A method and system for traumatic vascular injury prediction and risk assessment of internal bleeding

Abstract

The present application relates to the technical field of medical image processing, and discloses a method and system for predicting vascular injury and assessing internal hemorrhage risk, comprising: acquiring original data, wherein the original data includes body surface marker coordinate data, physiological state and trauma-related information of a critically injured trauma victim; preprocessing the body surface marker coordinate data in the original data to obtain standardized coordinate data; inputting the standardized coordinate data into a pre-trained graph neural network model for mapping processing to obtain vascular marker three-dimensional coordinate prediction data; calculating the spatial distance between the injury path and the vascular marker according to the vascular marker three-dimensional coordinate prediction data and the trauma-related information to obtain a quantitative evaluation result of vascular injury risk; and combining the quantitative evaluation result of vascular injury risk with the physiological state information of the victim to complete comprehensive quantitative assessment of the internal hemorrhage risk of the victim to obtain a corresponding risk level result.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, specifically to a method and system for predicting traumatic vascular injury and assessing the risk of internal bleeding. Background Technology

[0002] In recent years, with the development of artificial intelligence (AI), more and more research has focused on how to apply AI to clinical medical scenarios to improve the quality and efficiency of medical services. Currently, the application of AI combined with medical imaging mainly focuses on areas such as lung nodule detection and breast cancer screening, and significant results have been achieved. However, most of these studies are limited to the diagnosis of single diseases, and effective solutions are still lacking for the diagnosis of complex diseases influenced by multiple factors, such as trauma.

[0003] Currently, existing technologies have some limitations in identifying and assessing the risk of vascular injury and internal bleeding in critical trauma: they cannot achieve real-time and accurate spatial localization of in vivo vascular markers and quantitative assessment of internal bleeding risk in emergency scenarios where large medical imaging equipment such as multi-slice spiral CT (MDCT) is unavailable.

[0004] Specifically, existing vascular marker localization schemes heavily rely on in vivo scanning data from large imaging devices, making them unsuitable for pre-hospital emergency scenarios without such equipment and hindering early identification of vascular injuries. Furthermore, existing graph neural network-based anatomical marker prediction technologies cannot address the mapping requirements arising from the mismatch in the number of nodes between surface markers and in vivo vascular markers, and they fail to eliminate coordinate bias caused by differences in patient body size, making it impossible to accurately predict the location of blood vessels in vivo solely through non-invasively acquired surface marker data. In addition, current internal bleeding risk assessments largely rely on subjective clinical experience or single-dimensional parameter judgments, failing to integrate vascular spatial injury risk with multi-dimensional physiological information to achieve an objective and quantitative comprehensive assessment. The accuracy and reliability of these assessment results do not meet the clinical needs of emergency scenarios. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a method and system for predicting traumatic vascular injury and assessing the risk of internal bleeding, thereby solving the aforementioned technical problems.

[0006] Firstly, a method for predicting traumatic vascular injury and assessing the risk of internal bleeding is provided, including: Acquire raw data, which includes the coordinate data of surface landmarks, physiological status and trauma-related information of critically injured patients; The coordinate data of body surface markers in the original data are preprocessed to obtain standardized coordinate data; The standardized coordinate data is input into a pre-trained graph neural network model for mapping processing to obtain 3D coordinate prediction data of vascular markers. Based on the predicted three-dimensional coordinates of the vascular markers and trauma-related information, the spatial distance between the injured path and the vascular markers is calculated to obtain a quantitative assessment result of the risk of vascular injury. By combining the quantitative assessment results of the vascular injury risk with the patient's physiological status information, a comprehensive quantitative assessment of the patient's internal bleeding risk is completed, and the corresponding risk level result is obtained.

[0007] Furthermore, the surface marker coordinate data in the original data are preprocessed to obtain standardized coordinate data, including: The center point coordinates of the surface marker data in the original data are calibrated to obtain the three-dimensional center point coordinates of each surface marker. Select a preset central body surface marker, and perform a centering translation on the three-dimensional center point coordinates of all body surface markers to obtain the centered coordinate data. Extract the reference scale value corresponding to the body shape of the injured person, and perform scale normalization on the centered coordinate data to obtain standardized coordinate data.

