CN-121839177-B - Dynamic prevention and control method and system for chicken coccidiosis

CN121839177BCN 121839177 BCN121839177 BCN 121839177BCN-121839177-B

Abstract

The invention relates to the technical field of animal disease prevention and control, and discloses a dynamic prevention and control method and a system for chicken coccidiosis, wherein the method comprises the following steps of 1, obtaining environmental parameter data, excrement hyperspectral data and chicken crowd behavior image data; the method comprises the steps of (1) processing environmental parameter data, excrement hyperspectral data and chicken flock behavior image data to obtain a microecological state tensor, (3) dynamically embedding and reasoning by adopting a graph attention network to obtain a pathogen pathogenicity coefficient, a host susceptibility index and a drug resistance evolution factor, (4) inputting the pathogen pathogenicity coefficient, the host susceptibility index and the drug resistance evolution factor into a digital twin model, outputting an intestinal tract disease grading value, a drug resistance risk value and a predicted coccidian oocyst density value, and (5) inputting the intestinal tract disease grading value, the drug resistance risk value and the predicted coccidian oocyst density value into a machine learning algorithm to generate a prevention and control strategy.

Inventors

CAI HAIMING
ZHU YIBIN
ZHANG XIAOHUI
ZHANG JIANFEI
SUN MINGFEI
LIAO SHENQUAN
Qi nanshan
LI JUAN
LV MINNA
LIN XUHUI
Song yongle
CHEN XIANGJIE

Assignees

广东省农业科学院动物卫生研究所

Dates

Publication Date: 20260512
Application Date: 20260312

Claims (9)

