A pig farm introduction early warning method and system based on disease monitoring

Abstract

The application discloses a pig farm introduction early warning method and system based on disease monitoring. The method first acquires multi-dimensional pathogen monitoring data of the pig population to be introduced for a continuous preset time length, and extracts a pathogen carrying characteristic vector. Then, the pathogen carrying data of the pig population in the introduction area is matched across the field, and a matching deviation value is obtained. Based on the isolated forest algorithm, implicit deviations caused by the output field monitoring blind area are recognized and corrected, and a matching deviation benchmark value is obtained. The introduction of the epidemic input risk is evaluated and early warning is given according to the benchmark value, and the benchmark value is input into the risk-time coupling model to output the cumulative incidence probability curve corresponding to different isolation time lengths, and finally the introduction time safety window is determined. The method solves the problems of opaque health data of the seed source, inaccurate cross-field pathogen adaptability evaluation and lack of data support for isolation time length setting in the existing introduction management, can effectively reduce the introduction of major diseases such as blue ear and viral diarrhea, and ensures the stable production of the pig farm.

Description

Technical Field

[0001] This invention relates to the field of smart farming technology, and in particular to a method and system for early warning of pig farm breeding stock introduction based on disease monitoring. Background Technology

[0002] Currently, large-scale pig farms generally rely on external breeding stock to replace about 40% of their breeding herds annually. However, this importation has become a major route for the introduction of diseases. According to industry statistics, the proportion of major diseases such as porcine reproductive and respiratory syndrome (PRRS) and viral diarrhea introduced through breeding stock is as high as 70%-80%. Especially against the backdrop of the prevalence of highly pathogenic recombinant strains such as NADC30 / 34, the situation regarding the prevention and control of risks associated with imported breeding stock is severe.

[0003] Existing technologies have three main shortcomings: First, the transparency of breeding stock health data is low. Only a single pathogen test is conducted before introduction, lacking multi-dimensional monitoring data for continuous duration at the exporting farm, making it difficult to detect intermittent viral shedding or latent environmental infections. Second, cross-regional pathogen compatibility assessment is lacking. Differences in pathogen spectra between exporting and importing farms are not considered, making it impossible to quantify the mismatch of pathogen-carrying characteristics, which can easily lead to the misintroduction of wild-type PRRS virus or PED antigen-positive pigs. Third, the setting of quarantine duration lacks scientific basis. Existing quarantine periods are mostly fixed (e.g., 45 days), without considering the dynamic characteristics of pathogen incubation periods, leading to insufficient quarantine and the mixing of infected pigs into the breeding sow herd, causing outbreaks.

[0004] Therefore, there is an urgent need for an intelligent early warning method based on multi-dimensional disease monitoring data. This method can achieve accurate assessment of the risks of introducing new species and optimize the timing of introduction through cross-field pathogen feature matching, latent risk identification, and dynamic isolation decision-making, thereby ensuring the biosecurity of pig farms from the source. Summary of the Invention

[0005] To address at least one of the aforementioned technical problems, this invention proposes a method and system for early warning of pig farm breeding stock introduction based on disease monitoring.

[0006] The first aspect of this invention provides a method for early warning of pig farm breeding stock introduction based on disease monitoring, comprising: Acquire multi-dimensional pathogen monitoring data of the breeding pig herd to be introduced at the export farm for a continuous preset duration, and extract pathogen-carrying feature vectors of the breeding pig herd to be introduced based on the multi-dimensional pathogen monitoring data; Obtain pathogen-carrying data of pigs in the introduction farm area, and calculate the cross-field matching degree based on the pathogen-carrying feature vector and the pathogen-carrying data of pigs in the introduction farm area to obtain the cross-regional pathogen-carrying feature matching deviation value. Anomaly detection is performed on the cross-regional pathogen carrying feature matching deviation value based on the isolated forest algorithm. Latent deviation components caused by blind spots in the output field monitoring are identified. The cross-regional pathogen carrying feature matching deviation value is corrected based on the latent deviation components to obtain the corrected matching deviation benchmark value. The risk assessment of the introduction of diseases is carried out based on the corrected matching deviation value. The early warning of the introduction of diseases is carried out based on the risk of the introduction of diseases. A risk-time coupling model is constructed, and the corrected matching deviation benchmark value is input into the risk-time coupling model to output the cumulative incidence probability curves corresponding to different isolation durations. The safe window for introducing new pigs to the target pig farm is determined based on the cumulative incidence probability curves corresponding to different isolation durations.

[0007] In this solution, the step of acquiring multi-dimensional pathogen monitoring data of the pig herd to be introduced at the export farm for a continuous preset duration, and extracting pathogen-carrying feature vectors of the pig herd to be introduced based on the multi-dimensional pathogen monitoring data, specifically involves: Obtain environmental pathogen monitoring data of the breeding area where the pigs to be introduced are located in the export farm. The environmental pathogen monitoring data includes air aerosol pathogen detection data, pen surface swab sample pathogen detection data, drinking water and feed residue pathogen detection data, and excrement contamination area pathogen detection data. Based on the sampling location and sampling time information corresponding to different types of environmental pathogen monitoring data, the environmental pathogen monitoring data are spatially correlated and time-series arranged to construct an output field environmental pathogen distribution dataset; Feature extraction is performed on the detection results of various pathogens in the output field environmental pathogen distribution dataset to determine the environmental detection frequency, persistence time, spatial distribution range and environmental concentration change characteristics of different pathogen types in the breeding area. Based on the environmental concentration change characteristics and spatial distribution range, the diffusion status of pathogens in the export farm aquaculture environment was analyzed to determine the environmental pathogen exposure association characteristics; By fusing the environmental detection frequency, duration of presence, environmental pathogen exposure association features, and pathogen type features, a pathogen-carrying feature vector corresponding to the pig herd to be introduced is constructed.