[0008] Furthermore, the standardized coordinate data is input into a pre-trained graph neural network model for mapping processing to obtain 3D coordinate prediction data for vascular markers, including: Feature extraction is performed on the standardized coordinate data using graph convolutional layers to obtain the feature tensors of surface marker nodes. Perform a first tensor transpose on the node dimension and feature dimension of the feature tensor of the surface marker node to obtain the transposed feature tensor. By mapping the number of nodes in the transposed feature tensor through a fully connected layer, the transformation of surface marker node features to vascular marker node features is completed. Perform a second tensor transpose on the transformed feature tensor to restore the standard form of the graph node features; The feature tensors after morphology restoration are input to the output layer and dimension-mapped to obtain the three-dimensional coordinate prediction data of vascular markers.

[0009] Furthermore, based on the predicted three-dimensional coordinates of the vascular markers and trauma-related information, the spatial distance between the injury path and the vascular markers is calculated to obtain a quantitative assessment result of the vascular injury risk, including: Based on the trauma-related information of the injured person, a three-dimensional spatial model of the injury path is constructed; The shortest spatial Euclidean distance between the injured path and the coordinates of each vascular marker is calculated to obtain a quantitative assessment result of the vascular injury risk.

[0010] Furthermore, after obtaining the three-dimensional coordinate prediction data of vascular markers, the process also includes coordinate restoration and visualization of the prediction data, including: Based on the reverse operation of the preprocessing step, the coordinates of the predicted three-dimensional coordinates of vascular markers are restored to obtain the absolute coordinate data of the original acquisition space coordinate system corresponding to the body surface markers. Based on the topological connection relationship between surface markers and vascular markers, the absolute coordinate data is displayed in three-dimensional space to construct a three-dimensional vascular distribution topology map. Based on the obtained quantitative assessment results of vascular injury risk, the vascular sites with injury risk are highlighted in the three-dimensional vascular distribution topology map to complete the visualization processing of the prediction results.

[0011] Furthermore, by combining the quantitative assessment results of the vascular injury risk with the patient's physiological status information, a comprehensive quantitative assessment of the patient's internal bleeding risk is completed, yielding the corresponding risk level results, including: Simultaneously extract the physiological status information of the injured person and complete the classification and determination of abnormal physiological indicators; By combining the quantitative grading results of vascular injury risk with the grading results of abnormal physiological indicators, a comprehensive quantitative assessment of the risk of internal bleeding in injured individuals is completed, and the corresponding risk level results are obtained.

[0012] Furthermore, after obtaining the corresponding risk level result, it also includes: If a comprehensive quantitative assessment indicates that the injured person has a high risk of internal bleeding, a three-dimensional visual alarm information containing the location of high-risk vascular injury will be generated. The alarm information and risk assessment data of the injured person are automatically pushed to the medical terminal device; If the risk of intracranial hemorrhage in the injured person is comprehensively and quantitatively assessed, routine revenue monitoring alerts will be generated and simultaneously pushed to medical terminal devices.

[0013] Secondly, a system for predicting traumatic vascular injury and assessing the risk of internal bleeding is provided, based on a method for predicting traumatic vascular injury and assessing the risk of internal bleeding as described in any of the preceding claims, including: The acquisition module is configured to acquire raw data, which includes the coordinate data of surface landmarks, physiological status and trauma-related information of critically injured patients. The preprocessing module is configured to preprocess the body surface marker coordinate data in the original data to obtain standardized coordinate data; The mapping processing module is configured to input the standardized coordinate data into a pre-trained graph neural network model for mapping processing to obtain three-dimensional coordinate prediction data of vascular markers. The calculation module is configured to calculate the spatial distance between the injured path and the vascular marker based on the predicted three-dimensional coordinates of the vascular marker and trauma-related information, thereby obtaining a quantitative assessment result of the risk of vascular injury. The assessment module is configured to combine the quantitative assessment results of the vascular injury risk with the patient's physiological status information to complete a comprehensive quantitative assessment of the patient's internal bleeding risk and obtain the corresponding risk level result.

[0014] Thirdly, a terminal is provided, including a processor, an input device, an output device, and a memory, the processor, the input device, the output device, and the memory being interconnected, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is configured to invoke the program instructions to execute a method for predicting traumatic vascular injury and assessing the risk of internal bleeding as described in any of the preceding claims.

[0015] Fourthly, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform a method for predicting traumatic vascular injury and assessing the risk of internal bleeding as described in any of the preceding claims.

