Snakebite immunological rapid detection result, bite image and data evolution characteristic snakebite toxin type auxiliary judgment method and system

Abstract

This invention discloses a method and system for auxiliary determination of snakebite venom type by integrating rapid immune detection results, bite images, and data evolution characteristics. It relates to the field of medical auxiliary diagnostic technology and addresses the problems of false negatives in existing rapid biochemical tests and the inability of static detection to reflect pathological deterioration. First, it acquires regional prior probabilities and rapid detection characteristics of wound secretions, generating an immune response matrix through nonlinear time compensation to overcome sample attenuation errors. Then, based on the immune matrix, it extracts preliminary tendency anchor points and intelligently drives a dual-modal device to target and acquire wound images. It extracts swelling expansion and thermal gradient decay rates to construct multidimensional spatiotemporal evolution characteristics, transforming static appearances into dynamic pathological parameters. Finally, it performs consistency comparison between biochemical characteristics and physical evolution, triggering weight redistribution when factual conflicts occur, and using dynamic pathological facts to forcefully correct the output judgment result, significantly improving the error tolerance and accuracy of emergency diagnosis.

Description

Technical Field

[0001] This invention relates to the field of medical auxiliary diagnostic technology, specifically to a method and system for auxiliary determination of snake bite venom type by integrating rapid immune detection results of snake venom, bite images and data evolution characteristics. Background Technology

[0002] Snakebite is an acute and critical illness caused by a venomous snake biting a human skin, allowing its venom to penetrate and cause local and systemic poisoning. The venom produced by different snake species is mainly classified into cytotoxic, neurotoxic, hemotoxic, and mixed venom types, with significantly different pathogenic mechanisms. For example, hemotoxic venom can cause local swelling, pain, and poor clotting of bleeding, while neurotoxic venom can lead to serious consequences such as respiratory failure. In actual emergency situations, victims often cannot accurately identify the species of snake that caused the bite. Medical personnel must determine the type of venom as early as possible to administer specific antivenom serum, thereby reducing the patient's mortality and disability rates.

[0003] Current methods for identifying snakebite venom types primarily rely on immunophenotyping reagents to obtain samples from the wound's secretions for immunophenotyping. However, these rapid testing methods are highly susceptible to interference from differences in venom composition, antibody cross-reactivity, and the hook effect caused by excessively high venom concentrations in the sample, leading to false negatives or false positives. Furthermore, biochemical tests based solely on local wound secretions provide static results, failing to reflect the dynamic process of pathological deterioration of local tissue over time. This can easily delay medical intervention for severe tissue ischemia and necrosis or rapidly spreading swelling when false negatives occur in immunophenotyping, posing a significant risk of missed diagnoses and misdiagnoses. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for auxiliary determination of snake bite venom type by integrating rapid snake venom immune detection results, bite images, and data evolution characteristics, thus solving the problems mentioned in the background.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for auxiliary determination of snake bite venom type by integrating rapid snake venom immunoassay results, bite images, and data evolution characteristics, comprising the following steps: S1. Obtaining the GPS location coordinates of the patient's bite, retrieving and extracting the prior probability vector of regional candidate venom types from a pre-set geographic information database of venom spatial distribution; extracting the biochemical colorimetric optical characteristic values ​​and sampling interval time of the rapid snake venom immunochromatographic test strip based on the secretion sample at the wound site, and performing nonlinear weighting of the biochemical colorimetric optical characteristic values ​​using the sampling interval time through a preset biochemical concentration-time compensation function to generate an immune response feature matrix; S2. Inputting the prior probability vector of regional candidate venom types and the immune response feature matrix into a maximum likelihood estimation model to extract preliminary venom type anchor points, configuring the collaborative acquisition parameters of visible light and infrared thermal imaging dual-modal devices; acquiring images of the bite site at continuous time nodes according to the collaborative acquisition parameters, and obtaining visible light texture sequences and infrared... Temperature field sequence; S3. Calculate the wound swelling boundary expansion rate between adjacent time nodes in the visible light texture sequence using optical flow method, and extract the infrared thermal gradient attenuation rate of the necrotic center in the infrared temperature field sequence based on isotherm tracking; construct a multidimensional spatiotemporal evolution feature tensor by feature splicing the expansion rate and the infrared thermal gradient attenuation rate, calculate the time axis elastic alignment distance between the multidimensional spatiotemporal evolution feature tensor and the pre-set standard venom type pathological evolution curve, and generate a physical evolution matching degree matrix; S4. Perform consistency logic comparison between the time-compensated biochemical feature values ​​extracted from the immune response feature matrix and the evolution matching score in the physical evolution matching degree matrix; when the value of the immune response feature matrix is ​​lower than the negative judgment threshold and the value of the physical evolution matching degree matrix exceeds the high-risk deterioration threshold, trigger the weight redistribution logic, assigning the decision weight of the physical evolution matching degree matrix higher than that of the immune response feature matrix; perform weighted fusion calculation based on the redistributed decision weights, and output the snake bite venom type determination result.

[0006] Further, the specific process of obtaining the patient's bite location coordinates and retrieving the regional candidate venom type prior probability vector from the pre-set venomous snake spatial distribution geographic information database is as follows: Parse the GPS location coordinates to extract environmental constraint features such as altitude, latitude and longitude, and surface vegetation cover type; input the environmental constraint features into the pre-set venomous snake spatial distribution geographic information database to perform a spatial topological intersection query, extract the historical activity snake species list matching the target geographic grid, and construct a multidimensional probability distribution set by associating the venom type distribution frequency built into the historical activity snake species list; perform normalization processing on the multidimensional probability distribution set to generate a regional candidate venom type prior probability vector composed of prior confidence of different venom types.

[0007] Furthermore, the specific process of extracting biochemical colorimetric optical feature values ​​and sampling intervals from snake venom immunochromatographic rapid test strips based on wound secretion samples, and using a preset biochemical concentration-time compensation function to perform nonlinear weighting on the biochemical colorimetric optical feature values ​​using the sampling intervals to generate an immune response feature matrix is ​​as follows: Images of the colorimetric area of ​​the snake venom immunochromatographic rapid test strip are acquired and converted to the HSV color space; saturation and brightness components are extracted and fused to generate biochemical colorimetric optical feature values; the time span from the moment of the patient's bite to the moment the wound secretion sample is dropped onto the rapid test strip is recorded as the sampling interval; the biochemical colorimetric optical feature values ​​and sampling intervals are substituted into a preset snake venom local tissue absorption and diffusion dynamics time-series distribution model; compensation coefficients for corresponding time nodes are extracted based on the model's built-in local concentration decay curve; nonlinear exponential scaling transformation is performed on the biochemical colorimetric optical feature values ​​using the corresponding time node compensation coefficients; and the transformed values ​​are arrayed according to a preset standard venom type classification space dimension to generate an immune response feature matrix.

[0008] Furthermore, the specific process of inputting the prior probability vector of regional candidate virulence types and the immune response feature matrix into the maximum likelihood estimation model to extract preliminary virulence type anchor points, and configuring the collaborative acquisition parameters of the visible light and infrared thermal imaging dual-modal devices is as follows: The immune response feature matrix is ​​used as the observation parameter of the likelihood function, and expectation maximization iteration is performed in combination with the prior probability vector of regional candidate virulence types to extract the virulence type associated with the global maximum posterior probability value in the convergent state as the preliminary virulence type anchor point; the pre-set pathological characterization mapping knowledge base is retrieved to extract the typical local lesion area category features associated with the preliminary virulence type anchor point, and the spatial alignment parameters, sampling time frequency and infrared detector temperature resolution threshold of the visible light and infrared thermal imaging dual-modal devices are configured according to the typical local lesion area category features to generate collaborative acquisition parameters.

[0009] Furthermore, the specific process of acquiring visible light texture sequences and infrared temperature field sequences by acquiring images of the bite site at continuous time nodes according to the collaborative acquisition parameters is as follows: the visible light sensor and infrared detector are calibrated to coincide the center of view according to the spatial alignment parameters in the collaborative acquisition parameters; the shutter synchronization action of the dual-mode device is triggered according to the sampling time frequency in the collaborative acquisition parameters; multiple frames of original dual-spectral images are captured within the preset monitoring period; spatial registration and region of interest boundary cropping are performed on the multiple frames of original dual-spectral images; and visible light texture sequences and infrared temperature field sequences with the same timestamp and spatial resolution are separated and output.

[0010] Furthermore, the specific process of calculating the wound swelling boundary expansion rate between adjacent time nodes in the visible light texture sequence using the optical flow method, and extracting the infrared thermal gradient attenuation rate of the necrotic center in the infrared temperature field sequence based on isotherm tracking is as follows: Construct the pixel grayscale conservation equation for adjacent time nodes in the visible light texture sequence; apply the dense optical flow field algorithm to extract the local deformation velocity vector field; perform surface integration on the local deformation velocity vector field along the normal direction of the initial wound boundary contour to output the wound swelling boundary expansion rate; extract the lowest temperature extreme point representing the abnormally cold region of local ischemia in the infrared temperature field sequence to establish a reference center; apply the morphological threshold segmentation algorithm to generate multi-level isotherm topological contours; measure the spatial displacement of the same-level isotherm topological contours towards the reference center at adjacent time nodes; fuse the sampling time interval parameters of adjacent time nodes to perform differential derivation and output the infrared thermal gradient attenuation rate of the necrotic center.

