Intelligent evaluation system for skin graft color and temperature

Abstract

This invention relates to the field of medical monitoring technology and discloses a joint intelligent assessment system for skin flap color and temperature. The system includes a multimodal data acquisition module for the flap, a surface attribute analysis module for the flap, a deep pathological inference module, and a dynamic survival status prediction module. It generates a data package with a unified spatiotemporal reference by simultaneously acquiring and fusing color texture images and infrared thermal imaging data. The system quantifies and analyzes the boundaries and parameters of abnormal color and temperature regions from the fused data, then calculates the spatial overlap and correlation between abnormal color and temperature regions, and generates inference vectors representing deep blood supply and metabolic status based on rules. A state evolution model is constructed by combining historical vector data to achieve dynamic prediction of the flap's survival status. This system realizes the objective and synchronized analysis of multimodal information on flap appearance and inference of deep pathological correlations, improving the accuracy and early warning capabilities of postoperative monitoring.

Description

Technical Field

[0001] This invention relates to the field of medical monitoring technology, specifically to an intelligent assessment system for the combined color and temperature of skin grafts. Background Technology

[0002] In clinical flap transplantation postoperative monitoring, traditional methods primarily rely on medical staff to assess blood supply by visually observing flap color changes and manually sensing flap temperature. Existing technologies also include methods using digital cameras to photograph flap color or using infrared thermal imagers to separately acquire flap temperature distribution maps. However, these techniques are usually performed separately. Differences in acquisition time, shooting angle, and spatial resolution between color images and thermal imaging data make direct comparison at the precise same location and time difficult.

[0003] This isolated assessment method has flaws. Due to the lack of strict synchronization and spatial alignment, nurses cannot accurately determine whether a whitish area corresponds precisely to a temperature-lowered area, leading to superficial and speculative assessments of ischemia or congestion, which are prone to misjudgment. The assessment process heavily relies on nurses' personal experience, is highly subjective, and lacks objective, quantifiable, and unified standards. Current technology can only provide isolated surface information and cannot reveal the intrinsic spatial relationship between abnormal skin flap color and temperature. Therefore, it cannot effectively infer key pathological states such as deep blood perfusion and tissue metabolism within the skin flap, making early and accurate complication warnings difficult. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent evaluation system for skin graft color and temperature in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a combined intelligent evaluation system for skin graft color and temperature, the system comprising: The flap multimodal data acquisition module is used to simultaneously acquire color texture image data and spatial temperature field data of the transplanted flap area using high-resolution imaging equipment and infrared thermal imaging equipment, and to superimpose and fuse the two types of data according to a common spatial coordinate system and time reference to generate a flap status data package with fused spatiotemporal reference. The flap surface attribute analysis module receives the flap status data packet, identifies and quantifies the boundary and intensity parameters of ischemic pale areas, congested cyanotic areas and normal skin color areas from color texture image data, divides the boundary and temperature gradient parameters of abnormal high temperature areas, abnormal low temperature areas and transition temperature zones from spatial temperature field data, and summarizes the identified area boundaries, intensity parameters and temperature gradient parameters into a surface attribute parameter set. The deep-coupled pathological inference module retrieves the set of surface attribute parameters, calculates the spatial overlap between the ischemic pale area and the abnormal low temperature area, performs spatial correlation analysis between the congested cyanotic area and the abnormal high temperature area, and generates a coupled pathological inference vector characterizing the deep blood flow and metabolic state of the flap based on preset overlap correlation rules. The survival status dynamic simulation module receives the coupled pathological inference vector, combines it with similar vectors in the historical time series, constructs an evolution trajectory model of the flap survival status through time series analysis, and predicts the state vector at the next assessment time point based on the evolution trajectory model, generating a dynamic simulation report containing the current state and the predicted state.

[0006] Preferably, the process of generating the flap state data packet based on the fused spatiotemporal reference includes: The image pixel coordinate system of the high-resolution imaging device and the temperature data coordinate system of the infrared thermal imaging device are calibrated respectively, and an affine transformation matrix between the two coordinate systems is established. Using the affine transformation matrix, each temperature measurement point in the temperature data coordinate system is precisely mapped to the corresponding pixel position in the image pixel coordinate system; Each synchronously acquired image frame and temperature field data frame is assigned the same and unique timestamp code; Image data with completed coordinate mapping and timestamp encoding is packaged with temperature data to form the flap state data packet of the fused spatiotemporal reference; After generating the flap status data packet, the flap multimodal data capture module transmits it to the flap surface attribute parsing module and simultaneously records the data packet generation time and data size to the system log.

[0007] Preferably, the method for constructing the surface attribute parameter set includes: Define the reference value ranges for ischemic pallor and ecchymosis in the standard color space, as well as the color distribution model for normal skin color; The collected color texture image data is converted to the standard color space, and the matching degree of its value with the reference value range and the color distribution model is compared pixel by pixel to mark the pixel areas belonging to ischemic pallor, ecchymosis and normal skin color. The region growing algorithm merges adjacent pixel regions with the same attributes, and calculates the area, shape complexity, and average color intensity of each merged region as the intensity parameter of the color texture. In the spatial temperature field data collected by infrared thermal imaging equipment, abnormal high temperature thresholds and abnormal low temperature thresholds are set based on clinical experience. The value of each temperature measurement point is compared with the abnormal high temperature threshold and the abnormal low temperature threshold to divide the abnormal high temperature zone, the abnormal low temperature zone and the transition temperature zone between the two. Calculate the rate of temperature change within each region and at the region boundaries, and use it as the temperature gradient parameter; Finally, the intensity parameters of the color texture and the temperature gradient parameters are bound according to their respective flap space sub-regions to form the set of surface attribute parameters.

[0008] Preferably, the specific execution steps of the spatial overlap calculation and the spatial correlation analysis include: Extract the boundary coordinate set of the ischemic pale region and the boundary coordinate set of the abnormally low temperature region from the surface attribute parameter set; Calculate the area of ​​the intersection of the boundary coordinate set of the ischemic pale region and the boundary coordinate set of the abnormally low temperature region, and calculate the independent areas of the ischemic pale region and the abnormally low temperature region respectively; Divide the intersection area by the independent area of ​​the ischemic pale region to obtain the first overlap ratio; divide the intersection area by the independent area of ​​the abnormally low temperature region to obtain the second overlap ratio; take the weighted average of the first overlap ratio and the second overlap ratio as the final spatial overlap value. Extract the boundary coordinate set of the ecchymotic region and the boundary coordinate set of the abnormally high temperature region from the surface attribute parameter set; Calculate the shortest average distance between the boundary coordinate set of the congested cyanotic region and the boundary coordinate set of the abnormally high temperature region; According to the preset distance-association mapping table, the shortest average distance is converted into a spatial correlation strength value; The spatial overlap value and the spatial correlation strength value are combined into a multi-dimensional vector, and then weighted and fused according to a preset weight coefficient to generate the coupled pathological inference vector.