[0016] The invention employing the above technical solution has the following advantages: This invention acquires surface marker coordinate data, physiological status, and trauma-related information from critically injured patients. It preprocesses the surface marker coordinates to eliminate individual body size differences. A pre-trained graph neural network model maps the surface markers to vascular markers, obtaining 3D coordinate prediction data for the vascular markers. The spatial distance between the injury path and the vascular markers is then calculated to quantify the risk of vascular injury. Finally, combined with physiological status information, a comprehensive quantitative assessment of internal bleeding risk is completed. This invention enables vascular marker coordinate prediction based on surface data in scenarios without large medical imaging equipment, adapts to mapping requirements where the number of surface and vascular marker nodes is mismatched, reduces the impact of individual body size differences on coordinate prediction, and improves the objectivity of the assessment process by using quantitative calculations to complete the internal bleeding risk assessment. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention, the accompanying drawings used in the specific embodiments will be briefly described below. In all the drawings, the elements or parts are not necessarily drawn to scale.

[0018] Figure 1 This is a flowchart of the overall system for predicting the risk of traumatic vascular injury and internal bleeding using the method and system of the present invention, which is a prediction system for surface-vascular markers. Figure 2 This is a connection diagram of surface markers in a method and system for predicting traumatic vascular injury and assessing the risk of internal bleeding, as described in this invention. Figure 3 This is a connection diagram of vascular markers in a method and system for predicting traumatic vascular injury and assessing the risk of internal bleeding, as described in this invention. Figure 4 This is a schematic diagram of the attention network in the method and system for predicting traumatic vascular injury and assessing the risk of internal bleeding, as described in this invention. Figure 5 This is a schematic diagram of the results visualization in a method and system for predicting traumatic vascular injury and assessing the risk of internal bleeding according to the present invention. Figure 6 This invention provides a flowchart of a method and system for predicting traumatic vascular injury and assessing the risk of internal bleeding, which includes a flowchart for trauma management. Figure 7 This is a flowchart of a method and system for predicting traumatic vascular injury and assessing the risk of internal bleeding according to the present invention. Detailed Implementation

[0019] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0020] like Figures 1 to 7 As shown, a method for predicting traumatic vascular injury and assessing the risk of internal bleeding according to the present invention is executed by a computer device and includes: Step S01: Obtain raw data, which includes the coordinate data of surface landmarks, physiological status and trauma-related information of critically injured patients. Step S02: Preprocess the coordinate data of body surface markers in the original data to obtain standardized coordinate data; Step S03: Input the standardized coordinate data into the pre-trained graph neural network model for mapping processing to obtain the three-dimensional coordinate prediction data of vascular markers; Step S04: Based on the three-dimensional coordinate prediction data of vascular markers and trauma-related information, calculate the spatial distance between the injured path and the vascular markers to obtain a quantitative assessment result of the risk of vascular injury. Step S05: Combining the quantitative assessment results of vascular injury risk with the patient's physiological status information, complete the comprehensive quantitative assessment of the patient's internal bleeding risk and obtain the corresponding risk level result.

[0021] In this embodiment, the computer device acquires the raw data of the critically injured patient. The specific collection and implementation methods for various types of data are as follows: Coordinate data of body surface landmarks: In this embodiment, 14 body surface landmarks with anatomical stability were selected as the data collection objects, specifically including: midpoint of left and right clavicles, upper border of sternum, lower border of left and right scapulae, lower border of xiphoid process, left and right nipples, lower border of left and right costal arches, umbilicus, left and right anterior superior iliac spines, and pubic symphysis.

[0022] The original three-dimensional coordinate data of the above 14 body surface markers were obtained through a three-dimensional optical scanning device and a body surface positioning marker acquisition terminal, which served as the core input for model prediction. Physiological status information: The continuous physiological parameters of the injured are collected in real time through vital sign monitors and wearable medical devices, including core vital sign data such as respiratory rate, pulse rate, systolic blood pressure / diastolic blood pressure, and pulse pressure difference. The data is standardized into a structured JSON data stream for computer devices to read and call. Trauma-related information: Collect core information such as the type of trauma (car accident, fall from height, stab / gunshot wound), coordinates of the injury location, direction and depth of the injury, and extract trauma trajectory data that can be used to construct the injury path as the basic input for subsequent vascular injury risk quantification.

[0023] In this embodiment, the coordinate data of surface markers in the original data are preprocessed to obtain standardized coordinate data, including: The center point coordinates of the surface marker data in the original data were calibrated to obtain the three-dimensional center point coordinates of each surface marker. Select a preset central body surface marker, and perform a centering translation on the three-dimensional center point coordinates of all body surface markers to obtain the centered coordinate data. Extract the reference scale value corresponding to the body shape of the injured person, and perform scale normalization on the centered coordinate data to obtain standardized coordinate data.