[0011] Furthermore, the expansion rate and infrared thermal gradient decay rate are concatenated to construct a multidimensional spatiotemporal evolution feature tensor. The time axis elastic alignment distance between the multidimensional spatiotemporal evolution feature tensor and the pre-set standard toxicant pathological evolution curve is calculated to generate a physical evolution matching degree matrix. The specific process is as follows: The expansion rate of the wound swelling boundary and the infrared thermal gradient decay rate of the necrosis center are extracted and input into the maximum and minimum normalization functions to eliminate dimensional differences. The multidimensional spatiotemporal evolution feature tensor is constructed by cascading according to the time series dimension. The pre-set standard toxicant pathological evolution curve is retrieved to construct a time series reference sample. A local cost space matrix of the multidimensional spatiotemporal evolution feature tensor and the time series reference sample is established. The minimum cumulative deformation cost connected path is solved in the local cost space matrix through a dynamic programming path search algorithm. The terminal value of the connected path is extracted and mapped to the time axis elastic alignment distance. An exponential decay kernel function is introduced to perform confidence transformation on the time axis elastic alignment distance. The confidence transformation value is combined with the vector of the initial toxicant anchor point to generate a physical evolution matching degree matrix.

[0012] Furthermore, the logical process of performing a consistency logic comparison between the time-compensated biochemical feature values ​​in the immune response feature matrix and the evolutionary matching scores in the physical evolutionary matching degree matrix is ​​as follows: Separate the time-compensated biochemical feature values ​​within the immune response feature matrix, simultaneously extract the evolutionary matching scores within the physical evolutionary matching degree matrix, retrieve the mapping table of the preliminary tendency toxicity anchor points associated with the toxicological evolution stage, retrieve the time-compensated biochemical feature values ​​mapping to the early mild toxin accumulation label, and retrieve the evolutionary matching scores mapping to the mid-to-late stage severe tissue necrosis label; establish a Boolean logic conflict discriminant between the early mild toxin accumulation label and the mid-to-late stage severe tissue necrosis label, and trigger the inconsistency logic judgment flag when the Boolean logic conflict discriminant satisfies the condition of a cross-level collision in pathological deterioration.

[0013] Furthermore, when the value of the immune response feature matrix is ​​lower than the negative determination threshold and the value of the physical evolution matching degree matrix exceeds the high-risk deterioration threshold, the weight redistribution logic is triggered, assigning the decision weight of the physical evolution matching degree matrix higher than that of the immune response feature matrix. The specific process of weighted fusion calculation based on the redistributed decision weights and outputting the snakebite venom type determination result is as follows: The inconsistency logic determination identifier is parsed. Under the condition that the value of the immune response feature matrix falls below the set negative determination threshold boundary and the value of the physical evolution matching degree matrix crosses the set high-risk deterioration threshold boundary, the evolution feature priority logic mechanism is activated. A pre-set penalty attenuation factor is introduced to perform exponential dimensionality reduction modulation on the original weight parameters of the immune response feature matrix, and a pre-set gain amplification factor is introduced to perform logarithmic dimensionality increase modulation on the original weight parameters of the physical evolution matching degree matrix. These are then merged to generate a redistribution decision weight vector. The redistribution decision weight vector is applied to perform inner product space tensor fusion calculation on the immune response feature matrix and the physical evolution matching degree matrix. The classification dimension mapping standard snakebite venom type library with the largest response value in the inner product space tensor fusion calculation result is extracted, and the snakebite venom type determination result is output.

[0014] This snake bite venom type auxiliary determination system integrates snake venom immune rapid detection results, bite images, and data evolution characteristics, and includes the following modules: An immune response extraction module, used to obtain the patient's bite GPS location coordinates, retrieve and extract the prior probability vector of regional candidate venom types from a pre-set geographical information database of snake spatial distribution; extract biochemical colorimetric optical feature values ​​and sampling interval time from snake venom immunochromatographic rapid detection test strips based on wound secretion samples, and perform nonlinear weighting of the biochemical colorimetric optical feature values ​​using the sampling interval time through a preset biochemical concentration-time compensation function to generate an immune response feature matrix; a targeted image acquisition module, used to input the prior probability vector of regional candidate venom types and the immune response feature matrix into a maximum likelihood estimation model to extract preliminary venom type anchor points, configure the collaborative acquisition parameters of visible light and infrared thermal imaging dual-modal devices; acquire images of the bite area at continuous time nodes according to the collaborative acquisition parameters, and obtain visible light texture sequences and infrared temperature field sequences; spatiotemporal evolution matching. The module is used to calculate the expansion rate of wound swelling boundaries between adjacent time nodes in the visible light texture sequence using optical flow method, and to extract the infrared thermal gradient decay rate of the necrotic center in the infrared temperature field sequence based on isotherm tracking; the expansion rate and the infrared thermal gradient decay rate are concatenated to construct a multidimensional spatiotemporal evolution feature tensor, and the time axis elastic alignment distance between the multidimensional spatiotemporal evolution feature tensor and the pre-set standard venom type pathological evolution curve is calculated to generate a physical evolution matching degree matrix; the conflict correction judgment module is used to extract the time-compensated biochemical feature values ​​in the immune response feature matrix and compare them with the evolution matching scores in the physical evolution matching degree matrix using consistency logic; when the value of the immune response feature matrix is ​​lower than the negative judgment threshold and the value of the physical evolution matching degree matrix exceeds the high-risk deterioration threshold, the weight redistribution logic is triggered, and the physical evolution matching degree matrix is ​​given a higher decision weight than the immune response feature matrix; the weighted fusion calculation is performed based on the redistributed decision weights to output the snake bite venom type judgment result.

[0015] The present invention has the following beneficial effects:

[0016] (1) A method for assisting in the determination of snake bite venom type by integrating rapid snake venom immune test results, bite images and data evolution characteristics. By obtaining the GPS location coordinates of the patient's bite, the prior probability vector of regional candidate venom types is extracted. The biochemical colorimetric optical feature values ​​and sampling interval based on the secretion fluid sample at the wound site are extracted. The biochemical concentration time compensation function is used to perform nonlinear weighting to generate an immune response feature matrix, which effectively overcomes the hook effect and concentration decay error that are very easy to occur when relying on rapid test reagents. At the same time, the prior probability and immune response are combined to extract the preliminary venom type anchor point, and then the collaborative acquisition parameters of the visible light and infrared thermal imaging dual-mode equipment are accurately configured, realizing cross-modal linkage from biochemical prior to physical image targeted acquisition.

[0017] (2) The snake bite venom type auxiliary determination system integrates the results of rapid immune detection of snake venom, bite images and data evolution characteristics. It calculates the expansion rate of wound swelling boundary by optical flow method and extracts the infrared thermal gradient attenuation rate of necrotic center based on isotherm tracking. It constructs a multidimensional spatiotemporal evolution feature tensor and calculates the time axis elastic alignment distance, transforming static appearance into dynamic pathological deterioration time sequence features. At the same time, when the immune response feature value is lower than the negative determination threshold and the physical evolution value exceeds the high-risk deterioration threshold, the weight redistribution logic is triggered to give the physical evolution higher decision weight, which completely solves the misdiagnosis problem caused by false negatives of rapid immune detection. Through dynamic evolution facts forced correction, the error tolerance and accuracy of snake bite venom type determination in emergency environment are greatly improved.

[0018] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method for auxiliary determination of snake bite venom type that integrates rapid snake venom immune detection results, bite images, and data evolution characteristics.

[0020] Figure 2 This is a detailed flowchart of the physical evolution matching degree matrix generation of the present invention.

[0021] Figure 3 This is a detailed flowchart of the consistency logic comparison and weight redistribution of the present invention.

[0022] Figure 4 This is a flowchart of the snake bite venom type auxiliary determination system that integrates rapid snake venom immune detection results, bite images, and data evolution characteristics. Detailed Implementation

[0023] This application's embodiments solve the problems of existing snake venom immune rapid detection reagents being susceptible to cross-reaction and hook effect interference, resulting in false negative results, and single static biochemical detection failing to reflect the dynamic deterioration process of local pathology, leading to inaccurate venom type determination, by integrating snake venom immune rapid detection results, bite images, and data evolution characteristics into an auxiliary method and system for snake bite venom type determination.

[0024] The overall concept of the solution in this application embodiment is as follows:

[0025] First, spatial geographic information is used to obtain the prior probability of venomous snakes in the region. Combined with the rapid detection results of wound secretions after time decay compensation, preliminary venom type tendency anchor points are generated and a dual-modal image acquisition device is configured accordingly. Then, visible light and infrared images of the bite site are acquired at continuous time points to extract the spatiotemporal evolution dynamic rate reflecting the spread of swelling and tissue necrosis. Finally, the consistency of early biochemical rapid detection results and later pathological evolution characteristics is compared. When the biochemical results are negative but the physical evolution shows high-risk deterioration, the decision weight redistribution mechanism is automatically triggered. The dynamic pathological evolution facts are used as the main decision basis for forced correction, and finally, an accurate snake bite venom type determination result is output.