[0009] Preferably, the process of constructing and predicting the evolution trajectory model includes: From the system's historical database, retrieve the coupled pathological inference vectors generated for the same flap object at multiple consecutive historical assessment points, and arrange them in chronological order to form a time-series vector set; The time series vector set is smoothed and filtered to eliminate random fluctuations caused by acquisition noise; Analyze the changing trend of each vector dimension in the smoothed time series vector set, and use an autoregressive model to establish an independent trend prediction sub-model for each vector dimension; By integrating all the trend prediction sub-models of all vector dimensions, a complete evolution trajectory model is formed. The coupled pathological inference vector generated at the current assessment time point is input into the evolution trajectory model, and the model calculates the predicted value of the state vector at the next preset assessment time point. The coupled pathological inference vector at the current assessment time point is encapsulated together with the predicted value of the state vector to generate the dynamic inference report.

[0010] Preferably, the system further includes a confidence level calibration module, comprising: Receive the dynamic simulation report output by the survival status dynamic simulation module, and extract the state vector prediction value from the report; Waiting to reach the next actual assessment time point, obtain the actual coupled pathological inference vector obtained by the actual assessment time point and processed by the deep coupled pathological inference module; Calculate the Euclidean distance between the predicted state vector value and the actual coupled pathological inference vector, and use it as the deviation value for this prediction; The deviation value of this prediction is added to the historical prediction deviation queue of the flap object, and the moving average of all deviation values ​​in the historical prediction deviation queue is calculated. Based on the magnitude of the moving average, adjust the order of the autoregressive model or the strength parameter of the smoothing filter in the evolution trajectory model to complete the calibration of the model prediction confidence. The confidence assessment calibration module feeds back the calibrated model parameters to the survival status dynamic inference module to update the evolution trajectory model.

[0011] Preferably, the system further includes a regional microcirculation assessment submodule, the workflow of which includes: From the set of surface attribute parameters output by the flap surface attribute parsing module, the microvascular network pattern identified in the color texture image data is separated. From the spatial temperature field data, extract the subtle temperature change curves corresponding to the orientation of the microvascular network pattern; The connectivity and density parameters of the microvascular network pattern are analyzed in correspondence with the fluctuation frequency and amplitude parameters of the subtle temperature change curve. Establish matching pairs between microvascular pattern features and temperature fluctuation features, and generate regional microcirculation status scores based on the number and quality of matching pairs; The regional microcirculation assessment submodule appends the regional microcirculation status score to the coupled pathological inference vector generated by the deep coupled pathological inference module, forming an enhanced pathological inference vector.

[0012] Preferably, the generation of the regional microcirculation status score follows the following rules: Set minimum length thresholds and minimum density thresholds for the connectivity of microvascular patterns; Set the normal range threshold for the fluctuation frequency and the normal range threshold for the amplitude of the temperature fluctuation curve; For each identified microvascular network pattern segment, calculate its connectivity length and density per unit area; For each extracted segment of the minute temperature change curve, calculate its principal fluctuation frequency and average amplitude; When the connectivity length and density of a pattern segment are both greater than or equal to their corresponding thresholds, and the main fluctuation frequency and average amplitude of the corresponding temperature curve segment are both within their normal range thresholds, the matching pair is determined to be a high-quality match. Count the number of high-quality matching pairs among all matching pairs, and calculate the proportion of high-quality matching pairs to the total number of matching pairs; According to the preset ratio-score conversion table, the ratio is converted into the regional microcirculation status score.

[0013] Preferably, the system further integrates an intervention strategy mapping library, including: Establish a pathological state template library containing various typical postoperative complications of flap surgery. Each pathological state template is defined by a set of standardized coupled pathological inference vector ranges. The coupling pathological inference vector generated in real time by the deep coupling pathological inference module is compared with the similarity of all templates in the pathological state template library. Pathological state templates with similarity exceeding a preset matching threshold are selected as candidate matching templates; The intervention strategy mapping library pre-stores clinical intervention recommendations corresponding to each pathological state template; When a candidate matching template exists, the system automatically outputs clinical intervention recommendations associated with the candidate matching template and presents them in conjunction with the dynamic simulation report generated by the survival status dynamic simulation module.

[0014] Preferably, the intervention strategy mapping library has a self-updating mechanism, and its update triggering conditions and process are as follows: Continuously record the clinical intervention recommendations output by the system and the actual evolution of the flap status over a subsequent period of time; When the number of times the actual evolution direction deviates significantly from the direction expected based on the original template and dynamic inference report reaches a preset abnormal number threshold, the template update process is triggered. In the template update process, the coupled pathological inference vectors that cause deviations from the expected results and their subsequent actual evolution data are clustered to form new pathological state data clusters. The newly formed pathological state data clusters are reviewed and defined by clinical experts. After confirmation, they are added as new pathological state templates or used to correct the standardized vector range of the original pathological state templates. The updated pathological state template and associated intervention suggestions will be synchronized to the intervention strategy mapping library for subsequent matching and output.

[0015] Compared with the prior art, the beneficial effects of the present invention are: By simultaneously acquiring data using high-resolution imaging and infrared thermal imaging equipment, and mapping both types of data to a common spatial coordinate system and time reference for superposition and fusion, a data package with strictly spatially aligned flap color and temperature information at each time point was generated. This solved the problems of isolated multi-source information and spatiotemporal misalignment, providing a pixel-level precise data foundation for subsequent analysis. It enabled direct correlation analysis of color features and temperature fields at the same anatomical location, eliminating errors and subjectivity from human comparison.

[0016] This approach identifies and quantifies ischemic and congestive areas from color and texture images, delineates abnormally high and low temperature zones from temperature field data, calculates the spatial overlap between ischemic pale areas and abnormally low temperature areas, and analyzes the spatial correlation between congestive cyanotic areas and abnormally high temperature areas. Based on predefined rules, this spatial correlation is quantified into a coupled pathological inference vector. This scheme establishes a quantitative bridge between visible superficial signs and deep, invisible pathological states, transforming previously experience-based and vague correlation judgments into objective diagnostic indicators calculated from spatial data. It achieves a leap from phenomenological description to pathological inference, enabling earlier and more accurate revelation of deep blood flow obstructions or metabolic abnormalities in skin flaps.

