Wound healing progress prediction method and system based on 3D point cloud reconstruction, electronic device and storage medium

Abstract

This application provides a method, system, electronic device, and storage medium for predicting wound healing progress based on 3D point cloud reconstruction, relating to the field of 3D point cloud reconstruction technology. This application generates spectral feature vectors containing biological indicators such as hemoglobin concentration and tissue oxygenation status by acquiring the reflectance spectral data and initial three-dimensional point cloud data of the wound. The initial point cloud is processed to obtain target data that retains its spatial morphology. The spectral feature values ​​are then fused point-by-point with the spatial coordinates of the target point cloud to generate a three-dimensional point cloud data cube. A 3D convolutional neural network is used to extract the spatial structural features within the cube, and the changes in spectral features at adjacent time points are tracked using optical flow, coupling the two. Based on the coupling results, a cell migration map is constructed, and the changing trend of node connection strength in the map is analyzed to obtain the predicted wound healing progress, which can improve the accuracy and comprehensiveness of wound healing progress prediction.

Description

Technical Field

[0001] This application relates to the field of 3D point cloud reconstruction technology, and in particular to a method, system, electronic device and storage medium for predicting wound healing progress based on 3D point cloud reconstruction. Background Technology

[0002] In the field of wound healing progress prediction, especially in the treatment of chronic wounds such as diabetic foot ulcers and pressure ulcers, accurately grasping the healing process is crucial for adjusting treatment plans. The healing of these wounds involves not only the reduction of surface area but also is closely related to changes in the wound's three-dimensional morphology, such as edge smoothness, depth changes, and dynamic changes in internal tissue biomarkers, such as hemoglobin concentration and oxygenation status. Traditional 2D images struggle to fully capture the three-dimensional spatial structure of the wound, while single biomarker detection cannot correlate with its spatial distribution. Therefore, there is an urgent clinical need for a method that combines 3D point cloud reconstruction technology with spatial morphology and biological information to achieve the quantification and dynamic prediction of healing progress.

[0003] Currently, to address the aforementioned needs, existing solutions utilize 3D point cloud devices to acquire three-dimensional spatial coordinate data of the wound, while simultaneously employing spectral sensors to collect the wound's reflectance spectrum information. The spatial morphological features reflected in the 3D point cloud data are then simply fused with biological indicators extracted from the spectral data to construct a basic model and output predictions of healing progress. This solution attempts to compensate for the limitations of single-dimensional data by combining spatial and biological information.

[0004] However, the existing solution has significant drawbacks: First, the processing of 3D point cloud data only extracts the overall morphology, without considering the spatial structural differences at different levels, such as the small area at the edge of the wound and the medium area in the middle, resulting in insufficient refinement of spatial features. Second, the use of spectral data only focuses on the values ​​of biological indicators at a single time point, without dynamically tracking the feature changes at adjacent time points, making it difficult to reflect the dynamic trend of the healing process. Third, the fusion method of spatial morphological features and biological indicators is relatively simple, without establishing a deep correlation between the two in terms of spatial location and temporal changes, making it impossible for the fused information to accurately correspond to key healing mechanisms such as cell migration, ultimately affecting the accuracy of the prediction results. Summary of the Invention

[0005] The purpose of this application is to provide a method, system, electronic device and storage medium for predicting wound healing progress based on 3D point cloud reconstruction, so as to solve the problems of poor accuracy and comprehensiveness in wound healing progress prediction caused by imprecise spatial features, lack of spectral dynamics and insufficient fusion correlation in the prior art.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a method for predicting wound healing progress based on 3D point cloud reconstruction, comprising:

[0007] Acquire reflectance spectral data and initial three-dimensional point cloud data of wound tissue, and generate a spectral feature vector containing spectral feature values ​​of multiple biological indicators based on the reflectance spectral data, including hemoglobin concentration and tissue oxygenation status.

[0008] The initial three-dimensional point cloud data is processed to obtain target three-dimensional point cloud data that retains the spatial morphological features of the wound. The spectral feature values ​​in the spectral feature vector are mapped and fused with the spatial coordinates of the target three-dimensional point cloud data point by point to generate a three-dimensional point cloud data cube.

[0009] A 3D convolutional neural network is used to extract the spatial structure features in the three-dimensional point cloud data cube. The optical flow method is used to track the spectral feature changes corresponding to the spectral feature vectors at adjacent time points, and the spatial structure features and the spectral feature changes are coupled.

[0010] Based on the coupling processing results, a cell migration map of the wound tissue is constructed. By analyzing the changing trend of the connection strength of each node in the cell migration map, the healing progress prediction results in the wound healing process are obtained.

[0011] Optionally, the step of processing the initial three-dimensional point cloud data to obtain target three-dimensional point cloud data that retains the spatial morphological features of the wound, and mapping and fusing the spectral feature values ​​in the spectral feature vector with the spatial coordinates of the target three-dimensional point cloud data point by point to generate a three-dimensional point cloud data cube, includes:

[0012] The initial three-dimensional point cloud data is divided into multiple wound areas according to the point cloud density. Feature points reflecting spatial morphological changes are selected in each wound area. The feature points are combined to form target three-dimensional point cloud data.

[0013] The location markers of the reflectance spectral data are obtained, each spectral feature value in the spectral feature vector is associated with the corresponding location marker, and the spatial coordinates of each spectral feature value in the target three-dimensional point cloud data are determined.

[0014] All associated spectral feature values ​​are mapped and fused with the corresponding spatial coordinates point by point to generate a three-dimensional point cloud data cube.

[0015] Optionally, associating each spectral feature value in the spectral feature vector with the corresponding location marker and determining the spatial coordinates of each spectral feature value in the target 3D point cloud data includes:

[0016] Extract the spatial location information contained in the location marker, and associate each spectral feature value with the spatial location information to form an associated spectral feature value;

[0017] Determine the distribution range of all spatial coordinates and the relative positional relationship of each spatial coordinate in the target 3D point cloud data;

[0018] Based on the relative positional relationship, find the spatial coordinates with the smallest distance value from each of the spatial positional information within the distribution range; and determine each of the spatial coordinates as the spatial coordinates corresponding to the spectral feature value in the target three-dimensional point cloud data.

[0019] Optionally, based on the coupling processing results, a cell migration map of the wound tissue is constructed. By analyzing the changing trends of the connection strength of each node in the cell migration map, a prediction result of the healing progress in the wound healing process is obtained, including:

[0020] Multiple key locations in the wound tissue are selected from the coupling processing results, and each key location is used as a node in the cell migration map to determine the proximity relationship of each node in the wound tissue.

[0021] Based on the proximity relationship, connections are established between mutually adjacent nodes to generate a cell migration map;

[0022] Based on the first correlation value of the spatial structural features corresponding to each node and the second correlation value of the spectral feature change, the tightness of the connection between each pair of nodes is determined as the node connection strength.

[0023] Record the connection strength values ​​of each node at different time points to obtain the trend of the connection strength of each node;

[0024] Based on the changing trend, the healing progress prediction result in the wound healing process is determined using a preset prediction model.

[0025] Optionally, determining the degree of connection between any two nodes as the node connection strength based on the first correlation value of the spatial structural features corresponding to each node and the second correlation value of the spectral feature variation includes:

[0026] Extract the spatial structure features and spectral feature changes corresponding to each node;

[0027] By comparing the spatial structural features of every two nodes, a first distance similarity value is obtained from the edge range of the wound tissue to every two nodes, and a second distance similarity value is obtained from the middle range of the wound tissue.

[0028] Based on the first distance similarity value and the second distance similarity value, a first association degree value is determined;

[0029] By comparing the changes in spectral features between every two nodes, consistency information on changes in hemoglobin concentration and tissue oxygenation status between every two nodes is obtained, and a second correlation value is determined based on the consistency information.

[0030] Based on the first correlation degree value and the second correlation degree value, the tightness of the connection between each pair of nodes is determined, and the tightness is used as the node connection strength.

[0031] Optionally, the step of employing a 3D convolutional neural network to extract spatial structure features from the three-dimensional point cloud data cube, using optical flow to track the spectral feature changes corresponding to the spectral feature vectors at adjacent time points, and coupling the spatial structure features with the spectral feature changes includes:

[0032] A 3D convolutional neural network is used to divide the three-dimensional point cloud data cube into layers according to the layers from the edge of the wound tissue, the middle of the wound tissue, and the entire wound area, and process them separately to extract structural information at different layers. Based on the structural information at different layers, spatial structural features are generated.

[0033] The spectral feature vectors of each two adjacent time points are selected using the optical flow method. The corresponding spectral feature values ​​in the spectral feature vectors of each two adjacent time points are compared to obtain multiple spectral feature changes.

[0034] The spatial structural features are correlated and integrated with all the spectral feature variations to obtain the coupling processing result.

[0035] Optionally, generating a spectral feature vector containing spectral feature values ​​of multiple bioindicators based on the reflectance spectral data includes:

[0036] Determine the corresponding spectral segment for each biomarker in the reflectance spectral data, where hemoglobin concentration corresponds to the first spectral segment and tissue oxygenation status corresponds to the second spectral segment;

[0037] The numerical value reflecting the hemoglobin concentration is obtained based on the spectral information of the first spectral band and is used as the first spectral characteristic value;

[0038] The numerical values ​​reflecting the tissue oxygenation status are obtained based on the spectral information of the second spectral band, and are used as the second spectral characteristic values;

[0039] The first spectral feature value and the second spectral feature value are arranged in a preset order to form a spectral feature vector.

[0040] Secondly, this application provides a wound healing progress prediction system based on 3D point cloud reconstruction, including:

[0041] The acquisition module is used to acquire reflectance spectral data and initial three-dimensional point cloud data of wound tissue, and based on the reflectance spectral data, generate a spectral feature vector containing spectral feature values ​​of multiple biological indicators, including hemoglobin concentration and tissue oxygenation status.

