Virtual reality-based system for early assessment and management of severe trauma

Abstract

The present application relates to the technical field of emergency patient information processing, in particular to a severe trauma early assessment and processing system based on virtual-real combined interaction, which comprises a scene risk perception module, a dynamic physiological extraction module, an image feature analysis module, an abnormality level determination module and a priority scheduling classification module. The present application realizes continuous triggering of risk warning and hierarchical response by positioning abnormal period and synchronizing multi-end data, and realizes seamless information flow and task pushing between multi-terminal platforms by fusing multi-source data for feature analysis under unified time sequence, automatically generating trauma index and linking geographic location information, promoting automatic synchronization, intelligent classification and scheduling response of data from the scene to the hospital, establishing system advantages in multi-link cooperation, information closed loop and early trauma identification, further supporting physiological process visual feedback, hierarchical task closed loop and real-time resource matching, and improving the efficiency of early identification and processing of critically ill patients.

Description

Technical Field

[0001] This invention relates to the field of emergency patient information processing technology, and in particular to a system for early assessment and treatment of severe trauma based on virtual-real interaction. Background Technology

[0002] The field of emergency patient information processing mainly involves the collection of physiological parameters, data transmission, case information management, integration of diagnostic and treatment information, and multi-terminal information sharing in emergency scenarios. It covers multi-level medical collaboration scenarios including pre-hospital emergency care, emergency department treatment, and trauma centers. The focus is on achieving efficient acquisition, dynamic analysis, and real-time sharing of early information for critically ill patients with severe trauma, thereby improving the intelligence and informatization of clinical treatment processes. Traditional early assessment and treatment systems for severe trauma based on virtual-real interaction combine medical imaging, patient physiological parameter monitoring, and clinical consultation data. Through scenario simulation and linkage with actual clinical operations, these systems enable rapid assessment and triage of trauma patients, supporting medical staff in making preliminary judgments and formulating treatment plans.

[0003] Current technologies for emergency trauma assessment and triage still rely on medical staff manually entering vital sign data. Subjective judgment can easily lead to assessment delays and omissions of key information. There is a disconnect between pre-hospital and in-hospital data flow, and vital signs and injury images cannot be transmitted synchronously in real time, affecting the response of in-hospital teams. The triage mechanism mainly relies on manual operation, making it difficult to quickly prioritize the response to critically ill patients. Although data fusion achieves multi-source integration, the feedback method is basic, and the linkage between vision and touch is insufficient. The coordination of tasks and resources between the scene and the hospital is lagging behind, which limits the overall level of critical trauma identification and intelligent triage scheduling. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a system for early assessment and treatment of severe trauma based on a combination of virtual and real interaction.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a system for early assessment and treatment of severe trauma based on virtual-real interaction, the system comprising:

[0006] The scene risk perception module is based on IoT wearable devices. It analyzes changes in motion signals, lighting conditions and collision monitoring signals to determine the order of on-site risks, adjusts the acquisition frequency of each monitoring channel, and obtains the risk monitoring acquisition rate.

[0007] Based on the risk monitoring acquisition rate, the dynamic physiological extraction module identifies continuous fluctuations in electrocardiogram signals, analyzes heart rate trends, blood pressure changes, and blood oxygenation cycle fluctuations, locates abnormal moments of multiple physiological parameters, and obtains temporal fluctuation location features.

[0008] Based on the temporal fluctuation positioning features, the image feature analysis module analyzes the trauma images acquired by the portable imaging device, screens for missing data, supplements the data, filters texture parameters, compares the images with physiological changes, and obtains a set of linkage abnormality indicators.

[0009] The abnormality level determination module analyzes collision monitoring and blood pressure fluctuation parameters based on the aforementioned set of linked abnormal indicators, weights multiple data, and determines the patient's abnormal condition through the critical illness discrimination criteria to obtain the graded risk determination factor.

[0010] The priority scheduling and grading module analyzes the patient's geographical location and travel time based on the grading risk determination factors, and ranks the priority of grading patients according to the grading index, distance and travel time standardization to obtain the scheduling priority configuration.

[0011] The present invention improves upon the following: the risk monitoring acquisition rate includes scene category coding, channel priority sorting, and real-time adjustment markers; the temporal fluctuation positioning features include anomaly detection labels, change distribution curves, and key positioning moments; the linkage anomaly indicator set includes color saturation, average brightness, and texture grading factors; the graded risk judgment factors include risk classification codes, level assessment numbers, and reference comparison labels; and the scheduling priority configuration includes priority labels, scheduling object identifiers, and transfer recommendation schemes.

[0012] The present invention is improved in that the scene risk perception module includes:

[0013] The wearable signal acquisition submodule is based on IoT wearable devices. It uses data sources deployed on the patient's chest and wrist to organize the data streams of acceleration trajectory, direction offset channel and ambient illumination channel in a time synchronization manner, and matches the time tags output by each channel to obtain a set of multi-channel synchronous acquisition segments.

[0014] The motion trend extraction submodule compares the acceleration and direction offset channel data of each segment based on the multi-channel synchronous acquisition segment set, calculates the change amplitude of velocity and direction between segments, determines the consistency or turning point of the motion trend, and marks the trend change segment and duration in combination with the segment number to obtain the continuous trend change annotation group.

[0015] The channel frequency configuration submodule analyzes the illuminance channel data and collision signal trajectory within the corresponding segment based on the continuous trend change annotation group, compares the illuminance uniformity and the continuous characteristics of collision changes, and adjusts the data acquisition priority of each monitoring channel to obtain the risk monitoring acquisition rate.

[0016] The present invention is improved in that the dynamic physiological extraction module includes:

[0017] The ECG fluctuation identification submodule analyzes the ECG signal of the priority channel based on the risk monitoring acquisition rate, judges the waveform changes of continuous heartbeat cycles, identifies the rising and falling phases of each cycle, filters the intervals with consistent waveform stability and fluctuation trend, and adjusts the signal label distribution in combination with the sampling frequency to obtain a set of ECG fluctuation change indicators.

[0018] The parameter trend comparison submodule compares heart rate data and blood pressure data based on the ECG fluctuation change index set, analyzes the trend direction and change range of the two sets of data, and filters the intervals with synchronous trend changes and trend consistency to obtain the synchronous trend matching segment set.

[0019] The temporal feature localization submodule analyzes the fluctuation distribution of data in the blood oxygen channel within the cycle based on the synchronization trend matching segment set. By comparing the turning points of ECG and blood pressure parameters, it maps all fluctuation points to time coordinates, establishes the fluctuation correspondence of multi-channel physiological signals, and obtains temporal fluctuation localization features.

[0020] The present invention is improved in that the image feature parsing module includes:

[0021] Based on the temporal fluctuation positioning features, the image parameter screening submodule analyzes the distribution of color and brightness channels in the main region of the trauma image, determines whether the channels have a complete data structure, filters samples with missing or abnormal channel distribution, completes the channel content, and obtains the main region channel data integration structure.

[0022] The format standardization processing submodule calculates the spatial resolution and color space consistency of the image to be processed based on the main region channel data integration structure, identifies samples with different formats, adjusts the pixel structure and unifies the encoding method, and filters images that meet the unified standard to obtain a standardized pixel structure set.

[0023] The texture feature comparison submodule calculates the texture density and uniformity of each patch based on the standardized pixel structure set, combines the temporal fluctuation positioning features, compares the image texture index with the brightness sampling content under the same time sequence, obtains the texture-brightness response mapping offset, filters the delay structure of texture parameters, and obtains the linkage anomaly index set.

[0024] The present invention is improved in that the anomaly level determination module includes:

[0025] The collision feature extraction submodule analyzes the acceleration changes and impact duration in the collision monitoring signal based on the linkage anomaly index set, judges the trend of impact amplitude and direction changes within the sampling interval, identifies continuous signal segments and optimizes the occlusion interference judgment process, compares the differences in physical characteristics of each segment, calculates the motion performance of the impact segment, and obtains the collision physical feature index.

[0026] The physiological fluctuation identification submodule calculates the amplitude changes of systolic blood pressure, heart rate and blood oxygen signals within the corresponding acquisition period based on the collision physical characteristic index, judges the trend consistency of physiological parameters, filters the intervals where the change amplitude is higher than the amplitude threshold, and compares the fluctuation correlation between multi-channel signals to obtain the physiological abnormality response factor.