[0017] Based on the generated coupled pathological inference vector and its historical time series data, a trajectory model of flap survival status is constructed through time series analysis. This approach utilizes the continuous changing trend of the state vector for modeling and prediction, enabling dynamic deduction of the development direction of flap blood supply status. It upgrades single, static assessment to continuous, dynamic prognostic prediction, providing a prospective trend judgment basis for the selection of clinical intervention timing. Attached Figure Description

[0018] Figure 1 This is a timing diagram of the intelligent evaluation system for skin graft color and temperature described in this invention. Figure 2 A flowchart for generating flap state data packets that integrate spatiotemporal references; Figure 3 A flowchart for constructing a method for a set of surface attribute parameters; Figure 4 A trend graph showing the prediction bias and moving average bias of flap status; Figure 5The image shows a thermal analysis of the similarity of pathological conditions after skin flap transplantation. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 This invention provides a joint intelligent assessment system for skin flap color and temperature. The system includes: a multimodal data acquisition module for the flap, a surface attribute analysis module for the flap, a deep-coupled pathological inference module, and a dynamic survival status prediction module. The multimodal data acquisition module simultaneously acquires color texture image data and spatial temperature field data of the transplanted flap area using high-resolution imaging equipment and infrared thermal imaging equipment. By calibrating the coordinate systems of the two types of data and assigning a unified timestamp, a flap status data package with a fused spatiotemporal reference is generated. The surface attribute analysis module quantifies the boundary and intensity parameters of ischemic pale areas, congested cyanotic areas, and normal skin color areas from the color texture image data. Simultaneously, it delineates the boundary and temperature gradient parameters of abnormally high temperature areas, abnormally low temperature areas, and transitional temperature zones from the spatial temperature field data, forming a set of surface attribute parameters. The deep-coupled pathological inference module calculates the spatial overlap between the ischemic pale areas and the abnormally low temperature areas, and performs spatial correlation analysis between the congested cyanotic areas and the abnormally high temperature areas, generating a coupled pathological inference vector. The survival status dynamic simulation module constructs an evolution trajectory model based on the coupled pathological inference vectors of historical time series, predicts the state vector at the next assessment point, and outputs a dynamic simulation report.

[0021] In one embodiment of the present invention, see [reference] Figure 2 The image pixel coordinate system of the high-resolution imaging device and the temperature data coordinate system of the infrared thermal imaging device are calibrated separately, and the affine transformation matrix between the two coordinate systems is calculated through feature point registration. This affine transformation matrix is ​​used to map each temperature measurement point in the temperature data coordinate system to the corresponding pixel position in the image pixel coordinate system, achieving spatial alignment. Each set of synchronously acquired image frames and temperature field data frames is assigned the same and unique timestamp code to ensure time synchronization. The image data and temperature data after coordinate mapping and timestamp encoding are packaged into a flap status data packet, which simultaneously records the acquisition timestamp, data size, and coordinate system parameters. After generating the data packet, the flap multimodal data acquisition module transmits it to the flap surface attribute parsing module and records the generation time and size of the data packet in the system log.

[0022] In practice, generating a fused spatiotemporal reference flap state data package is a direct result of the collaborative work between high-resolution imaging equipment and infrared thermal imaging equipment. The high-resolution imaging equipment acquires color and texture image data of the transplanted flap region at a specific resolution, while the infrared thermal imaging equipment acquires temperature field data of the transplanted flap region at the corresponding spatial resolution. A single synchronous acquisition operation yields one frame of color and texture image data and one frame of spatial temperature field data. To establish the spatial correspondence between the two types of data, the image pixel coordinate system of the high-resolution imaging equipment and the temperature data coordinate system of the infrared thermal imaging equipment need to be calibrated. This process is accomplished by placing a set of calibration boards with a known spatial distribution within the common field of view of both devices. An affine transformation matrix is ​​calculated using the positions of feature points on the calibration boards in the two coordinate systems. This affine transformation matrix expresses the rotation, translation, and scaling relationships between the temperature data coordinate system and the image pixel coordinate system.

[0023] In some embodiments, an affine transformation matrix is ​​used to spatially map the spatial temperature field data. Each temperature measurement point in the temperature data coordinate system has coordinates. The affine transformation matrix is ​​applied to convert the coordinates of the temperature measurement points into coordinates in the image pixel coordinate system. The conversion formula is as follows: in: These are the coordinates of the temperature measurement point in the temperature data coordinate system. These are the coordinates in the image pixel coordinate system after mapping. It is the calculated affine transformation matrix. After mapping, the value of each temperature measurement point is associated with a specific pixel location in the color texture image, realizing the spatial superposition of temperature data and image data.

[0024] It is understandable that the fusion of time references is accomplished by assigning the same and unique timestamp code to each synchronously acquired data pair. The system's internal clock generates a timestamp code when triggering the acquisition command. This timestamp code is simultaneously appended to the metadata header of both the color texture image data and the spatial temperature field data frames. The color texture image data, spatial temperature field data, timestamp code, and affine transformation matrix parameters used for coordinate mapping are collectively packaged and encapsulated to form a flap state data packet that fuses the spatiotemporal reference. After successfully generating the flap state data packet, the flap multimodal data acquisition module immediately transmits the flap state data packet to the input buffer of the flap surface attribute parsing module via the system's internal data bus. In some embodiments, the tight coupling of spatial alignment and timestamp encoding ensures the consistency of multimodal data in subsequent processing. The flap state data packet, as a complete data unit, contains all the original spatiotemporal information required for flap state assessment. Upon receiving the flap status data packet, the flap surface attribute parsing module first parses the timestamp code and affine transformation matrix parameters in the packet header. Then, based on the affine transformation matrix parameters, it confirms that the spatial correspondence between temperature data and image data has been established. This confirmation step is a prerequisite for accurate region division and parameter extraction by the flap surface attribute parsing module. Optionally, the system log records the data packet generation time and compares it with the timestamp code within the flap status data packet to verify the integrity of the data transmission process. Inconsistencies in the comparison will trigger a data retransmission mechanism.