[0042] The generation module is used to process the initial three-dimensional point cloud data to obtain target three-dimensional point cloud data that retains the spatial morphological features of the wound. The spectral feature values ​​in the spectral feature vector are mapped and fused with the spatial coordinates of the target three-dimensional point cloud data point by point to generate a three-dimensional point cloud data cube.

[0043] The coupling module is used to extract the spatial structure features in the three-dimensional point cloud data cube using a 3D convolutional neural network, track the spectral feature changes corresponding to the spectral feature vectors at adjacent time points using optical flow, and perform coupling processing on the spatial structure features and the spectral feature changes.

[0044] The module is used to construct a cell migration map of wound tissue based on the coupling processing results. By analyzing the changing trend of the connection strength of each node in the cell migration map, the healing progress prediction results in the wound healing process are obtained.

[0045] Thirdly, this application provides an electronic device, comprising:

[0046] Memory, used to store computer programs;

[0047] A processor is configured to execute the computer program to implement the steps of the wound healing progress prediction method based on 3D point cloud reconstruction as described in the first aspect above.

[0048] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the wound healing progress prediction method based on 3D point cloud reconstruction as described in the first aspect above.

[0049] This application provides a method for predicting wound healing progress based on 3D point cloud reconstruction, comprising: acquiring reflectance spectral data and initial three-dimensional point cloud data of wound tissue; generating a spectral feature vector containing spectral feature values ​​of multiple biological indicators, including hemoglobin concentration and tissue oxygenation status, based on the reflectance spectral data; processing the initial three-dimensional point cloud data to obtain target three-dimensional point cloud data that retains the spatial morphological features of the wound; mapping and fusing the spectral feature values ​​in the spectral feature vector with the spatial coordinates of the target three-dimensional point cloud data point by point to generate a three-dimensional point cloud data cube; using a 3D convolutional neural network to extract the spatial structural features in the three-dimensional point cloud data cube; using optical flow to track the spectral feature changes corresponding to the spectral feature vectors at adjacent time points; and coupling the spatial structural features with the spectral feature changes; based on the coupling processing results, constructing a cell migration map of the wound tissue; and obtaining the prediction result of the healing progress in the wound healing process by analyzing the changing trend of the connection strength of each node in the cell migration map.

[0050] The wound healing progress prediction method based on 3D point cloud reconstruction provided in this application has the following advantages: By acquiring the reflectance spectral data and initial three-dimensional point cloud data of wound tissue and generating spectral feature vectors containing multiple biological indicators, it can provide a bioinformatics basis for wound healing analysis; by processing the initial three-dimensional point cloud data to obtain the target three-dimensional point cloud data, and by mapping and fusing the spectral feature values ​​with spatial coordinates point by point to generate a three-dimensional point cloud data cube, it can integrate the wound spatial morphology and biological indicator information; by using a 3D convolutional neural network to extract spatial structural features, using optical flow to track the changes in spectral features and performing coupling processing, it can capture the correlation between the wound spatial structure and the dynamic changes in biological indicators; by constructing a cell migration map and analyzing the trend of node connection strength changes to obtain the healing progress prediction results, it can achieve quantitative prediction of the wound healing process.

[0051] Furthermore, the initial 3D point cloud data was divided into wound regions according to density, and feature points were selected to form target 3D point cloud data. Position markers of the reflectance spectral data were extracted and associated with each spectral feature value. The spatial coordinates corresponding to each spectral feature value in the target 3D point cloud data were determined. The associated spectral feature values ​​and their corresponding spatial coordinates were then mapped and fused point-by-point to generate a 3D point cloud data cube. Through refined processing of the 3D point cloud data and precise association of spectral features with spatial location, the representativeness of the target 3D point cloud data to the wound morphology and the accuracy of spectral feature and spatial coordinate fusion were improved, providing a high-quality data foundation for subsequent feature extraction and coupling processing. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A flowchart illustrating a method for predicting wound healing progress based on 3D point cloud reconstruction, provided in an embodiment of this application;

[0054] Figure 2 A schematic diagram illustrating a specific implementation of a wound healing progress prediction method based on 3D point cloud reconstruction provided in this application embodiment;

[0055] Figure 3 This is a schematic diagram of a wound healing progress prediction system based on 3D point cloud reconstruction, provided in an embodiment of this application. Detailed Implementation

[0056] To address the issues of imprecise spatial features, lack of spectral dynamics, and insufficient fusion correlation in existing technologies, this application provides a method for predicting wound healing progress based on 3D point cloud reconstruction. This method employs the following design concept: First, acquire the reflectance spectral data and initial 3D point cloud data of the wound. Extract features from the spectral data, including biological indicators such as hemoglobin concentration and tissue oxygenation status. Then, process the 3D point cloud data to obtain target data reflecting the three-dimensional morphology of the wound. Merge the biological indicator features with the spatial locations of the 3D data to form comprehensive data containing both 3D structure and biological information. Next, use a network capable of processing 3D data to extract structural features from different parts of the wound. Record the changes in biological indicators over time using a change tracking method, and deeply integrate the structural features and indicator changes. Finally, construct a cell migration-related chart based on the combined results. By analyzing the changes in the connections between different parts of the chart, obtain the predicted wound healing progress. This approach not only meticulously captures the structural differences of different parts of the wound and tracks the dynamic changes of biological indicators, but also strengthens the connection between the two in terms of location and time, thereby overcoming the shortcomings of existing solutions and improving the accuracy of prediction.

[0057] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] The core of this application is to provide a method for predicting wound healing progress based on 3D point cloud reconstruction. A flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:

[0059] S11. Obtain the reflectance spectral data and initial three-dimensional point cloud data of the wound tissue. Based on the reflectance spectral data, generate a spectral feature vector containing spectral feature values ​​of multiple biological indicators, including hemoglobin concentration and tissue oxygenation status.

[0060] Among them, reflectance spectral data is data formed by the reflection of light of different wavelengths by wound tissue, which can reflect the biological characteristics of the tissue; initial three-dimensional point cloud data is a collection of a large number of points describing the three-dimensional position of the wound surface, and each point contains spatial coordinate information; biological indicators are characteristics that reflect the state of wound tissue, among which hemoglobin concentration reflects the blood supply to the wound and tissue oxygenation status reflects the oxygen supply to the tissue; spectral feature values ​​are the specific values ​​of the corresponding biological indicators; spectral feature vector is a sequence formed by arranging the spectral feature values ​​of multiple biological indicators in a certain order, which is used to integrate biological indicator information.

[0061] In this embodiment, the reflectance spectral data of the wound tissue is first obtained through a spectral detection device, and the initial three-dimensional point cloud data of the wound is obtained through a three-dimensional scanning device. For example, for a diabetic foot ulcer, the 400-1000nm wavelength light data reflected by the spectral instrument is collected, and the spatial coordinates of tens of thousands of points on the wound are collected by a three-dimensional scanner. Then, a spectral feature vector is generated based on the reflectance spectral data. Specifically, the spectral segment corresponding to the hemoglobin concentration (e.g., 500-600nm) and the spectral segment corresponding to the tissue oxygenation state (e.g., 700-800nm) are first determined. The feature values ​​of these two spectral segments are extracted from the reflectance spectral data as spectral feature values. Then, these two spectral feature values ​​are arranged in the order of "hemoglobin concentration, tissue oxygenation state" to form a spectral feature vector. For example, the value corresponding to the hemoglobin concentration of a certain point is extracted from the reflectance spectral data as 120, and the value corresponding to the tissue oxygenation state is 85. After arranging, a spectral feature vector of [120, 85] is formed.

[0062] S12. Process the initial three-dimensional point cloud data to obtain target three-dimensional point cloud data that retains the spatial morphological features of the wound. Map and fuse the spectral feature values ​​in the spectral feature vector with the spatial coordinates of the target three-dimensional point cloud data point by point to generate a three-dimensional point cloud data cube.

[0063] Among them, the target 3D point cloud data is the point cloud data that retains key spatial morphological features such as wound edge and depth after processing the initial 3D point cloud data; spatial coordinates are the coordinate values ​​that describe the position of each point in the target 3D point cloud data in 3D space; point-by-point mapping fusion is the process of combining each spectral feature value with its corresponding spatial coordinates; the 3D point cloud data cube is a 3D data structure that integrates spatial coordinates and corresponding spectral feature values, containing both spatial morphological information and biological indicator information, for subsequent feature extraction.

[0064] In this embodiment, the initial three-dimensional point cloud data is first processed to obtain the target three-dimensional point cloud data. The initial data is divided into regions such as the wound edge and center according to the point cloud density. Points reflecting morphological changes are selected from each region, and these points are combined to form the target data. For example, 200 turning points at the wound edge and 300 concave / convex points in the center are selected from the initial point cloud to form the target three-dimensional point cloud data. Secondly, the spectral feature values ​​are associated with spatial coordinates. The location markers during the acquisition of the reflectance spectral data are first obtained, such as "upper left corner of the wound" or "center of the wound." The spatial coordinates in the markers are then extracted. The location information is mapped to each spectral feature value. Then, the distribution range and relative position of all spatial coordinates in the target 3D point cloud data are determined. The spatial coordinates closest to each spatial location information are found as the coordinates of the corresponding spectral feature value. For example, the spectral feature value of "upper left corner of the wound" corresponds to the (x1, y1, z1) coordinates closest to that position in the target data. Finally, all the associated spectral feature values ​​and corresponding spatial coordinates are fused point by point to form a 3D point cloud data cube, that is, each spatial coordinate is accompanied by a corresponding spectral feature value, forming a 3D structure containing spatial and biological information.