[0024] Specifically, the core objective of this implementation step is to eliminate absolute coordinate offsets caused by differences in body shape such as height and weight among different injured individuals, avoid scale and positional shifts in the model, and improve the model's generalization ability. The specific implementation steps are as follows: Center point coordinate calibration: The center point coordinates of the surface marker coordinate data in the original data are calibrated. Outlier removal and smoothing are performed on the collected original coordinate data. The three-dimensional geometric center point of each surface marker ROI is recalculated to obtain the three-dimensional center point coordinates of each surface marker. Centralized translation processing: A preset central body surface landmark is selected. In this embodiment, the node corresponding to the lower edge of the xiphoid process, which is located at the relative center of the human body, is selected as the central coordinate. Its coordinates are: Three-dimensional center coordinates of all body surface markers To eliminate differences in absolute coordinate positions, a centering translation is performed to obtain the centered coordinate data. The translation formula is as follows: Scale normalization: Reference scale values ​​corresponding to the injured person's body shape are extracted. In this embodiment, the spatial Euclidean distance between the midpoint of the left clavicle and the midpoint of the right clavicle is extracted as the shoulder width reference value representing the injured person's body shape, denoted as... Using the shoulder width scalar as a benchmark, the centered coordinate data is scale-normalized to eliminate individual body shape differences, resulting in standardized coordinate data. The normalization formula is: The resulting standardized coordinate data serves as the standard input for the pre-trained graphical neural network model.

[0025] In this embodiment, the graph neural network model for predicting vascular marker coordinates is pre-trained through the following steps, the specific implementation of which is as follows: MDCT Database Recommendations and Marker Labeling: Retrospective clinical thoracic and abdominal MDCT impact data were collected in accordance with ethical approval. Inclusion criteria were: no severe skeletal deformities, image slice thickness ≤1.25mm, uniform pixel spacing, and exclusion of samples with large-area internal metal implants causing severe artifacts. Standardized Regions of Interest (ROI) delineation guidelines were jointly developed by senior radiologists and trauma surgeons, completing the delineation of ROIs for 14 surface markers and 15 vascular markers. For vascular markers, a spherical ROI with a radius r=2.0mm was delineated with the geometric center of the vascular branch opening as the origin. For surface markers, the outermost voxel of the corresponding bony structure or anatomical protrusion was used as the center. The three-dimensional geometric center coordinates of each ROI were extracted as label data (GroundTruth). Training dataset construction: All MDCT images were resampled to an isotropic 1mm×1mm×1mm voxel resolution using a trilinear interpolation algorithm, and the center point coordinates of the ROI were recalculated. The label data were subjected to the same centering translation and scale normalization preprocessing operations as in the prediction stage to eliminate individual position and body size differences. The training set, validation set and test set were divided in an 8:1:1 ratio to construct a complete dataset for model training.

[0026] Network architecture construction: Construct a graph attention network architecture consisting of two graph convolutional layers, a tensor transpose module, a fully connected layer, and an output layer. The number of input nodes is 14, the number of output nodes is 15, and the hidden layer feature dimension is 8. Model Training and Validation: The loss function is the error between the predicted coordinates of vascular markers and the actual label coordinates. In this embodiment, the mean squared error (MSE) is used as the loss function, and the formula is as follows: in, The coordinates are the normalized true label coordinates. The deep learning optimizer uses the Adam optimizer with momentum parameters set to β=0.9 and β=0.999. The training epochs are set to 100, the batch size to 32, and the initial learning rate to 0.01. A cosine annealing learning decay strategy is used to avoid local optima. In the validation and test sets, in addition to recording the MSE, mean absolute error (MAE) and root mean square error (RMSE) are simultaneously introduced to quantify and evaluate the physical error of the model's predictions. This completes the training, validation, and optimization of the model, ultimately yielding a pre-trained model that can be directly used for prediction. In this embodiment, standardized coordinate data is input into a pre-trained graph neural network model for mapping processing to obtain 3D coordinate prediction data of vascular markers, including: Feature extraction is performed on standardized coordinate data using graph convolutional layers to obtain the feature tensors of surface marker nodes. Perform the first tensor transpose on the node dimension and feature dimension of the feature tensor of the body surface marker node to obtain the transposed feature tensor. By mapping the number of nodes in the transposed feature tensor through a fully connected layer, the transformation of surface marker node features to vascular marker node features is completed. Perform a second tensor transpose on the transformed feature tensor to restore the standard form of the graph node features; The feature tensors after morphology restoration are input to the output layer and dimension-mapped to obtain the three-dimensional coordinate prediction data of vascular markers.