[0025] In one embodiment of the present invention, see [reference] Figure 3 Reference value ranges for ischemic pallor and ecchymotic cyanosis regions in a standard color space were defined, and a color distribution model for normal skin color was established. The acquired color texture image data was converted to this standard color space, and the matching degree between the values ​​and the reference ranges was compared pixel by pixel to mark the pixel regions of ischemic pallor, ecchymotic cyanosis, and normal skin color. Adjacent pixel regions with the same attributes were merged using a region growing algorithm, and the area, shape complexity, and average color intensity of each region were calculated as the intensity parameters of the color texture. In the spatial temperature field data, abnormal high temperature thresholds and abnormal low temperature thresholds were set based on clinical experience. After comparing each temperature measurement point with the thresholds, abnormal high temperature zones, abnormal low temperature zones, and transition temperature zones were defined. The rate of temperature change within and at the boundaries of each region was calculated as the temperature gradient parameter. Finally, the intensity parameters of the color texture and the temperature gradient parameters were bound to the sub-regions of the skin flap space to form a set of surface attribute parameters.

[0026] In practical implementation, the flap surface attribute parsing module receives flap state data packets fused with a spatiotemporal reference and separates color texture image data and spatially mapped spatial temperature field data from them. The process of constructing the surface attribute parameter set involves the independent parsing and final binding of color and temperature information. For the processing of color texture image data, the system pre-sets reference value ranges for ischemic pallor and congestive cyanosis in the CIELAB standard color space. The ischemic pallor area corresponds to a higher range of L values ​​and a and b values, while the congestive cyanosis area corresponds to a lower range of L values ​​and a and b values ​​that are biased towards blue. At the same time, a multidimensional Gaussian distribution model is established based on a large amount of sample data to describe the color distribution model of normal skin color. The specific implementation method of establishing the multidimensional Gaussian distribution model to describe the color distribution model of normal skin color includes the system first collecting a large amount of color texture image data of flap areas clinically diagnosed as having normal blood supply as a sample set. These sample data cover different skin colors, lighting conditions, and individual differences to ensure the generalization ability of the model. The flap surface attribute analysis module converts all color texture image data in the sample set to the CIELAB standard color space and extracts the L, a, and b channel values ​​of pixels labeled as normal skin color regions. Based on these value sets, the module calculates the mean vector and covariance matrix of normal skin color in the CIELAB space, thus constructing a multidimensional Gaussian probability model centered on the mean vector and with the covariance matrix describing the correlation of each dimension. The flap surface attribute analysis module converts the input color texture image data from the RGB color space to the CIELAB standard color space. Then, for each pixel in the image, it calculates the matching degree between its CIELAB value and the reference ranges for ischemic pallor, ecchymosis, and normal skin color distribution models. Based on the highest matching degree, the pixel is labeled with one of the following attributes: ischemic pallor, ecchymosis, or normal skin color.

[0027] In some embodiments, after marking, the flap surface attribute parsing module uses a region growing algorithm to aggregate pixels. The algorithm uses pixels with the same attributes as seeds, expands to their neighborhoods, and merges all adjacent pixels with the same attributes, thereby forming connected ischemic pale regions, congested cyanotic regions, and normal skin color regions. For each merged region, its area, shape complexity, and average color intensity within the region need to be calculated as intensity parameters of the color texture. Shape complexity can be quantified by the ratio of the region's perimeter to its area, and the calculation of the average color intensity involves the original RGB values ​​of all pixels within the region. The formula for calculating the intensity parameters of the color texture is: in: This represents the calculated average color intensity value of the region. This represents the total number of pixels in the area. Representing the first in each region The intensity values ​​of the red, green, and blue channels for each pixel. It can be understood that area, shape complexity, and average color intensity value together constitute the set of intensity parameters describing this color region.

[0028] Optionally, for processing the spatial temperature field data, the flap surface attribute analysis module divides the data based on pre-stored clinical experience thresholds, setting abnormal high temperature thresholds and abnormal low temperature thresholds. The module iterates through each mapped temperature measurement point in the spatial temperature field data, comparing the temperature measurement point's value with the abnormal high temperature threshold and abnormal low temperature threshold. Points with temperature measurement values ​​higher than the abnormal high temperature threshold are classified as abnormal high temperature zones, points with temperature measurement values ​​lower than the abnormal low temperature threshold are classified as abnormal low temperature zones, and points with temperature measurement values ​​between the two thresholds are classified as transitional temperature zones. After classification, the flap surface attribute analysis module further analyzes the temperature gradient parameters of each region, calculates the standard deviation of the temperature measurement point values ​​within the region as the internal temperature change rate, and simultaneously detects the numerical differences between adjacent temperature measurement points at the region boundary to calculate the boundary temperature change rate.

[0029] In some embodiments, the final set of surface attribute parameters is organized using flap spatial sub-regions as basic units. For each spatial sub-region divided from the flap image, the flap surface attribute parsing module binds the color attribute category, area, shape complexity, and average color intensity parameters of the sub-region obtained from the color texture image data to the temperature partition category and temperature gradient parameters of the sub-region obtained from the spatial temperature field data. The flap surface attribute parsing module generates a record for each spatial sub-region containing all the above-mentioned bound parameters, and the sum of the records of all spatial sub-regions constitutes the complete set of surface attribute parameters. Optionally, the precise spatial correspondence between color regions and temperature partitions depends on the coordinate mapping already completed in the flap state data packet fused with the spatiotemporal reference. The flap surface attribute parsing module ensures that the color parameters and temperature parameters extracted from the same spatial sub-region have geometric consistency based on the mapping relationship.

[0030] In one embodiment of the present invention, the boundary coordinate sets of the ischemic pallor region and the abnormal hypothermia region are extracted from the surface attribute parameter set. The intersection area and independent area of ​​the two sets are calculated, and the spatial overlap value is obtained by weighting the intersection area and independent area by ratio. For the congested cyanosis region and the abnormal hyperthermia region, the shortest average distance between their boundary coordinate sets is calculated, and converted into a spatial correlation strength value according to the distance-correlation mapping table. The spatial overlap value and the spatial correlation strength value are combined into a multi-dimensional vector, and a coupled pathological inference vector is generated by weighting according to a preset weight. The construction of the evolution trajectory model requires retrieving the coupled pathological inference vectors of multiple consecutive historical assessment time points of the same flap object to form a time series vector set. After smoothing and filtering, the changing trend of each vector dimension is analyzed, and an autoregressive model is used to establish a trend prediction sub-model for each dimension. After integration, a complete evolution trajectory model is formed. The current coupled pathological inference vector is input into the model to calculate the predicted value of the state vector at the next assessment time point, and a dynamic inference report is generated by encapsulation.