[0008] In this scheme, the acquisition of pathogen-carrying data of pigs in the introduction farm area, and the calculation of cross-regional matching degree based on the pathogen-carrying feature vector and the pathogen-carrying data of pigs in the introduction farm area to obtain the cross-regional pathogen-carrying feature matching deviation value, specifically involves: Obtain pathogen-carrying data corresponding to the pig herds in the introduction farm area, and perform feature analysis on the pathogen type information, pathogen detection status information, pathogen persistence information and pathogen spatial distribution information in the pathogen-carrying data to construct a pathogen-carrying feature set for the introduction farm area. Based on the pathogen-carrying feature vector of the pig herd to be introduced, the pathogen type features in the pathogen-carrying feature set of the introduction farm area are correlated accordingly, and the pathogen features with transmission correlation in different breeding farms are mapped to obtain cross-farm pathogen feature correspondence. Based on the cross-field pathogen feature correspondence, the differences between each pathogen type feature in the pathogen-carrying feature vector and the corresponding features in the pathogen-carrying feature set of the introduction field are calculated to determine the pathogen type difference, pathogen-carrying status difference, pathogen persistence difference, and pathogen spatial distribution difference. Based on the differences in pathogen species, pathogen carrying status, pathogen persistence, and pathogen spatial distribution, the consistency of pathogen carrying characteristics between the exporting and importing areas is analyzed to determine the cross-field pathogen carrying characteristic similarity. Based on the correlation weights corresponding to different pathogen feature dimensions in the pathogen carrying feature vector and the cross-field pathogen carrying feature similarity, the differences of each pathogen feature are fused and calculated to obtain the cross-regional pathogen carrying feature matching deviation value, which characterizes the degree of change in pathogen carrying status after the pig herd to be introduced is entered into the introduction farm area.

[0009] In this scheme, the isolated forest algorithm is used to detect anomalies in the cross-regional pathogen-carrying feature matching deviation value, identify latent bias components caused by blind spots in the output field monitoring, and correct the cross-regional pathogen-carrying feature matching deviation value based on the latent bias components to obtain the corrected matching deviation benchmark value, specifically as follows: An isolated forest model consisting of several binary trees is constructed. The cross-regional pathogen carrying feature matching deviation value is decomposed into a four-dimensional deviation feature vector according to the difference in pathogen type, the difference in pathogen carrying status, the difference in pathogen persistence, and the difference in pathogen spatial distribution, and then input into the model. For each binary tree, a feature dimension is randomly selected from the four-dimensional deviation feature vector. Within the value range of the feature dimension, a splitting value is randomly generated. Based on the splitting value, the vector corresponding to the selected feature dimension is assigned to the left and right branches. The random dimension selection and random splitting operations are recursively executed until the binary tree reaches a limited depth or the current node contains only a single vector. The average path length required for each four-dimensional deviation feature vector to reach a leaf node from the root node in all binary trees is calculated. The average path length is then compared with the average path length of all vectors in the isolated forest to obtain an anomaly score ranging from 0 to 1. The four-dimensional deviation feature vectors with abnormal scores exceeding a set threshold are marked as abnormal deviation vectors, and the feature dimensions of the abnormal deviation vectors that exceed the preset normal fluctuation range are extracted as candidate deviation components. The candidate deviation component is cross-validated with the environmental detection frequency, duration of persistence, spatial distribution range and environmental concentration change characteristics of the corresponding pathogen in the pathogen distribution dataset of the output farm environment. If the pathogen characteristics pointed to by the candidate deviation component have no detection record in the output farm environmental data, and the pathogen type corresponding to the candidate deviation component is consistent with the pathogen type in the previous disease records of the pig herd in the breeding farm area, and the duration corresponding to the difference in the persistence of the pathogen in the candidate deviation component is within the range of the incubation period of the same type of pathogen in the breeding farm area, then the candidate deviation component is determined to be a latent deviation component caused by the blind spot of the output farm monitoring. The feature dimension value corresponding to the latent bias component is replaced with the baseline carry-over feature value of the same type of pathogen in the introduction field under non-blind zone conditions. Based on the replaced value, the bias of this dimension is recalculated, and finally the corrected matching bias baseline value is output.

[0010] In this scheme, the risk assessment of imported diseases is performed based on the corrected matching deviation value; an early warning of imported diseases is issued based on the risk of imported diseases; and a risk-time coupled model is constructed. The corrected matching deviation benchmark value is input into the risk-time coupled model, and the cumulative incidence probability curves corresponding to different isolation durations are output. Specifically: The risk assessment of the introduction of diseases in the breeding farm is carried out based on the corrected matching deviation value. When the risk of the introduction of diseases is higher than the preset risk threshold, an early warning of the introduction of diseases is issued to the breeding farm. Obtain historical introduction records of the breeding farms, and extract the cross-regional pathogen carrying characteristic matching deviation value of the corresponding introduction batch, the first onset time of the corresponding introduction batch of pigs during the quarantine period, the number of sick pigs per day during the quarantine period, and the type identifier of the corresponding pathogen from each historical introduction record to construct a training sample set. Gradient boosting regression trees are used to iteratively train the training sample set. In each iteration, the residual between the predicted onset time and the actual onset time of the current model output is calculated. The sample weights of different matching deviation value intervals in the next iteration are dynamically adjusted according to the residual until the model converges, resulting in a risk-time coupled model composed of multiple regression trees. The corrected matching deviation baseline value is substituted into the converged risk-time coupled model. The split nodes of each regression tree are traversed sequentially. The branch to which the current input corrected matching deviation baseline value belongs is determined according to the matching deviation value range corresponding to the split node. The output values ​​of the branches to which all regression trees belong are accumulated to obtain the conditional incidence probability under different isolation durations. The conditional incidence probability under the continuous time dimension is integrated to generate the cumulative incidence probability curve corresponding to different isolation durations.