[0026] Please see Figure 1 This invention provides a technical solution: a method for assisting in the determination of snake bite venom type by integrating rapid snake venom immunoassay results, bite images, and data evolution characteristics, comprising the following steps: S1. Obtaining the GPS location coordinates of the patient's bite, retrieving and extracting the prior probability vector of regional candidate venom types from a pre-set geographic information database of venom spatial distribution; extracting the biochemical colorimetric optical feature values ​​and sampling interval time of the rapid snake venom immunochromatographic test strip based on the secretion sample at the wound site, and performing nonlinear weighting of the biochemical colorimetric optical feature values ​​using the sampling interval time through a preset biochemical concentration-time compensation function to generate an immune response feature matrix; S2. Inputting the prior probability vector of regional candidate venom types and the immune response feature matrix into a maximum likelihood estimation model to extract preliminary venom type anchor points, configuring the collaborative acquisition parameters of visible light and infrared thermal imaging dual-modal devices; acquiring images of the bite site at continuous time nodes according to the collaborative acquisition parameters, and obtaining visible light texture sequences and infrared temperature field sequences. S3. Calculate the wound swelling boundary expansion rate between adjacent time nodes in the visible light texture sequence using optical flow method, and extract the infrared thermal gradient attenuation rate of the necrotic center in the infrared temperature field sequence based on isotherm tracking; construct a multidimensional spatiotemporal evolution feature tensor by feature splicing the expansion rate and the infrared thermal gradient attenuation rate, calculate the time axis elastic alignment distance between the multidimensional spatiotemporal evolution feature tensor and the pre-set standard venom type pathological evolution curve, and generate a physical evolution matching degree matrix; S4. Perform consistency logic comparison between the time-compensated biochemical feature values ​​extracted from the immune response feature matrix and the evolution matching score in the physical evolution matching degree matrix; when the value of the immune response feature matrix is ​​lower than the negative judgment threshold and the value of the physical evolution matching degree matrix exceeds the high-risk deterioration threshold, trigger the weight redistribution logic, assigning the decision weight of the physical evolution matching degree matrix higher than that of the immune response feature matrix; perform weighted fusion calculation based on the redistributed decision weights, and output the snake bite venom type determination result.

[0027] In this implementation plan, step S1 is mainly used to acquire regional prior information and calibrate biochemical test results. First, the system extracts a regional candidate venom type prior probability vector from the database using the GPS coordinates of the patient's bite location. This vector refers to the initial probability set of various snake venom types that may appear in a specific area, quantified based on historical geographical distribution data and biostatistical laws. Simultaneously, the system extracts the biochemical colorimetric optical characteristic values ​​of the wound secretions on an immunochromatographic rapid test strip and, combined with the sampling interval, performs a nonlinear weighted calculation using a biochemical concentration and time compensation function. The biochemical concentration and time compensation function is a mathematical algorithm model specifically designed to compensate for the physical error caused by the gradual decrease in toxin concentration in the wound surface secretions as the venom diffuses into the deeper blood and lymphatic system over time. Through this step, the system can effectively overcome the false negative problem in early biochemical rapid tests caused by sampling time lag or abnormal local toxin concentrations, providing a precise immune response feature matrix with spatiotemporal dual calibration for subsequent auxiliary diagnosis. Step S2 is mainly used to drive the operation of the targeted image acquisition device based on the preliminary biochemical judgment results. The system inputs the acquired regional probability and immune response matrix into the maximum likelihood estimation model. The maximum likelihood estimation model is a mathematical algorithm based on statistical probability analysis, capable of reverse-engineering from the known observation data set to find the initial parameter combination most likely to lead to the observation result. Here, it is used to calculate the preliminary venom type anchor point with the highest probability, i.e., the initially determined high-suspected venom type target. Subsequently, based on this preliminary target, the system selectively adjusts the collaborative acquisition parameters of the visible light and infrared thermal imaging dual-modal devices, and acquires images of the wound area at continuous time points. This step changes the traditional static and blind wound photography mode in emergency medicine, realizing cross-modal linkage where early biochemical and geographical prior data guide medical hardware to perform specific spatial and frequency acquisition, laying a high-quality data foundation for subsequent extraction of objective pathological evolution features. Step S3 is mainly used to extract the temporal features of pathological deterioration from continuous images and align them with standard pathological evolution processes. The system employs optical flow to process visible light texture sequences. Optical flow is a low-level image processing algorithm in computer vision used to calculate the pixel motion velocity vector field between adjacent image frames. Here, it is used to accurately quantify the physical expansion rate of wound swelling as it spreads from the outer skin boundary. Simultaneously, it extracts the thermal gradient decay rate of the necrotic center from the infrared temperature field sequence based on isotherm tracking. Isotherm tracking refers to tracing the contraction or expansion trend of a specific temperature level contour in continuous infrared thermal imaging to characterize the rate of development and deterioration of local tissue ischemia and necrosis. After stitching these rates together, the system calculates the elastic alignment distance between the time axis and the standard evolution curve. This is a statistical algorithm that can stretch or compress the time axis to calculate the morphological similarity between two non-co-frequency time series, used to eliminate feature drift caused by different individual patient metabolic rates.This step transforms the static visual appearance of the wound into a dynamic parameter of pathological evolution speed, providing an objective physical verification dimension independent of biochemical reagents for toxicity determination. Step S4 is mainly used to perform conflict correction of global heterogeneous data and comprehensive determination of the final toxicity. The system extracts biochemical feature values ​​and performs consistency logic comparison with evolutionary matching scores. When a contradictory situation occurs where the immune feature value shows a low-risk negative range, while the physical evolutionary matching matrix value shows an extremely deteriorated high-risk state, the system will automatically trigger the weight redistribution logic. The weight redistribution logic is a conflict arbitration algorithm mechanism that dynamically adjusts the proportion of different heterogeneous evidence sources in the final decision-making system. In this conflict state, the system determines that the early rapid test strips have serious false negatives due to sample limitations, and thus forcibly reduces or even blocks the decision weight of biochemical results, and significantly increases the decision weight of true physical evolutionary features. Through this step, the system can intelligently identify and filter erroneous indicators and forcibly correct errors based on the most realistic dynamic evolution facts when there is a serious discrepancy between biochemical test results and local pathological deterioration. This completely solves the technical pain point that existing single-reagent rapid tests are prone to misdiagnosis and outputs highly reliable auxiliary determination results of snakebite venom type.

[0028] Specifically, the process of obtaining the GPS location coordinates of the patient's bite and retrieving the prior probability vector of regional candidate venom types from a pre-set geographic information database of venomous snake spatial distribution is as follows: Parse the GPS location coordinates to extract environmental constraint features such as altitude, latitude and longitude, and surface vegetation cover type; input the environmental constraint features into the pre-set geographic information database of venomous snake spatial distribution to perform a spatial topological intersection query, extract the historical activity snake species list matching the target geographic grid, and construct a multidimensional probability distribution set by associating the venom type distribution frequency built into the historical activity snake species list; perform normalization processing on the multidimensional probability distribution set to generate a prior probability vector of regional candidate venom types composed of prior confidence levels for different venom types.

[0029] In this implementation plan, the environmental constraint features of altitude, latitude and longitude, and surface vegetation cover type are extracted from the GPS coordinates of the patient's bite location. This is because the survival of different types of venomous snakes is highly dependent on specific microclimates and vegetation ecosystems. This step plays a technical role in narrowing down the scope of suspected venomous snake species in the initial stage of the identification process by utilizing geographical priors and objective natural laws. The environmental constraint features are then input into a geographic information database to perform a spatial topological intersection query, which calculates the geometric overlap between the injured person's spatial location and pre-defined polygon layers of historical habitats for various snake species, accurately filtering out historically active snake species within the target geographic grid. Subsequently, the system does not simply rely on frequency of occurrence but comprehensively associates specific environmental adaptation indicators within the area to construct a multidimensional probability distribution set, and performs normalization processing to generate confidence scores. The specific calculation formula is as follows: In the formula, Indicates that for the first The prior probability confidence values ​​generated from the candidate venom types are calculated. This represents the historical absolute observation frequency record value of the snake species corresponding to this venom type within the matched target geographic grid; This indicates that the extracted surface vegetation cover type of the crime scene is related to the first... Morphological overlap and matching degree of standard habitat vegetation characteristics for snake species with different venom types; This represents the probability density mapping coefficient of the extracted absolute altitude value onto the Gaussian distribution model of the suitable altitude for the survival of snake species with this venom type. This represents the traversal index parameter during the normalized cumulative summation process; This represents the total number of all possible venom types retrieved through spatial topological intersection within the target geographic grid. Through mathematical normalization in this step, the absolute historical frequency and environmental fit are transformed into a standard coarse feature vector with a constant sum, eliminating the dimensional inconsistencies caused by differences in the absolute number of basic venomous snakes in different regional ecosystems. This provides a spatial prior probability baseline with a unified data scale for subsequent cross-modal multi-source heterogeneous data fusion.

[0030] Specifically, the process of extracting biochemical colorimetric optical feature values ​​and sampling intervals from snake venom immunochromatographic rapid test strips based on wound secretion samples, and using a preset biochemical concentration-time compensation function to perform nonlinear weighting on the biochemical colorimetric optical feature values ​​using the sampling intervals to generate an immune response feature matrix is ​​as follows: The image of the colorimetric area of ​​the snake venom immunochromatographic rapid test strip is acquired and converted to the HSV color space; saturation and brightness components are extracted and fused to generate biochemical colorimetric optical feature values; the time span from the moment of the patient's bite to the moment the wound secretion sample is dropped onto the rapid test strip is recorded as the sampling interval; the biochemical colorimetric optical feature values ​​and sampling intervals are substituted into a preset snake venom local tissue absorption and diffusion dynamics time-series distribution model; compensation coefficients for corresponding time nodes are extracted based on the model's built-in local concentration decay curve; nonlinear exponential scaling transformation is performed on the biochemical colorimetric optical feature values ​​using the corresponding time node compensation coefficients; and the transformed values ​​are arrayed according to a preset standard venom type classification space dimension to generate an immune response feature matrix.