[0065] S13. Using a 3D convolutional neural network, spatial structure features in the three-dimensional point cloud data cube are extracted. The optical flow method is used to track the spectral feature changes corresponding to the spectral feature vectors at adjacent time points, and the spatial structure features and spectral feature changes are coupled.

[0066] Among them, 3D convolutional neural networks are networks that can process three-dimensional data and extract three-dimensional structural features; spatial structural features are information extracted from three-dimensional point cloud data cubes that describes the structural characteristics of different regions (such as edges and centers) of the wound; optical flow is a method for tracking data changes over time; spectral feature change is the difference between corresponding spectral feature values ​​in the spectral feature vectors of two adjacent time points; coupling processing is the process of associating and integrating spatial structural features with spectral feature change, used to combine spatial and dynamic biological information.

[0067] In this embodiment, a 3D convolutional neural network is first used to extract spatial structural features. A cube of three-dimensional point cloud data is input into the network, and the network extracts spatial structural features such as small-scale structural features at the edge of the wound and medium-scale structural features in the middle through layer-by-layer processing. For example, after network processing, a structural feature description of "rougher edges and flatter middle" is obtained. Secondly, the optical flow method is used to track the changes in spectral features. The spectral feature vectors of two adjacent time points are selected, and the difference between the corresponding spectral feature values ​​in the two vectors is calculated to obtain the changes in hemoglobin concentration and tissue oxygenation status. For example, the vector on the first day is [120, 85], and the vector on the second day is [110, 90]. The changes in hemoglobin concentration are calculated to be 110-120=-10, and the changes in tissue oxygenation status are 90-85=5. Finally, the spatial structural features and the changes in spectral features are coupled and integrated according to spatial location. For example, the structural feature of "rougher edges" is associated with the changes in hemoglobin concentration and tissue oxygenation status in the edge region.

[0068] S14. Based on the coupling processing results, a cell migration map of the wound tissue is constructed. By analyzing the changing trend of the connection strength of each node in the cell migration map, the healing progress prediction results in the wound healing process are obtained.

[0069] Among them, the cell migration map is a graph describing the association of key locations related to cell migration in wound tissue; nodes are points in the cell migration map that represent key locations of the wound (such as edge feature points, central feature points); connection strength is a description of the degree of connection between two nodes in the cell migration map; the trend of change is the change of connection strength over time; the healing progress prediction result is a judgment on the speed and trend of wound healing based on the analysis, used to assess the wound healing status.

[0070] In this embodiment, a cell migration map is first constructed based on the coupling processing results. Key locations of the wound are selected as nodes from the coupling results. Connections are established between adjacent nodes according to their spatial proximity to form an initial cell migration map. For example, five key locations are selected as nodes, and lines are connected between spatially adjacent nodes to form the map. Secondly, the node connection strength is determined. The connection strength between each pair of nodes is determined based on the correlation between the spatial structural features and spectral feature changes corresponding to each node. For example, if two adjacent nodes have similar structural features and consistent spectral change trends, the connection strength is set to 30, while the connection strength between nodes with large structural differences and opposite change trends is set to 10. Then, the values ​​of the connection strength of each node at different time points are recorded, and the change trends are obtained by comparing the values ​​at adjacent time points. For example, if a certain connection strength is 30 on the first day, 35 on the second day, and 40 on the third day, a "gradually increasing" trend is obtained. Finally, by analyzing the change trends of all connection strengths, the healing progress prediction results are obtained. For example, if most connection strengths show an increasing trend, it is judged that wound healing is progressing.

[0071] Figure 2 This is a schematic diagram illustrating a specific implementation of a wound healing progress prediction method based on 3D point cloud reconstruction, provided in an embodiment of this application. This specific implementation method is similar to... Figure 1 The process of the method described in [the previous section] is similar and will not be repeated here. Based on Figure 1 and Figure 2The following specific example is provided in this application: In a medical center, the pressure ulcer wound of patient A is examined: First, reflectance spectral data of the wound is acquired using a C-brand spectrometer, and initial three-dimensional point cloud data is acquired using a D-brand 3D scanner. From the reflectance spectral data, spectral feature values ​​of hemoglobin concentration are extracted from the 500-600nm spectral band, and spectral feature values ​​of tissue oxygenation status are extracted from the 700-800nm ​​spectral band. These are then combined sequentially to form a spectral feature vector. Next, the initial three-dimensional point cloud data is divided into edge and central regions according to density. 150 feature points are selected from the edge region, and 250 feature points are selected from the central region to form target data. The spectral feature values ​​are associated with the acquisition location markers, the spatial location information of the markers is extracted, and the distance between the markers and each spatial coordinate in the target data is calculated. The coordinates with the smallest distance are selected as the coordinates of the corresponding spectral feature values. After point-by-point fusion, a three-dimensional point cloud data cube is generated. Then, the cube was input into a 3D convolutional neural network to extract the spatial structural features of "obvious undulations" in the edge region and "relatively flat" in the central region. Spectral feature vectors from the first and second days were selected, and the change in hemoglobin concentration was calculated to be 117-125=-8, and the change in tissue oxygenation status was calculated to be 88-82=6. The spatial structural features and corresponding changes were then coupled and integrated. Finally, three nodes (1, 2, 3) at the edge and two nodes (4, 5) in the central region were selected, and connections were established between adjacent nodes to form a cell migration map. The connection strength between nodes 1 and 2 was determined to be 25, between nodes 2 and 3 to be 28, between nodes 3 and 4 to be 20, and between nodes 4 and 5 to be 22. The strength was recorded for five consecutive days: nodes 1 and 2 were 25, 27, 29, 31, and 33 respectively; nodes 2 and 3 were 28, 30, 32, 34, and 36 respectively. The strength of other nodes also showed an increasing trend, leading to the prediction that the wound healing process was accelerating.

[0072] By executing steps S11-S14, this embodiment acquires the reflectance spectral data and initial three-dimensional point cloud data of the wound surface, generating a spectral feature vector containing key biological indicators, providing a bioinformatics foundation for analysis. By processing the point cloud data and fusing spectral features with spatial coordinates, a three-dimensional point cloud data cube integrating spatial morphology and biological information is generated, providing high-quality data for subsequent processing. Fine spatial structural features are extracted using a 3D convolutional neural network, and the dynamic changes of biological indicators are captured and coupled using optical flow, achieving a deep integration of spatial and dynamic biological information. Finally, by constructing a cell migration map and analyzing changes in connection strength, the integrated information is transformed into a quantitative prediction of healing progress. Overall, this embodiment achieves a complete logic from data acquisition to accurate prediction, effectively capturing the spatial morphological differences of the wound surface and the dynamic changes of biological indicators, establishing a deep correlation between the two, and providing a comprehensive and accurate basis for predicting wound healing progress.

[0073] In one possible embodiment, S12, the initial three-dimensional point cloud data is processed to obtain target three-dimensional point cloud data that retains the spatial morphological features of the wound. The spectral feature values ​​in the spectral feature vector are then mapped and fused point-by-point with the spatial coordinates of the target three-dimensional point cloud data to generate a three-dimensional point cloud data cube, including:

[0074] Step 121: Divide the initial three-dimensional point cloud data into multiple wound areas according to the point cloud density, select feature points that reflect spatial morphological changes in each wound area, and combine the feature points to form the target three-dimensional point cloud data.

[0075] The initial 3D point cloud data consists of a large collection of points describing the three-dimensional position of the wound surface, with each point containing spatial coordinate information; the point cloud density is the number of points per unit space; the wound region is the different parts of the initial 3D point cloud data divided according to the point cloud density; the feature points are the key points in each wound region that can reflect the spatial morphological changes of the wound edge, such as the undulations and concavities in the middle; the target 3D point cloud data is the point cloud data that retains the key spatial morphological features of the wound after combining the feature points of each region.

[0076] In this embodiment, the initial three-dimensional point cloud data is first divided into multiple wound regions according to point cloud density. The density is determined by calculating the number of points per unit volume, and points with similar densities are grouped into one region. For example, in the initial three-dimensional point cloud data of diabetic foot ulcer wounds, the edge part has a large number of points and high density per unit volume, while the middle part has a small number of points and low density, thus dividing it into edge region and middle region. Secondly, feature points reflecting spatial morphological changes are selected in each wound region. In the edge region, turning points reflecting the degree of edge curvature are selected, and in the middle region, the highest and lowest points reflecting the surface unevenness are selected. Finally, all feature points are combined to form target three-dimensional point cloud data. For example, 180 turning points are selected from the edge region, and 220 uneven points are selected from the middle region to form target data containing 400 feature points.

[0077] Step 122: Obtain the location markers of the reflectance spectral data, associate each spectral feature value in the spectral feature vector with the corresponding location marker, and determine the spatial coordinates of each spectral feature value in the target 3D point cloud data.

[0078] Among them, the location marker of the reflectance spectral data is the identification information that records the location of the reflectance spectral data acquisition, indicating that the spectral data comes from the specific location of the wound; the spectral feature vector is a sequence of spectral feature values ​​containing multiple biological indicators, such as hemoglobin concentration and tissue oxygenation status; the spectral feature value is the specific value of the corresponding biological indicator; the spatial coordinate is the position information of each feature point in the target 3D point cloud data in 3D space; the association is to establish a correspondence between each spectral feature value and the corresponding location marker, and determining the corresponding spatial coordinate is to find the spatial coordinate in the target 3D point cloud data that is closest to the location marker.