[0011] In this plan, determining the safe window for introducing new pigs to the target pig farm based on the cumulative incidence probability curves corresponding to different isolation durations specifically involves: The cumulative incidence probability values ​​corresponding to each discrete time point on the cumulative incidence probability curve are iterated through, and each cumulative incidence probability value is compared with the preset acceptable risk threshold for disease in the introduction area. All time points whose cumulative incidence probability values ​​are not greater than the acceptable risk threshold for disease are selected. The selected time points are arranged in chronological order to construct a candidate safe duration set. The time interval between two adjacent time points in the candidate safe duration set is extracted. Continuous time intervals with time intervals greater than the shortest operation time required for a single introduction operation in the introduction site are marked as preliminary safe windows. Obtain feeding and management plan data for the introduction site within the corresponding time period of the initial safety window, identify the operation periods involving vaccination, group transfer and mixed rearing, and environmental disinfection in the feeding and management plan, and remove the intervals in the initial safety window that overlap with the operation periods to obtain the airspace safety window; The start time of the net air safety window is aligned with the expected slaughter time of the pig herd to be introduced to the output farm. Net air safety windows with start times earlier than the expected slaughter time are eliminated. The start time of the remaining net air safety windows is used as the trigger node for the introduction operation, and the end time is used as the latest cutoff node for isolation and observation after the introduction, thus obtaining the introduction timing safety window for the target pig farm.

[0012] A second aspect of the present invention also provides a pig farm introduction early warning system based on disease monitoring. The system includes a memory and a processor. The memory includes a pig farm introduction early warning method program based on disease monitoring. When the pig farm introduction early warning method program based on disease monitoring is executed by the processor, it implements the steps of the pig farm introduction early warning method based on disease monitoring as described in any of the above claims.

[0013] This invention discloses a method and system for early warning of pig farm introduction based on disease monitoring. The method first acquires multi-dimensional pathogen monitoring data of the pig herd to be introduced for a continuously preset duration, extracting pathogen-carrying feature vectors. Then, it performs cross-field matching with pathogen-carrying data of pig herds in the introduction area to obtain a matching deviation value. Based on the isolated forest algorithm, it identifies and corrects latent deviations caused by blind spots in the output farm's monitoring, obtaining a matching deviation benchmark value. Based on this, it assesses the risk of disease input from the introduced pigs and issues an early warning. Simultaneously, the benchmark value is input into a risk-time coupling model, outputting cumulative disease probability curves corresponding to different isolation durations, ultimately determining the safe window for introduction. This method solves the problems of opaque health data of breeding stock, inaccurate cross-farm pathogen compatibility assessment, and lack of data support for setting isolation durations in existing introduction management. It can effectively reduce the risk of major diseases such as porcine reproductive and respiratory syndrome (PRRS) and viral diarrhea introduced through breeding, ensuring stable pig farm production. Attached Figure Description

[0014] Figure 1 A flowchart of a pig farm introduction early warning method based on disease monitoring according to the present invention is shown; Figure 2 The flowchart of the present invention for extracting pathogen-carrying feature vectors from a pig herd to be introduced is shown; Figure 3 The flowchart illustrating the present invention for obtaining cross-regional pathogen-carrying feature matching deviation values ​​is shown. Figure 4 A block diagram of a pig farm introduction early warning system based on disease monitoring according to the present invention is shown. Detailed Implementation

[0015] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0016] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0017] Figure 1 The flowchart of a pig farm introduction early warning method based on disease monitoring according to the present invention is shown.

[0018] like Figure 1 As shown, the first aspect of the present invention provides a method for early warning of pig farm introduction based on disease monitoring, comprising: Acquire multi-dimensional pathogen monitoring data of the breeding pig herd to be introduced at the export farm for a continuous preset duration, and extract pathogen-carrying feature vectors of the breeding pig herd to be introduced based on the multi-dimensional pathogen monitoring data; Obtain pathogen-carrying data of pigs in the introduction farm area, and calculate the cross-field matching degree based on the pathogen-carrying feature vector and the pathogen-carrying data of pigs in the introduction farm area to obtain the cross-regional pathogen-carrying feature matching deviation value. Anomaly detection is performed on the cross-regional pathogen carrying feature matching deviation value based on the isolated forest algorithm. Latent deviation components caused by blind spots in the output field monitoring are identified. The cross-regional pathogen carrying feature matching deviation value is corrected based on the latent deviation components to obtain the corrected matching deviation benchmark value. The risk assessment of the introduction of diseases is carried out based on the corrected matching deviation value. The early warning of the introduction of diseases is carried out based on the risk of the introduction of diseases. A risk-time coupling model is constructed, and the corrected matching deviation benchmark value is input into the risk-time coupling model to output the cumulative incidence probability curves corresponding to different isolation durations. The safe window for introducing new pigs to the target pig farm is determined based on the cumulative incidence probability curves corresponding to different isolation durations.

[0019] It should be noted that by acquiring multi-dimensional pathogen monitoring data of the pig herd to be introduced at the export farm for a continuously preset duration and extracting pathogen-carrying feature vectors, the dynamic distribution patterns of pathogens in the breeding environment can be comprehensively captured, effectively overcoming the limitations of the randomness of a single detection and significantly improving the accuracy of the health status characterization of the breeding stock. By calculating the cross-field matching degree between the pathogen-carrying feature vectors and the pathogen-carrying data of the pig herd in the import farm, the differences in pathogen spectrum and the degree of deviation in carrying status between the two farms can be accurately quantified. Based on the isolated forest algorithm, anomaly detection of matching deviation values ​​and identification of latent deviation components caused by blind spots in the export farm's monitoring can effectively uncover environmental samples that are not covered by routine detection. To mitigate potential infection risks and avoid missed detections or misjudgments due to monitoring blind spots, a more accurate matching deviation baseline value is obtained through correction. Based on the corrected matching deviation baseline value, risk assessment and early warning of disease introduction are conducted. Combined with a risk-time coupling model, cumulative incidence probability curves corresponding to different isolation durations are output, realizing a dynamic correlation between risk level and isolation period, overcoming the blindness of traditional fixed isolation durations. Finally, by analyzing the cumulative incidence probability curve, the safe window for introducing new breeds is determined, which can accurately lock in the optimal introduction period while ensuring biosecurity, reducing the risk of disease introduction from the source and improving the scientific nature and control effectiveness of pig farm introduction management.