[0031] In practical implementation, the deep-coupled pathological inference module receives a set of surface attribute parameters from the flap surface attribute parsing module. The module's execution logic revolves around calculating the spatial overlap between the ischemic pallor region and the abnormal hypothermia region, and analyzing the spatial correlation between the congested cyanosis region and the abnormal hyperthermia region. The deep-coupled pathological inference module extracts the boundary coordinate set of the ischemic pallor region from the surface attribute parameter set. This boundary coordinate set consists of the coordinates of all pixels constituting the region's outline; it also extracts the boundary coordinate set of the abnormal hypothermia region. The module calculates the intersection area of ​​the regions enclosed by the boundary coordinate sets of the ischemic pallor region and the abnormal hypothermia region. This intersection area represents the area of ​​the common region with the color attribute of ischemic pallor and the temperature attribute of abnormal hypothermia. The module also calculates the independent areas of the ischemic pallor region and the abnormal hypothermia region separately. The spatial overlap value is obtained through the following weighted calculation: in: This represents the final spatial overlap value. Represents the area of ​​intersection. The independent area representing the ischemic pallor region, The independent area representing the abnormally low temperature region, and The weighting coefficients are preset and satisfy the following conditions: .

[0032] In some embodiments, for the analysis of the congested cyanotic region and the abnormally high temperature region, the deep-coupled pathological inference module extracts the boundary coordinate sets of the congested cyanotic region and the abnormally high temperature region from the surface attribute parameter set. The module calculates the shortest distance from each point in the boundary coordinate set of the congested cyanotic region to the boundary coordinate set of the abnormally high temperature region, and averages all the shortest distances to obtain the shortest average distance. The module internally stores a distance-association mapping table, which defines the non-linear transformation relationship from the shortest average distance to the spatial correlation strength value. The module queries the distance-association mapping table based on the shortest average distance and outputs the corresponding spatial correlation strength value. It can be understood that the spatial overlap value and the spatial correlation strength value are combined into a multi-dimensional vector, such as a two-dimensional vector. ,in The value represents the spatial correlation strength. The deep coupled pathological inference module performs weighted fusion on this multidimensional vector according to the preset weight coefficients to generate the final coupled pathological inference vector, which serves as a quantitative representation of the deep blood flow and metabolic state of the flap.

[0033] Optionally, the survival status dynamic simulation module receives coupled pathological inference vectors. This module retrieves coupled pathological inference vectors generated for the same skin flap object at multiple consecutive historical assessment points from the system's historical database, arranging them in chronological order of timestamps to form a time-series vector set. The survival status dynamic simulation module performs smoothing filtering on the time-series vector set, using a moving average method to suppress random fluctuations caused by data acquisition noise. The module analyzes the changing trend of each vector dimension in the smoothed time-series vector set, and independently establishes a trend prediction sub-model using an autoregressive model for each vector dimension. The order of the autoregressive model is determined by the information criterion. In its implementation, the survival status dynamic simulation module analyzes the smoothed time series vector set, processing each dimension of the coupled pathological inference vector independently. The module first extracts the numerical sequence of each dimension at consecutive time points, observing its pattern of change over time, such as upward, downward, or stable trends. For each dimension sequence, the module uses an autoregressive model, representing the current value as a linear combination of values ​​from several past times. The model parameters are obtained by fitting historical data using the least squares method. The order of the autoregressive model is selected using an information criterion. The module calculates the goodness of fit and complexity penalty term for different orders, and after comprehensive comparison, selects the order that minimizes the information criterion value as the optimal model order. Finally, an independent trend prediction sub-model is established for each vector dimension. These sub-models collectively form the basis of the evolution trajectory model for subsequent state prediction. All trend prediction sub-models for all vector dimensions are integrated to form a complete evolution trajectory model. The survival status dynamic simulation module inputs the newly generated coupled pathological inference vector at the current assessment time point into the evolution trajectory model. The evolution trajectory model runs the autoregressive models for each dimension to calculate the predicted state vector value for the next preset assessment time point.

[0034] In some embodiments, the survival status dynamic extrapolation module encapsulates the coupled pathological inference vector at the current assessment time point with the predicted state vector output by the model. The encapsulation also includes the timestamp information and flap identification information corresponding to the vector, generating a dynamic extrapolation report. This report is presented as a structured document, clearly distinguishing between current state data and future prediction data. Optionally, during operation, if the evolution trajectory model receives a new coupled pathological inference vector, it will incorporate it into the time series vector set and update the parameters of the autoregressive model in real time, achieving adaptive adjustment of the model. The output of the survival status dynamic extrapolation module provides a trend reference based on time series analysis for clinical judgment.

[0035] In one embodiment of the present invention, the confidence calibration module receives the predicted state vector value from the dynamic inference report. After obtaining the actual coupled pathological inference vector at the actual evaluation time, it calculates the Euclidean distance between the predicted value and the actual vector as the deviation value. This deviation value is added to the historical prediction deviation queue and a moving average is calculated. Based on the moving average, the order of the autoregressive model or the smoothing filter intensity parameter in the evolution trajectory model is adjusted to complete the model prediction confidence calibration. The regional microcirculation assessment submodule separates the microvascular network pattern from the color texture image data from the surface attribute parameter set and extracts the subtle temperature change curves corresponding to the pattern direction. The connectivity and density parameters of the microvascular pattern are analyzed, along with the frequency and amplitude parameters of the temperature fluctuation curve, to establish feature matching pairs. When the pattern connectivity length and density meet the threshold and the corresponding temperature fluctuation parameter is within the normal range, it is determined to be a high-quality match. After statistically analyzing the proportion of high-quality matching pairs, a regional microcirculation status score is generated according to the proportion-score conversion table. This score is attached to the coupled pathological inference vector to form an enhanced pathological inference vector.

[0036] In practice, the confidence assessment calibration module receives the dynamic simulation report output by the survival status dynamic simulation module and extracts the predicted state vector value from the report. The module waits for the system to reach the next preset actual assessment time point, which is triggered by the system scheduling module to initiate a new round of data acquisition and analysis. At this time point, the module obtains the actual coupled pathological inference vector, which is actually collected at the assessment time point and processed by the deep coupled pathological inference module. The module calculates the Euclidean distance between the predicted state vector value and the actual coupled pathological inference vector, using this distance as the deviation value for the current prediction. This deviation value is a scalar value. The module adds this deviation value to the historical prediction deviation queue corresponding to the flap object. This queue is a fixed-length first-in-first-out queue used to store the deviation values ​​from the most recent predictions. The module calculates the moving average of all deviation values ​​in the historical prediction deviation queue, which reflects the overall level of recent prediction deviation.