[0027] The critical illness risk scoring submodule, based on the aforementioned physiological abnormality response factors, jointly analyzes the fluctuation range of blood pressure sequence, the magnitude of collision impact, and the characteristics of heart rate regulation, calculates the abnormality dependent variable, locates the proportion of sampling time covered by its continuous offset segment, and obtains the graded risk determination factor.

[0028] The present invention is improved in that the priority scheduling hierarchical module includes:

[0029] The location analysis submodule analyzes the spatial coordinates between the patient's location information and the hospital's location information based on the graded risk judgment factor, compares the geographical relative relationship between the two, determines the spatial distribution of the current location and the target hospital, filters the location parameters, and obtains the geographical distance calculation result.

[0030] The traffic time assessment submodule optimizes the path data obtained from the geographical distance calculation results, analyzes the road grade and traffic information of each transfer route, judges the connectivity and traffic trend of each path node, filters continuous traffic routes, and obtains path traffic assessment indicators.

[0031] The priority sorting configuration submodule arranges the graded priority parameters of each patient based on the route accessibility assessment index, adjusts the correspondence between patients and emergency vehicles, establishes the priority numbering order, and obtains the dispatch priority configuration.

[0032] The present invention is improved in that the IoT wearable device refers to an Internet of Things device that can be worn on the patient to collect vital signs and on-site environmental data in real time; the physiological change refers to the moment when the abnormal physiological parameters are extracted, which corresponds one-to-one with the image acquisition time; and the patient's abnormal condition refers to the patient's risk level or degree of urgency under the current classification.

[0033] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0034] This invention relies on the collaboration of multiple types of sensing devices and terminals to dynamically allocate acquisition channels and automatically acquire environmental and motion states, collision signals, physiological data, and image information. Through abnormal time period positioning and multi-terminal data synchronization, multi-source data is integrated for feature analysis under a unified time sequence, enabling continuous triggering of risk warning and graded response. It automatically generates a trauma index and links it with geographical location information, achieving seamless information flow and task push between multiple terminal platforms. This promotes automatic synchronization, intelligent grading, and scheduling response of data from the field to the hospital, establishing system advantages in multi-stage collaboration, information closure, and early trauma identification. It further supports visual feedback of physiological processes, graded task closure, and real-time resource matching, improving the efficiency of early identification and treatment of critically ill patients. Attached Figure Description

[0035] Figure 1 This is a system flowchart of the present invention;

[0036] Figure 2 This is a flowchart of the scene risk perception module in this invention;

[0037] Figure 3 This is a flowchart of the dynamic physiological extraction module in this invention;

[0038] Figure 4 This is a flowchart of the image feature parsing module in this invention;

[0039] Figure 5 This is a flowchart of the anomaly level determination module in this invention;

[0040] Figure 6 This is a flowchart of the priority scheduling hierarchical module in this invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0042] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Example

[0043] Please see Figure 1 This invention provides a technical solution: a severe trauma early assessment and treatment system based on virtual-real interaction, comprising:

[0044] The scene risk perception module is based on IoT wearable devices to judge the changing trend of motion signals collected by the vehicle terminal, identify the rescue scene with insufficient lighting, compare the change amplitude of continuous motion data, analyze the dynamic changes of collision monitoring signals, and adjust the acquisition frequency of each monitoring channel according to the risk sequence at the scene to obtain the risk monitoring acquisition rate.

[0045] The dynamic physiological extraction module judges the continuous fluctuation of the ECG signal in the priority channel based on the risk monitoring acquisition rate, analyzes the parameter change trend of the heart rate data during the sampling period, compares it with the blood pressure change time series, identifies the fluctuation range within the blood oxygen cycle, and performs time positioning based on the fluctuation points of multiple physiological parameters to obtain the temporal fluctuation positioning characteristics.

[0046] The image feature analysis module analyzes trauma images acquired by portable imaging devices at a specified time based on temporal fluctuation positioning features, judges the integrity of parameters of color and brightness distribution in the main region, screens for missing images, and supplements missing items by the terminal. It unifies the image data format and filters texture parameters, compares image features with physiological change points, and obtains a set of linkage abnormality indicators.

[0047] The abnormality level determination module analyzes the changing parameters of collision monitoring based on the linkage abnormality index set, combines the abnormal fluctuation of blood pressure, calculates the weighted total performance of multiple parameters, judges it through the critical illness discrimination benchmark, determines the patient's current abnormal condition, and obtains the graded risk determination factor.

[0048] The priority dispatching and grading module analyzes the patient's current location information based on the grading risk judgment factor, determines the traffic time parameters of the emergency transport process, performs standardized processing according to the grading index, geographical distance and time, arranges the priority of each patient in order, identifies the number of high-priority patients and emergency vehicles, and obtains the dispatching priority configuration.

[0049] Risk monitoring and data collection rates include scene category coding, channel priority sorting, and real-time adjustment markers. Temporal fluctuation positioning features include anomaly detection labels, change distribution curves, and key positioning moments. Linked anomaly indicator sets include color saturation, average brightness, and texture grading factors. Graded risk judgment factors include risk classification codes, level assessment numbers, and reference comparison labels. Scheduling priority configuration includes priority labels, scheduling object identifiers, and transfer recommendation schemes.

[0050] The main objective is to address the three core issues in severe trauma care: delayed assessment, fragmented information, and inaccurate triage. This aims to shorten the critical "golden hour" treatment window by integrating data from wearable devices (heart rate, blood oxygen), mobile monitors, and portable ultrasound / imaging equipment to dynamically generate a trauma index. This enables seamless data transfer between pre-hospital emergency care, ambulances, and hospital emergency departments, while simultaneously initiating in-hospital resource preparation (such as operating rooms and blood banks). The general process is as follows: multi-source data collection → data integrity assessment → incomplete (supplementary) / complete data processing → data standardization → data evaluation → trauma index calculation → triage decision-making (pre-defined plan) → cross-terminal execution.

[0051] More specifically:

[0052] Vital signs (ECG, BP, SpO2), images / videos of the trauma site, and on-site environmental data (such as collision intensity) are acquired through IoT devices. Heterogeneous data are normalized into standard feature vectors, images are processed using a pre-trained ResNet-50, and time-series physiological data are analyzed using LSTM to output a comprehensive trauma index (0-10 points). This index is then used for tiered processing (automatic triggering of contingency plans based on the trauma index).

[0053] ≥8 points (red): Notify the trauma center to activate the multidisciplinary team and reserve an operating room;

[0054] 5-7 points (yellow): Push the emergency medical checklist to the emergency department tablet;

[0055] ≤4 points (green): Generate basic treatment guidelines to the first responder's handheld terminal.

[0056] ResNet-50 and LSTM can be replaced by the lightweight model MobileNetV3+Transformer, which is suitable for low-computing-power terminals. In addition to the trauma index, location weights (such as distance from the hospital) can be added to dynamically adjust the grading threshold.

[0057] In the scenario risk perception module, IoT wearable devices refer to IoT devices that can be worn on patients to collect vital signs and environmental data in real time, such as smart wristbands, smart patches, and portable sensors, commonly used to record data such as heart rate, blood oxygen, and body temperature; vehicle-mounted terminals refer to the integrated hardware system installed in ambulances, responsible for collecting, storing, and forwarding various physiological, environmental, and location information, and can receive data from wearable devices and other sensors in the ambulance; motion signals refer to the motion data of patients or vehicles detected by devices such as accelerometers and gyroscopes, including speed, acceleration, and changes in direction of movement, used to determine whether there is violent activity or abnormal movement at the treatment site; and insufficiently lit treatment sites refer to areas where the light intensity detected by ambient light sensors is lower than the preset treatment standard. This condition typically affects the accuracy of on-site emergency procedures and image acquisition; continuous motion data refers to the sequence of motion parameters continuously collected by motion sensors (such as accelerometers) within a certain time range, used to analyze activity trends and sudden changes; dynamic changes in collision monitoring signals refer to the physical impact signals and their changing trajectories detected in real time by collision sensors (such as force sensors and gyroscopes), which can be used to identify the patient's stress condition or the severity of the accident; on-site risk priority refers to the prioritization of different risk factors based on the collected risk-related parameters, facilitating subsequent adjustments to the acquisition strategy and response priority; each monitoring channel refers to the various data acquisition lines and types in the system, such as heart rate acquisition channels, blood pressure acquisition channels, ambient light channels, collision sensing channels, etc., each responsible for the acquisition and monitoring of different types of data.