[0031] In this implementation plan, the image of the colorimetric area of ​​the snake venom immunochromatographic rapid test strip is acquired and converted from the conventional red-green-blue color space to a hue-saturation-brightness color space. This is because lighting conditions in wilderness emergency rescue scenarios are extremely variable and complex. Extracting the saturation and brightness components separately can effectively eliminate the physical interference of ambient light and shadow and reflection phenomena on the colorimetric concentration recognition of the test strip. The extracted components are then fused to generate biochemical colorimetric optical feature values. The specific calculation model is as follows: In the formula, This represents the biochemical colorimetric optical characteristic value output after fusion processing; This represents the average purity and saturation components of the colorimetric area image extracted after removing hue and ambient brightness. This represents the average value of the brightness component of the extracted colorimetric area image on the test strip; This represents the color channel fusion weighting coefficient. The specific method for determining this weighting coefficient is to pre-collect a standard concentration snake venom test strip test image set containing multiple light interferences, calculate the information entropy of the saturation channel and the information entropy of the brightness channel separately, and take the empirical ratio of the information entropy of the saturation channel to the total information entropy of the two as the determined weighting coefficient value. Obtaining the sampling interval time and substituting it into a pre-set snake venom local tissue absorption and diffusion dynamics time-series distribution model for weighted compensation is because after venom is injected into the local human tissue, it gradually diffuses into the deeper layers of the patient's body over time through microvascular and lymphatic fluid circulation, causing the concentration of toxins remaining in the surface secretions of the wound to exhibit a dynamic nonlinear decay. If the time factor is not compensated for, simply relying on the color depth of the test strip at the current moment can easily lead to serious false negatives. After extracting the time node compensation coefficient based on the model's built-in local concentration decay curve mapping, the system applies this coefficient to perform a nonlinear exponential scaling transformation on the biochemical colorimetric optical characteristic values. The specific formula is as follows: In the formula, This represents the true immunobiochemical response mapping characteristic value after nonlinear weighted compensation and restoration along the time axis, where e represents the base constant of the natural logarithm. It represents the fluid permeation compensation coefficient extracted from the built-in local concentration decay curve for a specific time point, used to characterize the average diffusion decay rate of venom absorbed by local tissue microcirculation during that time period. This indicates the sampling interval from the initial moment of the patient's bite injury to the actual moment when the secretion sample at the wound site was dripped onto the rapid test strip; This is represented by a pre-defined biochemical regulatory constant characterizing the transmembrane diffusion damping ability of snake venom proteins of different molecular weights. Finally, according to the multidimensional spatial dimensions of the pre-defined standard venom type classification, the system sequentially maps the calculated transformed values ​​corresponding to various specific toxins into an empty feature tensor and performs array expansion, thereby generating an immune response feature matrix covering all identification dimensions. Through this core pre-processing step, mathematical derivation reconstructs the initial state of the true biochemical response concentration before large-scale diffusion of venom upon initial entry into wound tissue, completely solving the problem of concentration decay distortion and judgment blind spots in rapid surface detection reagents caused by delays in the emergency window period.

[0032] Specifically, the process of inputting the prior probability vector of regional candidate virulence types and the immune response feature matrix into the maximum likelihood estimation model to extract preliminary virulence type anchor points, and configuring the collaborative acquisition parameters of the visible light and infrared thermal imaging dual-modal devices is as follows: The immune response feature matrix is ​​used as the observation parameter of the likelihood function, and expectation maximization iteration is performed in combination with the prior probability vector of regional candidate virulence types to extract the virulence type associated with the global maximum posterior probability value in the convergent state as the preliminary virulence type anchor point; the pre-set pathological characterization mapping knowledge base is retrieved to extract the typical local lesion area category features associated with the preliminary virulence type anchor point, and the spatial alignment parameters, sampling time frequency and infrared detector temperature resolution threshold of the visible light and infrared thermal imaging dual-modal devices are configured according to the typical local lesion area category features to generate collaborative acquisition parameters.

[0033] In this implementation scheme, the immune response feature matrix is ​​used as the observation parameter of the likelihood function and combined with the prior probability vector of regional candidate venom types to perform expectation-maximization iteration. This is because early rapid immune detection, despite time compensation, may still have slight observation noise caused by environmental interference. Introducing maximum likelihood estimation and expectation-maximization iteration allows for the deep mathematical fusion of prior macro-probabilities in spatial regions and micro-level biochemical observation data using Bayesian posterior logic, enabling the search for the snake species most closely resembling the actual culprit from incomplete observation data. The system continuously updates the posterior probabilities of various possible venom types during the iteration process until the state converges. The specific calculation formula is as follows: In the formula, : indicates the first The calculation of the first iteration step The posterior probability value after the venom type is updated; : Represents the corresponding digit extracted from the prior probability vector of regional candidate virulence types. Prior probability confidence values ​​for different venom types; : Represents the i-th element in the immune response feature matrix Time-compensated biochemical characteristic observations in each dimension; : Indicates the total number of feature dimensions contained within the immune response feature matrix; : indicates the first In the nth iteration Model distribution state parameters corresponding to different venom types; : Represents the value of the conditional likelihood function for observing a specific immune response characteristic under given venom type distribution state parameters; : Indicates the total number of candidate venom types covered in the pre-set geographic information database and pathological characterization mapping knowledge base; This represents the index variable that iterates through all candidate venom types when calculating the denominator of the global posterior probability. Through iterative convergence of this step, the system extracts the venom type associated with the global maximum posterior probability value as an initial venom anchor point. Subsequently, a pre-set pathological characterization mapping knowledge base is retrieved to extract the typical local lesion area category features associated with this anchor point. Because the pathological changes caused by different venom types vary greatly—for example, hemotoxic venoms easily cause large-area, rapidly spreading swelling, while cytotoxic venoms easily cause local deep tissue necrosis—the system configures the spatial alignment parameters, sampling time frequency, and infrared detector temperature resolution threshold of the dual-modal visible light and infrared thermal imaging devices differently based on these lesion characteristics. The infrared detector temperature resolution threshold is determined by extracting the root mean square deviation of the local skin temperature gradient in the early stages of the disease in historical confirmed cases of the same venom type, and then floating it downwards by one standard deviation as the detection sensitivity boundary. By combining these collaborative acquisition parameters, subsequent medical imaging hardware can possess targeted capture capabilities, avoiding the omission of pathological details caused by blindly taking pictures.

[0034] Please see Figure 2 Specifically, the process of acquiring images of the bite site at continuous time points according to the collaborative acquisition parameters to obtain the visible light texture sequence and the infrared temperature field sequence is as follows: The visible light sensor and the infrared detector are calibrated to coincide the center of view according to the spatial alignment parameters in the collaborative acquisition parameters; the dual-mode device shutter synchronization is triggered according to the sampling time frequency in the collaborative acquisition parameters; multiple frames of original dual-spectral images are captured within a preset monitoring period; spatial registration and region of interest boundary cropping are performed on the multiple frames of original dual-spectral images; and the visible light texture sequence and the infrared temperature field sequence with the same timestamp and spatial resolution are output separately.

[0035] In this implementation scheme, the visible light sensor and infrared detector are calibrated to coincide the center of view based on the spatial alignment parameters in the collaborative acquisition parameters. This is because the dual-modal device has two physical lenses, and the acquired images inevitably have spatial parallax. Direct superposition would cause the heat source position to misalign with the actual skin texture. The system uses calibration parameters to eliminate parallax and triggers the shutter synchronization action of the dual-modal device according to the sampling time frequency to ensure that the multiple frames of original dual-spectral images captured within the preset monitoring period are absolutely consistent in timestamps. After acquiring the original images, the system performs spatial registration and region of interest boundary cropping on the multiple frames of original dual-spectral images to filter out irrelevant clutter pixels in the background environment and achieve accurate mapping of the dual spectra in the two-dimensional coordinate system. The specific spatial registration and mapping calculation formula is as follows: In the formula, : Represents the horizontal coordinate parameter of a pixel in a visible light texture sequence image after registration and mapping; : Represents the vertical ordinate parameter of a pixel in a visible light texture sequence image after registration and mapping; : Represents the homography geometric perspective transformation matrix composed of spatial alignment parameters extracted from the collaborative acquisition parameters, used to perform translation, rotation, and scaling correction in image space. This represents the horizontal coordinate parameter of the feature pixels extracted from the original infrared temperature field sequence image. This represents the vertical ordinate parameter of the feature pixels extracted from the original infrared temperature field sequence image. Through the above homography spatial mapping and edge cropping processing, the system completely eliminates the spatiotemporal asynchrony bias caused by the device hardware architecture, and finally separates and outputs visible light texture sequences and infrared temperature field sequences with the same timestamp and spatial resolution. This step eliminates physical interference at the image level, and delivers the purest and highly aligned visual texture and thermodynamic distribution data of skin lesions to the lower-level algorithm, so that the subsequent optical flow method for edge tracking and isotherm extraction has a solid and rigorous data foundation.