[0037] In some embodiments, the confidence assessment calibration module adjusts the parameters of the evolution trajectory model within the survival dynamic inference module based on the magnitude of the moving average. If the moving average exceeds a preset high threshold, the confidence assessment calibration module increases the order of the autoregressive model in the evolution trajectory model or strengthens the intensity parameter of the smoothing filter. If the moving average is below a preset low threshold, the confidence assessment calibration module decreases the order of the autoregressive model in the evolution trajectory model or weakens the intensity parameter of the smoothing filter. After adjustment, the confidence assessment calibration module feeds back the calibrated model parameters to the survival dynamic inference module, which updates the evolution trajectory model using the new parameters, thus completing the calibration of the model prediction confidence. The workflow of the regional microcirculation assessment submodule is independent of the above calibration process. The regional microcirculation assessment submodule separates the microvascular network pattern identified in the color texture image data from the set of surface attribute parameters output by the flap surface attribute analysis module. The identification of the microvascular network pattern is based on image enhancement and linear structure detection algorithms.

[0038] The regional microcirculation assessment submodule calculates the connectivity and density parameters of the microvascular network pattern. Connectivity is assessed by whether the vessel segments in the pattern are continuous and uninterrupted, while density is calculated by the total length of the vessel pattern per unit area. The submodule also calculates the fluctuation frequency and amplitude parameters of the subtle temperature change curve. The fluctuation frequency is obtained through spectral analysis to determine the dominant frequency, and the amplitude is obtained by calculating the average difference between the peak and trough values ​​of the curve. Furthermore, the submodule performs a correspondence analysis between microvascular pattern features and temperature fluctuation features, establishing matching pairs between them. A matching pair means that a segment of microvascular pattern and its corresponding temperature curve segment are considered to originate from the same microscopic physiological structure. Based on the number and quality of matching pairs, the submodule generates a regional microcirculation status score. This score is then appended to the coupled pathology inference vector generated by the deep coupled pathology inference module, forming an enhanced pathology inference vector.

[0039] Optionally, the generation of regional microcirculation status scores follows preset rules. These rules set minimum length thresholds and minimum density thresholds for the connectivity of microvascular patterns, as well as normal range thresholds for the fluctuation frequency and amplitude of temperature fluctuation curves. For each identified microvascular network pattern segment, the regional microcirculation assessment submodule calculates its connectivity length and density per unit area; for each extracted minute temperature change curve segment, the regional microcirculation assessment submodule calculates its dominant fluctuation frequency and average amplitude. When the connectivity length and density of a pattern segment are both greater than or equal to their corresponding thresholds, and the dominant fluctuation frequency and average amplitude of its corresponding temperature curve segment are both within their normal range thresholds, the regional microcirculation assessment submodule determines this match as a high-quality match. The regional microcirculation assessment submodule counts the number of high-quality matches among all matches and calculates the proportion of high-quality matches to the total number of matches. See Table 1.

[0040] Table 1: Proportion-Rating Conversion Table In some embodiments, the quantitative calculation of regional microcirculation status scores can be expressed as follows: in: Represents the regional microcirculation status score. This represents the proportion of high-quality matching pairs to the total number of matching pairs. The mapping function defined by the ratio-rating conversion table represents the input ratio value. Mapped to a discrete rating scale It can be understood that the enhanced pathological inference vector adds a dimension to the original coupled pathological inference vector; this new dimension stores the regional microcirculation status score. .

[0041] See Figure 4 During the confidence calibration process for flap state prediction, the variation characteristics of prediction bias at different assessment time points were presented. In the figure, the "single prediction Euclidean distance bias" (red line) represents the Euclidean distance between the predicted and actual values ​​of the state vector at each assessment time point, and its fluctuation reflects the fluctuation in the accuracy of a single prediction; the "moving average bias (window size = 8)" (blue line) is the average of the most recent 8 biases, reflecting the overall trend of recent prediction bias. Meanwhile, the figure sets "high bias threshold (0.4)" and "low bias threshold (0.2)" as triggers for model parameter adjustments: when the moving average bias exceeds 0.4, the autoregressive order or smoothing filter strength of the model needs to be increased; when it is below 0.2, the corresponding parameters are weakened, thereby achieving dynamic calibration of prediction confidence.

[0042] In one embodiment of the present invention, a pre-strategy mapping library pre-stores pathological state templates for various typical postoperative complications of flap surgery. Each template is defined by a standardized coupled pathological inference vector range. The real-time vectors generated by the deep coupled pathological inference module are compared with the template library for similarity calculation. Templates with similarity exceeding a threshold are selected as candidate matching templates, and associated clinical intervention suggestions are automatically output and presented in conjunction with a dynamic deduction report. The self-updating mechanism of the intervention strategy mapping library continuously records the intervention measures output by the system and the subsequent actual evolution direction of the flap. When the cumulative number of deviations between the actual evolution and the expected outcome reaches a threshold, a template update is triggered. During the update process, vectors that cause deviations and their subsequent data are clustered to form new data clusters. After review and definition by clinical experts, these clusters are added as new pathological state templates or the vector range of the original templates is corrected. The updated templates and intervention suggestions are synchronized to the mapping library.

[0043] In practical implementation, the system integrates an intervention strategy mapping library, which includes a pathological state template library. This library stores pathological state templates for various typical post-flap complications. Each pathological state template is defined by a standardized range of coupled pathological inference vectors, defining the lower and upper limits of each vector dimension's value. The intervention strategy mapping library receives coupled pathological inference vectors generated in real-time by the deep coupled pathological inference module. It then calculates the similarity between these real-time coupled pathological inference vectors and all pathological state templates in the pathological state template library. Similarity calculation can employ methods such as cosine similarity or the reciprocal of Euclidean distance. The intervention strategy mapping library presets a matching threshold, selecting pathological state templates whose similarity calculation results exceed this threshold as candidate matching templates. In some embodiments, the intervention strategy mapping library pre-stores clinical intervention recommendations corresponding to each pathological state template. These recommendations exist in text or structured instruction form. When one or more candidate matching templates exist, the intervention strategy mapping library automatically outputs the clinical intervention recommendation associated with the candidate matching template with the highest similarity. The output clinical intervention recommendations are presented in conjunction with the dynamic simulation report generated by the survival status dynamic simulation module. This in conjunction means that the clinical intervention recommendations are presented to the user as an independent chapter or additional note in the dynamic simulation report.