[0058] In the dynamic physiological extraction module, the priority channel ECG signal refers to the ECG signal monitoring line that is given higher priority due to risk prediction among all physiological parameter acquisition channels, and its acquisition frequency can be higher than other channels; the parameter change trend refers to the trend and fluctuation pattern of parameters such as heart rate and blood pressure over time during the data acquisition period, used to determine whether there are abnormal changes; the blood pressure change time series refers to the data sequence composed of continuously acquired blood pressure readings sorted by time, which can be used to dynamically analyze blood pressure fluctuations; the fluctuation interval refers to the specific time period in which physiological parameters show significant changes during the analysis process, used to locate abnormal moments; the time positioning of fluctuation points refers to the specific time marking of data points with abnormal changes within the above fluctuation intervals, which facilitates subsequent linkage analysis with other modal data.

[0059] In the image feature analysis module, portable imaging equipment refers to portable medical imaging devices used on-site that can acquire static images or dynamic videos of the trauma site, such as handheld ultrasound, portable cameras, and head-mounted cameras; parameter integrity refers to whether the key parameters (such as color, brightness, resolution, etc.) in the acquired image data are complete and can meet the requirements of subsequent feature analysis; terminal supplementation refers to the manual entry of relevant images or parameters through manual intervention or terminal devices when automatically acquired data is missing or of substandard quality, to ensure data integrity; image data format refers to the data standards and format specifications (such as JPEG, PNG, DICOM, etc.) used by the acquired image or video files to ensure data compatibility and smooth subsequent processing; texture parameters refer to the numerical indicators (such as texture density, directionality, uniformity, etc.) extracted from the details of the edge and surface of the trauma area during image analysis, used to assist in lesion identification and grading; physiological change points refer to the moments of abnormal physiological parameters extracted in conjunction with the dynamic physiological extraction module, which correspond one-to-one with the image acquisition time and are used for multimodal linkage interpretation.

[0060] In the abnormality level determination module, the collision monitoring change parameters refer to the physical quantities such as mechanical changes and acceleration changes extracted during the continuous monitoring of collision events by the system, which are used to reflect the force and severity of injury to the patient; abnormal fluctuation performance refers to the drastic changes in physiological parameters (such as blood pressure) in a short period of time, which are often used to help identify the severity of the patient's condition; the weighted total performance of multiple parameters refers to the comprehensive evaluation data obtained by synthesizing multiple key risk parameters (such as imaging features, collision intensity, blood pressure abnormalities, etc.) after assigning weights; the critical illness discrimination criteria refer to the preset quantifiable standards used to distinguish between ordinary cases and severe or critical cases, providing a criterion for subsequent grading and resource allocation; the patient's current abnormal condition refers to the risk level or severity of the patient under the grading system at the current moment after comprehensive determination with the discrimination criteria.

[0061] In the priority dispatching and grading module, the patient's current location information refers to the geographic coordinates of the patient or ambulance's location obtained in real time through positioning technologies such as GPS, which are used for subsequent resource dispatching; the traffic time parameter refers to the estimated transfer time from the scene to the target hospital based on map systems or traffic prediction tools; standardization processing refers to the normalization or standard score conversion of raw parameters (such as grading index, distance, and time) of different dimensions and sources to make them comparable; patient grading priority refers to the priority order number of each patient in the process of emergency resource allocation, ambulance dispatching, etc., based on the system evaluation results; high priority refers to patients or tasks that have obtained earlier or more resource response rights after comprehensive evaluation, usually critically ill or those that are close to the hospital and have short time requirements.

[0062] Please see Figure 2 The scenario risk perception module includes:

[0063] The wearable signal acquisition submodule is based on IoT wearable devices. It uses data sources deployed on the patient's chest and wrist to organize the data streams of acceleration trajectory, direction offset channel and ambient illumination channel in a time synchronization manner, and matches the time tags output by each channel to obtain a set of multi-channel synchronous acquisition segments.

[0064] Basic sensor signals are acquired using IoT wearable devices deployed on the patient's chest and wrist. The chest node collects acceleration information representing the stable state of the upper body and records changes in the direction of force. The wrist node collects the acceleration trajectory and angular velocity information caused by arm swinging and turning. Both nodes are equipped with ambient light sensors to obtain local illumination information. During data acquisition, the data acquired by each node is timestamped. All data is organized in chronological order. First, the acceleration data is axially corrected, and then the overall acceleration output value of the chest and wrist at each moment is calculated, generating a continuous frame data stream. For the directional offset, angular velocity information in three directions is extracted, and the angular velocity information at adjacent time points is calculated. The velocity value is calculated to determine its change range. The illuminance channel collects the raw illuminance value at each moment and compares the difference between the illuminance values ​​at the chest and wrist nodes. Each sampling point constitutes a complete channel segment containing acceleration value, angular velocity change value, and illuminance difference value. This set of segments is continuously combined in time to form a collection of acquisition segments several seconds long. In a patient fall event, the acceleration value recorded at the wrist node increased from 1.2 to 3.8, the yaw value change of the directional angular velocity increased to more than ten times the original state, and the illuminance difference between the chest and wrist also decreased from the usual 5 to less than 1. This time segment is classified into the key segment set. Finally, the data from all channels are aggregated according to a unified time label to form a multi-channel synchronous acquisition segment set.

[0065] The motion trend extraction submodule is based on a multi-channel synchronous acquisition of a set of segments. It compares the acceleration and direction offset channel data of each segment, calculates the change in velocity and direction between segments, determines the consistency or turning point of the motion trend, and marks the trend change segments and durations in combination with the segment sequence number to obtain a continuous trend change annotation group.

[0066] The system processes each segment from a multi-channel synchronously acquired data set, performing differential analysis on the acceleration of the chest and wrist in each segment. It determines the presence of inflection points in the motion trend by analyzing the average difference trends between segment sequences. The system compares the mean change amplitude of the segment sequences, identifying the start or end point of a trend change when the average acceleration difference between segments exceeds a set threshold. Directional offset data is compared based on the angular velocity changes in three directions, constructing a directional vector for each segment. If the similarity of directional vectors between consecutive segments decreases significantly, it is marked as an inflection point in the directional change. In one patient fall, the average acceleration difference from segment 40 to segment 58 increased by more than 0.8, and the directional vector change was significantly different from the preceding and following stable segments. At this time point, the trend was marked as a sudden change, with a duration exceeding 3 seconds. This segment was assigned a special label, recording its start and end segment numbers and duration for subsequent dynamic adjustment of risk channel priorities. Each label element in the generated trend change annotation group contains four types of information: trend start and end positions, change type, change amplitude, and trend duration.

[0067] The channel frequency configuration submodule analyzes the illuminance channel data and collision signal trajectory within the corresponding segment based on the continuous trend change annotation group, compares the illuminance uniformity and the continuous characteristics of collision changes, and adjusts the data acquisition priority of each monitoring channel to obtain the risk monitoring acquisition rate.

[0068] Taking each segment in the trend change annotation group as a unit, extract its corresponding illuminance data sequence and collision signal data. Calculate the difference between the maximum and minimum illuminance values. When the illuminance value change within a segment is greater than 15 and the overall average is less than 20, it is judged as uneven and dark. Further analyze the collision signal within that segment. Calculate the signal amplitude change for each frame and accumulate the total vibration of that segment. When the accumulated vibration amplitude exceeds 50 and this state is maintained continuously for more than 20 frames, it is judged as a continuous collision state. If any of the above conditions are met, enter the channel priority reconfiguration process. The sampling frequency of the acceleration signal channel is increased by one level from the current base value. At the same time, the sampling frequencies of the direction offset channel and the illuminance channel are also increased in order of priority. In a certain car accident scene, the difference in illuminance between the patient's chest and wrist fluctuated by 22 in this section, the average illuminance of the chest was as low as 12, the cumulative change of the collision signal reached 75 and lasted for 4 seconds, which was identified as a high-risk period. The acquisition frequency of the three types of channels was increased from the low-speed state to the high-frequency acquisition state, and the corresponding channel was marked with a high-priority label in the channel management table. After this operation is completed, the risk monitoring acquisition rate of each channel during this period is output.