[0079] In this embodiment, the location markers of the reflectance spectral data are first obtained. These markers are used to record the wound location when the spectrum is collected, such as "wound edge area M1" and "wound center area M2". Secondly, each spectral feature value in the spectral feature vector is associated with the corresponding location marker to clarify which marker each spectral feature value of hemoglobin concentration and tissue oxygenation status comes from. For example, the spectral feature value 120 of hemoglobin concentration is associated with "edge area M1", and the spectral feature value 85 of tissue oxygenation status is associated with "center area M2". Finally, the spatial coordinates corresponding to each spectral feature value in the target three-dimensional point cloud data are determined, the spatial position information of the location marker is extracted, the distance between this information and all spatial coordinates in the target data is calculated, and the coordinate with the smallest distance is selected. For example, the spatial position information of "edge area M1" has the smallest distance to (x2, y2, z2) in the target data, and this coordinate is the coordinate of the corresponding spectral feature value.

[0080] Step 123: Map and fuse all associated spectral feature values ​​with their corresponding spatial coordinates point by point to generate a three-dimensional point cloud data cube.

[0081] Among them, the associated spectral feature value is the spectral feature value that establishes a correspondence with the location marker; the corresponding spatial coordinate is the spatial coordinate in the target 3D point cloud data that matches the spectral feature value; point-by-point mapping fusion is the process of combining each associated spectral feature value with its corresponding spatial coordinate; the 3D point cloud data cube is a 3D data structure generated through this fusion process, in which each spatial coordinate is accompanied by a corresponding spectral feature value, and also contains the spatial morphology information and biological indicator information of the wound.

[0082] In this embodiment, each associated spectral feature value and its corresponding spatial coordinates are first mapped and fused point by point. A corresponding spectral feature value is added to each spatial coordinate as additional information. For example, the spatial coordinates (x3, y3, z3) are combined with the associated hemoglobin concentration spectral feature value 120, and (x4, y4, z4) are combined with the associated tissue oxygenation status spectral feature value 85. Then, all the points that have been mapped and fused are integrated together to form a three-dimensional point cloud data cube. Each point in the cube contains both spatial coordinates describing the location and spectral feature values ​​reflecting the biological state, integrating the spatial morphology and biological indicator information of the wound.

[0083] This application provides the following specific example: The initial three-dimensional point cloud data of the pressure ulcer wound of patient A is processed to calculate the number of points per unit area to determine the point cloud density. There are 80 points per square centimeter in the edge part and 40 points per square centimeter in the middle part, which are then divided into edge region and middle region. 160 turning points reflecting the irregular shape of the edge are selected in the edge region, and 240 concave and convex points reflecting the surface undulation are selected in the middle region, which are combined to form target three-dimensional point cloud data containing 400 feature points. The location markers "Edge Area Marker A1" and "Central Area Marker A2" for the wound reflectance spectral data are obtained. The hemoglobin concentration spectral feature value generated in S11 is associated with "Edge Area Marker A1", and the tissue oxygenation status spectral feature value is associated with "Central Area Marker A2". The spatial location information of A1 and A2 is extracted, and their distances to each spatial coordinate in the target data are calculated. The distance between A1 and coordinates (x3, y3, z3) is 0.2 cm (less than the distance to other coordinates), and the distance between A2 and coordinates (x4, y4, z4) is 0.3 cm (less than the distance to other coordinates). These two coordinates are determined to be the spatial coordinates of the corresponding spectral feature values. The hemoglobin concentration spectral feature value associated with "Edge Area Marker A1" is fused with (x3, y3, z3), and the tissue oxygenation status spectral feature value associated with "Central Area Marker A2" is fused with (x4, y4, z4). The feature values ​​and coordinates associated with other markers are processed in the same way. Finally, all fused points are integrated to generate a three-dimensional point cloud data cube of the wound.

[0084] By executing steps 121-123, this embodiment of the application divides the wound area according to density and selects feature points, reducing redundant data while preserving key spatial morphological features, thus making the target three-dimensional point cloud data more representative. By associating spectral feature values ​​with location markers and matching spatial coordinates, it achieves a precise correspondence between biological indicator information and spatial location. Finally, by point-by-point mapping and fusion, a three-dimensional point cloud data cube is generated, organically integrating the spatial morphological information and biological indicator information of the wound. The synergistic effect of these three steps provides a unified, complete, and high-quality data foundation for subsequent extraction of spatial structural features and analysis of dynamic changes in biological indicators, improving the effectiveness and accuracy of data processing.

[0085] In one possible embodiment, step 122, associating each spectral feature value in the spectral feature vector with its corresponding location marker and determining the spatial coordinates of each spectral feature value in the target 3D point cloud data, includes:

[0086] a1. Extract the spatial location information contained in the location markers, and map each spectral feature value to the spatial location information to form associated spectral feature values.

[0087] Among them, the location marker is an identifier that records the location of the reflectance spectral data acquisition, including the spatial location information of the location in the wound; the spatial location information is data describing the specific location of the acquisition point in three-dimensional space; the spectral feature value is a specific value reflecting biological indicators such as hemoglobin concentration and tissue oxygenation status; the associated spectral feature value is data formed by establishing a correspondence between each spectral feature value and its corresponding spatial location information, which includes both biological indicator information and its acquisition location information.

[0088] In this embodiment, the spatial location information contained in the location markers is first extracted. This information usually records the location of the spectral data acquisition point in coordinate form. For example, the spatial location information (x5, y5, z5) is extracted from the marker "edge C point" of the diabetic foot ulcer wound. Secondly, each spectral feature value is mapped to the corresponding spatial location information, that is, it is determined which spectral feature value comes from which location. For example, the spectral feature value 115 of hemoglobin concentration is mapped to (x5, y5, z5) of "edge C point", and the spectral feature value 82 of tissue oxygenation status is mapped to (x6, y6, z6) of "middle D point", forming the associated spectral feature values.

[0089] a2. Determine the distribution range of all spatial coordinates and the relative positional relationship of each spatial coordinate in the target 3D point cloud data.

[0090] Among them, the target three-dimensional point cloud data is point cloud data that preserves the key spatial morphological features of the wound and is composed of multiple spatial coordinates; spatial coordinates are numerical values ​​that describe the position of a point in three-dimensional space; the distribution range refers to the range formed by the maximum and minimum values ​​of all spatial coordinates in the three dimensions of x, y, and z; the relative positional relationship refers to the distance and direction relationship between each spatial coordinate, which is used to describe the positional relationship between points.

[0091] In this embodiment, the distribution range of all spatial coordinates in the target 3D point cloud data is first determined by finding the maximum and minimum values ​​in the three dimensions of x, y, and z. For example, the maximum value of x in the target data is 15 and the minimum value is 0, the maximum value of y is 12 and the minimum value is 1, and the maximum value of z is 5 and the minimum value is 0, thus determining the distribution range. Secondly, the relative positional relationship of each spatial coordinate is determined by calculating the distance and direction between any two coordinates. For example, the distance between coordinates (2,3,1) and (6,5,1) is calculated to be 4.47, and the direction is 4 units along the positive x-axis and 2 units along the positive y-axis from the former to the latter.

[0092] a3. Based on the relative positional relationship, find the spatial coordinates with the smallest distance value from each spatial positional information within the distribution range, and determine each spatial coordinate as the spatial coordinates corresponding to the spectral feature value in the target three-dimensional point cloud data.

[0093] Among them, relative positional relationship refers to the distance and direction relationship between spatial coordinates in the target 3D point cloud data; distribution range refers to the maximum and minimum value range of all spatial coordinates in the x, y, and z dimensions; spatial position information refers to the spectral acquisition point position data extracted from the position markers; distance value refers to the distance between the spatial position information and each spatial coordinate in the target 3D point cloud data; the spatial coordinate with the minimum distance value refers to the spatial coordinate that is closest to the spatial position information within the distribution range, and is used as the corresponding coordinate of the spectral feature value in the target data.

[0094] In this embodiment, firstly, based on the relative positional relationship determined in a2, the distance between the spatial position information and each spatial coordinate in the target 3D point cloud data is calculated within the distribution range. For example, the distance between (2,3,1) of "edge region Q1" and the coordinate (2,3,1) in the target data is 0, and the distance between (3,4,1) and the coordinate is 1.41. Secondly, the smallest distance value is found among all distance values, and the corresponding spatial coordinate is the matching result. For example, (2,3,1) has the smallest distance to (2,3,1) in the target data, so this coordinate is determined as the spatial coordinate corresponding to the spectral feature value associated with "edge region Q1". Finally, the above process is repeated for each spatial position information to obtain the spatial coordinates corresponding to all spectral feature values.

[0095] This application provides the following specific example: The location markers of the reflectance spectral data of the pressure ulcer wound of patient A are processed. Spatial location information (2,3,1) is extracted from the "edge region Q1" and (6,5,1) is extracted from the "central region Q2". The spectral feature values ​​of hemoglobin concentration and tissue oxygenation status are respectively mapped to these two pieces of information to form associated spectral feature values. The target three-dimensional point cloud data of the wound is analyzed to determine the distribution range of spatial coordinates as x (0-14), y (0-11), and z (0-4), and the coordinates (...) are calculated. The relative distance between (2,3,1) and (6,5,1) is 4.47. For the spatial location information (2,3,1) of “edge region Q1”, the distance between it and each spatial coordinate is calculated within the distribution range of the target three-dimensional point cloud data. The distance to coordinate (2,3,1) is 0, and the distance to other coordinates is greater than 0. Therefore, (2,3,1) is determined as the spatial coordinate of the corresponding spectral feature value. The same calculation is performed on (6,5,1) of “central region Q2”, and the coordinate with the smallest distance (6,5,1) is found as the corresponding coordinate.