[0036] Specifically, the process of calculating the wound swelling boundary expansion rate between adjacent time nodes in the visible light texture sequence using optical flow method, and extracting the infrared thermal gradient attenuation rate of the necrotic center in the infrared temperature field sequence based on isotherm tracking, is as follows: Construct the pixel grayscale conservation equation for adjacent time nodes in the visible light texture sequence; apply the dense optical flow field algorithm to extract the local deformation velocity vector field; perform surface integration on the local deformation velocity vector field along the normal direction of the initial wound boundary contour to output the wound swelling boundary expansion rate; extract the lowest temperature extreme point representing the abnormally cold area of ​​local ischemia in the infrared temperature field sequence to establish a reference center; apply the morphological threshold segmentation algorithm to generate multi-level isotherm topological contours; measure the spatial displacement of the same-level isotherm topological contours towards the reference center at adjacent time nodes; fuse the sampling time interval parameters of adjacent time nodes to perform differential derivation and output the infrared thermal gradient attenuation rate of the necrotic center.

[0037] In this implementation scheme, the gray-level conservation equation for adjacent time nodes of the visible light texture sequence is constructed because the optical reflection characteristics of the wound area remain essentially unchanged over a short period of time. By assuming that the gray-level of the same physical point is constant in adjacent time frames, the system can apply a dense optical flow field algorithm to track the displacement trajectory of the skin texture pixel by pixel, thereby extracting a fine local deformation velocity vector field. Considering that blood toxins can cause rapid centripetal spread and severe swelling and expansion of the wound, the system performs an area integral operation on the local deformation velocity vector field along the normal direction of the initial boundary contour of the wound to quantify this macroscopic pathological diffusion trend and output the wound swelling boundary expansion rate. The specific calculation formula is as follows: In the formula, : Indicates the calculated rate of expansion of the wound swelling boundary; : Represents the total number of pixel sampling points after discretizing the initial boundary contour of the wound; : Represents the actual sampling time interval parameter between adjacent time nodes in the visible light texture sequence; q: Represents the iteration index variable for traversing the initial boundary contour pixel sampling points of the wound; : Represents the coordinates of the point calculated using the dense optical flow field algorithm. Local deformation velocity vector field parameters at the location; : Indicates the coordinate point The unit normal vector radiates outward from the initial boundary contour of the wound towards the surrounding normal tissue. Simultaneously, cytotoxicity can cause myocardial damage and induce local microvascular thrombosis and severe tissue ischemia, which manifests as abnormally cold areas in the infrared temperature field image. The system extracts the lowest temperature extreme point representing the abnormally cold area of ​​local ischemia to establish a reference center, and applies a morphological thresholding segmentation algorithm to generate multi-level isothermal topological contours with a set temperature gradient step size. By measuring the spatial displacement of the topological contours of the same level of isotherms contracting towards the reference center at adjacent time points, and fusing the sampling time interval parameter of adjacent time points, the system performs differential derivation of spatial features on time variables, outputting the infrared thermal gradient attenuation rate of the necrotic center. This process transforms the microscopic pathological deterioration characteristics that are difficult for doctors to accurately quantify with the naked eye in emergency scenarios into objective physical change data, providing a core dynamic evolution criterion for distinguishing between high-risk blood toxins and deep cytotoxicity.

[0038] Specifically, the process of constructing a multidimensional spatiotemporal evolution feature tensor by concatenating the expansion rate and the infrared thermal gradient decay rate, and calculating the time axis elastic alignment distance between the multidimensional spatiotemporal evolution feature tensor and the pre-set standard toxicant pathological evolution curve to generate a physical evolution matching degree matrix is ​​as follows: The expansion rate of the wound swelling boundary and the infrared thermal gradient decay rate of the necrosis center are extracted and input into a maximum-minimum normalization function to eliminate dimensional differences. A positional cascade is performed according to the time series dimension to construct the multidimensional spatiotemporal evolution feature tensor. A time series reference sample is constructed by retrieving the pre-set standard toxicant pathological evolution curve. A local cost space matrix is ​​established between the multidimensional spatiotemporal evolution feature tensor and the time series reference sample. A dynamic programming path search algorithm is used to solve for the minimum cumulative deformation cost connected path in the local cost space matrix. The endpoint value of the connected path is extracted and mapped to the time axis elastic alignment distance. An exponential decay kernel function is introduced to perform confidence transformation on the time axis elastic alignment distance. The confidence transformation value is combined with the vector of the initial toxicant type anchor point to generate the physical evolution matching degree matrix.

[0039] In this implementation scheme, the expansion rate of the wound swelling boundary and the attenuation rate of the infrared thermal gradient at the necrosis center are extracted and input into the maximum-minimum normalization function to eliminate the dimensional difference. This is because the expansion rate reflects the magnitude of spatial area change over time, while the attenuation rate reflects the temperature gradient decrease over time. Since their physical dimensions are different, they cannot be directly compared and calculated in the same vector space. By normalizing and performing positional cascading according to the time series dimension to construct a multidimensional spatiotemporal evolution feature tensor, the system can fuse multidimensional pathological features to form a complete dynamic evolution trajectory. Due to significant differences in age, physical condition, and basal metabolic rate among different patients, the absolute time axis of the same toxin attack may exhibit stretching or compression. To solve this problem, the system retrieves a pre-set standard toxicity pathological evolution curve to construct a time series reference sample, establishes a local cost space matrix of the multidimensional spatiotemporal evolution feature tensor and the time series reference sample, and solves for the minimum cumulative deformation cost connected path in the local cost space matrix using a dynamic programming path search algorithm. The core logic of the dynamic programming calculation process lies in finding the optimal alignment method with the minimum cumulative error at each time point. The specific recursive formula is as follows: In the formula, : indicates the coordinate position in the local cost space matrix. The minimum value of the cumulative deformation cost calculated at that point; : Represents the multidimensional spatiotemporal evolution feature tensor at time step The feature vector at the specified location and the time series reference sample at the specified time step The cost of the local Euclidean distance between the standard vectors at a given location; : Represents the time series recursive index variable of the multidimensional spatiotemporal evolution feature tensor; : Represents the time series recursive index variable of the time series reference sample; The function operation represents the minimum value among the three parameters within the parentheses. The system maps the final endpoint value of the completed connected path to the time axis flexible alignment distance, which effectively overcomes the matching misalignment error caused by individual patient metabolic and immune differences. Subsequently, the system introduces an exponential decay kernel function to perform confidence transformation on the time axis flexible alignment distance. The smoothing penalty coefficient inside the decay kernel function is determined by extracting a set of historically diagnosed snake bite samples with typical evolutionary characteristics, and calculating the inverse of the distribution variance of the time axis flexible alignment distance of all samples in this set as the smoothing penalty coefficient. The system combines the calculated confidence transformation value with the initial venom type anchor point execution vector to finally generate a physical evolution matching degree matrix. This provides a highly reliable and quantitative physical evolution verification basis for subsequent intelligent screening of false negative defects in immunoassay test strips and forced weight correction.

[0040] Please see Figure 3Specifically, the logical process of comparing the time-compensated biochemical feature values ​​in the immune response feature matrix with the evolutionary matching scores in the physical evolutionary matching degree matrix is ​​as follows: Separate the time-compensated biochemical feature values ​​in the immune response feature matrix, simultaneously extract the evolutionary matching scores in the physical evolutionary matching degree matrix, retrieve the mapping table of the preliminary tendency toxicity anchor points associated with the toxicological evolution stage, retrieve the time-compensated biochemical feature values ​​to map the early mild toxin accumulation label, and retrieve the evolutionary matching scores to map the mid-to-late stage severe tissue necrosis label; establish a Boolean logic conflict discriminant between the early mild toxin accumulation label and the mid-to-late stage severe tissue necrosis label, and trigger the inconsistency logic judgment flag when the Boolean logic conflict discriminant satisfies the condition of cross-level collision of pathological deterioration.