[0069] Please see Figure 3 The dynamic physiological extraction module includes:

[0070] The ECG fluctuation identification submodule analyzes the ECG signal of the priority channel based on the risk monitoring acquisition rate, judges the waveform changes of continuous heartbeat cycles, identifies the rising and falling phases of each cycle, filters the intervals with consistent waveform stability and fluctuation trend, and adjusts the signal label distribution in combination with the sampling frequency to obtain a set of ECG fluctuation change indicators.

[0071] Based on the high-priority ECG channel set by the risk monitoring acquisition rate, signal data analysis of continuous heartbeat cycles is performed. The voltage amplitude sequence representing the ECG waveform in the acquired signal is called up. The waveform within each cycle is segmented according to the time label, and the rising and falling segments in each heartbeat cycle are extracted. The starting position of the rising segment is determined as the first rapid rise point after the baseline, and the falling segment as the first rapid decline point after the peak. Each cycle consists of a start point, peak point, and end point, forming a complete waveform band. In the example, if the sampling frequency is 250Hz, 250 data points are collected per second. In the ECG data of healthy adults, the average duration of each cycle is 800ms. The waveform is divided into segments according to this cycle length. The maximum voltage change value within each segment is extracted and compared with the direction of change in adjacent segments. If the direction of change of the maximum value in two consecutive segments is... If the waveform is the same and the variation amplitude is less than 0.2 mV, the segment is considered a stable waveform segment. At the same time, if the peak interval is within the normal range (600 to 1000 ms) without significant shortening or lengthening, it is considered periodically stable. Each period segment is traversed and ECG segments that meet the stable trend are selected according to the waveform direction change pattern. In a certain actual acquisition, the patient's ECG signal had a continuous stable signal segment with peak fluctuation of less than 0.1 mV and interval fluctuation of less than 50 ms for 12 consecutive seconds. This segment was recorded as a high-stability segment. Then, the signals in this segment were relabeled according to the sampling frequency. A corresponding time label was assigned to each sampling point and the index table was updated. Finally, complete fluctuation feature information containing indicators such as cycle start point, peak position, fall time, and cycle length was generated, forming a set of ECG fluctuation change indicators.

[0072] The parameter trend comparison submodule compares heart rate data and blood pressure data based on the ECG fluctuation change index set, analyzes the trend direction and change range of the two sets of data, and filters the intervals with synchronous trend changes and trend consistency to obtain the synchronous trend matching segment set.

[0073] The system receives structured time-series data from a set of electrocardiogram (ECG) fluctuation indicators and compares it point-by-point with data collected from the blood pressure monitoring channel within the same time period. First, the heart rate data is processed, converting each cycle length into heart rate beats per minute to construct a heart rate trend sequence. For the corresponding time period, blood pressure data is used to construct independent time series for systolic and diastolic blood pressure. The trend directions of the two physiological indicators are matched. If, in five consecutive sampling cycles, the heart rate value increases and the systolic blood pressure also increases concurrently, with the average change exceeding a set threshold, then the trend direction is considered consistent. Further screening is then performed. Select the variation amplitude of two sets of data, set the threshold for heart rate trend amplitude to 5 beats per minute, and the threshold for blood pressure trend amplitude to 5 mmHg. If any trend change meets the above amplitude conditions in three sampling segments, the segment is identified as a trend-matching segment. In practical applications, if the heart rate rises from 78 to 93 within 10 seconds, and the systolic blood pressure rises from 115 to 128, and both sequences meet the requirements for increasing direction and fluctuation range between every two sampling segments, then the segment is classified as a synchronous trend segment. Record the start time, end time, trend direction, and rate of change of the segment, and include it in the synchronous trend-matching segment set.

[0074] The temporal feature localization submodule analyzes the fluctuation distribution of data in the blood oxygen channel within the cycle based on the synchronous trend matching segment set. By comparing the turning points of ECG and blood pressure parameters, it maps all fluctuation points to time coordinates, establishes the fluctuation correspondence of multi-channel physiological signals, and obtains temporal fluctuation localization features.

[0075] Based on the time range recorded in the synchronization trend matching segment set, the blood oxygen saturation data sequence of the corresponding time period is called in the blood oxygen channel. The mean of the data for each second in the sequence is calculated, and the maximum and minimum fluctuation points within that second are marked and their corresponding time labels are recorded. The peaks and trend turning points previously extracted from ECG and blood pressure data are compared, and the time points are mapped to the blood oxygen time axis. A one-to-one time pairing operation is performed on all fluctuation points to construct a multi-channel fluctuation point reference table. At the same time, the time difference between each reference point is calculated to identify delayed or premature response phenomena. In a monitoring of a patient with high altitude hypoxia, the ECG peak appeared at 21.2 seconds, the blood pressure turning point appeared at 21.5 seconds, and the blood oxygen change point appeared at 21.9 seconds. The cross-channel delays of this set of data were recorded as 0.3 seconds and 0.7 seconds, respectively. The time synchronization table of the three types of channel data is updated according to the mapping relationship of all point pairs, and a label field is added to record its turning direction and change amplitude, forming a fluctuation correspondence table of multi-channel physiological signals under a unified time axis, and obtaining the temporal fluctuation localization characteristics.

[0076] Please see Figure 4 The image feature analysis module includes:

[0077] The image parameter screening submodule analyzes the distribution of color and brightness channels in the main region of trauma images based on temporal fluctuation localization features, determines whether the channels have a complete data structure, filters samples with missing or abnormal channel distribution, completes the channel content, and obtains the integrated data structure of the main region channels.

[0078] The trauma images captured at the corresponding time points are extracted and processed in the main region analysis process. The original matrices of the image's color and luminance channels are accessed. For the color channels, the pixel density distribution of the red, green, and blue channels is extracted, and the effective distribution ratio of pixel values ​​within each channel is calculated. If the area with zero or constant pixel values ​​in any channel exceeds 25% of the entire image area, the data structure of that channel is considered incomplete. Simultaneously, luminance gradient detection is performed on the luminance channel, calculating the degree of luminance jump between adjacent pixels. If the luminance jump frequency continuously exceeds a threshold area less than 15% of the total image area, the luminance distribution is considered discontinuous. [The process is then repeated.] The two criteria are used to screen out image samples with missing data or structural anomalies. For missing color channel content, the average value of pixels at the same position in adjacent channels and temporally adjacent frames is used for interpolation. For example, in an image taken by a portable camera, there is a region in the green channel with a constant pixel value of 0, covering 28% of the total pixels. This is judged as a missing channel, and the region is interpolated by the average value of pixels at the same position in the red and blue channels. For areas with discontinuous brightness, the pixel adjustment is completed by spatial local mean expansion. The color channel and brightness channel are integrated into the same structure, and a channel integrity identifier field is added to form the main region channel data integration structure.

[0079] The format standardization processing submodule is based on the main region channel data integration structure. It calculates the spatial resolution and color space consistency of the image to be processed, judges the samples with different formats, adjusts the pixel structure and unifies the encoding method, and filters the images that meet the unified standard to obtain a standardized pixel structure set.

[0080] Based on the main region channel data integration structure as input, each image undergoes dual-dimensional standardization processing of spatial and color attributes. First, the number of rows and columns of pixels in the image is extracted to calculate the resolution and determine if it falls within a standard preset range. The effective resolution range is set to 640×480 to 1920×1080. If the image pixel size exceeds or falls below this range, it is marked as an image with inconsistent resolution. Second, the uniformity of the color space structure is analyzed. By reading the image encoding header information, the current encoding space type used by the image is extracted. The standard definition is RGB color space. If the image is detected to use YUV or CMYK space, it is marked as an image with inconsistent color space. For images with resolution differences, their pixel arrays are resampled and adjusted using bilinear interpolation. The method unifies the image size to 1280×720. For samples with inconsistent color spaces, all pixels are spatially mapped and converted channel by channel. When mapping CMYK to RGB, the values ​​are replaced according to the color conversion matrix. At the same time, the RGB three-channel data is normalized by 8 bits to ensure that all image data are consistent in terms of color bit width, encoding space, and resolution. In an actual image acquisition, a total of 85 images were acquired, of which 11 images used YUV space and had a resolution of 352×288. These images were found to be inconsistent with the standard. After being uniformly adjusted, they were encoded into RGB structure, the size was adjusted to 1280×720, and they were marked as images with normalized format. Images that passed the above verification process and conformed to the unified format standard were retained to form a standardized pixel structure set.