[0027] Specifically, in this embodiment, the pre-trained graph neural network model is an improved graph attention network. The model input is the standardized three-dimensional coordinate data of 14 physical markers, and the output is the three-dimensional coordinate prediction data of 15 vascular markers. The 15 vascular markers specifically include: the openings of the abdominal aorta, celiac trunk, left and right renal arteries, proper hepatic artery, superior mesenteric artery, splenic artery, left and right internal iliac arteries, left and right common iliac arteries, left and right external iliac arteries, and left and right femoral arteries.

[0028] Graph Convolution Feature Extraction: Features are extracted from standardized coordinate data using two graph convolutional layers. In this embodiment, the initial feature dimension of the nodes is 3-dimensional, corresponding to the standardized three-dimensional coordinates of the surface markers. The feature dimension of the hidden layer of the nodes is set to 8-dimensional. After two layers of GAT processing, the feature tensor of the surface marker nodes is obtained, denoted as: Where B is the batch size, N is the number of nodes of the surface markers (N=14 in this implementation), and F is the feature dimension of the hidden layer (F=8 in this implementation). First tensor transpose: Perform the first tensor transpose on the feature tensor of the surface marker nodes, swapping the node dimension and feature dimension from (B, N, F) to (B, F, N), resulting in the transposed feature tensor, denoted as: Fully connected layer node mapping and bed-specific attention weighting: Through a feature mapping module with anatomical priors, high-fidelity conversion of surface marker node features to vascular marker node features is achieved.

[0029] Considering the differences in anatomical stability and importance of different surface landmarks in critical trauma prediction (for example, the upper edge of the sternum and the lower edge of the xiphoid process are core bony landmarks of the human midline, and their spatial coordinates have a much stronger constraint on the localization of major blood vessels than the peripheral nodes of the limbs), this invention innovatively introduces a surface landmark attention weighting matrix and a feature dimension scaling factor.

[0030] The specific conversion formula is as follows: in, To initialize the physiological attention diagonal matrix of surface biomarkers, its diagonal elements Learnable importance weights representing various body surface markers enable the model to adaptively focus on key anatomical nodes when mapping across domain nodes. For the feature dimension vector, use the Hadamard product. For the transposed feature tensor Perform channel-level dynamic scaling to avoid gradient vanishing during feature mapping and enhance key feature dimensions; The affine transformation weight matrix is ​​the mapping of the number of nodes. (In this embodiment, the input node...) Output node ); This is the bias term for the fully connected layer. The final output tensor... The shape becomes (B,F,M). Second tensor transpose processing: Perform a second tensor transpose processing on the transformed feature tensor, swapping the feature dimensions from (B,F,M) to (B,M,F), restoring the standard shape of the graph node features, and obtaining the vascular marker node feature tensor, denoted as: Output layer dimension mapping: The feature tensor after morphological restoration is input into the output layer for dimension mapping, using a weight matrix. Dimensionality reduction is performed to obtain the final 3D coordinate prediction data of vascular markers. The dimensionality reduction formula is as follows: in, The bias term for the output layer is (B, M, 3), and the final output tensor size is (B, M, 3), corresponding to the normalized three-dimensional coordinate prediction values ​​of 15 vascular markers.

[0031] In this embodiment, based on the three-dimensional coordinate prediction data of vascular markers and trauma-related information, the spatial distance between the injured path and the vascular markers is calculated to obtain a quantitative assessment result of the vascular injury risk, including: Based on the trauma-related information of the injured person, a three-dimensional spatial model of the injury path is constructed; The shortest spatial Euclidean distance between the injured path and the coordinates of each vascular marker is calculated to obtain a quantitative assessment of the risk of vascular injury.

[0032] Specifically, the construction of the three-dimensional spatial model of the injury path: based on the injury location, direction and depth of the injury in the injured person's trauma-related information, the injury path is modeled as a ray model in three-dimensional space, thus completing the three-dimensional spatial modeling of the injury path; Shortest spatial Euclidean distance calculation: Calculate the shortest spatial Euclidean distance between the injured path ray and the center point of the predicted three-dimensional coordinates of each vascular marker, denoted as distance D; Vascular injury risk quantification and grading: Vascular injury risk is quantified and graded based on a distance threshold. In this embodiment, the distance threshold is set to 5mm. If D < 5mm, the vascular site is determined to be in a high-risk state for vascular injury. If 5mm ≤ D < 10mm, it is determined to be in a medium-risk state. If D ≥ 10mm, it is determined to be in a low-risk state. Finally, the quantitative assessment result of vascular injury risk is output.