[0044] Optionally, the intervention strategy mapping library has a self-updating mechanism. This mechanism continuously records the clinical intervention recommendations output by the system and the actual evolution direction of the flap status over a subsequent period. The actual evolution direction is characterized by the changing trend of the actual coupled pathological inference vector obtained at subsequent time points. The intervention strategy mapping library compares the expected evolution direction based on the original pathological state template and dynamic inference report with the actual evolution direction. When the cumulative number of times the actual evolution direction deviates significantly from the expected evolution direction reaches a preset abnormality threshold, the self-updating mechanism triggers the template update process.

[0045] In some embodiments, during the template update process, the self-updating mechanism collects the actual evolution data of the coupled pathological inference vectors that lead to expected deviations and their subsequent time points, forming a set of historical data sequences. The self-updating mechanism performs cluster analysis on this set of historical data sequences, using algorithms such as K-means or hierarchical clustering to aggregate vectors with similar characteristics into new pathological state data clusters. Clinical experts review and define the newly formed pathological state data clusters, confirming the clinicopathological significance represented by each cluster and defining a new set of standardized coupled pathological inference vector ranges for them. After review and confirmation, the self-updating mechanism adds the newly defined pathological state template to the pathological state template library, or uses the newly defined standardized coupled pathological inference vector ranges to correct the vector range of the original pathological state template.

[0046] It is understandable that a specific implementation of similarity calculation is as follows: in: This represents the calculated similarity value. The total number of dimensions representing the coupled pathological inference vector. The real-time coupled pathological inference vector represents the first... The value of dimension, Representing the The standardized vector range of the pathological state template is in the th... The central value of the dimension, Representing the The standardized vector range of the pathological state template is in the th... The radius and width of the dimension. Optional, anomaly count threshold. It is an integer, when the number of deviations is counted. satisfy Updates are triggered at specific times.

[0047] See Figure 5In the process of matching pathological states in the intervention strategy mapping library, the similarity distribution between real-time coupled pathological inference vectors and typical pathological state templates at different assessment time points is presented. Specifically, the figure uses "pathological state templates" as the row dimension (covering templates for typical complications such as arterial insufficiency and venous congestion) and "assessment time point" as the column dimension. The matching degree between the real-time vector and each template is quantified by a color gradient (corresponding to the similarity value scale on the right): the closer the similarity value is to 1.0, the darker the corresponding color (e.g., the similarity of the arterial insufficiency template at time point 1 is 0.96, and the similarity of the venous congestion template at time point 5 is 0.98), indicating a higher degree of fit between the real-time state and the pathological template. This heat map distribution can directly support the screening of candidate pathological templates (selecting templates with similarity exceeding the threshold), thereby associating and outputting corresponding clinical intervention suggestions, and providing data basis for subsequent updates of pathological state templates (based on the deviation between similarity matching and actual evolution).

[0048] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0049] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A combined intelligent assessment system for transplanted skin flap color and temperature, characterized in that, include: The flap multimodal data acquisition module is used to simultaneously acquire color texture image data and spatial temperature field data of the transplanted flap area using high-resolution imaging equipment and infrared thermal imaging equipment, and to superimpose and fuse the two types of data according to a common spatial coordinate system and time reference to generate a flap status data package with fused spatiotemporal reference. The flap surface attribute analysis module receives the flap status data packet, identifies and quantifies the boundary and intensity parameters of ischemic pale areas, congested cyanotic areas and normal skin color areas from color texture image data, divides the boundary and temperature gradient parameters of abnormal high temperature areas, abnormal low temperature areas and transition temperature zones from spatial temperature field data, and summarizes the identified area boundaries, intensity parameters and temperature gradient parameters into a surface attribute parameter set. The deep-coupled pathological inference module retrieves the set of surface attribute parameters, calculates the spatial overlap between the ischemic pale area and the abnormal low temperature area, performs spatial correlation analysis between the congested cyanotic area and the abnormal high temperature area, and generates a coupled pathological inference vector characterizing the deep blood flow and metabolic state of the flap based on preset overlap correlation rules. The survival status dynamic simulation module receives the coupled pathological inference vector, combines it with similar vectors in the historical time series, constructs an evolution trajectory model of the flap survival status through time series analysis, and predicts the state vector at the next assessment time point based on the evolution trajectory model, generating a dynamic simulation report containing the current state and the predicted state.

2. The intelligent assessment system for combined color and temperature of transplanted skin flaps according to claim 1, characterized in that, The process of generating the flap state data packet based on the fused spatiotemporal reference includes: The image pixel coordinate system of the high-resolution imaging device and the temperature data coordinate system of the infrared thermal imaging device are calibrated respectively, and an affine transformation matrix between the two coordinate systems is established. Using the affine transformation matrix, each temperature measurement point in the temperature data coordinate system is precisely mapped to the corresponding pixel position in the image pixel coordinate system; Each synchronously acquired image frame and temperature field data frame is assigned the same and unique timestamp code; Image data with completed coordinate mapping and timestamp encoding is packaged with temperature data to form the flap state data packet of the fused spatiotemporal reference; After generating the flap status data packet, the flap multimodal data capture module transmits it to the flap surface attribute parsing module and simultaneously records the data packet generation time and data size to the system log.

3. The intelligent assessment system for combined color and temperature of transplanted skin flaps according to claim 1, characterized in that, The method for constructing the surface attribute parameter set includes: Define the reference value ranges for ischemic pallor and ecchymosis in the standard color space, as well as the color distribution model for normal skin color; The collected color texture image data is converted to the standard color space, and the matching degree of its value with the reference value range and the color distribution model is compared pixel by pixel to mark the pixel areas belonging to ischemic pallor, ecchymosis and normal skin color. The region growing algorithm merges adjacent pixel regions with the same attributes, and calculates the area, shape complexity, and average color intensity of each merged region as the intensity parameter of the color texture. In the spatial temperature field data collected by infrared thermal imaging equipment, abnormal high temperature thresholds and abnormal low temperature thresholds are set based on clinical experience. The value of each temperature measurement point is compared with the abnormal high temperature threshold and the abnormal low temperature threshold to divide the abnormal high temperature zone, the abnormal low temperature zone and the transition temperature zone between the two. Calculate the rate of temperature change within each region and at the region boundaries, and use it as the temperature gradient parameter; Finally, the intensity parameters of the color texture and the temperature gradient parameters are bound according to their respective flap space sub-regions to form the set of surface attribute parameters.

