A method and system for collecting and managing information on clinical emergency medical equipment

Abstract

This application relates to the field of computer technology and discloses a method and system for collecting and managing information on clinical emergency medical equipment. The method includes: acquiring video frame sequences through a visual perception unit, constructing a three-dimensional spatial map, and locating the patient and equipment in real time; acquiring vital sign data streams from the equipment; and automatically binding the equipment and patient identities based on dual criteria of spatial proximity and trajectory correlation to generate structured emergency records. The system includes a visual perception unit, a three-dimensional mapping and positioning unit, a data receiving unit, a spatial behavior analysis unit, and an intelligent binding decision-making unit. This application significantly improves the accuracy, timeliness, and data integrity of emergency medical information collection through non-intrusive and automated association.

Description

Technical Field

[0001] This invention belongs to the field of computer technology, specifically relating to a method and system for information collection and management of clinical emergency equipment. Background Technology

[0002] With the rapid development of smart healthcare systems, clinical emergency scenarios place higher demands on the real-time performance, accuracy, and automation of equipment information management. In high-pressure, fast-paced emergency environments such as emergency rooms and ambulances, medical staff need to quickly link various life support devices (such as ventilators, defibrillators, and infusion pumps) to specific patients to ensure the accurate collection of vital sign data and the reliability of subsequent treatment decisions. Traditional methods rely on manual operation, including manually scanning device QR codes, entering device numbers, or selecting associated patients via touchscreens. This process not only consumes valuable rescue time but is also prone to mismatches, omissions, or duplicate bindings under high-intensity workloads, leading to chaotic data flow, distorted medical records, and even medical errors.

[0003] Automatic device identification technology based on the Internet of Things (IoT) and intelligent sensing is gradually becoming a research hotspot. Existing solutions attempt to introduce methods such as radio frequency identification (RFID), Bluetooth beacons, or Wi-Fi positioning to achieve device identification and rough location tracking. However, due to limitations such as signal interference, multipath effects, and complex indoor environments, their positioning accuracy is insufficient to support the requirement of precise one-to-one binding between "device and patient." Especially in emergency situations where multiple people are present and equipment is moving intensively, matching mechanisms that rely solely on signal strength cannot effectively distinguish between spatially adjacent but logically unrelated devices and patients, often resulting in misbinding or binding delays.

[0004] Existing technologies generally suffer from the following shortcomings: First, they lack the ability to dynamically model the three-dimensional spatial structure of emergency scenarios, making it impossible to accurately capture the relative positional relationship between devices and patients. Second, the device identification and data binding processes are disconnected, failing to integrate spatial proximity, movement trajectory correlation, and device status information for judgment. Third, existing systems mostly employ rigid threshold matching strategies, which are ill-suited to adapting to non-ideal operating conditions such as frequent device movement, obstruction, and signal fluctuations during emergency procedures. These problems result in insufficient robustness of existing solutions in real clinical environments, hindering the achievement of highly reliable device-patient intelligent binding without manual intervention, severely restricting the efficiency improvement and safety closed-loop construction of emergency information systems. Therefore, there is an urgent need for a clinical emergency device information collection and management method that integrates spatial computing and fuzzy matching algorithms to solve the technical challenges of automatic device perception, precise positioning, and intelligent data binding. Summary of the Invention

[0005] This invention provides a method and system for collecting and managing information on clinical emergency medical equipment, aiming to solve technical problems such as low efficiency, high error rate, and fragmented data flow caused by the complexity of equipment types and the tight operating environment at emergency sites, which often result in manual binding of equipment to patients. By constructing an automated identification and binding mechanism based on visual perception and spatial computing, real-time, accurate, and seamless association is achieved between emergency medical equipment, patient identity, and vital signs data flow, thereby improving the informatization level of the emergency medical process and the efficiency of clinical response.

[0006] This invention provides a method for collecting and managing information on clinical emergency medical devices, comprising: collecting a continuous video frame sequence containing a patient and emergency medical devices through at least one visual perception unit deployed at the emergency scene; constructing a three-dimensional spatial map of the emergency scene based on the video frame sequence, and performing real-time semantic segmentation and spatial positioning of the patient and each emergency medical device in the three-dimensional spatial map to obtain the patient's spatial coordinates and posture information, as well as the spatial coordinates, device type, and operating status of each emergency medical device; acquiring the vital signs data stream output by each emergency medical device, and establishing a mapping relationship between the vital signs data stream and the corresponding device physical identifier; monitoring the spatial movement trajectory of the patient and each emergency medical device within a preset time window, and calculating the spatial proximity index and trajectory correlation coefficient between any emergency medical device and the patient; when the spatial proximity index is lower than a first preset threshold and the trajectory correlation coefficient is higher than a second preset threshold, determining that the emergency medical device is in use with the patient, and automatically binding the physical identifier, device type, operating status, and corresponding vital signs data stream of the emergency medical device with the patient's identity information to generate a structured emergency event record.

[0007] Preferably, the visual perception unit is a binocular camera built into augmented reality glasses worn by medical staff, or a multi-view wide-angle camera array fixedly installed on the ceiling or wall of the emergency room; the visual perception unit has a frame rate of not less than 30 frames per second, a resolution of not less than 1.92 million pixels, and has an infrared fill light function to adapt to low-light environments.