[0081] The texture feature comparison submodule, based on a standardized pixel structure set, calculates the texture density and uniformity of each patch. Combining temporal fluctuation localization features, it compares the image texture indices with the brightness sampling content at the same time interval, using the following formula:

[0082] ;

[0083] Get texture-luminance response map offset By filtering the delay structure of texture parameters, a set of linkage anomaly indicators is obtained, among which... Indicates the first Texture density factor of each tile, Indicates the first Texture uniformity factor of each tile Indicates the first Each time-series point corresponds to a brightness mean sampling unit. Indicates the total number of texture tiles. Indicates the number of brightness sampling points;

[0084] The texture-luminance response mapping offset refers to the process of dividing the main region of a trauma image, calculating the texture density and texture uniformity of each patch, using these as local structural features, and comparing them with the mean luminance samples obtained at the same time point. By comparing the overall structure synthesized from the texture features of each patch with the average state of a set of mean luminance samples in terms of numerical "offset" (i.e., the mean of the absolute values ​​of the differences), it reflects the consistency or difference in the mapping between the texture structure response of the image and the luminance temporal samples.

[0085] If the texture-luminance response mapping offset is small, it indicates that the texture distribution of the image region at that moment is relatively consistent with the trend of the luminance sampling distribution, and the local texture features are closely related to the luminance response. If the offset is large, it indicates that there is a difference or delay between the changes in texture features and the luminance response, suggesting that structural or tissue abnormalities have occurred in the trauma area. This can serve as a numerical representation for multimodal data linkage, used to comprehensively judge the correlation and anomalies between image features and physiological time-series signals, providing a data foundation for automatically identifying potential anomalies in trauma areas in subsequent stages such as anomaly index screening and risk level determination.

[0086] The image is divided into several patches of fixed size. The texture density factor and texture uniformity factor of each patch are extracted as structural feature parameters. Texture density is obtained by averaging the gray-level differences between adjacent pixels in the patch, in gray levels. Texture uniformity is calculated based on the proportion of the main diagonal elements of the patch's gray-level co-occurrence matrix, in percentage. Simultaneously, the average brightness is extracted from the image brightness channel at the corresponding time point as a reference, also in gray levels, and linear normalization is applied. )Will , Mapped to Range, brightness parameters Since all three have the same value, there is no space for normalization processing. Therefore, constant values ​​are maintained. The original parameters and normalized values ​​of the three tiles are as follows:

[0087] Tile T1: , , The corresponding normalization result is , , (Set constants);

[0088] Tile T2: , , The normalization result is , , ;

[0089] Tile T3: , , The normalization result is , , Substitute the above data into the formula:

[0090] ;

[0091] The structural strength calculations and offset analyses are performed sequentially as follows:

[0092] Plot 1 The offset is ;

[0093] Plot 2 The offset is ;

[0094] Plot 3 The offset is After averaging the three, we get:

[0095] ,

[0096] Based on the statistical results of the linkage characteristics of multiple sets of trauma images and physiological signal co-acquisition data, a texture-luminance response offset was set. The reference range for determining structural consistency is as follows:

[0097] when When the time is right, it indicates that there is a high degree of consistency between changes in image texture structure and brightness temporal response, and the features of each channel show coordinated changes with no obvious structural delay or difference;

[0098] when When the response is moderate, it indicates that there is a certain degree of temporal lag or local structural mismatch between the texture intensity and brightness changes of the image structure, which requires further screening.

[0099] when When the relationship between texture structure and brightness response is weak or misalignment occurs, it indicates structural abnormalities caused by factors such as blurred tissue boundaries, uneven lighting, or image defocus within the trauma area, and should be treated as an abnormal patch.

[0100] The calculation result is Located in the second interval mentioned above The results indicate that at least some patches in the main region of the image exhibit asynchronous relationships between the combined structural strength of their texture density and texture uniformity and the timing of brightness sampling. This suggests a certain degree of difference in organizational structure and illumination response within the image, which can be considered as a region with structural response delay. This meets the criteria for screening abnormal regions in linkage analysis. Therefore, this offset result is directly used as a standard to screen out patch numbers and corresponding feature vectors that meet the response offset criteria from the standardized pixel structure set. The formula converts two different texture attributes into a unified metric by square rooting the sum of their squares and constructs a response consistency mapping between texture and brightness by subtracting from the average structure of brightness response. This allows for multi-channel linkage alignment detection of fine structures without the need for direct comparison of global pixels in the image.

[0101] Please see Figure 5 The anomaly level determination module includes:

[0102] The collision feature extraction submodule analyzes the acceleration changes and impact duration in the collision monitoring signal based on the linkage anomaly index set, judges the trend of impact amplitude and direction changes within the sampling interval, identifies continuous signal segments and optimizes the occlusion interference judgment process, compares the differences in physical characteristics of each segment, calculates the motion performance of the impact segment, and obtains the collision physical feature index.

[0103] Based on the time points and image anomaly indicators recorded by the linkage anomaly indicators, collision monitoring signals within the corresponding time period are retrieved, and acceleration data is extracted frame by frame. The maximum acceleration change value and direction change angle within each sampling interval are statistically analyzed. In operation, the three-axis acceleration components are first extracted from the acceleration channel, and the acceleration change between consecutive frames is calculated. When the change value is greater than 2.5 in two consecutive sampling segments, it is judged that a strong impact has occurred. At the same time, the angle channel information is extracted to identify the direction deflection. The direction data within five frames before and after the impact time point is compared and analyzed. If the direction deflection amplitude continuously exceeds 15 degrees and the number of consecutive frames is not less than 3 frames, it is considered that there is a trend of rapid direction shift. Several impact segments are identified based on the two conditions of impact amplitude and direction change. Continuity detection is performed on each impact segment, and the maximum acceleration change value and direction change angle within each segment are calculated. The validity of a signal segment is determined by counting the number of consecutive frames. If the number of valid frames is less than 80% of the total number of frames in the segment, it is marked as an obstructed segment. Based on this, the obstructed frames are compensated using the average of adjacent frame data to complete the interference optimization operation. Subsequently, the physical parameters extracted from each segment are analyzed for differences. The maximum acceleration value, direction change angle, and impact duration of the segment are used to calculate the impact intensity level, direction change coefficient, and sustained stability index of the segment. In a set of actual data, segment A recorded an acceleration peak of 4.2, a direction change angle of 28 degrees, and an impact duration of 1.6 seconds. The corresponding data for segment B were 2.9, 12 degrees, and 0.8 seconds. After comparing the two with multiple indicators, they were classified into different levels. Segment A was classified as a high-intensity impact segment, and segment B as a medium-intensity impact segment, forming a structured set of collision physical characteristic indicators.

[0104] The physiological fluctuation identification submodule calculates the amplitude changes of systolic blood pressure, heart rate and blood oxygen signals within the corresponding acquisition period based on collision physical feature indicators, judges the trend consistency of physiological parameters, filters the intervals where the change amplitude is higher than the amplitude threshold, and compares the fluctuation correlation between multi-channel signals to obtain physiological abnormality response factors.

[0105] After receiving the collision physical characteristic indicators, systolic blood pressure, heart rate, and blood oxygen data sequences within the corresponding acquisition period are extracted according to the time range of the collision segment. The maximum and minimum values ​​within the same period in each channel are statistically analyzed, and the amplitude variation value of each physiological parameter is calculated. The systolic blood pressure change is judged based on whether the difference is greater than 10 mmHg; the heart rate change is judged based on whether it exceeds 8 beats per minute; and the blood oxygen saturation change threshold is set at 2 percentage points. During processing, the amplitude variation of each channel is compared with the corresponding threshold. If the change amplitude exceeds the threshold, it is considered a high-fluctuation segment. Further analysis is conducted to determine whether high fluctuations occur simultaneously in multiple physiological channels. Three channels are extracted separately. By comparing the start and end times of fluctuations within the same time period and the direction of fluctuation, if the median values ​​of each channel deviate from the steady-state value in the same direction and the time interval is less than 2 seconds, it is determined to be a segment with a consistent trend. In a practical application, within a certain period of time, the patient's heart rate increased from 82 to 96, systolic blood pressure increased from 120 to 137, and blood oxygen decreased from 97 to 94. The magnitudes of these three changes were 14, 17, and 3, respectively, all of which exceeded the set thresholds. Moreover, the directions of change of the three values ​​occurred simultaneously within 1 second after the start of sampling. Based on this, the segment was marked as a physiological synchronous fluctuation segment, and the magnitude of change, trend direction, and fluctuation delay time of each channel were structurally registered and extracted as a component of the physiological abnormal response factor.