[0096] By executing steps a1 to a3, this embodiment of the application extracts spatial information of location markers and associates them with spectral feature values, thus assigning clear location attributes to biometrics; by determining the spatial range and coordinate relationships of the target 3D point cloud data, it provides a spatial reference for coordinate matching; and by calculating the minimum distance, it achieves a precise correspondence between spectral feature values ​​and spatial coordinates. These three steps work synergistically to ensure the accurate association between biometric information and spatial morphological information, laying a reliable foundation for subsequent data fusion processing and improving the accuracy and effectiveness of overall data integration.

[0097] In one possible embodiment, S14, based on the coupling processing results, a cell migration map of the wound tissue is constructed. By analyzing the changing trends of the connection strength of each node in the cell migration map, the healing progress prediction results in the wound healing process are obtained, including:

[0098] Step 141: Select multiple key locations of the wound tissue from the coupling processing results, and use each key location as a node in the cell migration map to determine the proximity relationship of each node in the wound tissue.

[0099] Among them, the coupling processing result is the information after the correlation and integration of spatial structural features and spectral feature changes; key locations are locations in the wound tissue that are of great significance to healing analysis, such as edge turning points, central feature points, etc.; nodes are points representing key locations in the cell migration map; proximity relationship is the degree of proximity of each node in the wound space, used to describe the spatial distance relationship between nodes.

[0100] In this embodiment, firstly, multiple key locations of the wound tissue are selected from the coupling processing results. These locations are typically typical areas that reflect healing changes. For example, in the coupling results of a diabetic foot ulcer wound, three inflection points at the edge and two concave-convex points in the middle are selected as key locations. Secondly, each key location is treated as a node in a cell migration map, and each node is assigned a label for differentiation. Finally, the proximity relationship of each node in the wound tissue is determined by calculating the spatial distance between nodes. For example, the distance between node 1 and node 2 is measured to be 5 mm, and the distance between node 2 and node 3 is 4 mm, thereby determining which node locations are closer.

[0101] Step 142: Based on proximity relationships, establish connections between mutually adjacent nodes to generate a cell migration map.

[0102] Among them, proximity refers to the degree of closeness between nodes in the wound space; connection is the linking of adjacent nodes with lines in the cell migration diagram; the cell migration diagram is a graphic that shows the relationship between key locations in the wound through nodes and connections, and is used to intuitively present the connection between locations.

[0103] In this embodiment of the application, firstly, based on the proximity relationship determined in step 141, it is determined which nodes are adjacent to each other. For example, nodes with a distance of less than 7 mm are set as adjacent nodes, and nodes 1 and 2 (distance 6 mm) and nodes 2 and 3 (distance 5 mm) are adjacent nodes. Secondly, connections are established between adjacent nodes, and these nodes are connected in the diagram with lines. Finally, a cell migration diagram is generated, which contains all nodes and the connections between them. For example, nodes 1 to 2 to 3 form a series of connections, and nodes 3 to 4 and 4 to 5 form another series of connections.

[0104] Step 143: Determine the degree of connection between each pair of nodes as the node connection strength based on the first correlation value of the spatial structural features corresponding to each node and the second correlation value of the spectral feature change.

[0105] Among them, spatial structural features are information reflecting the structural characteristics of each location of the wound; the first correlation value is a numerical value describing the similarity of the spatial structural features corresponding to two nodes; the spectral feature change is the difference in spectral features at adjacent time points; the second correlation value is a numerical value describing the consistency of the spectral feature change corresponding to two nodes; the node connection strength is a numerical value reflecting the tightness of the connection between two nodes, obtained by combining the first correlation value and the second correlation value.

[0106] In this embodiment, the first correlation value of the spatial structural features corresponding to each node is calculated firstly by comparing the similarity of the structural features. For example, the spatial structural features of node 1 and node 2 are both "rough edges", and the first correlation value is 8 (range 0-10). The structural features of node 3 and node 4 are "rough edges" and "flat center", respectively, and the first correlation value is 3. Secondly, the second correlation value of the spectral feature change is calculated by comparing the consistency of the change trend. For example, the hemoglobin concentration changes of node 1 and node 2 are both decreasing, and the second correlation value is 7 (range 0-10). The changes of node 3 and node 4 are decreasing and increasing, respectively, and the second correlation value is 2. Finally, the first correlation value and the second correlation value are combined. For example, the two are added together and averaged to obtain the node connection strength. The strength from node 1 to node 2 is (8+7)÷2=7.5, and the strength from node 3 to node 4 is (3+2)÷2=2.5.

[0107] Step 144: Record the values ​​of the connection strength of each node at different time points to obtain the changing trend of the connection strength of each node.

[0108] Among them, different time points refer to multiple times when wound data is collected, such as the same time every day; the node connection strength value is a specific value describing the tightness of the connection between nodes; the trend of change is the direction and magnitude of the increase or decrease of the connection strength value over time, such as gradually increasing or gradually decreasing.

[0109] In this embodiment, the connection strength values ​​of each node at different time points are first recorded in chronological order, for example, once a day. The strength of node 1 to node 2 is 7.5 on day 1, 8.0 on day 2, and 8.5 on day 3. Secondly, the values ​​of adjacent time points are compared to calculate the changes. For example, the strength increases by 0.5 from day 1 to day 2, and by 0.5 from day 2 to day 3. Finally, the trend of the connection strength of each node is obtained. For example, the strength of node 1 to node 2 shows a gradual increasing trend, and the strength of node 3 to node 4 changes from 3.5 to 3.8 and then to 4.0, also showing an increasing trend.

[0110] Step 145: Based on the changing trend, use a preset prediction model to determine the predicted healing progress in the wound healing process.

[0111] Among them, the trend of change is the direction and magnitude of the increase or decrease of the node connection strength over time; the preset prediction model is a model set in advance to calculate the healing progress based on the trend of change, and the model is built based on known healing case data; the healing progress prediction result is a judgment on the speed and stage of wound healing, such as accelerated healing, stable healing, etc.

[0112] In this embodiment, the changing trends of the connection strength of each node are first collected and organized into trend data. For example, all connection strengths show a gradual increasing trend, and the increase is relatively stable. Secondly, these trend data are input into a preset prediction model. The model is based on the characteristics of the trend, such as an increase indicating the progress of healing. Finally, the model outputs the prediction result of the healing progress in the wound healing process. For example, if the model determines that the trend meets the characteristics of accelerated healing, it outputs the result "the healing progress is accelerating".

[0113] This application provides the following specific example: The pressure ulcer wound of patient A is treated. From the coupling results, three edge nodes (1, 2, 3) and two middle nodes (4, 5) are selected. Measurements determine that the distance between nodes 1 and 2 is 6 mm, between nodes 2 and 3 is 5 mm, between nodes 3 and 4 is 8 mm, and between nodes 4 and 5 is 7 mm, thus determining the proximity relationships. Based on these proximity relationships, connections are established between nodes 1 and 2, between nodes 2 and 3, between nodes 3 and 4, and between nodes 4 and 5, generating a cell migration map. The connection strength is calculated: for nodes 1 and 2, it is (8+7)÷2=7.5; for nodes 2 and 3, it is (7+6)÷2. =6.5, nodes 3 to 4 are (4+3)÷2=3.5, nodes 4 to 5 are (5+4)÷2=4.5; the intensity was recorded for 3 consecutive days: day 1 (7.5, 6.5, 3.5, 4.5), day 2 (8.0, 7.0, 3.8, 4.8), day 3 (8.5, 7.5, 4.0, 5.0). By comparison, it was found that all intensities showed an increasing trend. The trend was input into the preset model, and the model combined the pattern of "continuous increase in connection strength corresponding to healing progress" in historical cases to output the prediction result that "the healing progress of the wound is accelerating and is in the active healing stage".

[0114] By executing steps 141-145, this embodiment of the application provides basic nodes and spatial connections for cell migration mapping by selecting key locations as nodes and determining proximity relationships; it visualizes the spatial associations of key locations on the wound by establishing connections to generate graphics; it quantifies the connections between nodes by calculating connection strength, facilitating dynamic comparison; it captures the dynamic trends of connections between nodes by recording strength changes, providing dynamic data for judging the healing process; and finally, it transforms the trends into intuitive healing progress prediction results through a preset model. These five steps work synergistically to integrate the spatial and biological information of the wound into analyzable and predictable intuitive conclusions, achieving effective tracking and prediction of the wound healing process and providing valuable reference for clinical treatment.

[0115] In one possible embodiment, step 143, determining the node connection strength based on the first correlation value of the spatial structural features corresponding to each node and the second correlation value of the spectral feature change, includes:

[0116] b1. Extract the spatial structure features and spectral feature changes corresponding to each node.

[0117] In this context, nodes are points representing key locations of the wound in the cell migration map; spatial structural features are information reflecting the structural characteristics of the wound at the node's location, such as edge smoothness and central unevenness; spectral feature changes are the differences between spectral feature values ​​at two adjacent time points, reflecting changes in biological indicators such as hemoglobin concentration and tissue oxygenation status; extraction is the process of obtaining the spatial structural features and spectral feature changes corresponding to each node from the coupling processing results, generating two types of feature data corresponding to each node.

[0118] In this embodiment, firstly, for each node in the cell migration map, its corresponding spatial structural features are extracted from the coupling processing results. For example, in the coupling results of diabetic foot ulcer wounds, the spatial structural features of node 1 are extracted as "rough edges and obvious undulations", and those of node 2 are "rough edges and small undulations". Secondly, the spectral feature change corresponding to each node is extracted, that is, the difference in spectral feature values ​​between adjacent time points is calculated. For example, the hemoglobin concentration of node 1 is 120 on day 1 and 110 on day 2, and the change is 110 minus 120 equals -10. The tissue oxygenation status is 85 on day 1 and 90 on day 2, and the change is 90 minus 85 equals +5. Therefore, the spectral feature change of node 1 is (-10, +5). Similarly, the spectral feature change of node 2 is extracted as (-8, +6).