[0041] In this implementation plan, the time-compensated biochemical feature values ​​within the immune response feature matrix are separated, and evolutionary matching scores are extracted simultaneously for consistency logic comparison. This is because, in real-world emergency rescue and clinical treatment scenarios, while snake venom immunoassay reagents offer the advantage of rapid screening, they are highly susceptible to cross-reaction interference caused by differences in venom composition across different regions. Furthermore, excessively high toxin concentrations at the wound site can easily induce the hook effect. These objective limitations of biochemical testing can lead to false negatives. If the auxiliary judgment system blindly relies on a single biochemical indicator, it can easily mask the true trend of disease deterioration and delay the golden period for serum injection. Therefore, the system retrieves the toxicological evolution stage mapping table, mapping the extracted biochemical feature values ​​to early-stage mild toxin accumulation labels and the extracted evolutionary matching scores to mid-to-late-stage severe tissue necrosis labels, thereby discretizing the continuous-dimensional collected values ​​into medical pathological stage status labels. The Boolean logic conflict discriminant is established to accurately quantify the contradictions between the visual appearance and biochemical substance of the wound. The specific logic discriminant conditions and state activation calculation formulas are as follows: In the formula, : This indicates the triggering inconsistency logic judgment flag. A value of 1 indicates that there is a collision condition of the pathological deterioration level, and a value of 0 indicates that the physical pathological manifestations and biochemical and immunological indicators are in a consistent state. : Represents the time-compensated biochemical feature values ​​extracted from the immune response feature matrix; : This represents the preset negative threshold. The method for determining this threshold is to collect a large number of subcutaneous tissue fluid samples from healthy volunteers who have been confirmed not to have been bitten by venomous snakes, perform rapid immunochromatographic testing, extract the upper limit of the normal distribution of biochemical characteristic values ​​of all healthy samples, and add a 5% tolerance margin to establish this threshold parameter. This represents the Boolean logical AND operator, used to determine the mathematical logical boundary where two conditions are simultaneously true. : Represents the evolution matching score extracted from the physical evolution matching degree matrix; : This represents the set high-risk deterioration threshold. The method for determining this threshold is to statistically analyze the image evolution feature dataset of historically diagnosed cases that were severely hemotoxic or cytotoxic and had already experienced irreversible necrosis of deep tissues, and take the lower quartile of its evolution score distribution as the warning boundary threshold parameter. This represents a Boolean logic OR operator, used to characterize non-conflicting mutually exclusive states where biochemical characteristics have not fallen below the negative lower limit or physical evolution has not exceeded the high-risk upper limit. Through this step, the system can utilize a cross-modal data cross-validation mechanism to keenly detect extreme contradictory cases where biochemical tests show no toxicity but the wound exhibits rapid swelling and expansion with decreasing thermal gradient. This triggers the corresponding inconsistency logic judgment flag, providing an indispensable pure mathematical logic decision-making pre-instruction for subsequent forced correction and weight redistribution mechanisms, completely avoiding the ambiguity brought about by natural language judgment.

[0042] Specifically, when the value of the immune response feature matrix is ​​lower than the negative determination threshold and the value of the physical evolution matching degree matrix exceeds the high-risk deterioration threshold, the weight redistribution logic is triggered, assigning the decision weight of the physical evolution matching degree matrix higher than that of the immune response feature matrix. The specific process of weighted fusion calculation based on the redistributed decision weights and outputting the snakebite venom type determination result is as follows: The inconsistency logic determination identifier is parsed. Under the condition that the value of the immune response feature matrix falls below the set negative determination threshold boundary and the value of the physical evolution matching degree matrix crosses the set high-risk deterioration threshold boundary, the evolution feature priority logic mechanism is activated. A pre-set penalty attenuation factor is introduced to perform exponential dimensionality reduction modulation on the original weight parameters of the immune response feature matrix, and a pre-set gain amplification factor is introduced to perform logarithmic dimensionality increase modulation on the original weight parameters of the physical evolution matching degree matrix. These are then merged to generate a redistribution decision weight vector. The redistribution decision weight vector is applied to perform inner product space tensor fusion calculation on the immune response feature matrix and the physical evolution matching degree matrix. The classification dimension mapping standard snakebite venom type library with the largest response value in the inner product space tensor fusion calculation result is extracted, and the snakebite venom type determination result is output.

[0043] In this implementation scheme, when the inconsistency logic judgment flag is detected as active, the system determines that the early immune response feature matrix has been severely affected by biochemical detection window interference or hook effect, resulting in data distortion. At this time, the system forcibly abandons the excessive reliance on biochemical indicators and activates the evolutionary feature priority logic mechanism to ensure judgment security. Introducing a pre-set penalty attenuation factor to perform exponential dimensionality reduction control on the original weight parameters of the immune response feature matrix is ​​to significantly weaken or even shield the decision-making proportion of ineffective biochemical indicators. Simultaneously, introducing a pre-set gain amplification factor to perform logarithmic dimensionality increase control on the original weight parameters of the physical evolution matching degree matrix is ​​to smoothly and significantly enhance the core decision-making status of objective medical pathological evolution facts. The specific calculation formula for merging and generating the redistributed decision weight vector is as follows: ; In the formula, : Indicates the redistribution of decision weights in the immune response feature matrix generated after exponential dimensionality reduction regulation; : Represents the native weight parameters of the immune response feature matrix set during the initial environment calibration phase of the system; : Represents an exponentially decaying arithmetic function with the natural logarithm as its base; : Indicates a preset penalty decay factor, used to control the steep slope of the weight drop of biochemical results when a cross-level collision occurs; : This represents the redistribution decision weights of the physical evolution matching degree matrix generated after logarithmic-level dimensionality increase control; : Represents the native weight parameters of the physical evolution matching degree matrix set during the initial environment calibration phase of the system; : Indicates the preset gain amplification factor, used to control the magnitude of weight increase of physical evolution factual evidence in conflict states; : Represents the natural logarithm operation function; : This represents a positive smoothing constant set to prevent singularities of zero values ​​from occurring within the logarithmic function. The inner product space tensor fusion calculation is performed on the immune response feature matrix and the physical evolution matching degree matrix using a redistribution decision weight vector. The classification dimension mapping standard snakebite venom type library with the largest response value is extracted, and the final safe snakebite venom type determination result is output. The specific tensor fusion decision process is as follows: In the formula, : Represents the classification index of the final snakebite venom type determination result extracted and mapped from the classification dimension with the largest response value; : Indicates the complete set of candidate venomous snake species covered in the pre-set standard snake bite venom type library; : Indicates the category dimension as an independent parameter for performing a multi-class traversal search on the entire set of candidate venomous snake species; : This represents the extreme value extraction function that searches for the maximum response value in the inner product space tensor fusion calculation result among all candidate category dimensions; : Represents the total number of tensor depth dimensions when the feature matrix is ​​unfolded in the multilayer perceptual space; : Represents the cumulative traversal index of the tensor's depth dimension; : Represents the immune response feature matrix at the th Category Dimension and the The values ​​of the elements in the tensor slice of the depth dimension; : indicates that the physical evolution matching degree matrix is ​​at the th Category Dimension and the The system utilizes deep-dimensional tensor slice element values. Through this dynamic correction and tensor fusion step, the invention completely solves the technical pain point of existing auxiliary judgment systems, which are prone to serious missed diagnoses or fatal misdiagnoses when faced with conflicts in multi-source heterogeneous data. Triggered by logical rules, the system automatically deprives false negative biochemical results of their decision-making power, achieving a weight transfer from static biochemical screening to dynamic, objective pathological evolution facts. This allows the system to output highly robust and error-tolerant toxicity auxiliary judgment conclusions under high-pressure environments such as emergency rooms and bedside rapid testing, effectively guiding medical personnel to administer antivenom serum as early and accurately as possible for rescue intervention.

[0044] Please see Figure 4 This snake bite venom type auxiliary determination system integrates snake venom immune rapid detection results, bite images, and data evolution characteristics. It includes the following modules: an immune response extraction module, used to obtain the patient's bite GPS location coordinates and retrieve the prior probability vector of regional candidate venom types from a pre-set geographical information database of snake spatial distribution; extracting the biochemical colorimetric optical characteristic values ​​and sampling interval of the snake venom immune chromatography rapid detection test strip based on wound secretion samples; using a preset biochemical concentration-time compensation function to perform nonlinear weighting on the biochemical colorimetric optical characteristic values ​​with the sampling interval time to generate an immune response feature matrix; a targeted image acquisition module, used to input the prior probability vector of regional candidate venom types and the immune response feature matrix into a maximum likelihood estimation model to extract preliminary venom type anchor points; configuring the collaborative acquisition parameters of visible light and infrared thermal imaging dual-modal devices; acquiring images of the bite area at continuous time nodes according to the collaborative acquisition parameters to obtain visible light texture sequences and infrared temperature field sequences; and spatiotemporal evolution matching. The module is used to calculate the expansion rate of wound swelling boundaries between adjacent time nodes in the visible light texture sequence using optical flow method, and to extract the infrared thermal gradient decay rate of the necrotic center in the infrared temperature field sequence based on isotherm tracking; the expansion rate and the infrared thermal gradient decay rate are concatenated to construct a multidimensional spatiotemporal evolution feature tensor, and the time axis elastic alignment distance between the multidimensional spatiotemporal evolution feature tensor and the pre-set standard venom type pathological evolution curve is calculated to generate a physical evolution matching degree matrix; the conflict correction judgment module is used to extract the time-compensated biochemical feature values ​​in the immune response feature matrix and compare them with the evolution matching scores in the physical evolution matching degree matrix using consistency logic; when the value of the immune response feature matrix is ​​lower than the negative judgment threshold and the value of the physical evolution matching degree matrix exceeds the high-risk deterioration threshold, the weight redistribution logic is triggered, and the physical evolution matching degree matrix is ​​given a higher decision weight than the immune response feature matrix; the weighted fusion calculation is performed based on the redistributed decision weights to output the snake bite venom type judgment result.