1. A dynamic prevention and control method for chicken coccidiosis is characterized by comprising the following steps: step 1, acquiring environmental parameter data, excrement hyperspectral data and chicken flock behavior image data; Step 2, analyzing, processing and splicing environmental parameter data, excrement hyperspectral data and chicken flock behavior image data to obtain a microecological state tensor, and performing time sequence expansion on the microecological state tensor to obtain a time sequence tensor; Step 3, constructing a pathology knowledge graph, and dynamically embedding and reasoning the pathology knowledge graph by adopting a graph attention network based on a microecological state tensor to obtain a pathogen pathogenicity coefficient, a host susceptibility index, coccidium outbreak risk probability and a drug resistance evolution factor; inputting the microecological state tensor, the time sequence tensor, the pathogenic pathogenicity coefficient, the host susceptibility index and the drug resistance evolution factor into a digital twin model, and outputting an intestinal tract lesion grading value, a drug resistance risk value and a predicted coccidian oocyst density value by the digital twin model; Inputting the intestinal tract lesion grading value, the drug resistance risk value and the predicted coccidian oocyst density value into a machine learning algorithm, and generating a prevention and control strategy by the machine learning algorithm; the digital twin model comprises a coccidium-flora-immune cell interaction sub-model and an intestinal substance transmission sub-model based on computational fluid dynamics; The coccidian-flora-immune cell interaction submodel comprises: the coccidian proliferation equation is as follows: Wherein, the Is the first The endogenous proliferation rate of the coccidian species, Is the first Biomass of the species coccidium in the intestinal tract, For the environmental load capacity of the intestinal tract to coccidium, Is the first Biomass of the species coccidium in the intestinal tract, For the immune system to the first The killing coefficient of the coccidium species, For the local immunity intensity of the host, Is the first Pairs of species group Antagonism or synergy coefficient of coccidian, Is the first The abundance of the dominant intestinal flora, Is the killing or inhibiting coefficient of the medicine to coccidium, Is the real-time concentration distribution of the medicine in the intestinal tract; immune response equation, the formula is: Wherein, the For the response coefficient of the immune system to coccidian stimulation, Is the first Biomass of the species coccidium in the intestinal tract, As a natural decay factor of the intensity of the immunity, Local immunity intensity for the host; The flora succession equation is as follows: Wherein, the Is the first The rate of endogenous proliferation of the species of bacteria, Is the first The abundance of the dominant intestinal flora, For the environmental load capacity of the intestinal tract to the flora, Is the first Pairs of coccidium The coefficient of influence of the species of bacteria, Is the first Biomass of the species coccidium in the intestinal tract, Is the killing or inhibiting coefficient of the drug to the flora, Is the real-time concentration distribution of the medicine in the intestinal tract; the score mapping formula of intestinal lesions is: Wherein, the As a Sigmoid function, the coefficient 4 is a scaling factor, As a normalized value of coccidian biomass relative to environmental load bearing capacity, In order to provide an immune-pathogen interaction term, For clostridium perfringens abundance normalization values, 、、 Are all the weight coefficients of the two-dimensional space model, Is a threshold offset; The drug resistance risk calculation formula is: Wherein, the As an evolution factor of the drug resistance, Is the drug resistance variation coefficient driven by population scale, For the reference population size to be a reference, The pressure-driven drug resistance accumulation coefficient is selected for the drug, Is a time integral of the concentration of the drug, As a genetic coefficient of drug resistance recorded by historical medication, For the historical medication strength index, Is time of Time (1) Coccidian biomass.
2. The dynamic chicken coccidiosis prevention and control method as recited in claim 1, wherein said step 2 comprises the sub-steps of: A1, carrying out Min-Max normalization processing on environmental parameter data to obtain an environmental feature vector; a2, calculating to obtain a coccidian oocyst index and a bleeding index by adopting a coccidian oocyst index formula and a bleeding index formula based on the excreta hyperspectral data, inputting the coccidian oocyst index and the bleeding index into a pre-trained convolutional neural network regression model to obtain a current coccidian oocyst density value, carrying out logarithmic normalization processing on the current coccidian oocyst density value to obtain a coccidian density estimated value, and constructing a spectral feature vector based on the coccidian oocyst index, the bleeding index and the coccidian density estimated value; Step A3, analyzing the chicken flock behavior image data by adopting a target detection algorithm, carrying out ratio normalization processing to obtain the ratio of the feeding activity, the movement activity and the rest duration, and constructing a behavior feature vector based on the ratio of the feeding activity, the movement activity and the rest duration; step A4, vector splicing is carried out on the environmental feature vector, the spectral feature vector and the behavior feature vector to obtain a microecological state tensor; And A5, performing time series expansion on the microecological state tensor, and forming the time series tensor based on the microecological state tensor within 24 hours.
3. The dynamic chicken coccidiosis prevention and control method according to claim 2, wherein in the step A2, the coccidian oocyst index formula is: Wherein, the 、 Is the characteristic reflection wavelength of the lipid and the protein of the oocyst wall of the coccidian, R is the reflectivity for background interference wavelength; The bleeding index formula is: Wherein, the As a function of the wavelength(s), Is wavelength of The first derivative of the reflectivity at that point, For the wavelength of The integration is performed and the integration is performed, 、 The start wavelength and the end wavelength of the absorption band characteristic of oxyhemoglobin, respectively.