[0119] b2. Compare the spatial structural features of every two nodes to obtain the first distance similarity value from the edge range of the wound tissue to every two nodes, and the second distance similarity value from the middle range of the wound tissue.

[0120] Among them, spatial structural features are information reflecting the characteristics of the wound structure at the location of the node; each pair of nodes is any pair of adjacent nodes in the cell migration map; the edge range is the spatial range of the edge region of the wound; the first distance similarity value is a numerical value describing the degree of similarity in distance from two nodes to the edge range of the wound; the central range is the spatial range of the central region of the wound; and the second distance similarity value is a numerical value describing the degree of similarity in distance from two nodes to the central range of the wound. Both are used to quantify the similarity of spatial structural features.

[0121] In this embodiment, the spatial structural features of each pair of nodes are first compared, with a focus on analyzing their distances to the edge of the wound. By calculating the actual distance difference between the two nodes and the edge, a first distance similarity value is obtained. For example, if node 1 is 5 mm from the edge and node 2 is 6 mm from the edge, the distance difference is 6 minus 5 equals 1 mm. The similarity value is set to a range of 0 to 1, with the smaller the difference, the closer the similarity value is to 1. The calculation method is 1 minus the difference value divided by the larger distance, i.e., 1 minus 1 divided by 6 is approximately equal to 0.83. Therefore, the first distance similarity value is 0.83. Next, the distances of these two nodes to the middle of the wound are analyzed, and the distance difference is calculated to obtain a second distance similarity value. For example, if node 1 is 10 mm from the middle and node 2 is 12 mm from the middle, the difference is 12 minus 10 equals 2 mm. The calculation method is 1 minus 2 divided by 12, which is approximately equal to 0.83. Therefore, the second distance similarity value is 0.83.

[0122] b3. Determine the first degree of association value based on the first distance similarity value and the second distance similarity value.

[0123] The first distance similarity value describes the degree of similarity between the distances of two nodes to the edge of the wound; the second distance similarity value describes the degree of similarity between the distances of two nodes to the middle of the wound; and the first correlation value is obtained by combining the first and second distance similarity values, reflecting the degree of correlation between the spatial structural features of the two nodes.

[0124] In this embodiment of the application, the first distance similarity value and the second distance similarity value obtained in step b2 are first obtained. For example, the first distance similarity value of two nodes is 0.83 and the second distance similarity value is 0.83. Then, the first correlation value is obtained by combining the two values ​​in a preset way. For example, the two similarity values ​​are added together and the average value is taken. The calculation process is (0.83 plus 0.83) divided by 2 equals 0.83. Therefore, the first correlation value of the two nodes is 0.83.

[0125] b4. Compare the changes in spectral characteristics between each pair of nodes to obtain consistency information on changes in hemoglobin concentration and tissue oxygenation status between each pair of nodes, and determine the second correlation degree value based on the consistency information.

[0126] Among them, the spectral feature change is the difference in spectral feature values ​​between adjacent time points, including changes in hemoglobin concentration and tissue oxygenation status; each pair of nodes is any pair of adjacent nodes in the cell migration map; consistency information is information describing whether the spectral feature changes of two nodes are consistent in terms of trend and magnitude; the second correlation value is obtained based on the consistency information and reflects the degree of correlation between the spectral feature changes of two nodes.

[0127] In this embodiment, the spectral characteristic changes of every two nodes are first compared to analyze their consistency in hemoglobin concentration changes. For example, the hemoglobin concentration change of node 1 is -10 (decreasing) and that of node 2 is -8 (decreasing), showing a consistent trend and similar magnitude. Secondly, the consistency of tissue oxygenation status changes is analyzed. For example, the change of node 1 is +5 (increasing) and that of node 2 is +6 (increasing), showing a consistent trend and magnitude. Then, the consistency information from these two aspects is combined to obtain the consistency level. For example, a consistent trend is scored as 0.5 points, similar magnitudes as 0.4 points, and the total consistency information corresponds to 0.9 points. Finally, the consistency information is converted into a second correlation degree value, for example, the score of 0.9 is directly used as the second correlation degree value.

[0128] b5. Based on the first and second correlation values, determine the tightness of the connection between each pair of nodes, and use the tightness as the node connection strength.

[0129] Among them, the first correlation degree value is a numerical value that reflects the correlation degree of the spatial structural features of two nodes; the second correlation degree value is a numerical value that reflects the correlation degree of the spectral feature changes of two nodes; the tightness is an attribute that describes the closeness of the connection between two nodes; and the node connection strength is a numerical value that quantifies the tightness and is used to represent the closeness of the connection between two nodes.

[0130] In this embodiment, firstly, the first correlation degree value obtained in step b3 and the second correlation degree value obtained in step b4 are obtained. For example, the first correlation degree value of two nodes is 0.83 and the second correlation degree value is 0.9. Secondly, the two values ​​are combined in a preset way to determine the tightness of the connection between the two nodes. For example, the two values ​​are added together and the average value is taken. The calculation process is (0.83 plus 0.9) divided by 2 equals 0.865. Finally, the value of the tightness is determined as the node connection strength, that is, the connection strength between the two nodes is 0.865.

[0131] This application provides the following specific example: Nodes 1 and 2 in the cell migration map of patient A's pressure ulcer wound are processed: First, the spatial structural features of node 1 are extracted from the coupling results as "irregular edges with protrusions". The spectral feature changes are obtained by calculating the values ​​at adjacent time points: hemoglobin concentration: 115 on day 1, 108 on day 2, change: 108-115=-7; tissue oxygenation status: 80 on day 1, 84 on day 2, change: 84-80=+4. The spatial structural features of node 2 are "irregular edges with depressions". The spectral feature changes are: hemoglobin concentration: 112 on day 1, 106 on day 2, change: 106-112=-6; tissue oxygenation status: 81 on day 1, 86 on day 2, change: 86-81=+5. Next, spatial structural features were compared. Node 1's distance to the edge was 4 mm, and Node 2's distance to the edge was 5 mm, a difference of 1 mm, resulting in a first distance similarity value of 0.8. Node 1's distance to the center was 9 mm, and Node 2's distance to the center was 11 mm, a difference of 2 mm, resulting in a second distance similarity value of 0.82. The combined first correlation score was 0.81. Then, spectral feature changes were compared. Both nodes showed a decrease in hemoglobin concentration with similar magnitudes, and an increase in tissue oxygenation with similar magnitudes, corresponding to a consistency score of 0.85, thus determining a second correlation score of 0.85. Finally, combining the two correlation scores, the connection strength between Node 1 and Node 2 was determined to be 0.83.

[0132] By executing steps b1 to b5, this embodiment of the application provides basic data for node association analysis by extracting the spatial structural features and spectral feature changes of nodes; it quantifies the degree of spatial structural association by calculating the similarity values ​​of distances to the edge and central regions; it obtains the degree of association of biological indicator changes by analyzing the consistency of spectral changes; and finally, it determines the node connection strength by combining the two types of association strength. These five steps work synergistically to comprehensively quantify the association between nodes from both spatial and biological dimensions, making the calculation of connection strength more scientific and comprehensive, and providing a reliable quantitative basis for subsequent analysis of connection strength change trends and prediction of wound healing progress.

[0133] In one possible embodiment, S13, a 3D convolutional neural network is used to extract spatial structure features from the 3D point cloud data cube, and optical flow is used to track the spectral feature changes corresponding to the spectral feature vectors at adjacent time points. The spatial structure features and spectral feature changes are coupled, including:

[0134] Step 131: Using a 3D convolutional neural network, the three-dimensional point cloud data cube is divided into layers according to the level from the edge of the wound tissue, the middle of the wound tissue, and the entire wound area, and processed separately to extract structural information at different levels. Based on the structural information at different levels, spatial structural features are generated.

[0135] Among them, 3D convolutional neural networks are networks that can process three-dimensional data and extract its features; three-dimensional point cloud data cubes are three-dimensional data structures that integrate spatial coordinates and spectral feature values; the edge of the wound tissue is the outer part of the wound, the middle part is the middle part of the wound, and the entire wound range is the complete area including the edge and the middle; hierarchical division is to partition the data cube in the order of edge, middle, and the entire range; structural information is the details reflecting the wound morphology in each level: such as the smoothness of the edge and the concavity and convexity of the middle; spatial structural features are the features that describe the overall spatial morphology of the wound after integrating the structural information of each level.

[0136] In this embodiment, a 3D convolutional neural network is first used to divide the three-dimensional point cloud data cube into layers from the edge and center of the wound to the entire area. For example, for the three-dimensional point cloud data cube of a diabetic foot ulcer, the edge region, the center region, and the entire wound area are first divided. Then, the 3D convolutional neural network is used to process the partitioned data of each layer and extract the structural information of each layer. For example, the "edge undulation degree" is extracted from the edge region, the "center depression depth" is extracted from the center region, and the "overall area change" is extracted from the entire area. Finally, the structural information of each layer is integrated to generate spatial structural features, such as combining them into a feature description of "obvious edge undulation, shallow center depression, and reduced overall area".

[0137] Step 132: Use the optical flow method to select the spectral feature vectors of each two adjacent time points, compare the corresponding spectral feature values ​​in the spectral feature vectors of each two adjacent time points, and obtain multiple spectral feature changes.

[0138] Among them, optical flow is a method for tracking data changes over time and used to determine the correspondence between data at adjacent time points. Adjacent time points refer to two consecutive times when data is collected, such as the first day and the second day. The spectral feature vector is a sequence containing biological indicators such as hemoglobin concentration and tissue oxygenation status. The spectral feature value is the specific value of each biological indicator in the vector. The spectral feature change is the difference in the spectral feature values ​​corresponding to adjacent time points, reflecting the changes in biological indicators.