[0045] In this implementation scheme, the immune response extraction module is mainly used to obtain the prior probability of geographical space and to perform time decay calibration on the biochemical rapid test results. This module first analyzes the GPS coordinates of the patient's bite to extract the prior probability vector of regional candidate venom types. This vector refers to the initial probability set of various snake venom types appearing in a specific area, quantified based on historical geographical distribution data and biostatistical laws. Simultaneously, it extracts the biochemical colorimetric optical characteristic values ​​of the rapid test strip based on the secretion sample from the wound, and substitutes them into the biochemical concentration and time compensation function to perform nonlinear weighted calculations. The biochemical concentration and time compensation function is a mathematical algorithm specifically designed to compensate for the natural decay error of toxin concentration in the wound surface secretion caused by the diffusion of venom into the deeper blood and lymphatic networks over time. This module effectively overcomes the false negative defect of rapid test strips caused by sampling time lag or local venom diffusion in field emergency rescue, providing the system with a data foundation of immune response feature matrix that has undergone precise spatiotemporal calibration. The targeted image acquisition module is mainly used to drive cross-modal medical imaging hardware to perform targeted imaging according to preliminary biochemical diagnostic instructions. This module inputs the immune response matrix and regional prior probabilities into the maximum likelihood estimation model. The maximum likelihood estimation model is a statistical algorithm that, based on a statistical probability framework, reverse-engineers the parameters of the most likely pathogenic fluid type from existing observational data, thereby extracting the initial propensity virulence anchor point with the highest probability confidence. Subsequently, the system retrieves the pathological characterization mapping knowledge base and, based on the typical local lesion characteristics corresponding to different virulence types, differentiates the collaborative acquisition parameters of the visible light and infrared thermal imaging dual-modal devices, performing synchronous imaging at continuous time points. This module completely changes the traditional static and blind routine wound imaging mode in emergency departments, realizing cross-modal linkage control where biochemical test results and geographical prior data intelligently guide the underlying hardware sensors to perform specific spatial and frequency acquisition, laying an unbiased high-definition dynamic imaging foundation for subsequent capture of objective pathological evolution characteristics. The spatiotemporal evolution matching module is mainly used to extract multidimensional pathological deterioration features from the continuous image stream and perform temporal alignment matching with standard pathological evolution models. This module employs optical flow to process visible light texture sequences. Optical flow is a fundamental technique in computer vision used to estimate the velocity vector field of pixel motion between adjacent image frames. Here, it is used to accurately quantify the physical centripetal expansion rate of wound swelling boundaries. Simultaneously, it extracts the infrared thermal gradient decay rate of the necrotic center from the infrared temperature field sequence based on isotherm tracking. Isotherm tracking refers to continuously tracking the contraction spatial displacement of the temperature layer contour of a specific cold region in an infrared thermal image to characterize the deterioration rate of deep tissue ischemia and necrosis. This module then concatenates the above rate features and calculates their elastic alignment distance with the time axis of a preset standard curve. This is a statistical algorithm that can dynamically stretch or compress the time dimension to calculate the morphological similarity between two non-co-frequency time series.This module extracts static visual representations of wounds into dynamic pathological evolution characteristic parameters, eliminating feature offset matching errors caused by differences in individual patient metabolic rates, and providing a solid purely physical verification benchmark for judgment. The conflict correction judgment module is mainly used to arbitrate contradictions in global multi-source heterogeneous data and make comprehensive decisions on the final toxicity type. This module extracts time-compensated biochemical feature values ​​and performs consistency logic comparison with evolutionary matching scores. Consistency logic comparison refers to the verification mechanism in the medical auxiliary judgment system to verify whether there is a contradiction between the colorimetric results of biochemical reagents and the objective fact of local pathological deterioration. When the immune response value is found to be below the negative judgment threshold while the physical evolution matching value exceeds the high-risk deterioration threshold, the module automatically triggers the weight redistribution logic. This is a conflict arbitration mathematical mechanism that dynamically adjusts the proportion of different objective evidence sources in the final decision-making process. At this time, the module determines that the rapid biochemical test is limited by the hook effect of the sample test strip or cross-reaction, resulting in serious false negatives and missed diagnoses. Therefore, it forcibly reduces or even blocks the decision weight of the biochemical results and assigns a very high decision weight to the objective physical pathological evolution. This module completely solves the industry pain point that existing single-modality biochemical detection reagents are prone to misdiagnosis. When the detection indicators deviate significantly, it relies on real dynamic pathological evolution facts to perform forced intelligent correction, thereby outputting toxicity type auxiliary judgment results with extremely high robustness and high fault tolerance.

[0046] In summary, this application has at least the following effects:

[0047] This paper presents a method and system for auxiliary determination of snakebite venom type by integrating rapid snake venom immunoassay results, bite imagery, and data evolution characteristics. By introducing geospatial prior probability and a biochemical concentration time compensation function, it effectively overcomes the initial false negative error caused by sampling time lag, local venom diffusion, and hook effect interference in traditional rapid wound secretion test reagents. Furthermore, it utilizes a preliminary biochemical tendency anchor point to intelligently drive a dual-modal device of visible light and infrared thermal imaging for targeted acquisition, and constructs a multi-dimensional spatiotemporal evolution model based on wound swelling expansion and deep ischemia necrosis rates obtained through optical flow and isotherm tracking. This technology enables a cross-modal leap from static visual representation to dynamic pathological evolution speed. Ultimately, it innovatively constructs a consistency logic comparison and weight redistribution conflict correction mechanism. When biochemical test indicators show safety but objective local pathology shows a high-risk deterioration, which is a serious deviation, it automatically filters out invalid biochemical test results and assigns the weight of real physical evolution data to dominate the decision-making process. This completely solves the pain points of fatal missed diagnoses and misdiagnoses that are prone to occur in complex emergency and bedside rapid testing scenarios, significantly improving the error tolerance, robustness, and timeliness of comprehensive determination of snakebite venom type.

[0048] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0049] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0050] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0051] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0052] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0053] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for assisting in the determination of the type of snakebite by fusing the results of rapid immunological detection of snake venom, images of the bite, and data evolution characteristics, characterized in that, Includes the following steps: S1. Obtain the GPS location coordinates of the patient's bite wound, and retrieve and extract the prior probability vector of the regional candidate venom type from the pre-set geographical information database of venom spatial distribution; extract the biochemical colorimetric optical feature value and sampling interval time of the snake venom immunochromatographic rapid test strip based on the secretion fluid sample at the wound site, and perform nonlinear weighting on the biochemical colorimetric optical feature value using the sampling interval time through the preset biochemical concentration-time compensation function to generate an immune response feature matrix; S2. Input the prior probability vector of regional candidate venom types and the immune response feature matrix into the maximum likelihood estimation model to extract preliminary venom type anchor points, and configure the collaborative acquisition parameters of visible light and infrared thermal imaging dual-modal devices; at continuous time nodes, perform image acquisition on the bite site according to the collaborative acquisition parameters to obtain visible light texture sequence and infrared temperature field sequence. S3. The expansion rate of wound swelling boundary between adjacent time nodes in the visible light texture sequence is calculated by optical flow method, and the infrared thermal gradient decay rate of the necrotic center in the infrared temperature field sequence is extracted based on isotherm tracking; the expansion rate and the infrared thermal gradient decay rate are concatenated to construct a multidimensional spatiotemporal evolution feature tensor, and the time axis elastic alignment distance between the multidimensional spatiotemporal evolution feature tensor and the pre-set standard toxicity pathological evolution curve is calculated to generate a physical evolution matching degree matrix; S4. Extract the time-compensated biochemical feature values ​​from the immune response feature matrix and perform a consistency logic comparison with the evolutionary matching score from the physical evolution matching degree matrix; When the value of the immune response feature matrix is ​​lower than the negative determination threshold and the value of the physical evolution matching degree matrix exceeds the high-risk deterioration threshold, the weight redistribution logic is triggered, and the decision weight of the physical evolution matching degree matrix is ​​given higher than that of the immune response feature matrix. The weighted fusion calculation is performed based on the redistributed decision weights to output the snake bite venom type determination result.

2. The fusion snake venom immunological rapid detection result, bite image and data evolution characteristic snakebite injury toxin type auxiliary judgment method according to claim 1, characterized in that: The specific process of obtaining the GPS location coordinates of the patient's bite and retrieving and extracting the prior probability vector of regional candidate venom types from a pre-established geographic information database of venomous snake spatial distribution is as follows: Analyze GPS positioning coordinates to extract environmental constraint features such as altitude, latitude and longitude, and surface vegetation cover type; Input environmental constraint features into a pre-set geographic information database of venomous snake spatial distribution, perform spatial topological intersection query, extract the list of historical active snake species matching the target geographic grid, and construct a multidimensional probability distribution set by associating the venom type distribution frequency built into the historical active snake species list. Normalization is performed on the set of multidimensional probability distributions to generate a prior probability vector of regional candidate venom types composed of prior confidence of different venom types.

3. The fusion snake venom immunological rapid detection result, bite image and data evolution characteristic snakebite injury toxin type auxiliary judgment method according to claim 1, characterized in that: The specific process of extracting biochemical colorimetric optical characteristic values ​​and sampling intervals from snake venom immunochromatographic rapid test strips based on wound secretion samples, and then using a preset biochemical concentration-time compensation function to perform nonlinear weighting on the biochemical colorimetric optical characteristic values ​​with the sampling interval time to generate an immune response feature matrix is ​​as follows: Images of the color development area of ​​snake venom immunochromatographic rapid test strips were acquired and converted to the HSV color space. Saturation and lightness components were extracted and fused to generate biochemical colorimetric optical characteristic values. The time span from the moment the patient was bitten to the moment the secretion sample was dropped onto the rapid test strip was recorded as the sampling interval. The biochemical colorimetric optical characteristic value and the sampling interval were substituted into the pre-set snake venom local tissue absorption and diffusion dynamics time-series distribution model. The corresponding time node compensation coefficient was extracted based on the local concentration decay curve built into the model. The corresponding time node compensation coefficient is applied to perform a nonlinear exponential scaling transformation on the biochemical colorimetric optical characteristic values, and the transformed values ​​are arrayed according to the preset standard toxicity classification spatial dimension to generate an immune response feature matrix.