4. The intelligent assessment system for combined color and temperature of transplanted skin flaps according to claim 3, characterized in that, The specific execution steps for the spatial overlap calculation and the spatial correlation analysis include: Extract the boundary coordinate set of the ischemic pale region and the boundary coordinate set of the abnormally low temperature region from the surface attribute parameter set; Calculate the area of ​​the intersection of the boundary coordinate set of the ischemic pale region and the boundary coordinate set of the abnormally low temperature region, and calculate the independent areas of the ischemic pale region and the abnormally low temperature region respectively; Divide the intersection area by the independent area of ​​the ischemic pale region to obtain the first overlap ratio; divide the intersection area by the independent area of ​​the abnormally low temperature region to obtain the second overlap ratio; take the weighted average of the first overlap ratio and the second overlap ratio as the final spatial overlap value. Extract the boundary coordinate set of the ecchymotic region and the boundary coordinate set of the abnormally high temperature region from the surface attribute parameter set; Calculate the shortest average distance between the boundary coordinate set of the congested cyanotic region and the boundary coordinate set of the abnormally high temperature region; According to the preset distance-association mapping table, the shortest average distance is converted into a spatial correlation strength value; The spatial overlap value and the spatial correlation strength value are combined into a multi-dimensional vector, and then weighted and fused according to a preset weight coefficient to generate the coupled pathological inference vector.

5. The intelligent assessment system for combined color and temperature of transplanted skin flaps according to claim 1, characterized in that, The process of constructing and predicting the evolution trajectory model includes: From the system's historical database, retrieve the coupled pathological inference vectors generated for the same flap object at multiple consecutive historical assessment points, and arrange them in chronological order to form a time-series vector set; The time series vector set is smoothed and filtered to eliminate random fluctuations caused by acquisition noise; Analyze the changing trend of each vector dimension in the smoothed time series vector set, and use an autoregressive model to establish an independent trend prediction sub-model for each vector dimension; By integrating all the trend prediction sub-models of all vector dimensions, a complete evolution trajectory model is formed. The coupled pathological inference vector generated at the current assessment time point is input into the evolution trajectory model, and the model calculates the predicted value of the state vector at the next preset assessment time point. The coupled pathological inference vector at the current assessment time point is encapsulated together with the predicted value of the state vector to generate the dynamic inference report.

6. The intelligent assessment system for combined color and temperature of transplanted skin flaps according to claim 5, characterized in that, The system also includes a confidence level calibration module, comprising: Receive the dynamic simulation report output by the survival status dynamic simulation module, and extract the state vector prediction value from the report; Waiting to reach the next actual assessment time point, obtain the actual coupled pathological inference vector obtained by the actual assessment time point and processed by the deep coupled pathological inference module; Calculate the Euclidean distance between the predicted state vector value and the actual coupled pathological inference vector, and use it as the deviation value for this prediction; The deviation value of this prediction is added to the historical prediction deviation queue of the flap object, and the moving average of all deviation values ​​in the historical prediction deviation queue is calculated. Based on the magnitude of the moving average, adjust the order of the autoregressive model or the strength parameter of the smoothing filter in the evolution trajectory model to complete the calibration of the model prediction confidence. The confidence assessment calibration module feeds back the calibrated model parameters to the survival status dynamic inference module to update the evolution trajectory model.

7. The intelligent assessment system for combined color and temperature of transplanted skin flaps according to claim 1, characterized in that, The system further includes a regional microcirculation assessment submodule, the workflow of which includes: From the set of surface attribute parameters output by the flap surface attribute parsing module, the microvascular network pattern identified in the color texture image data is separated. From the spatial temperature field data, extract the subtle temperature change curves corresponding to the orientation of the microvascular network pattern; The connectivity and density parameters of the microvascular network pattern are analyzed in correspondence with the fluctuation frequency and amplitude parameters of the subtle temperature change curve. Establish matching pairs between microvascular pattern features and temperature fluctuation features, and generate regional microcirculation status scores based on the number and quality of matching pairs; The regional microcirculation assessment submodule appends the regional microcirculation status score to the coupled pathological inference vector generated by the deep coupled pathological inference module, forming an enhanced pathological inference vector.

8. The intelligent assessment system for combined color and temperature of transplanted skin flaps according to claim 7, characterized in that, The generation of the regional microcirculation status score follows these rules: Set minimum length thresholds and minimum density thresholds for the connectivity of microvascular patterns; Set the normal range threshold for the fluctuation frequency and the normal range threshold for the amplitude of the temperature fluctuation curve; For each identified microvascular network pattern segment, calculate its connectivity length and density per unit area; For each extracted segment of the minute temperature change curve, calculate its principal fluctuation frequency and average amplitude; When the connectivity length and density of a pattern segment are both greater than or equal to their corresponding thresholds, and the main fluctuation frequency and average amplitude of the corresponding temperature curve segment are both within their normal range thresholds, the matching pair is determined to be a high-quality match. Count the number of high-quality matching pairs among all matching pairs, and calculate the proportion of high-quality matching pairs to the total number of matching pairs; According to the preset ratio-score conversion table, the ratio is converted into the regional microcirculation status score.

9. The intelligent assessment system for combined color and temperature of transplanted skin flaps according to claim 1, characterized in that, The system also integrates an intervention strategy mapping library, including: Establish a pathological state template library containing various typical postoperative complications of flap surgery. Each pathological state template is defined by a set of standardized coupled pathological inference vector ranges. The coupling pathological inference vector generated in real time by the deep coupling pathological inference module is compared with the similarity of all templates in the pathological state template library. Pathological state templates with similarity exceeding a preset matching threshold are selected as candidate matching templates; The intervention strategy mapping library pre-stores clinical intervention recommendations corresponding to each pathological state template; When a candidate matching template exists, the system automatically outputs clinical intervention recommendations associated with the candidate matching template and presents them in conjunction with the dynamic simulation report generated by the survival status dynamic simulation module.

10. The intelligent assessment system for combined color and temperature of transplanted skin flaps according to claim 9, characterized in that, The intervention strategy mapping library has a self-updating mechanism, and its update triggering conditions and process are as follows: Continuously record the clinical intervention recommendations output by the system and the actual evolution of the flap status over a subsequent period of time; When the number of times the actual evolution direction deviates significantly from the direction expected based on the original template and dynamic inference report reaches a preset abnormal number threshold, the template update process is triggered. In the template update process, the coupled pathological inference vectors that cause deviations from the expected results and their subsequent actual evolution data are clustered to form new pathological state data clusters. The newly formed pathological state data clusters are reviewed and defined by clinical experts. After confirmation, they are added as new pathological state templates or used to correct the standardized vector range of the original pathological state templates. The updated pathological state template and associated intervention suggestions will be synchronized to the intervention strategy mapping library for subsequent matching and output.

Images