[0020] Figure 2 The flowchart of the present invention for extracting pathogen-carrying feature vectors from a pig herd to be introduced is shown.

[0021] According to an embodiment of the present invention, the step of acquiring multi-dimensional pathogen monitoring data of the breeding pig herd to be introduced at the export farm for a continuous preset duration, and extracting pathogen-carrying feature vectors of the breeding pig herd to be introduced based on the multi-dimensional pathogen monitoring data, specifically includes: Obtain environmental pathogen monitoring data of the breeding area where the pigs to be introduced are located in the export farm. The environmental pathogen monitoring data includes air aerosol pathogen detection data, pen surface swab sample pathogen detection data, drinking water and feed residue pathogen detection data, and excrement contamination area pathogen detection data. Based on the sampling location and sampling time information corresponding to different types of environmental pathogen monitoring data, the environmental pathogen monitoring data are spatially correlated and time-series arranged to construct an output field environmental pathogen distribution dataset; Feature extraction is performed on the detection results of various pathogens in the output field environmental pathogen distribution dataset to determine the environmental detection frequency, persistence time, spatial distribution range and environmental concentration change characteristics of different pathogen types in the breeding area. Based on the environmental concentration change characteristics and spatial distribution range, the diffusion status of pathogens in the export farm aquaculture environment was analyzed to determine the environmental pathogen exposure association characteristics; By fusing the environmental detection frequency, duration of presence, environmental pathogen exposure association features, and pathogen type features, a pathogen-carrying feature vector corresponding to the pig herd to be introduced is constructed.

[0022] It should be noted that by collecting multi-dimensional environmental pathogen monitoring data, including airborne aerosols, pen surface wiping, drinking water and feed residues, and excrement contamination areas, and constructing a spatiotemporally correlated dataset using sampling location and time information, it is possible to comprehensively reconstruct the pathogen contamination status of the breeding environment from a macroscopic perspective. This overcomes the representativeness limitations of traditional methods that rely solely on individual blood sampling tests and significantly improves the detection capability for latent infections in the herd. Because pigs have high-frequency contact with environmental media during the breeding process, their respiration, feeding, and excretion behaviors continuously release pathogens into the environment. At the same time, aerosols and pollutants in the environment also have a reciprocal effect on the pig herd, resulting in continuous exposure. This two-way dynamic exchange makes the distribution characteristics of environmental pathogens highly positively correlated and causally linked with the actual carrier status of the pig herd. Therefore, through in-depth analysis of environmental detection frequency, persistence time, spatial diffusion range, and concentration change trends, the degree of pathogen stress and potential carrier characteristics of the pig herd can be accurately deduced. Finally, a pathogen carrier feature vector that can objectively reflect the true health level of the entire herd can be constructed.

[0023] Figure 3 The flowchart illustrating the cross-regional pathogen-carrying feature matching deviation value obtained by the present invention is shown.

[0024] According to an embodiment of the present invention, the step of obtaining pathogen-carrying data of pig herds in the introduction farm area, and calculating the cross-regional matching degree based on the pathogen-carrying feature vector and the pathogen-carrying data of pig herds in the introduction farm area to obtain the cross-regional pathogen-carrying feature matching deviation value, specifically involves: Obtain pathogen-carrying data corresponding to the pig herds in the introduction farm area, and perform feature analysis on the pathogen type information, pathogen detection status information, pathogen persistence information and pathogen spatial distribution information in the pathogen-carrying data to construct a pathogen-carrying feature set for the introduction farm area. Based on the pathogen-carrying feature vector of the pig herd to be introduced, the pathogen type features in the pathogen-carrying feature set of the introduction farm area are correlated accordingly, and the pathogen features with transmission correlation in different breeding farms are mapped to obtain cross-farm pathogen feature correspondence. Based on the cross-field pathogen feature correspondence, the differences between each pathogen type feature in the pathogen-carrying feature vector and the corresponding features in the pathogen-carrying feature set of the introduction field are calculated to determine the pathogen type difference, pathogen-carrying status difference, pathogen persistence difference, and pathogen spatial distribution difference. Based on the differences in pathogen species, pathogen carrying status, pathogen persistence, and pathogen spatial distribution, the consistency of pathogen carrying characteristics between the exporting and importing areas is analyzed to determine the cross-field pathogen carrying characteristic similarity. Based on the correlation weights corresponding to different pathogen feature dimensions in the pathogen carrying feature vector and the cross-field pathogen carrying feature similarity, the differences of each pathogen feature are fused and calculated to obtain the cross-regional pathogen carrying feature matching deviation value, which characterizes the degree of change in pathogen carrying status after the pig herd to be introduced is entered into the introduction farm area.