4. The method for dynamically preventing and controlling chicken coccidiosis according to claim 1, wherein in the step 3, the graph attention network is adopted to dynamically embed and infer a pathological knowledge graph, and the steps of obtaining a pathogenic pathogenicity coefficient, a host susceptibility index, a coccidian outbreak risk probability and a drug resistance evolution factor comprise the following sub-steps: step B1, constructing a heterogeneous graph, wherein the heterogeneous graph comprises an environment entity node, a pathogen entity node, a host entity node, a flora entity node and a drug entity node; step B2, calculating attention coefficients of an environment entity node, a pathogen entity node, a host entity node, a flora entity node and a medicine entity node by adopting a graph attention network, and normalizing by a Softmax function to obtain attention weights; step B3, updating the node characteristic vector by a weighted summation mode based on the attention weight; And B4, mapping the updated node characteristic vector through a full-connection layer by the graph annotation force network to obtain a pathogen pathogenicity coefficient, a host susceptibility index, a coccidium outbreak risk probability and a drug resistance evolution factor.
5. The dynamic chicken coccidiosis prevention and control method of claim 4, wherein the attention coefficient is calculated by the formula: Wherein, the And Respectively the characteristic vectors of the environmental entity node, the pathogenic entity node, the host entity node, the flora entity node or the drug entity node, As a function of the non-linear activation, In order to share the weight matrix, For the attention weight vector, ||represents a stitching operation; The calculation formula of the attention weight is as follows: Wherein, the Is a node Is defined by a set of neighboring nodes of the network, To be with natural constant An exponential function of the base; the calculation formula for updating the node characteristic vector by a weighted summation mode is as follows: Wherein, the In order for the updated node feature vector to be updated, As a function of the non-linear activation, In order for the attention to be weighted, In order to share the weight matrix, Is a node Is used for the feature vector of (a), Is a node Is described herein).
6. The dynamic chicken coccidiosis prevention and control method of claim 1, wherein the step 4 comprises the following sub-steps: Step C1, determining the real-time concentration distribution of the medicine in the intestinal canal by adopting an intestinal substance transmission submodel; Step C2, inputting a microecological state tensor, a time sequence tensor, a pathogenic pathogenicity coefficient, a host susceptibility index and a drug resistance evolution factor into a coccidian-flora-immune cell interaction submodel based on the real-time concentration distribution of the drug in the intestinal tract to obtain a coccidian biomass evolution curve, an immune intensity evolution curve and a flora abundance evolution curve; Step C3, calculating to obtain an intestinal lesion score evolution curve by adopting an intestinal lesion score mapping formula based on the coccidium biomass evolution curve, the immunity intensity evolution curve and the flora abundance evolution curve; Step C4, calculating to obtain a drug resistance development trend by adopting a drug resistance risk calculation formula based on a coccidium biomass evolution curve and a historical drug use record; And step C5, obtaining an intestinal tract lesion score value, a drug resistance risk value and a predicted coccidian oocyst density value according to the coccidian biomass evolution curve, the intestinal tract lesion score evolution curve and the drug resistance development trend.
7. The method according to claim 1, wherein in the step 5, the machine learning algorithm is a near-end strategy optimization PPO algorithm, a genetic algorithm, a particle swarm optimization algorithm, or a simulated annealing algorithm.
8. The dynamic chicken coccidiosis prevention and control method according to claim 1, wherein in the step 5, a multi-objective optimization function is adopted to perform optimization training on a machine learning algorithm; the multi-objective optimization function is: Wherein, the In order to predict the daily gain of the body, The value of the score for the intestinal lesions, In order to be economical and cost-effective, As a value of the risk of resistance to the drug, 、、 And Are weight coefficients.
9. A chicken coccidiosis dynamic prevention and control system, which is characterized by being used for realizing the chicken coccidiosis dynamic prevention and control method according to any one of claims 1-8, and comprising the following units: The data acquisition module is used for acquiring environmental parameter data, excrement hyperspectral data and chicken flock behavior image data; The state tensor construction module is used for analyzing, processing and splicing the environmental parameter data, the excrement hyperspectral data and the chicken flock behavior image data to obtain a microecological state tensor, and then carrying out time sequence expansion on the microecological state tensor to obtain a time sequence tensor; The knowledge calculation module is used for constructing a pathological knowledge graph, and dynamically embedding and reasoning the pathological knowledge graph by adopting a graph attention network based on the microecological state tensor to obtain a pathogen pathogenicity coefficient, a host susceptibility index, coccidium outbreak risk probability and a drug resistance evolution factor; The digital twin simulation module is used for inputting the microecological state tensor, the time sequence tensor, the pathogenic pathogenicity coefficient, the host susceptibility index and the drug resistance evolution factor into a digital twin model, and outputting an intestinal tract lesion grading value, a drug resistance risk value and a predicted coccidian oocyst density value by the digital twin model; The intelligent decision module is used for inputting the intestinal tract lesion grading value, the drug resistance risk value and the predicted coccidian oocyst density value into a machine learning algorithm, and the machine learning algorithm generates a prevention and control strategy.