4. The fusion snake venom immunological rapid detection result, bite image and data evolution characteristic snakebite injury toxin type auxiliary judgment method according to claim 1, characterized in that: The specific process of inputting the prior probability vector of regional candidate virulence types and the immune response feature matrix into the maximum likelihood estimation model to extract preliminary virulence type anchor points, and configuring dual-modal equipment of visible light and infrared thermal imaging to collaboratively acquire parameters is as follows: The immune response feature matrix is ​​used as the observation parameter of the likelihood function. Combined with the prior probability vector of regional candidate toxic types, the expectation maximization iteration is performed. The toxic type with the global maximum posterior probability value in the convergent state is extracted as the initial toxic type anchor point. The system retrieves a pre-set pathological characterization mapping knowledge base, extracts preliminary toxicity-related anchor points associated with typical local lesion area category characteristics, and configures the spatial alignment parameters of the visible light and infrared thermal imaging dual-modal devices, sampling time frequency, and infrared detector temperature resolution threshold based on the typical local lesion area category characteristics, and combines them to generate collaborative acquisition parameters.

5. The fusion snake venom immunological rapid detection result, bite image and data evolution characteristic snakebite injury toxin type auxiliary judgment method according to claim 4, characterized in that: The specific process of acquiring visible light texture sequences and infrared temperature field sequences by performing image acquisition on the bite site at continuous time nodes according to collaborative acquisition parameters is as follows: Based on the spatial alignment parameters in the collaborative acquisition parameters, the visible light sensor and the infrared detector are calibrated to coincide in the field of view center. Based on the sampling time frequency in the collaborative acquisition parameters, the shutter synchronization action of the dual-mode device is triggered, and multiple frames of original dual-spectral images are captured within the preset monitoring period. Spatial registration and region of interest boundary cropping are performed on multiple frames of original dual-spectral images to separate and output visible light texture sequences and infrared temperature field sequences with the same timestamp and spatial resolution.

6. The fusion snake venom immunological rapid detection result, bite image and data evolution characteristic snakebite injury toxin type auxiliary judgment method according to claim 1, characterized in that: The specific process of calculating the wound swelling boundary expansion rate between adjacent time nodes in the visible light texture sequence using optical flow method, and extracting the infrared thermal gradient attenuation rate of the necrosis center in the infrared temperature field sequence based on isotherm tracking is as follows: Construct a grayscale conservation equation for adjacent time nodes of the visible light texture sequence, apply a dense optical flow field algorithm to extract the local deformation velocity vector field, perform surface integral operation on the local deformation velocity vector field along the normal direction of the initial boundary contour of the wound, and output the swelling boundary expansion rate of the wound. The lowest temperature extreme point representing the abnormal cold zone of local ischemia in the infrared temperature field sequence is extracted to establish a reference center. A morphological threshold segmentation algorithm is applied to generate multi-level isotherm topological profiles. The spatial displacement of the topological profiles of the same level of isotherms at adjacent time nodes towards the reference center is measured. The sampling time interval parameters of adjacent time nodes are fused to perform differential derivation and output the infrared thermal gradient attenuation rate of the necrotic center.

7. The fusion snake venom immunological rapid detection result, bite image and data evolution characteristic snakebite injury toxin type auxiliary judgment method according to claim 6, characterized in that: The specific process of constructing a multidimensional spatiotemporal evolution feature tensor by concatenating the expansion rate and the infrared thermal gradient attenuation rate, and calculating the time axis elastic alignment distance between the multidimensional spatiotemporal evolution feature tensor and the pre-set standard toxicity pathological evolution curve to generate the physical evolution matching degree matrix is ​​as follows: Extract the expansion rate of wound swelling boundary and the attenuation rate of infrared thermal gradient at the necrosis center, input them into the maximum and minimum normalization functions to eliminate dimensional differences, and perform positional cascading according to the time series dimension to construct a multidimensional spatiotemporal evolution feature tensor; A time series reference sample is constructed by retrieving the pathological evolution curve of a pre-set standard toxic type. A multidimensional spatiotemporal evolution feature tensor and a local cost space matrix of the time series reference sample are established. The minimum cumulative deformation cost connected path is solved in the local cost space matrix by a dynamic programming path search algorithm. The terminal value of the connected path is extracted and mapped to the time axis elastic alignment distance. An exponential decay kernel function is introduced to perform confidence transformation on the time axis elastic alignment distance. The confidence transformation value is then combined with the initial tendency toxic anchor point execution vector to generate a physical evolution matching degree matrix.

8. The fusion snake venom immunological rapid detection result, bite image and data evolution characteristic snakebite injury toxin type auxiliary judgment method according to claim 1, characterized in that: The logical process of performing a consistency comparison between the time-compensated biochemical feature values ​​in the immune response feature matrix and the evolutionary matching score in the physical evolution matching degree matrix is ​​as follows: Separate time-compensated biochemical feature values ​​within the immune response feature matrix, simultaneously extract evolution matching scores within the physical evolution matching degree matrix, retrieve the mapping table of preliminary tendency toxicity anchor points associated with toxicological evolution stages, retrieve time-compensated biochemical feature values ​​to map early mild toxin accumulation labels, and retrieve evolution matching scores to map mid-to-late stage severe tissue necrosis labels. Establish a Boolean logic conflict discriminant between early mild toxin accumulation labels and mid-to-late stage severe tissue necrosis labels. When the Boolean logic conflict discriminant satisfies the condition of a cross-level collision in pathological deterioration, trigger an inconsistency logic judgment flag.

9. The fusion snake venom immunological rapid detection result, bite image and data evolution characteristic snakebite injury toxin type auxiliary judgment method according to claim 8, characterized in that: When the value of the immune response feature matrix is ​​lower than the negative determination threshold and the value of the physical evolution matching degree matrix exceeds the high-risk deterioration threshold, a weight reallocation logic is triggered, assigning a decision weight to the physical evolution matching degree matrix that is higher than that to the immune response feature matrix. The specific process of performing weighted fusion calculation based on the reallocated decision weights and outputting the snakebite venom type determination result is as follows: The inconsistency logic judgment flag is analyzed, and the evolutionary feature priority logic mechanism is activated when the value of the immune response feature matrix falls below the set negative judgment threshold boundary and the value of the physical evolution matching degree matrix crosses the set high-risk deterioration threshold boundary. A pre-set penalty attenuation factor is introduced to perform exponential dimensionality reduction modulation on the original weight parameters of the immune response feature matrix, and a pre-set gain amplification factor is introduced to perform logarithmic dimensionality increase modulation on the original weight parameters of the physical evolution matching degree matrix. The two are then combined to generate a redistribution decision weight vector. The inner product space tensor fusion calculation is performed on the immune response feature matrix and the physical evolution matching degree matrix by applying the redistribution decision weight vector. The snake bite venom type library with the largest response value classification dimension mapping is extracted from the inner product space tensor fusion calculation result, and the snake bite venom type determination result is output.

10. The venom type auxiliary judgment system for fusion of snakebite immunological rapid detection results, bite image and data evolution characteristics, applied to the venom type auxiliary judgment method for fusion of snakebite immunological rapid detection results, bite image and data evolution characteristics according to any one of claims 1-9, characterized in that, Includes the following modules: The immune response extraction module is used to obtain the GPS location coordinates of the patient's bite, retrieve and extract the prior probability vector of regional candidate venom types from the pre-set geographical information database of venom spatial distribution, and extract the biochemical colorimetric optical feature values ​​and sampling interval time of the snake venom immunochromatographic rapid test strip based on the secretion fluid sample at the wound site. The biochemical colorimetric optical feature values ​​are nonlinearly weighted by the sampling interval time using the preset biochemical concentration-time compensation function to generate an immune response feature matrix. The targeted image acquisition module is used to input the prior probability vector of regional candidate virulence types and the immune response feature matrix into the maximum likelihood estimation model to extract preliminary virulence type anchor points, and configure the collaborative acquisition parameters of the visible light and infrared thermal imaging dual-modal devices; at continuous time nodes, images of the bite site are acquired according to the collaborative acquisition parameters to obtain visible light texture sequences and infrared temperature field sequences. The spatiotemporal evolution matching module is used to calculate the expansion rate of wound swelling boundaries between adjacent time nodes in the visible light texture sequence using optical flow method, and to extract the infrared thermal gradient decay rate of the necrotic center in the infrared temperature field sequence based on isotherm tracking; the expansion rate and the infrared thermal gradient decay rate are concatenated to construct a multidimensional spatiotemporal evolution feature tensor, and the time axis elastic alignment distance between the multidimensional spatiotemporal evolution feature tensor and the pre-set standard toxicity pathological evolution curve is calculated to generate a physical evolution matching degree matrix; The conflict correction and determination module is used to extract the time-compensated biochemical feature values ​​in the immune response feature matrix and compare them with the evolutionary matching scores in the physical evolution matching degree matrix for consistency logic comparison. When the value of the immune response feature matrix is ​​lower than the negative determination threshold and the value of the physical evolution matching degree matrix exceeds the high-risk deterioration threshold, the weight redistribution logic is triggered, and the decision weight of the physical evolution matching degree matrix is ​​given higher than that of the immune response feature matrix. The weighted fusion calculation is performed based on the redistributed decision weights to output the snake bite venom type determination result.

Images