[0139] In this embodiment, firstly, the spectral feature vectors of each two adjacent time points are selected using the optical flow method. For example, the spectral feature vectors of the first and second days of the diabetic foot ulcer wound are selected to ensure that the vectors of the two time points correspond in spatial location. Secondly, the corresponding spectral feature values ​​in these two vectors are compared, that is, the values ​​of the same biological indicator are compared. For example, the hemoglobin concentration in the vector on the first day is 120 and the tissue oxygenation status is 85, while on the second day they are 110 and 90, respectively. Finally, the differences in the corresponding feature values ​​are calculated to obtain multiple spectral feature changes. For example, the change in hemoglobin concentration is 110 minus 120, which equals -10, and the change in tissue oxygenation status is 90 minus 85, which equals +5.

[0140] Step 133: Correlate and integrate the spatial structure features with all spectral feature variations to obtain the coupling processing result.

[0141] Among them, spatial structural features reflect the characteristics of different layers of wound morphology; spectral feature changes are the differences in biological indicators at adjacent time points; correlation integration is the process of combining spatial structural features and spectral feature changes according to spatial location; the coupling processing result is data that includes both spatial morphological information and dynamic changes in biological indicators after integration, which is used for subsequent analysis of the relationship between the two.

[0142] In this embodiment, the spatial location of each layer (edge, center) in the spatial structural features is first determined. For example, the edge region corresponds to the range of x1 to y1 in three-dimensional coordinates, and the center region corresponds to the range of x2 to y2. Secondly, the spectral feature changes are associated with each layer according to their spatial location. For example, the spectral changes in the edge region correspond to the edge structural information, and the changes in the center correspond to the center structural information. Finally, the spatial structural features and spectral feature changes at the corresponding locations are associated and integrated to form a coupled processing result, such as the integrated data of "obvious edge undulations (spatial features), hemoglobin concentration decreased by 5 (spectral changes)" and "shallow central depression (spatial features), tissue oxygenation status increased by 3 (spectral changes)".

[0143] This application provides the following specific example: Treatment of pressure ulcer wounds in patient A: First, a 3D convolutional neural network is used to divide the three-dimensional point cloud data cube into layers according to the edge (4 mm wide), center, and the entire wound surface. Structural information is extracted from the edge ("irregular with 3 protrusions"), the center ("2 shallow depressions"), and the overall wound surface ("elliptical shape with medium area"), and integrated into spatial structural features. Second, optical flow is used to select spectral feature vectors for the first and second days. The vector for the edge region on the first day is [hemoglobin concentration 115, tissue oxygenation status 82], and the vector for the center region is... The values ​​are [112, 81]. On the second day, the edge region is [108, 86] and the middle region is [106, 86]. The calculated spectral changes in the edge region are 108 minus 115 equals -7 and 86 minus 82 equals +4. In the middle region, the changes are 106 minus 112 equals -6 and 86 minus 81 equals +5. Finally, the spatial structure features are correlated with the spectral changes in the corresponding regions. The edge region has "irregular 3 protrusions" corresponding to (-7, +4), and the middle region has "2 shallow depressions" corresponding to (-6, +5), thus obtaining the coupling processing result.

[0144] By executing steps 131 to 133, this embodiment of the application extracts spatial structural features hierarchically using a 3D convolutional neural network, comprehensively capturing the morphological details of the wound edge, center, and overall structure, making the spatial information more detailed. By accurately matching spectral data at adjacent time points using optical flow, the calculated changes in spectral features clearly reflect the dynamic changes in biological indicators, providing dynamic biological information for analysis. Finally, by integrating the two through spatial location association, the generated coupled processing result simultaneously contains morphological and biological dynamic information, realizing the organic combination of spatial and biological information, and providing a comprehensive and integrated data foundation for subsequent analysis of node association and prediction of healing progress.

[0145] In one possible embodiment, S11, based on reflectance spectral data, generating a spectral feature vector containing spectral feature values ​​of multiple biological indicators, including:

[0146] Step 111: Determine the corresponding spectral segment for each biomarker in the reflectance spectral data, where hemoglobin concentration corresponds to the first spectral segment and tissue oxygenation status corresponds to the second spectral segment.

[0147] Among them, bioindicators are characteristics reflecting the state of wound tissue, including hemoglobin concentration and tissue oxygenation status; reflectance spectral data are data formed by the reflection of light of different wavelengths by wound tissue; spectral segments are specific wavelength ranges in the reflectance spectral data; the first spectral segment is the wavelength range corresponding to hemoglobin concentration; the second spectral segment is the wavelength range corresponding to tissue oxygenation status; the result generated in this step is two definite spectral segments, which are used to extract the feature values ​​of the corresponding bioindicators in subsequent steps.

[0148] In this embodiment, the spectral segment corresponding to each biomarker in the reflectance spectral data is first determined. Based on the different reflectance characteristics of different biomarkers, for example, the reflectance of hemoglobin in the wavelength range of 500-600 nm is closely related to its concentration, and the tissue oxygenation state has a specific reflectance pattern in the wavelength range of 700-800 nm. Therefore, 500-600 nm is determined as the first spectral segment (corresponding to hemoglobin concentration), and 700-800 nm is determined as the second spectral segment (corresponding to tissue oxygenation state). For example, the reflectance spectral data of diabetic foot ulcers is divided into two spectral segments according to this pattern.

[0149] Step 112: Obtain the numerical value reflecting the hemoglobin concentration based on the spectral information of the first spectral band, and use it as the first spectral characteristic value.

[0150] Among them, the first spectral band is the wavelength range corresponding to the hemoglobin concentration; the spectral information is the light reflection data within the first spectral band; the value reflecting the hemoglobin concentration is a value calculated from the spectral information that can reflect the amount of hemoglobin; the first spectral characteristic value is a specific name for this value, used to represent the characteristics of the hemoglobin concentration.

[0151] In this embodiment, the spectral information of the first spectral band, namely the reflectance data in the range of 500-600 nanometers, such as the reflectance of each wavelength in this range, is first obtained; then the characteristics of this spectral information are calculated, such as calculating the average reflectance of all wavelengths in this range to obtain the average reflectance; then, combined with the known correspondence between reflectance and hemoglobin concentration, the average reflectance is converted into a value reflecting the hemoglobin concentration. For example, when the average reflectance is 0.28, the corresponding concentration value is 115, which is the first spectral characteristic value. For example, the first spectral characteristic value is obtained by following this process for the first spectral band data of diabetic foot ulcers.

[0152] Step 113: Obtain the numerical value reflecting the tissue oxygenation status based on the spectral information of the second spectral band, and use it as the second spectral characteristic value.

[0153] The second spectral band is the wavelength range corresponding to the tissue oxygenation state; the spectral information is the light reflection data within the second spectral band; the value reflecting the tissue oxygenation state is a value calculated from the spectral information that reflects the tissue oxygen supply; the second spectral characteristic value is a specific name for this value, used to represent the characteristics of the tissue oxygenation state.

[0154] In this embodiment, the spectral information of the second spectral band, namely the reflectance data in the range of 700-800 nanometers, such as the reflectance of each wavelength in this range, is first obtained; then, the characteristics of this spectral information are calculated, such as calculating the average reflectance of all wavelengths in this range to obtain the average reflectance; then, combined with the known correspondence between reflectance and tissue oxygenation status, the average reflectance is converted into a value reflecting the tissue oxygenation status. For example, when the average reflectance is 0.39, the corresponding oxygenation value is 82, which is the second spectral characteristic value. For example, the second spectral characteristic value is obtained by following this process for the second spectral band data of diabetic foot ulcers.

[0155] Step 114: Arrange the first spectral feature value and the second spectral feature value in a preset order to form a spectral feature vector.

[0156] The first spectral feature value reflects the hemoglobin concentration; the second spectral feature value reflects the tissue oxygenation status; the preset order is a pre-defined rule for arranging the feature values, such as hemoglobin concentration first and then tissue oxygenation status; the spectral feature vector is a sequence of two spectral feature values ​​arranged in a preset order, used to integrate biometric information.

[0157] In this embodiment, the preset order is first determined as "first spectral feature value (hemoglobin concentration) first, second spectral feature value (tissue oxygenation status) second"; then the first spectral feature value and the second spectral feature value are arranged in this order, for example, the first feature value is 115 and the second is 82, and the arrangement forms a sequence of [115, 82]; this sequence is the spectral feature vector, which is used to centrally reflect the information of two biological indicators, such as the feature values ​​of diabetic foot ulcers, which are arranged in this order to form a spectral feature vector.

[0158] This application provides the following specific example: Processing the wound reflectance spectral data of patient A: First, determine the spectral ranges corresponding to the biomarkers, with hemoglobin concentration corresponding to 500-600 nm (first spectral range) and tissue oxygenation status corresponding to 700-800 nm (second spectral range); extract spectral information from the first spectral range, where the reflectance at each wavelength is 0.26, 0.27, 0.29, and 0.30, and calculate the average reflectance as (0.26 + 0.27 + 0.29 + 0.30). ÷ 4 = 0.28, and the first spectral feature value 115 is obtained according to the corresponding relationship; spectral information is extracted from the second spectral band, and the reflectance of each wavelength is 0.38, 0.39, 0.40, and 0.39. The average reflectance is calculated as (0.38+0.39+0.40+0.39)÷4 = 0.39, and the second spectral feature value 82 is obtained; according to the preset order (first hemoglobin concentration, then tissue oxygenation status), the two feature values ​​are arranged as [115, 82] to form a spectral feature vector.