[0033] In this embodiment, after obtaining the three-dimensional coordinate prediction data of vascular markers, the method further includes coordinate restoration and visualization processing of the prediction data, including: Based on the reverse operation of the preprocessing step, the coordinates of the predicted three-dimensional coordinates of vascular markers are restored to obtain the absolute coordinate data of the original acquisition space coordinate system corresponding to the body surface markers. Based on the topological connection relationship between surface markers and vascular markers, the absolute coordinate data is displayed in three-dimensional space to construct a three-dimensional vascular distribution topology map. Based on the obtained quantitative assessment results of vascular injury risk, the vascular sites with injury risk are highlighted in the three-dimensional vascular distribution topology map to complete the visualization processing of the prediction results.

[0034] Specifically, the purpose of this embodiment is to restore the normalized coordinates output by the model to the original acquisition space coordinate system of the injured person, so as to provide an intuitive visual reference for clinical treatment. The specific implementation steps are as follows: Coordinate Restoration Processing: Based on the inverse operation of the preprocessing steps, the coordinates of the predicted three-dimensional coordinates of vascular markers are restored to obtain the absolute coordinate data of the original acquisition space coordinate system corresponding to the surface markers. The inverse operation formula is as follows: in, The actual shoulder width scalar of the injured person calculated during the pretreatment phase. The original coordinates of the central surface markers selected for the preprocessing stage; 3D spatial display and topology map construction: Based on the preset topological connection relationship between surface markers and vascular markers, the restored absolute coordinate data is displayed as a 3D point cloud and connection. The coordinates of surface markers and predicted vascular markers are presented in 3D space to construct a 3D vascular distribution topology map. Highlighting of high-risk areas: Based on the quantitative assessment results of vascular injury risk, high- and medium-risk vascular areas are highlighted and marked with corresponding levels in the three-dimensional vascular distribution topology map to complete the visualization of the prediction results.

[0035] In this embodiment, by combining the quantitative assessment results of vascular injury risk with the patient's physiological state information, a comprehensive quantitative assessment of the patient's internal bleeding risk is completed, resulting in a corresponding risk level result, including: Simultaneously extract the physiological status information of the injured person and complete the classification and determination of abnormal physiological indicators; By combining the quantitative grading results of vascular injury risk with the grading results of abnormal physiological indicators, a comprehensive quantitative assessment of the risk of internal bleeding in injured individuals is completed, and the corresponding risk level results are obtained.

[0036] Specifically, the physiological indicators abnormality status classification judgment: the physiological status information of the injured person is extracted simultaneously to complete the classification judgment of the physiological indicators abnormality status. In this embodiment, three core physiological abnormality judgment indicators are set: ① Systolic blood pressure is below 90 mmHg; ② Pulse rate exceeding 120 beats / min; ③ Continuous monitoring shows a pulse pressure difference change greater than 20 mmHg; meeting one indicator is considered a mild abnormality, and meeting two or more indicators is considered a severe abnormality. Comprehensive quantitative assessment and risk level output: Combining the quantitative grading results of vascular injury risk with the grading results of abnormal physiological indicators, a comprehensive quantitative assessment of the risk of internal bleeding in the injured person is completed, and the corresponding risk level result is obtained. The specific judgment rules are as follows: High risk of internal bleeding: High risk of vascular injury, and at least two physiological indicators are abnormal at the same time; Medium risk of internal bleeding: Medium risk of vascular injury, accompanied by at least one abnormal physiological indicator; Low risk of internal bleeding / no risk of internal bleeding: low risk of vascular injury, no abnormal physiological indicators or no risk of vascular injury, regardless of the presence or absence of any single abnormal physiological indicator.

[0037] In this embodiment, after obtaining the corresponding risk level result, the method further includes: If a comprehensive quantitative assessment indicates that the injured person has a high risk of internal bleeding, a three-dimensional visual alarm information containing the location of high-risk vascular injury will be generated. Automatically push alarm information and risk assessment data of the injured to medical terminal devices; If the risk of intracranial hemorrhage in the injured person is comprehensively and quantitatively assessed, routine revenue monitoring alerts will be generated and simultaneously pushed to medical terminal devices.