[0025] It should be noted that by converting the pathogen-carrying feature vectors of the pig herd to be introduced from the export farm and the pathogen-carrying data of the pig herd in the import farm into comparable data in a unified feature space, quantitative analysis of the pathogen ecological status between different breeding areas can be achieved. First, the pathogen-carrying feature vectors of the pig herd to be introduced are subjected to dimensional analysis to extract feature parameters such as pathogen type, detection status, persistence time, spatial distribution range, and exposure correlation. Simultaneously, the pathogen-carrying data of the pig herd in the import farm are subjected to the same dimensional feature processing to construct a corresponding pathogen-carrying feature set. Since different breeding farms may have different pathogen names but similar transmission attributes, or the same pathogen type but different prevalence status, a cross-farm pathogen feature mapping relationship is established based on pathogen transmission correlation, and pathogen features with transmission correlation are correspondingly associated. Subsequently, based on the mapped pathogen characteristics, the differences between the exporting and importing farms in terms of pathogen species, carrier status, duration of presence, and spatial distribution were calculated. The difference in pathogen species reflects the risk of new or missing pathogens; the difference in carrier status characterizes the difference in pathogen infection levels; the difference in duration of presence measures the difference in the long-term survival ability of pathogens; and the difference in spatial distribution reflects changes in the spread of pathogens. Furthermore, a pathogen characteristic difference matrix was constructed based on the differences in each dimension, and a weighted fusion calculation was performed using the association weights corresponding to different pathogen characteristic dimensions. This ensures that differences in pathogens with strong transmissibility, long duration of presence, and large spread have a greater impact on the results, ultimately yielding the cross-regional pathogen carrier characteristic matching deviation value. Because different pig farms are affected by their breeding environment, immunization levels, and historical disease pressures, their pathogen spectrum composition and carrier status vary. Therefore, relying solely on a single pathogen detection result is insufficient to accurately reflect the risk of disease transmission after introduction. By calculating the cross-regional pathogen-carrying feature matching deviation value, the degree of inconsistency in the pathogen ecological state between two farms can be determined. When the matching deviation value is small, it indicates that the pathogen-carrying state of the pig herd to be introduced is similar to that of the target farm, and the risk of disease importation is low. When the matching deviation value is large, it indicates that the pig herd to be introduced may carry pathogens missing from the target farm, or have a stronger ability to continuously spread, and have a higher risk of disease importation. Therefore, this matching deviation value can provide a quantitative basis for subsequent latent deviation identification, risk warning, and determination of isolation safety windows, improving the accuracy of disease prevention and control in pig farms. The feature mapping of pathogens with transmission correlation in different farms refers to establishing a correspondence between the pathogen characteristics of the exporting farm and the importing farm based on the transmission routes, infected objects, epidemic patterns, and environmental adaptability relationships between different pathogens, so that pathogen information with potential transmission impact in the two farms can be compared and analyzed under a unified feature dimension.

[0026] According to an embodiment of the present invention, the method of anomaly detection of the cross-regional pathogen carrying feature matching deviation value based on the isolated forest algorithm, identifying the latent deviation component caused by the blind spot of the output field monitoring, and correcting the cross-regional pathogen carrying feature matching deviation value according to the latent deviation component to obtain the corrected matching deviation benchmark value, specifically: An isolated forest model consisting of several binary trees is constructed. The cross-regional pathogen carrying feature matching deviation value is decomposed into a four-dimensional deviation feature vector according to the difference in pathogen type, the difference in pathogen carrying status, the difference in pathogen persistence, and the difference in pathogen spatial distribution, and then input into the model. For each binary tree, a feature dimension is randomly selected from the four-dimensional deviation feature vector. Within the value range of the feature dimension, a splitting value is randomly generated. Based on the splitting value, the vector corresponding to the selected feature dimension is assigned to the left and right branches. The random dimension selection and random splitting operations are recursively executed until the binary tree reaches a limited depth or the current node contains only a single vector. The average path length required for each four-dimensional deviation feature vector to reach a leaf node from the root node in all binary trees is calculated. The average path length is then compared with the average path length of all vectors in the isolated forest to obtain an anomaly score ranging from 0 to 1. The four-dimensional deviation feature vectors with abnormal scores exceeding a set threshold are marked as abnormal deviation vectors, and the feature dimensions of the abnormal deviation vectors that exceed the preset normal fluctuation range are extracted as candidate deviation components. The candidate deviation component is cross-validated with the environmental detection frequency, duration of persistence, spatial distribution range and environmental concentration change characteristics of the corresponding pathogen in the pathogen distribution dataset of the output farm environment. If the pathogen characteristics pointed to by the candidate deviation component have no detection record in the output farm environmental data, and the pathogen type corresponding to the candidate deviation component is consistent with the pathogen type in the previous disease records of the pig herd in the breeding farm area, and the duration corresponding to the difference in the persistence of the pathogen in the candidate deviation component is within the range of the incubation period of the same type of pathogen in the breeding farm area, then the candidate deviation component is determined to be a latent deviation component caused by the blind spot of the output farm monitoring. The feature dimension value corresponding to the latent bias component is replaced with the baseline carry-over feature value of the same type of pathogen in the introduction field under non-blind zone conditions. Based on the replaced value, the bias of this dimension is recalculated, and finally the corrected matching bias baseline value is output.

[0027] It should be noted that, due to the influence of factors such as the breeding environment, pathogen transmission cycle, and detection coverage on the pathogen-carrying status of the pig herd to be introduced, some pathogens may still not be captured in time during continuous monitoring at the exporting farm. For example, low-abundance pathogens may not reach the detection threshold within the sampling period, pathogens in the latent stage may not produce obvious environmental release characteristics, or the pathogen distribution area may not be covered by the sampling points. This can lead to a latent deviation between the pathogen monitoring data at the exporting farm and the actual carrier status of the pig herd. When these undetected pathogen characteristics are matched with the existing pathogen-carrying characteristics in the importing farm, abnormal changes may occur in the amount of difference in pathogen species, carrier status, persistence, or spatial distribution, thereby affecting the accuracy of the cross-regional pathogen carrier characteristic matching deviation value. Therefore, by decomposing the cross-regional pathogen-carrying feature matching deviation value into a four-dimensional deviation feature vector composed of pathogen species differences, carrying status differences, persistence differences, and spatial distribution differences, and using the isolated forest algorithm to detect anomalies in this feature vector, this method effectively identifies anomalous deviation vectors that deviate from normal pathogen change patterns. Since the isolated forest algorithm randomly selects feature dimensions and constructs multiple binary trees, it judges anomalies based on the isolation degree of samples in the feature space. Latent deviation components caused by blind spots in the output field monitoring usually only show significant deviations in some pathogen feature dimensions, allowing them to be quickly isolated with a shorter path during random partitioning. Furthermore, by combining pathogen distribution data from the output field environment, historical disease records from the introduction site, and pathogen incubation periods to cross-validate the anomalous features, it can be determined whether the anomalous deviation is caused by monitoring omissions, and the corresponding latent deviation components can be extracted and corrected. This reduces the impact of monitoring blind spots on cross-regional pathogen matching results and improves the accuracy of disease risk assessment for introduced species. The cross-validation refers to comparing candidate bias components identified in isolated forests with pathogen monitoring data from the exporting area, historical disease records from the introducing area, and pathogen incubation periods to determine whether abnormal biases are caused by monitoring blind spots. By analyzing the detection status, persistence time, spatial distribution range, and matching relationship between the pathogen characteristics corresponding to the candidate biases in the exporting area and the past pathogen prevalence in the introducing area, if the pathogen is not detected in the exporting area, but its characteristic changes conform to the transmission pattern and incubation period of similar pathogens in the introducing area, then the abnormal bias is determined to be a latent bias component, thereby avoiding misjudgments caused by normal fluctuations.