[0159] By executing steps 111 to 114, this embodiment of the application ensures the targeting and accuracy of feature extraction by determining the spectral bands corresponding to the bioindicators; by calculating the average reflectance from the spectral information and converting it into specific bioindicator values, abstract spectral data is transformed into quantifiable features; and by forming spectral feature vectors through orderly arrangement, structured integration of bioindicator information is achieved. These four steps work synergistically to accurately extract and integrate key biological information reflecting the wound state, providing a reliable bioindicator foundation for subsequent fusion with spatial information and analysis of the wound healing process.

[0160] Figure 3 A schematic diagram of a wound healing progress prediction system based on 3D point cloud reconstruction provided in this application embodiment is shown below. Figure 3 As shown, the system includes:

[0161] The acquisition module 31 is used to acquire the reflectance spectral data and initial three-dimensional point cloud data of the wound tissue. Based on the reflectance spectral data, a spectral feature vector containing spectral feature values ​​of multiple biological indicators is generated. The multiple biological indicators include hemoglobin concentration and tissue oxygenation status.

[0162] The generation module 32 is used to process the initial three-dimensional point cloud data to obtain target three-dimensional point cloud data that retains the spatial morphological features of the wound. The spectral feature values ​​in the spectral feature vector are mapped and fused with the spatial coordinates of the target three-dimensional point cloud data point by point to generate a three-dimensional point cloud data cube.

[0163] The coupling module 33 is used to extract the spatial structure features in the three-dimensional point cloud data cube using a 3D convolutional neural network, track the spectral feature changes corresponding to the spectral feature vectors at adjacent time points using optical flow method, and couple the spatial structure features with the spectral feature changes.

[0164] Module 34 is used to construct a cell migration map of wound tissue based on the coupling processing results. By analyzing the changing trend of the connection strength of each node in the cell migration map, the healing progress prediction results in the wound healing process are obtained.

[0165] The wound healing progress prediction system based on 3D point cloud reconstruction in this application is used to implement the aforementioned wound healing progress prediction method based on 3D point cloud reconstruction. Therefore, the specific implementation of the wound healing progress prediction system based on 3D point cloud reconstruction can be found in the embodiment section of the wound healing progress prediction method based on 3D point cloud reconstruction above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0166] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the wound healing progress prediction method based on 3D point cloud reconstruction described above.

[0167] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described methods for predicting wound healing progress based on 3D point cloud reconstruction.

[0168] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0169] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the wound healing progress prediction method based on 3D point cloud reconstruction.

[0170] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0171] The foregoing has provided a detailed description of a wound healing progress prediction method, system, electronic device, and storage medium based on 3D point cloud reconstruction provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for predicting wound healing progress based on 3D point cloud reconstruction, characterized in that, include: Acquire reflectance spectral data and initial three-dimensional point cloud data of wound tissue, and generate a spectral feature vector containing spectral feature values ​​of multiple biological indicators based on the reflectance spectral data, including hemoglobin concentration and tissue oxygenation status. The initial three-dimensional point cloud data is processed to obtain target three-dimensional point cloud data that retains the spatial morphological features of the wound. The spectral feature values ​​in the spectral feature vector are mapped and fused with the spatial coordinates of the target three-dimensional point cloud data point by point to generate a three-dimensional point cloud data cube. A 3D convolutional neural network is used to extract the spatial structure features in the three-dimensional point cloud data cube. The optical flow method is used to track the spectral feature changes corresponding to the spectral feature vectors at adjacent time points, and the spatial structure features and the spectral feature changes are coupled. Based on the coupling processing results, a cell migration map of the wound tissue is constructed. By analyzing the changing trend of the connection strength of each node in the cell migration map, the healing progress prediction results in the wound healing process are obtained. The process of processing the initial three-dimensional point cloud data to obtain target three-dimensional point cloud data that retains the spatial morphological features of the wound, and mapping and fusing the spectral feature values ​​in the spectral feature vector with the spatial coordinates of the target three-dimensional point cloud data point by point to generate a three-dimensional point cloud data cube, includes: The initial 3D point cloud data is divided into multiple wound areas according to point cloud density. Feature points reflecting spatial morphological changes are selected in each wound area, and these feature points are combined to form target 3D point cloud data. The position markers of the reflectance spectral data are obtained, and each spectral feature value in the spectral feature vector is associated with the corresponding position marker. The spatial coordinates corresponding to each spectral feature value in the target 3D point cloud data are determined. All associated spectral feature values ​​and their corresponding spatial coordinates are mapped and fused point by point to generate a 3D point cloud data cube. The process involves employing a 3D convolutional neural network to extract spatial structure features from the 3D point cloud data cube, using optical flow to track the spectral feature changes corresponding to the spectral feature vectors at adjacent time points, and coupling the spatial structure features with the spectral feature changes, including: A 3D convolutional neural network is used to divide the 3D point cloud data cube into layers, from the edge of the wound tissue to the middle of the wound tissue and to the entire wound area, and process them separately to extract structural information at different levels. Based on the structural information at different levels, spatial structural features are generated. The optical flow method is used to select the spectral feature vectors of each two adjacent time points, and the corresponding spectral feature values ​​in the spectral feature vectors of each two adjacent time points are compared to obtain multiple spectral feature changes. The spatial structural features are associated and integrated with all the spectral feature changes to obtain the coupled processing result.

2. The method according to claim 1, characterized in that, Associating each spectral feature value in the spectral feature vector with the corresponding location marker, and determining the spatial coordinates of each spectral feature value in the target 3D point cloud data, includes: Extract the spatial location information contained in the location marker, and associate each spectral feature value with the spatial location information to form an associated spectral feature value; Determine the distribution range of all spatial coordinates and the relative positional relationship of each spatial coordinate in the target 3D point cloud data; Based on the relative positional relationship, find the spatial coordinates with the smallest distance value from each of the spatial positional information within the distribution range, and determine each of the spatial coordinates as the spatial coordinates corresponding to the spectral feature value in the target three-dimensional point cloud data.

3. The method according to claim 1, characterized in that, Based on the coupling processing results, a cell migration map of the wound tissue is constructed. By analyzing the changing trends of the connection strength of each node in the cell migration map, the healing progress prediction results in the wound healing process are obtained, including: Multiple key locations in the wound tissue are selected from the coupling processing results, and each key location is used as a node in the cell migration map to determine the proximity relationship of each node in the wound tissue. Based on the proximity relationship, connections are established between mutually adjacent nodes to generate a cell migration map; Based on the first correlation value of the spatial structural features corresponding to each node and the second correlation value of the spectral feature change, the tightness of the connection between each pair of nodes is determined as the node connection strength. Record the connection strength values ​​of each node at different time points to obtain the trend of the connection strength of each node; Based on the changing trend, the healing progress prediction result in the wound healing process is determined using a preset prediction model.

4. The method according to claim 3, characterized in that, The determination of the connection strength between any two nodes, based on the first correlation value of the spatial structural features corresponding to each node and the second correlation value of the spectral feature variation, includes: Extract the spatial structure features and spectral feature changes corresponding to each node; By comparing the spatial structural features of every two nodes, a first distance similarity value is obtained from the edge range of the wound tissue to every two nodes, and a second distance similarity value is obtained from the middle range of the wound tissue. Based on the first distance similarity value and the second distance similarity value, a first association degree value is determined; By comparing the changes in spectral features between every two nodes, consistency information on changes in hemoglobin concentration and tissue oxygenation status between every two nodes is obtained, and a second correlation value is determined based on the consistency information. Based on the first correlation degree value and the second correlation degree value, the tightness of the connection between each pair of nodes is determined, and the tightness is used as the node connection strength.

5. The method according to claim 1, characterized in that, The step of generating a spectral feature vector containing spectral feature values ​​of multiple biological indicators based on the reflectance spectral data includes: Determine the corresponding spectral segment for each biomarker in the reflectance spectral data, where hemoglobin concentration corresponds to the first spectral segment and tissue oxygenation status corresponds to the second spectral segment; The numerical value reflecting the hemoglobin concentration is obtained based on the spectral information of the first spectral band and is used as the first spectral characteristic value; The numerical values ​​reflecting the tissue oxygenation status are obtained based on the spectral information of the second spectral band, and are used as the second spectral characteristic values; The first spectral feature value and the second spectral feature value are arranged in a preset order to form a spectral feature vector.

6. A wound healing progress prediction system based on 3D point cloud reconstruction, used to execute the wound healing progress prediction method based on 3D point cloud reconstruction as described in any one of claims 1 to 5, characterized in that, include: The acquisition module is used to acquire reflectance spectral data and initial three-dimensional point cloud data of wound tissue, and based on the reflectance spectral data, generate a spectral feature vector containing spectral feature values ​​of multiple biological indicators, including hemoglobin concentration and tissue oxygenation status. The generation module is used to process the initial three-dimensional point cloud data to obtain target three-dimensional point cloud data that retains the spatial morphological features of the wound. The spectral feature values ​​in the spectral feature vector are mapped and fused with the spatial coordinates of the target three-dimensional point cloud data point by point to generate a three-dimensional point cloud data cube. The coupling module is used to extract the spatial structure features in the three-dimensional point cloud data cube using a 3D convolutional neural network, track the spectral feature changes corresponding to the spectral feature vectors at adjacent time points using optical flow, and perform coupling processing on the spatial structure features and the spectral feature changes. The module is used to construct a cell migration map of wound tissue based on the coupling processing results. By analyzing the changing trend of the connection strength of each node in the cell migration map, the healing progress prediction results in the wound healing process are obtained.

7. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the wound healing progress prediction method based on 3D point cloud reconstruction as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the method for predicting wound healing progress based on 3D point cloud reconstruction as described in any one of claims 1 to 5.

Images