[0038] Specifically, if a comprehensive quantitative assessment indicates that an injured person has a high risk of internal bleeding, a three-dimensional visual alarm message is generated, which includes the location of the dangerous blood vessel injury, the risk level, and a three-dimensional blood vessel distribution topology map. The alarm message, as well as the injured person's risk assessment data and real-time physiological parameters, are automatically pushed to trauma treatment medical terminal equipment and medical staff's handheld terminals, and priority treatment scheduling actions are executed simultaneously. If the comprehensive quantitative assessment indicates that the injured person has a low to medium risk of internal bleeding, a corresponding level of trauma monitoring alert information will be generated and simultaneously pushed to the medical terminal device to remind medical staff to continuously monitor the injured person's vital signs and changes in the injury. If a comprehensive quantitative assessment indicates that the injured person has no risk of internal bleeding, routine trauma monitoring alerts are generated and simultaneously pushed to medical terminal devices, and corresponding injury monitoring and triage are carried out according to the routine trauma treatment process.

[0039] In other embodiments, a system for predicting traumatic vascular injury and assessing the risk of internal bleeding is provided, based on a method for predicting traumatic vascular injury and assessing the risk of internal bleeding according to any of the preceding embodiments, comprising: The acquisition module is configured to acquire raw data, which includes the coordinate data of surface landmarks, physiological status and trauma-related information of critically injured patients. The preprocessing module is configured to preprocess the coordinate data of body surface markers in the raw data to obtain standardized coordinate data; The mapping processing module is configured to input standardized coordinate data into a pre-trained graph neural network model for mapping processing to obtain 3D coordinate prediction data of vascular markers. The calculation module is configured to calculate the spatial distance between the injured path and the vascular markers based on the predicted data of the three-dimensional coordinates of vascular markers and trauma-related information, so as to obtain a quantitative assessment result of the risk of vascular injury. The assessment module is configured to combine the quantitative assessment results of vascular injury risk with the patient's physiological status information to complete a comprehensive quantitative assessment of the patient's internal bleeding risk and obtain the corresponding risk level result.

[0040] In other embodiments, a terminal is provided, including a processor, an input device, an output device, and a memory, which are interconnected. The memory is used to store a computer program, which includes program instructions. The processor is configured to invoke the program instructions to execute a method for predicting traumatic vascular injury and assessing the risk of internal bleeding, as described above.

[0041] In other embodiments, a computer-readable storage medium is provided that stores a computer program including program instructions that, when executed by a processor, cause the processor to perform a method for predicting traumatic vascular injury and assessing the risk of internal bleeding as described above.

[0042] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for predicting traumatic vascular injury and assessing the risk of internal bleeding, characterized in that, include: Acquire raw data, which includes the coordinate data of surface landmarks, physiological status and trauma-related information of critically injured patients; The coordinate data of body surface markers in the original data are preprocessed to obtain standardized coordinate data; The standardized coordinate data is input into a pre-trained graph neural network model for mapping processing to obtain 3D coordinate prediction data of vascular markers. Based on the predicted three-dimensional coordinates of the vascular markers and trauma-related information, the spatial distance between the injured path and the vascular markers is calculated to obtain a quantitative assessment result of the risk of vascular injury. By combining the quantitative assessment results of the vascular injury risk with the patient's physiological status information, a comprehensive quantitative assessment of the patient's internal bleeding risk is completed, and the corresponding risk level result is obtained.

2. The method for predicting traumatic vascular injury and assessing the risk of internal bleeding according to claim 1, characterized in that, The surface marker coordinate data in the original data are preprocessed to obtain standardized coordinate data, including: The center point coordinates of the surface marker data in the original data are calibrated to obtain the three-dimensional center point coordinates of each surface marker. Select a preset central body surface marker, and perform a centering translation on the three-dimensional center point coordinates of all body surface markers to obtain the centered coordinate data. Extract the reference scale value corresponding to the body shape of the injured person, and perform scale normalization on the centered coordinate data to obtain standardized coordinate data.