[0106] The critical illness risk scoring submodule, based on physiological abnormality response factors, jointly analyzes the fluctuation range of blood pressure sequences, impact amplitude, and heart rate regulation characteristics, using the following formula:

[0107] ;

[0108] Calculate the dependent variable for abnormal performance, locate the proportion of sampling time covered by its continuous offset segments, and obtain the graded risk determination factor. ,in, This represents the maximum observed data in the blood pressure sequence, indicating the highest blood pressure data detected within the specified sampling period. This represents the average data in the blood pressure sequence, indicating the average level of all blood pressure data within the sampling period. This is the coefficient of variation of the collision amplitude, used to reflect the variation amplitude or fluctuation intensity of the collision signal within the sampling period. The root mean square statistical results of the blood pressure sequence reflect the overall energy level of the blood pressure signal. The peak amplitude of the blood pressure sequence is the difference between the largest and smallest data points in the blood pressure sequence.

[0109] The risk assessment factor is an assessment quantification formed by normalizing and comprehensively calculating the abnormal manifestations of patients in trauma treatment scenarios using extreme fluctuations in blood pressure sequences (such as the deviation between the maximum observed data and the average data), changes in collision signals (reflected by a dimensionless coefficient of change), and the overall intensity and amplitude of blood pressure signal fluctuations. The larger the factor, the more obvious the combined abnormal manifestations of physiology and collision related to the patient in the current sampling period, reflecting a higher level of trauma risk, which helps in automatic triage and priority dispatch of emergency care.

[0110] The maximum observed data in a blood pressure series is defined as Its value is obtained by comparing all instantaneous blood pressure data within the sampling period one by one. In this embodiment, the sampling sequence is: ,but mmHg; mean of the sampling sequence as The calculation method is to sum all samples and then divide by the sample size. In this example, it is... mmHg, using the above two items, the maximum deviation of blood pressure can be calculated. mmHg; coefficient of variation of the impact amplitude This is a dimensionless parameter, its value is obtained by dividing the standard deviation of the impact acceleration by the average acceleration. In this example, the monitoring equipment records a standard deviation of 0.74g and an average acceleration of 0.9g. Therefore... Root mean square of blood pressure sequence The blood pressure signal energy level is represented by the square root of the average of the squares of all sampling points. In this sequence, the root mean square is... mmHg; peak blood pressure amplitude This is the difference between the maximum and minimum values ​​in the blood pressure sequence; in this example, it is... mmHg. Linear normalization was applied to... The parameters are processed using an interval approach, with the normalized interval set based on clinical range standards. The reference minimum value for blood pressure parameters is set at 90 mmHg, and the maximum value at 180 mmHg. The normalized interval for peak-to-peak amplitude is set to 20 mmHg to 80 mmHg. The normalized parameters are shown below:

[0111] ;

[0112] ;

[0113] (dimensionless)

[0114] ;

[0115] ;

[0116] Substitute the normalized parameters into the scoring formula:

[0117] ;

[0118] The following operations are performed:

[0119] Calculate the deviation term: ;

[0120] Multiply by the collision coefficient: ;

[0121] Calculation of squared terms: , ;

[0122] Combine denominators and take the square root: ;

[0123] Substituting into the formula, the calculation result is:

[0124] ;

[0125] The scope is divided into three risk level segments to guide subsequent scheduling priority decisions, as defined below:

[0126] when Defined as a low-risk zone, it means that the patient's current physiological fluctuations and collision characteristics are generally within a controllable range, do not trigger resource priority response, and only maintain the normal treatment path.

[0127] when Defined as a medium-risk zone, indicating that the patient's condition is significantly fluctuating, the system should activate the early warning mechanism and prepare transfer resources, entering a waiting-for-schedule state.

[0128] when Defined as a high-risk zone, it indicates that physiological and shock signals are simultaneously and violently linked, requiring immediate execution of high-priority scheduling strategies and coordination with the trauma center response.

[0129] This result indicates that the risk assessment factor for tiered risk is... , belongs to the interval The low-risk segment indicates that the patient's blood pressure sequence has a small deviation and limited collision signal fluctuation within the current sampling period, and is generally in a stable edge region. In subsequent scheduling stages, this patient will be marked as a regular queuing task and will not enter the red or yellow priority channels. This value is directly used as a risk classification factor input into the standardized calculation of the scheduling module, forming the basis for ranking with other patient results and used to generate scheduling priority configuration. The formula introduces the product between the blood pressure fluctuation deviation term and the collision fluctuation coefficient, and normalizes it by the square root of the square of the blood pressure energy and the fluctuation amplitude, constructing a classification factor system that takes into account both the absolute degree of deviation and the ability to change relative, which has strong discriminative power and scalability.

[0130] Please see Figure 6 The priority scheduling hierarchy module includes:

[0131] The location parsing submodule analyzes the spatial coordinates between patient location information and hospital location information based on the graded risk judgment factor, compares the geographical relative relationship between the two, determines the spatial distribution of the current location and the target hospital, filters the location parameters, and obtains the geographical distance calculation results.

[0132] The system extracts the patient's real-time location data, obtaining the longitude and latitude coordinates of the current location via GPS. Simultaneously, it retrieves the geographic coordinates of the target hospital from the hospital resource library. The patient's coordinates and the hospital's coordinates are treated as a pair of spatial points for two-dimensional spatial distance analysis. First, a coordinate transformation is performed, converting the original GPS format to a unified geodetic coordinate system. Based on this, the straight-line distance between the two points is calculated as a basic geographic reference value. Then, a road offset correction parameter is introduced, using local map service data to adjust the straight-line distance for route accessibility. Finally, the existence and accessibility of a road between the two points are determined, specifically whether at least one city road exists. The main road connecting these two points serves as the benchmark for accessibility. In this example, the patient is located at latitude and longitude (121.4512, 31.2234), and the hospital is located at coordinates (121.4778, 31.2305). The straight-line distance between the two points is calculated to be 3.2 kilometers. The map service indicates that there are three road routes, none of which include highway entrances. The urban expressway is selected as the primary recommended route, and the road grade label of each segment in the route is used for scoring. After eliminating hospitals with no accessible routes, the hospital target with the shortest local distance and road grade that meets the accessibility requirements is finally selected, and the calculated distance of this route segment is output as the geographic distance calculation result.

[0133] The traffic time assessment submodule optimizes the path data obtained from the geographical distance calculation results, analyzes the road grade and traffic information of each transfer route, judges the connectivity and traffic trend of each path node, filters continuous travel routes, and obtains path traffic assessment indicators.

[0134] Further, real-time and historical traffic information for each road node in the path is obtained from the map service interface. First, each road segment on the path is traversed, and its road level field is extracted, assigning a weighted score to each level. For example, urban expressways have a weight of 1.0, arterial roads 0.8, secondary roads 0.5, and highways 1.2. Next, the traffic status of each node in the current time period is extracted. The current congestion level is determined by collecting road segment traffic speed data. A speed below 30% of the road segment's speed limit is marked as severely congested, below 60% as moderately congested, and above as smooth. Then, connectivity is assessed for each node in the entire path to determine whether there are medium-to-high-speed connections between nodes. The decision is based on whether a path is interrupted or requires a detour. If any node is interrupted, the path is removed. A traffic score is evaluated for each complete path. The score is calculated by combining the path length, average speed, and node connectivity. In one evaluation task, path A is an urban expressway with a length of 4.1 kilometers and an average speed of 28 kilometers per hour. There are no interruptions. Path B is a main road with a length of 4.6 kilometers and an average speed of 22 kilometers per hour. There is a closure at one intersection. Path B is determined to be a discontinuous path and is removed. Only path A is retained and its traffic score is recorded as 0.92. This path's traffic evaluation index is selected as the corresponding path for this patient.

[0135] The priority sorting configuration submodule arranges the graded priority parameters of each patient based on the route access assessment index, adjusts the correspondence between patients and emergency vehicles, establishes the priority numbering order, and obtains the dispatch priority configuration.