[0028] According to an embodiment of the present invention, the step of assessing the risk of imported diseases based on the corrected matching deviation value, issuing an early warning for imported diseases based on the risk of imported diseases, and constructing a risk-time coupled model, inputting the corrected matching deviation benchmark value into the risk-time coupled model, and outputting the cumulative incidence probability curves corresponding to different isolation durations, specifically includes: The risk assessment of the introduction of diseases in the breeding farm is carried out based on the corrected matching deviation value. When the risk of the introduction of diseases is higher than the preset risk threshold, an early warning of the introduction of diseases is issued to the breeding farm. Obtain historical introduction records of the breeding farms, and extract the cross-regional pathogen carrying characteristic matching deviation value of the corresponding introduction batch, the first onset time of the corresponding introduction batch of pigs during the quarantine period, the number of sick pigs per day during the quarantine period, and the type identifier of the corresponding pathogen from each historical introduction record to construct a training sample set. Gradient boosting regression trees are used to iteratively train the training sample set. In each iteration, the residual between the predicted onset time and the actual onset time of the current model output is calculated. The sample weights of different matching deviation value intervals in the next iteration are dynamically adjusted according to the residual until the model converges, resulting in a risk-time coupled model composed of multiple regression trees. The corrected matching deviation baseline value is substituted into the converged risk-time coupled model. The split nodes of each regression tree are traversed sequentially. The branch to which the current input corrected matching deviation baseline value belongs is determined according to the matching deviation value range corresponding to the split node. The output values ​​of the branches to which all regression trees belong are accumulated to obtain the conditional incidence probability under different isolation durations. The conditional incidence probability under the continuous time dimension is integrated to generate the cumulative incidence probability curve corresponding to different isolation durations.

[0029] It should be noted that the risk of pathogen input between the introduced pig herd and the target pig farm is accurately quantified based on the corrected matching deviation value. A risk-time coupled model is established by combining historical differences in pathogen carriage, onset time, and disease variation patterns during the quarantine period, thus dynamically linking pathogen carriage risk with the quarantine observation period. By using gradient boosting regression trees to learn from multidimensional historical introduction data, the nonlinear relationship between pathogen carriage characteristics and disease occurrence time can be explored, improving the risk prediction model's adaptability to complex disease transmission patterns. By inputting the real-time obtained matching deviation benchmark value into the model, the trend of disease probability changes under different quarantine durations can be predicted, and a cumulative disease probability curve can be generated, avoiding the problems of insufficient quarantine or resource waste caused by relying solely on a fixed quarantine period.

[0030] According to an embodiment of the present invention, determining the safe window for introducing new pigs to the target pig farm based on the cumulative morbidity probability curves corresponding to different isolation durations specifically involves: The cumulative incidence probability values ​​corresponding to each discrete time point on the cumulative incidence probability curve are iterated through, and each cumulative incidence probability value is compared with the preset acceptable risk threshold for disease in the introduction area. All time points whose cumulative incidence probability values ​​are not greater than the acceptable risk threshold for disease are selected. The selected time points are arranged in chronological order to construct a candidate safe duration set. The time interval between two adjacent time points in the candidate safe duration set is extracted. Continuous time intervals with time intervals greater than the shortest operation time required for a single introduction operation in the introduction site are marked as preliminary safe windows. Obtain feeding and management plan data for the introduction site within the corresponding time period of the initial safety window, identify the operation periods involving vaccination, group transfer and mixed rearing, and environmental disinfection in the feeding and management plan, and remove the intervals in the initial safety window that overlap with the operation periods to obtain the airspace safety window; The start time of the net air safety window is aligned with the expected slaughter time of the pig herd to be introduced to the output farm. Net air safety windows with start times earlier than the expected slaughter time are eliminated. The start time of the remaining net air safety windows is used as the trigger node for the introduction operation, and the end time is used as the latest cutoff node for isolation and observation after the introduction, thus obtaining the introduction timing safety window for the target pig farm.

[0031] It should be noted that, based on the cumulative incidence probability curves corresponding to different isolation durations, the dynamic analysis of the disease risk changes over time after the introduction of the species is conducted, and a safe window for introducing the species that meets actual production needs is selected from the time intervals that meet the risk constraints. By comparing the cumulative incidence rate at each time point with a preset acceptable risk threshold for diseases, this invention avoids risk bias caused by determining the introduction time solely based on fixed quarantine periods or experience, enabling dynamic determination of the quarantine period based on the actual pathogen transmission probability. Furthermore, by combining the time required for a single introduction operation to screen continuous safe intervals, it ensures that the determined safe window not only meets disease risk requirements but also possesses practical feasibility for introduction operations. By incorporating target pig farm feeding and management plan data, periods of production activities that may affect pig herd health or increase pathogen transmission risk, such as vaccination, herd transfer and mixed rearing, and environmental disinfection, can be eliminated, reducing the interference of production operation factors on disease risk assessment results and improving the reliability of the safe window. Furthermore, by temporally matching the safe window with the expected slaughter time of the pig herd awaiting introduction at the exporting farm, it ensures that the introduction time matches the breeding stock supply cycle, achieving synergistic optimization of risk control and production planning. Therefore, this invention can accurately determine the optimal introduction time and quarantine observation period while ensuring biosecurity during introduction, reducing the risk of exogenous pathogen input and improving the scientific nature and flexibility of pig farm introduction management.

[0032] Figure 4 A block diagram of a pig farm introduction early warning system based on disease monitoring according to the present invention is shown.