3. The method for predicting traumatic vascular injury and assessing the risk of internal bleeding according to claim 1, characterized in that, The standardized coordinate data is input into a pre-trained graph neural network model for mapping processing to obtain 3D coordinate prediction data for vascular markers, including: Feature extraction is performed on the standardized coordinate data using graph convolutional layers to obtain the feature tensors of surface marker nodes. Perform a first tensor transpose on the node dimension and feature dimension of the feature tensor of the surface marker node to obtain the transposed feature tensor. By mapping the number of nodes in the transposed feature tensor through a fully connected layer, the transformation of surface marker node features to vascular marker node features is completed. Perform a second tensor transpose on the transformed feature tensor to restore the standard form of the graph node features; The feature tensors after morphology restoration are input to the output layer and dimension-mapped to obtain the three-dimensional coordinate prediction data of vascular markers.

4. The method for predicting traumatic vascular injury and assessing the risk of internal bleeding according to claim 1, characterized in that, Based on the predicted three-dimensional coordinates of the vascular markers and trauma-related information, the spatial distance between the injury path and the vascular markers is calculated to obtain a quantitative assessment of the risk of vascular injury, including: Based on the trauma-related information of the injured person, a three-dimensional spatial model of the injury path is constructed; The shortest spatial Euclidean distance between the injured path and the coordinates of each vascular marker is calculated to obtain a quantitative assessment result of the vascular injury risk.

5. The method for predicting traumatic vascular injury and assessing the risk of internal bleeding according to claim 1, characterized in that, After obtaining the 3D coordinate prediction data of vascular markers, the process also includes coordinate restoration and visualization of the prediction data, including: Based on the reverse operation of the preprocessing step, the coordinates of the predicted three-dimensional coordinates of vascular markers are restored to obtain the absolute coordinate data of the original acquisition space coordinate system corresponding to the body surface markers. Based on the topological connection relationship between surface markers and vascular markers, the absolute coordinate data is displayed in three-dimensional space to construct a three-dimensional vascular distribution topology map. Based on the obtained quantitative assessment results of vascular injury risk, the vascular sites with injury risk are highlighted in the three-dimensional vascular distribution topology map to complete the visualization processing of the prediction results.

6. The method for predicting traumatic vascular injury and assessing the risk of internal bleeding according to claim 1, characterized in that, By combining the quantitative assessment results of the vascular injury risk with the patient's physiological status information, a comprehensive quantitative assessment of the patient's internal bleeding risk is completed, yielding the corresponding risk level results, including: Simultaneously extract the physiological status information of the injured person and complete the classification and determination of abnormal physiological indicators; By combining the quantitative grading results of vascular injury risk with the grading results of abnormal physiological indicators, a comprehensive quantitative assessment of the risk of internal bleeding in injured individuals is completed, and the corresponding risk level results are obtained.

7. The method for predicting traumatic vascular injury and assessing the risk of internal bleeding according to claim 1, characterized in that, After obtaining the corresponding risk level result, it also includes: If a comprehensive quantitative assessment indicates that the injured person has a high risk of internal bleeding, a three-dimensional visual alarm information containing the location of high-risk vascular injury will be generated. The alarm information and risk assessment data of the injured person are automatically pushed to the medical terminal device; If the risk of intracranial hemorrhage in the injured person is comprehensively and quantitatively assessed, routine revenue monitoring alerts will be generated and simultaneously pushed to medical terminal devices.

8. A system for predicting traumatic vascular injury and assessing the risk of internal bleeding, characterized in that, A method for predicting traumatic vascular injury and assessing the risk of internal bleeding, based on any one of claims 1 to 7, includes: The acquisition module is configured to acquire raw data, which includes the coordinate data of surface landmarks, physiological status and trauma-related information of critically injured patients. The preprocessing module is configured to preprocess the body surface marker coordinate data in the original data to obtain standardized coordinate data; The mapping processing module is configured to input the standardized coordinate data into a pre-trained graph neural network model for mapping processing to obtain three-dimensional coordinate prediction data of vascular markers. The calculation module is configured to calculate the spatial distance between the injured path and the vascular marker based on the predicted three-dimensional coordinates of the vascular marker and trauma-related information, thereby obtaining a quantitative assessment result of the risk of vascular injury. The assessment module is configured to combine the quantitative assessment results of the vascular injury risk with the patient's physiological status information to complete a comprehensive quantitative assessment of the patient's internal bleeding risk and obtain the corresponding risk level result.

9. A terminal, characterized in that, The device includes a processor, an input device, an output device, and a memory, which are interconnected. The memory stores a computer program, which includes program instructions. The processor is configured to invoke the program instructions to execute a method for predicting traumatic vascular injury and assessing the risk of internal bleeding as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform a method for predicting traumatic vascular injury and assessing the risk of internal bleeding as described in any one of claims 1 to 7.

Images