[0136] Based on the calculated route accessibility assessment indicators, and combined with the risk level and corresponding time response indicators in each patient's risk assessment factors, the system first extracts the geographical route accessibility score, the patient's current risk level, and the distance information between the patient's location and the emergency vehicle from the list of all patients currently awaiting treatment. Then, parameter normalization is performed. Risk levels are set on a 10-level scale: values ​​of 8 to 10 represent high risk, 5 to 7 represent medium risk, and 4 and below represent low risk. Accessibility scores are set between 0 and 1, and distances are assigned as 1 for the shortest distance and 0 for the longest. All normalized parameters are summarized into a sorting parameter matrix. The system prioritizes all patients by assigning a risk level weight of 0.5, a passage score weight of 0.3, and a distance weight of 0.2. All patient scores are then ranked from highest to lowest to form a priority list. The system then retrieves the location data of currently available emergency vehicles and matches the vehicle closest to the patient in front of it as the dispatch target. In one actual dispatch task, patient A has a score of 0.91, patient B has a score of 0.82, and vehicle V3 is identified as being only 1.1 kilometers away from patient A. Therefore, a dispatch relationship between V3 and A is established, and subsequent vehicle numbers are sequentially configured to correspond to them, outputting the dispatch priority configuration.

[0137] In addition, the system mainly consists of CAVE immersive interaction, virtual-real integration technology, AI intelligent recognition technology, and physiological driving technology. The virtual-real integration features scene reconstruction (simulating real trauma emergency scenarios such as multiple injuries, massive bleeding, and airway obstruction, supporting highly immersive operational training; physical simulation equipment (such as intelligent bionic patient models) interacts with the virtual scene, providing tactile feedback to enhance the realism of operations); multi-role collaborative drills (supporting multiple people to collaborate online simultaneously, covering roles such as attending physicians, nurses, and surgeons, recreating the complete emergency process (assessment-treatment-transfer); intelligent task allocation; real-time monitoring of team cooperation efficiency; strengthening awareness of the "golden hour" of treatment); intelligent assessment and debriefing (automatically recording key indicators such as operation time, process standardization, and team communication, generating visual assessment reports; supporting 3D scene playback for easy team review of errors and targeted improvements); and a standardized course library (built-in international standard course modules such as "Advanced Trauma Life Support (ATLS)" and "Emergency Trauma Team Collaboration Guidelines," supporting custom case editing).

[0138] The process includes the following steps:

[0139] Step 1, Physical Operation Sensing and Data Acquisition:

[0140] 1. Operation execution:

[0141] Trainees use real clinical instruments (such as cervical collars, endotracheal tubes, tourniquets, etc.) to perform trauma care procedures (such as cervical spine fixation, airway establishment, pressure hemostasis, etc.) on polymer simulation mannequins.

[0142] The system supports multi-role collaborative drills, allowing multiple trainees, including emergency physicians, nurses, and surgeons, to collaborate in the same CAVE immersive space to complete steps such as "assessment-treatment-transfer".

[0143] The CAVE immersive display system provides a 360° three-dimensional clinical scene, allowing trainees to operate in a highly realistic emergency environment and gain a stronger sense of immersion.

[0144] 2. Data Collection:

[0145] Simulating a multimodal intelligent sensor array integrated inside the human body (such as position sensors, pressure sensors, and optical sensors), it monitors changes in operational posture and force in real time, accurately collects key physical parameters, and automatically judges the accuracy of the operation through AI algorithms. For example:

[0146] ① Check whether the endotracheal tube has passed through the glottis correctly and has not been mistakenly inserted into the esophagus;

[0147] ② Determine whether the tourniquet pressure has reached the arterial hemostasis threshold and maintained for an effective time;

[0148] ③ Identify whether the intravenous puncture needle has accurately entered the vascular model and maintained a stable access.

[0149] The system can also record the operation trajectory, force distribution, and duration, providing complete raw data for subsequent review and evaluation.

[0150] 3. Signal conversion:

[0151] The multi-channel physical signals (such as pressure, angle, position, conductivity, etc.) collected by the sensor are converted into a standardized digital signal stream by the signal conversion module;

[0152] The system performs real-time filtering, noise suppression, and feature extraction on the raw data, and uses an AI recognition model to determine the current operation type (such as intubation, compression, bandaging, infusion, etc.).

[0153] The processed signal stream is input into the virtual patient model, achieving high-precision data mapping from the physical world to the virtual world.

[0154] Step 2, Virtual Patient Status Update and Physiological Driven Calculation:

[0155] 1. Status Update:

[0156] The collected standardized operational data is input into the virtual patient model, triggering dynamic updates to the patient's status, for example:

[0157] ① If step one detects that the catheter has correctly entered the trachea, the model updates the "airway status" to "artificial airway established";

[0158] ② If the tourniquet pressure is detected to be effective in step one, the model updates "Right leg artery bleeding" in "Circulation Status" to "Controlled";

[0159] If a cardiopulmonary resuscitation (CPR) procedure is detected in step one, the "Resuscitation Status" flag is activated, and the system enters the spontaneous circulation recovery simulation phase.

[0160] The system will comprehensively consider the timing, intensity, and duration of the operation to automatically evaluate the quality of the operation and match it with preset physiological thresholds.

[0161] 2. Physiological calculations:

[0162] The physiological drive engine is triggered, and based on the updated patient status, it performs real-time calculations using pre-set physiological models (such as cardiovascular, respiratory, and metabolic models) to deduce the dynamic changes in the patient's vital signs. This process is dynamic and continuous, rather than relying on pre-set scripts, ensuring scientific, continuous, and reliable feedback. For example:

[0163] Once the major bleeding points are effectively controlled, the engine will calculate and present the process of decreased heart rate, increased blood pressure, and improved perfusion.

[0164] Once an intravenous access is established and intravenous infusion begins, the system will automatically adjust indicators such as blood volume and central venous pressure.

[0165] If an error is made (such as the tube being accidentally inserted into the esophagus), the system will simulate reactions such as respiratory distress and decreased blood oxygenation.

[0166] Step 3, Generation and Presentation of Multimodal Feedback:

[0167] (1) Rendering generation:

[0168] The 3D rendering engine receives vital sign data and virtual patient status information output by the physiological driving engine, and generates visual, auditory and tactile feedback data streams in real time. All feedback is based on real-time calculation results to ensure that the scene and physiological performance are completely synchronized.

[0169] (2) Visual feedback:

[0170] The CAVE immersive system presents the physiological changes of a virtual patient in dynamic visuals:

[0171] The changes in trauma at the corresponding site of the virtual patient (such as cessation of bleeding and the effect of wound dressing);

[0172] Real-time changes in electrocardiogram, blood pressure, and blood oxygen waveforms and values ​​on a simulated monitor;

[0173] Vital signs fluctuate in real time during operation, and trainees can judge the effectiveness of the operation through visual signals.

[0174] (3) Auditory feedback:

[0175] The spatial audio system outputs multiple sound sources, including ECG monitoring alarm sounds, breathing sounds, air passing through water sounds, and blood pressure cuff deflation sounds, creating a realistic clinical environment soundscape.

[0176] (4) Immersive presentation:

[0177] The generated multimedia data is output through CAVE panoramic projection to construct a 360° immersive training environment;

[0178] The fusion of visual, auditory, and tactile senses allows trainees to be fully immersed in a virtual ward, achieving a realistic interactive experience where "what you see is what you do, and what you do is what you change," significantly improving training focus and operational feedback sensitivity.

[0179] Step 4, Human-Computer Interaction and Closed-Loop Training Cycle:

[0180] (1) Trainee decision-making and re-operation:

[0181] Trainees make clinical judgments based on feedback from the CAVE system and perform new procedures (returning to step one), forming a closed-loop training mechanism of "operation - feedback - re-judgment - re-operation";

[0182] The AI ​​guidance module can provide intelligent prompts or intervention options at key points (such as airway management failure or unstable circulation) to help trainees optimize their decision-making logic.

[0183] (2) Multi-person collaboration:

[0184] It supports multiple trainees entering the CAVE environment at the same time, identifies the position and role of different trainees through spatial tracking technology, and allows them to perform different tasks in the ABCDE process in parallel (such as one person being responsible for airway management while another person controls bleeding at the same time), and to communicate and collaborate naturally in a shared field of vision.