[0033] A second aspect of the present invention also provides a pig farm introduction early warning system based on disease monitoring. The system includes: a memory 401, a processor 402, and a communication interface 403. The memory includes a pig farm introduction early warning method program based on disease monitoring. The communication interface is used for data connection and communication between the memory and the processor. When the pig farm introduction early warning method program based on disease monitoring is executed by the processor, it implements the steps of the pig farm introduction early warning method based on disease monitoring as described in any of the above claims.

[0034] This invention discloses a method and system for early warning of pig farm introduction based on disease monitoring. The method first acquires multi-dimensional pathogen monitoring data of the pig herd to be introduced for a continuously preset duration, extracting pathogen-carrying feature vectors. Then, it performs cross-field matching with pathogen-carrying data of pig herds in the introduction area to obtain a matching deviation value. Based on the isolated forest algorithm, it identifies and corrects latent deviations caused by blind spots in the output farm's monitoring, obtaining a matching deviation benchmark value. Based on this, it assesses the risk of disease input from the introduced pigs and issues an early warning. Simultaneously, the benchmark value is input into a risk-time coupling model, outputting cumulative disease probability curves corresponding to different isolation durations, ultimately determining the safe window for introduction. This method solves the problems of opaque health data of breeding stock, inaccurate cross-farm pathogen compatibility assessment, and lack of data support for setting isolation durations in existing introduction management. It can effectively reduce the risk of major diseases such as porcine reproductive and respiratory syndrome (PRRS) and viral diarrhea introduced through breeding, ensuring stable pig farm production.

[0035] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0036] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0037] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for early warning of pig farm introduction based on disease monitoring, characterized in that, Includes the following steps: Acquire multi-dimensional pathogen monitoring data of the breeding pig herd to be introduced at the export farm for a continuous preset duration, and extract pathogen-carrying feature vectors of the breeding pig herd to be introduced based on the multi-dimensional pathogen monitoring data; Obtain pathogen-carrying data of pigs in the introduction farm area, and calculate the cross-field matching degree based on the pathogen-carrying feature vector and the pathogen-carrying data of pigs in the introduction farm area to obtain the cross-regional pathogen-carrying feature matching deviation value. Anomaly detection is performed on the cross-regional pathogen carrying feature matching deviation value based on the isolated forest algorithm. Latent deviation components caused by blind spots in the output field monitoring are identified. The cross-regional pathogen carrying feature matching deviation value is corrected based on the latent deviation components to obtain the corrected matching deviation benchmark value. The risk assessment of the introduction of diseases is carried out based on the corrected matching deviation value. The early warning of the introduction of diseases is carried out based on the risk of the introduction of diseases. A risk-time coupling model is constructed, and the corrected matching deviation benchmark value is input into the risk-time coupling model to output the cumulative incidence probability curves corresponding to different isolation durations. The safe window for introducing new pigs to the target pig farm is determined based on the cumulative incidence probability curves corresponding to different isolation durations.

2. The method for early warning of pig farm introduction based on disease monitoring according to claim 1, characterized in that, The process of acquiring multi-dimensional pathogen monitoring data of the breeding pig herd to be introduced at the export farm for a continuous preset duration, and extracting pathogen-carrying feature vectors of the breeding pig herd based on the multi-dimensional pathogen monitoring data, specifically involves: Obtain environmental pathogen monitoring data of the breeding area where the pigs to be introduced are located in the export farm. The environmental pathogen monitoring data includes air aerosol pathogen detection data, pen surface swab sample pathogen detection data, drinking water and feed residue pathogen detection data, and excrement contamination area pathogen detection data. Based on the sampling location and sampling time information corresponding to different types of environmental pathogen monitoring data, the environmental pathogen monitoring data are spatially correlated and time-series arranged to construct an output field environmental pathogen distribution dataset; Feature extraction is performed on the detection results of various pathogens in the output field environmental pathogen distribution dataset to determine the environmental detection frequency, persistence time, spatial distribution range and environmental concentration change characteristics of different pathogen types in the breeding area. Based on the environmental concentration change characteristics and spatial distribution range, the diffusion status of pathogens in the export farm aquaculture environment was analyzed to determine the environmental pathogen exposure association characteristics; By fusing the environmental detection frequency, duration of presence, environmental pathogen exposure association features, and pathogen type features, a pathogen-carrying feature vector corresponding to the pig herd to be introduced is constructed.

3. The method for early warning of pig farm introduction based on disease monitoring according to claim 1, characterized in that, The process involves acquiring pathogen-carrying data from the pig herds in the introduction farm area, and calculating the cross-regional matching degree based on the pathogen-carrying feature vector and the pathogen-carrying data of the pig herds in the introduction farm area to obtain the cross-regional pathogen-carrying feature matching deviation value. Specifically: Obtain pathogen-carrying data corresponding to the pig herds in the introduction farm area, and perform feature analysis on the pathogen type information, pathogen detection status information, pathogen persistence information and pathogen spatial distribution information in the pathogen-carrying data to construct a pathogen-carrying feature set for the introduction farm area. Based on the pathogen-carrying feature vector of the pig herd to be introduced, the pathogen type features in the pathogen-carrying feature set of the introduction farm area are correlated accordingly, and the pathogen features with transmission correlation in different breeding farms are mapped to obtain cross-farm pathogen feature correspondence. Based on the cross-field pathogen feature correspondence, the differences between each pathogen type feature in the pathogen-carrying feature vector and the corresponding features in the pathogen-carrying feature set of the introduction field are calculated to determine the pathogen type difference, pathogen-carrying status difference, pathogen persistence difference, and pathogen spatial distribution difference. Based on the differences in pathogen species, pathogen carrying status, pathogen persistence, and pathogen spatial distribution, the consistency of pathogen carrying characteristics between the exporting and importing areas is analyzed to determine the cross-field pathogen carrying characteristic similarity. Based on the correlation weights corresponding to different pathogen feature dimensions in the pathogen carrying feature vector and the cross-field pathogen carrying feature similarity, the differences of each pathogen feature are fused and calculated to obtain the cross-regional pathogen carrying feature matching deviation value, which characterizes the degree of change in pathogen carrying status after the pig herd to be introduced is entered into the introduction farm area.