Description

Dynamic prevention and control method and system for chicken coccidiosis Technical Field The application belongs to the technical field of animal disease prevention and control, and particularly relates to a dynamic prevention and control method and system for chicken coccidiosis. Background Chicken coccidiosis (Coccidiosis) is a parasitic disease that is extremely harmful to the poultry industry worldwide caused by protozoa of the genus Eimeria that colonize chicken intestinal epithelial cells. The disease has high incidence rate, mainly causes the bloody stool, slow growth and even death of chickens, and the feed conversion rate of the resistant chickens is greatly reduced due to the damage of intestinal tracts. In the prior art, the prevention and control of chicken coccidiosis mainly characterized by the following pain points: 1. Monitoring hysteresis traditional monitoring means rely on manual collection of fresh feces for microscopic oocyst count (OPG) or section-examined dead chicken observation intestinal lesions Score (Lesion Score). These methods are all "post hoc" diagnostics, and often when found, coccidia have completed multiple breeding cycles in the chicken flock, causing irreversible intestinal damage. 2. The problem of drug resistance is that the coccidium is extremely easy to generate drug resistance due to long-term and blindly use of ionophores or chemical synthesis anticoccidial drugs (such as monensin, diclazuril and the like). The lack of alternative or shuttle drug guidelines based on scientific data makes the development of new drugs far from the rate of drug resistance. 3. The existing prevention and control system usually only focuses on killing coccidia, but neglects the synergic pathogenic effect of intestinal flora (such as clostridium perfringens) and coccidia (such as coccidiosis induced necrotic enteritis). The lack of a systematic perspective allows for comprehensive consideration of the complex dynamic relationships between environment, host immunity, flora and pathogen. 4. Decision-making lacks previewing, namely that a specific effect of a certain prevention and control scheme cannot be known before the farm uses the scheme, and the specific effect can be tested by experience only in the current specific chicken flock state and environmental conditions. Therefore, the application solves the technical problem of how to realize accurate prevention and control of chicken coccidiosis. Disclosure of Invention The application mainly aims to provide a dynamic prevention and control method for chicken coccidiosis, which constructs microecological state tensor of multisource data, embeds and infers a pathological knowledge graph by adopting a graph attention network, and combines a digital twin technology to realize quantitative simulation of a dynamic game process of coccidiosis, flora and immunity, thereby realizing accurate prevention and control of chicken coccidiosis. In order to achieve the above purpose, the present application adopts the following technical scheme: a dynamic prevention and control method for chicken coccidiosis comprises the following steps: step 1, acquiring environmental parameter data, excrement hyperspectral data and chicken flock behavior image data; Step 2, analyzing, processing and splicing environmental parameter data, excrement hyperspectral data and chicken flock behavior image data to obtain a microecological state tensor, and performing time sequence expansion on the microecological state tensor to obtain a time sequence tensor; Step 3, constructing a pathology knowledge graph, and dynamically embedding and reasoning the pathology knowledge graph by adopting a graph attention network based on a microecological state tensor to obtain a pathogen pathogenicity coefficient, a host susceptibility index, coccidium outbreak risk probability and a drug resistance evolution factor; inputting the microecological state tensor, the time sequence tensor, the pathogenic pathogenicity coefficient, the host susceptibility index and the drug resistance evolution factor into a digital twin model, and outputting an intestinal tract lesion grading value, a drug resistance risk value and a predicted coccidian oocyst density value by the digital twin model; And 5, inputting the intestinal tract lesion grading value, the drug resistance risk value and the predicted coccidian oocyst density value into a machine learning algorithm, and generating a prevention and control strategy by the machine learning algorithm. Preferably, the step 2 includes the following substeps: A1, carrying out Min-Max normalization processing on environmental parameter data to obtain an environmental feature vector; a2, calculating to obtain a coccidian oocyst index and a bleeding index by adopting a coccidian oocyst index formula and a bleeding index formula based on the excreta hyperspectral data, inputting the coccidian oocyst index and the bleeding index into a pre-trained convolutional n