[0185] It can automatically assign tasks, monitor team coordination in real time, and assess the efficiency of the "golden hour" rescue.

[0186] Breaking through the limitations of traditional individual training, it promotes the improvement of team emergency collaboration and communication capabilities.

[0187] (3) System evaluation:

[0188] The entire training process is monitored in real time. The intelligent scoring function in the system backend records the entire training process in real time, including operation time, sequence, quality and physiological results; it automatically generates quantitative evaluation reports and personalized review videos, supports 3D scene playback and error step analysis; after training, it can output comprehensive scores for individuals and teams, and provide targeted improvement suggestions for skill gaps, realizing a learning closed loop of "visual evaluation + intelligent review + continuous optimization".

[0189] Through the four closely linked steps described above, an intelligent closed loop of "physical operation - state perception - physiological calculation - immersive feedback" is constructed, successfully integrating real tactile operation, dynamic physiological simulation and immersive visual performance. This solves key problems in existing technologies such as sensory fragmentation, rigid response and difficulty in collaboration, and provides a new methodological path for achieving efficient and standardized emergency team training for severe trauma.

[0190] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A system for early assessment and treatment of severe trauma based on virtual-real interaction, characterized in that, The system includes: The scene risk perception module is based on IoT wearable devices. It analyzes changes in motion signals, lighting conditions and collision monitoring signals to determine the order of on-site risks, adjusts the acquisition frequency of each monitoring channel, and obtains the risk monitoring acquisition rate. Based on the risk monitoring acquisition rate, the dynamic physiological extraction module identifies continuous fluctuations in electrocardiogram signals, analyzes heart rate trends, blood pressure changes, and blood oxygenation cycle fluctuations, locates abnormal moments of multiple physiological parameters, and obtains temporal fluctuation location features. Based on the temporal fluctuation positioning features, the image feature analysis module analyzes the trauma images acquired by the portable imaging device, screens for missing data, supplements the data, filters texture parameters, compares the images with physiological changes, and obtains a set of linkage abnormality indicators. The image feature analysis module includes: Based on the temporal fluctuation positioning features, the image parameter screening submodule analyzes the distribution of color and brightness channels in the main region of the trauma image, determines whether the channels have a complete data structure, filters samples with missing or abnormal channel distribution, completes the channel content, and obtains the main region channel data integration structure. The format standardization processing submodule calculates the spatial resolution and color space consistency of the image to be processed based on the main region channel data integration structure, identifies samples with different formats, adjusts the pixel structure and unifies the encoding method, and filters images that meet the unified standard to obtain a standardized pixel structure set. The texture feature comparison submodule calculates the texture density and uniformity of each patch based on the standardized pixel structure set, combines the temporal fluctuation positioning features, compares the image texture index with the brightness sampling content under the same time sequence, obtains the texture-brightness response mapping offset, filters the delay structure of texture parameters, and obtains the linkage anomaly index set. The abnormality level determination module analyzes collision monitoring and blood pressure fluctuation parameters based on the aforementioned set of linked abnormal indicators, weights multiple data, and determines the patient's abnormal condition through the critical illness discrimination criteria to obtain the graded risk determination factor. The anomaly level determination module includes: The collision feature extraction submodule analyzes the acceleration changes and impact duration in the collision monitoring signal based on the linkage anomaly index set, judges the trend of impact amplitude and direction changes within the sampling interval, identifies continuous signal segments and optimizes the occlusion interference judgment process, compares the differences in physical characteristics of each segment, calculates the motion performance of the impact segment, and obtains the collision physical feature index. The physiological fluctuation identification submodule calculates the amplitude changes of systolic blood pressure, heart rate and blood oxygen signals within the corresponding acquisition period based on the collision physical characteristic index, judges the trend consistency of physiological parameters, filters the intervals where the change amplitude is higher than the amplitude threshold, and compares the fluctuation correlation between multi-channel signals to obtain the physiological abnormality response factor. The critical illness risk scoring submodule, based on the aforementioned physiological abnormality response factors, jointly analyzes the fluctuation range of blood pressure sequence, the amplitude of collision impact, and the characteristics of heart rate regulation, calculates the abnormality dependent variable, locates the proportion of sampling time covered by its continuous offset segment, and obtains the graded risk judgment factor. The priority scheduling and grading module analyzes the patient's geographical location and travel time based on the grading risk determination factors, and ranks the priority of grading patients according to the grading index, distance and travel time standardization to obtain the scheduling priority configuration.

2. The early assessment and treatment system for severe trauma based on virtual-real interaction according to claim 1, characterized in that, The risk monitoring and acquisition rate includes scene category coding, channel priority sorting, and real-time adjustment markers. The temporal fluctuation positioning features include anomaly detection labels, change distribution curves, and key positioning moments. The linkage anomaly indicator set includes color saturation, average brightness, and texture grading factors. The graded risk judgment factors include risk classification codes, level assessment numbers, and reference comparison labels. The scheduling priority configuration includes priority labels, scheduling object identifiers, and transfer recommendation schemes.

3. The early assessment and treatment system for severe trauma based on virtual-real interaction according to claim 1, characterized in that, The scenario risk perception module includes: The wearable signal acquisition submodule is based on IoT wearable devices. It uses data sources deployed on the patient's chest and wrist to organize the data streams of acceleration trajectory, direction offset channel and ambient illumination channel in a time synchronization manner, and matches the time tags output by each channel to obtain a set of multi-channel synchronous acquisition segments. The motion trend extraction submodule compares the acceleration and direction offset channel data of each segment based on the multi-channel synchronous acquisition segment set, calculates the change amplitude of velocity and direction between segments, determines the consistency or turning point of the motion trend, and marks the trend change segment and duration in combination with the segment number to obtain the continuous trend change annotation group. The channel frequency configuration submodule analyzes the illuminance channel data and collision signal trajectory within the corresponding segment based on the continuous trend change annotation group, compares the illuminance uniformity and the continuous characteristics of collision changes, and adjusts the data acquisition priority of each monitoring channel to obtain the risk monitoring acquisition rate.

4. The early assessment and treatment system for severe trauma based on virtual-real interaction according to claim 1, characterized in that, The dynamic physiological extraction module includes: The ECG fluctuation identification submodule analyzes the ECG signal of the priority channel based on the risk monitoring acquisition rate, judges the waveform changes of continuous heartbeat cycles, identifies the rising and falling phases of each cycle, filters the intervals with consistent waveform stability and fluctuation trend, and adjusts the signal label distribution in combination with the sampling frequency to obtain a set of ECG fluctuation change indicators. The parameter trend comparison submodule compares heart rate data and blood pressure data based on the ECG fluctuation change index set, analyzes the trend direction and change range of the two sets of data, and filters the intervals with synchronous trend changes and trend consistency to obtain the synchronous trend matching segment set. The temporal feature localization submodule analyzes the fluctuation distribution of data in the blood oxygen channel within the cycle based on the synchronization trend matching segment set. By comparing the turning points of ECG and blood pressure parameters, it maps all fluctuation points to time coordinates, establishes the fluctuation correspondence of multi-channel physiological signals, and obtains temporal fluctuation localization features.

5. The early assessment and treatment system for severe trauma based on virtual-real interaction according to claim 1, characterized in that, The priority scheduling hierarchy module includes: The location analysis submodule analyzes the spatial coordinates between the patient's location information and the hospital's location information based on the graded risk judgment factor, compares the geographical relative relationship between the two, determines the spatial distribution of the current location and the target hospital, filters the location parameters, and obtains the geographical distance calculation result. The traffic time assessment submodule optimizes the path data obtained from the geographical distance calculation results, analyzes the road grade and traffic information of each transfer route, judges the connectivity and traffic trend of each path node, filters continuous traffic routes, and obtains path traffic assessment indicators. The priority sorting configuration submodule arranges the graded priority parameters of each patient based on the route accessibility assessment index, adjusts the correspondence between patients and emergency vehicles, establishes the priority numbering order, and obtains the dispatch priority configuration.

6. The early assessment and treatment system for severe trauma based on virtual-real interaction according to claim 1, characterized in that, The IoT wearable device refers to an Internet of Things device that can be worn on the patient to collect vital signs and environmental data in real time. The physiological change refers to the moment when abnormal physiological parameters are extracted, which corresponds one-to-one with the image acquisition time. The patient's abnormal condition refers to the patient's risk level or degree of urgency under the current classification.

Images