
CN106682451B (zh) 2019-05-14

Similar Documents

Publication Publication Date Title
WO2018103131A1 (zh) 一种生物组织模拟材料的配方比例确定方法及系统
CN106530367B (zh) 一种基于Firm阈值迭代的电学层析成像稀疏重建方法
Shen et al. A macaque connectome for large-scale network simulations in TheVirtualBrain
CN107247259B (zh) 基于神经网络的k分布海杂波形状参数估计方法
CN110610242A (zh) 一种联邦学习中参与者权重的设置方法及装置
CN102496156B (zh) 基于协同量子粒子群算法的医学图像分割方法
CN105574191B (zh) 在线社会网络多源点信息溯源系统及其方法
Tidholm et al. Comparisons of 2‐and 3‐dimensional echocardiographic methods for estimation of left atrial size in dogs with and without myxomatous mitral valve disease
CN103237320B (zh) 无线传感器网络基于混合量化卡尔曼融合的目标跟踪方法
Kumar et al. IoT-driven experimental framework for advancing electrical impedance tomography
WO2020007026A1 (zh) 分割模型训练方法、装置及计算机可读存储介质
CN110825583A (zh) 一种针对云数据中心多指标融合的能效定性评估技术
Liang et al. Dynamic imaging for time-varying distribution using electrical/ultrasonic dual-modality tomography
Yu et al. Determination of Fetal Left Ventricular Volume Based on Two‐Dimensional Echocardiography
CN114519722A (zh) 一种基于卷积神经网络的颈动脉提取方法
CN109342367B (zh) 一种基于控制蒙特卡罗方法的扩散光学成像方法及系统
Biradar et al. Estimation of the population mean based on extremes ranked set sampling
Yang et al. Improving the particle swarm algorithm and optimizing the network intrusion detection of neural network
CN106056632A (zh) 基于动态重采样粒子滤波的雷达目标跟踪方法
Ricordeau et al. Exploring 3d-structure analysis tools for a simulated bone remodelling process
Brenden et al. Simulation-based evaluation of the accuracy in indirectly estimating gillnet selectivity
Kubikova et al. Morphometry and reconstruction of hepatic lobules in pig based on serial histological sections
Wu et al. A 3D ultrasound model of red blood cell aggregation
Panichkitkosolkul An Improved Confidence Interval for the Difference between Standard Deviations of Normal Distributions Using a Ranked Set Sampling
Muhammed et al. The Inverse Lomax Chen Distribution: Properties and Applications

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16923567

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16923567

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 02/10/2019)

122 Ep: pct application non-entry in european phase

Ref document number: 16923567

Country of ref document: EP

Kind code of ref document: A1