4. The method for early warning of pig farm introduction based on disease monitoring according to claim 1, characterized in that, The isolated forest algorithm is used to detect anomalies in the cross-regional pathogen-carrying feature matching deviation value, identify latent bias components caused by blind spots in the output field monitoring, and correct the cross-regional pathogen-carrying feature matching deviation value based on the latent bias components to obtain the corrected matching deviation benchmark value, specifically as follows: An isolated forest model consisting of several binary trees is constructed. The cross-regional pathogen carrying feature matching deviation value is decomposed into a four-dimensional deviation feature vector according to the difference in pathogen type, the difference in pathogen carrying status, the difference in pathogen persistence, and the difference in pathogen spatial distribution, and then input into the model. For each binary tree, a feature dimension is randomly selected from the four-dimensional deviation feature vector. Within the value range of the feature dimension, a splitting value is randomly generated. Based on the splitting value, the vector corresponding to the selected feature dimension is assigned to the left and right branches. The random dimension selection and random splitting operations are recursively executed until the binary tree reaches a limited depth or the current node contains only a single vector. The average path length required for each four-dimensional deviation feature vector to reach a leaf node from the root node in all binary trees is calculated. The average path length is then compared with the average path length of all vectors in the isolated forest to obtain an anomaly score ranging from 0 to 1. The four-dimensional deviation feature vectors with abnormal scores exceeding a set threshold are marked as abnormal deviation vectors, and the feature dimensions of the abnormal deviation vectors that exceed the preset normal fluctuation range are extracted as candidate deviation components. The candidate deviation component is cross-validated with the environmental detection frequency, duration of persistence, spatial distribution range and environmental concentration change characteristics of the corresponding pathogen in the pathogen distribution dataset of the output farm environment. If the pathogen characteristics pointed to by the candidate deviation component have no detection record in the output farm environmental data, and the pathogen type corresponding to the candidate deviation component is consistent with the pathogen type in the previous disease records of the pig herd in the breeding farm area, and the duration corresponding to the difference in the persistence of the pathogen in the candidate deviation component is within the range of the incubation period of the same type of pathogen in the breeding farm area, then the candidate deviation component is determined to be a latent deviation component caused by the blind spot of the output farm monitoring. The feature dimension value corresponding to the latent bias component is replaced with the baseline carry-over feature value of the same type of pathogen in the introduction field under non-blind zone conditions. Based on the replaced value, the bias of this dimension is recalculated, and finally the corrected matching bias baseline value is output.

5. The method for early warning of pig farm introduction based on disease monitoring according to claim 1, characterized in that, The process involves assessing the risk of disease introduction based on the corrected matching deviation value, issuing early warnings for disease introduction based on the risk, and constructing a risk-time coupled model. The corrected matching deviation baseline value is input into the risk-time coupled model, and the cumulative incidence probability curves corresponding to different isolation durations are output. Specifically: The risk assessment of the introduction of diseases in the breeding farm is carried out based on the corrected matching deviation value. When the risk of the introduction of diseases is higher than the preset risk threshold, an early warning of the introduction of diseases is issued to the breeding farm. Obtain historical introduction records of the breeding farms, and extract the cross-regional pathogen carrying characteristic matching deviation value of the corresponding introduction batch, the first onset time of the corresponding introduction batch of pigs during the quarantine period, the number of sick pigs per day during the quarantine period, and the type identifier of the corresponding pathogen from each historical introduction record to construct a training sample set. Gradient boosting regression trees are used to iteratively train the training sample set. In each iteration, the residual between the predicted onset time and the actual onset time of the current model output is calculated. The sample weights of different matching deviation value intervals in the next iteration are dynamically adjusted according to the residual until the model converges, resulting in a risk-time coupled model composed of multiple regression trees. The corrected matching deviation baseline value is substituted into the converged risk-time coupled model. The split nodes of each regression tree are traversed sequentially. The branch to which the current input corrected matching deviation baseline value belongs is determined according to the matching deviation value range corresponding to the split node. The output values ​​of the branches to which all regression trees belong are accumulated to obtain the conditional incidence probability under different isolation durations. The conditional incidence probability under the continuous time dimension is integrated to generate the cumulative incidence probability curve corresponding to different isolation durations.

6. The method for early warning of pig farm introduction based on disease monitoring according to claim 1, characterized in that, The step of determining the safe window for introducing new pigs to the target pig farm based on the cumulative incidence probability curves corresponding to different isolation durations is as follows: The cumulative incidence probability values ​​corresponding to each discrete time point on the cumulative incidence probability curve are iterated through, and each cumulative incidence probability value is compared with the preset acceptable risk threshold for disease in the introduction area. All time points whose cumulative incidence probability values ​​are not greater than the acceptable risk threshold for disease are selected. The selected time points are arranged in chronological order to construct a candidate safe duration set. The time interval between two adjacent time points in the candidate safe duration set is extracted. Continuous time intervals with time intervals greater than the shortest operation time required for a single introduction operation in the introduction site are marked as preliminary safe windows. Obtain feeding and management plan data for the introduction site within the corresponding time period of the initial safety window, identify the operation periods involving vaccination, group transfer and mixed rearing, and environmental disinfection in the feeding and management plan, and remove the intervals in the initial safety window that overlap with the operation periods to obtain the airspace safety window; The start time of the net air safety window is aligned with the expected slaughter time of the pig herd to be introduced to the output farm. Net air safety windows with start times earlier than the expected slaughter time are eliminated. The start time of the remaining net air safety windows is used as the trigger node for the introduction operation, and the end time is used as the latest cutoff node for isolation and observation after the introduction, thus obtaining the introduction timing safety window for the target pig farm.

7. A pig farm introduction early warning system based on disease monitoring, characterized in that, The disease monitoring-based early warning system for pig farm introduction includes a memory and a processor. The memory includes a program for a disease monitoring-based early warning method for pig farm introduction. When the program is executed by the processor, it implements the steps of the disease monitoring-based early warning method for pig farm introduction as described in any one of claims 1 to 6.

Images