[0008] Preferably, the construction of the three-dimensional spatial map of the emergency rescue scenario specifically includes: extracting and matching feature points from a continuous video frame sequence, and generating a sparse point cloud map using a simultaneous localization and mapping algorithm; based on this, performing pixel-level semantic segmentation on each frame image using a deep neural network to identify the patient's torso region, head region, and the outlines of various emergency rescue equipment; combining the sparse point cloud and semantic segmentation results, generating a dense three-dimensional mesh model containing semantic labels using a dense reconstruction algorithm, which serves as the three-dimensional spatial map.

[0009] Preferably, the real-time semantic segmentation and spatial positioning of the patient and each emergency medical device specifically includes: fitting the patient region in the semantic segmentation result into an ellipsoidal geometric model, with its centroid coordinates serving as the patient's spatial coordinates; fitting the region of each identified emergency medical device into a cuboid bounding box, with its geometric center coordinates serving as the device's spatial coordinates; and simultaneously, extracting the device's unique physical identifier and device type information from the device's nameplate or display screen using device appearance template matching or optical character recognition technology.

[0010] Preferably, the emergency medical equipment includes one or more of a ventilator, defibrillator, electrocardiogram monitor, infusion pump, and blood oxygen saturation monitor; the operating status includes power-on, standby, working, fault, and low power status; the operating status is determined by analyzing the content of the device display screen, the color and flashing frequency of the indicator lights, or the vibration characteristics of the device casing.

[0011] Preferably, the acquisition of vital sign data streams output by each emergency medical device specifically includes: receiving physiological parameter data packets transmitted from the built-in sensors of the emergency medical device via a wireless communication module; the physiological parameter data packets include a timestamp, device physical identifier, parameter type, and parameter value; the parameter type includes one or more of heart rate, respiratory rate, blood oxygen saturation, non-invasive blood pressure, body temperature, and ventilation volume.

[0012] Preferably, the calculation of the spatial proximity index between any emergency medical device and the patient specifically involves: within a preset time window, taking the Euclidean distance sequence between the spatial coordinates of the device and the spatial coordinates of the patient, calculating its time-weighted average value, and using it as the spatial proximity index; the time weighting adopts an exponential decay function, with the distance weight being higher the closer to the current time.

[0013] Preferably, the calculation of the trajectory correlation coefficient specifically includes: extracting the three-dimensional motion trajectory vector sequences of the patient and the emergency equipment within a preset time window; performing dynamic time warping and alignment on the two trajectory vector sequences; and calculating the Pearson correlation coefficient of the aligned sequence as the trajectory correlation coefficient.

[0014] Preferably, the duration of the preset time window is 10 to 60 seconds; the first preset threshold is 1.5 meters; and the second preset threshold is 0.7 meters.

[0015] Preferably, after the binding of the device and the patient is completed, the method further includes: pushing the binding result to the hospital information system in real time and updating the device usage record in the electronic medical record; if it is detected that the same device meets the binding conditions with multiple patients at the same time in a short period of time, a conflict verification mechanism is triggered, and arbitration is carried out based on whether the device type supports multiple patients sharing, patient priority score and trajectory stability index, so as to retain a unique valid binding relationship.

[0016] This invention provides a clinical emergency equipment information collection and management system, comprising: a visual perception unit for collecting a continuous video frame sequence containing a patient and emergency equipment; a three-dimensional scene mapping and target localization unit for constructing a three-dimensional spatial map of the emergency scene based on the video frame sequence, and performing real-time semantic segmentation and spatial localization of the patient and each emergency equipment in the three-dimensional spatial map to obtain the patient's spatial coordinates, posture information, and the spatial coordinates, equipment type, and operating status of each emergency equipment; a vital signs data receiving unit for acquiring the vital signs data stream output by each emergency equipment and establishing a mapping relationship between the vital signs data stream and the corresponding physical identifier of the equipment; a spatial behavior analysis unit for monitoring the spatial movement trajectory of the patient and each emergency equipment within a preset time window, and calculating the spatial proximity index and trajectory correlation coefficient between any emergency equipment and the patient; and an intelligent binding decision unit for determining that the emergency equipment is in use with the patient when the spatial proximity index is lower than a first preset threshold and the trajectory correlation coefficient is higher than a second preset threshold, and automatically binding the physical identifier, equipment type, operating status, and corresponding vital signs data stream of the emergency equipment with the patient's identity information to generate a structured emergency event record.

[0017] Preferably, the 3D scene mapping and target localization unit includes a feature extraction subunit, a simultaneous localization and mapping subunit, a semantic segmentation subunit, and a dense reconstruction subunit; the feature extraction subunit uses a scale-invariant feature transformation algorithm or an ORB feature algorithm to extract key points in the image; the simultaneous localization and mapping subunit uses a graph-optimized visual inertial odometry framework; the semantic segmentation subunit uses a convolutional neural network with an encoder-decoder structure, and its training dataset contains emergency scene images labeled with patient and various emergency equipment categories; the dense reconstruction subunit uses a multi-view stereo vision algorithm or a neural radiation field reconstruction algorithm.

[0018] Preferably, the spatial behavior analysis unit includes a trajectory extraction subunit, a dynamic time warping subunit, and a correlation calculation subunit; the trajectory extraction subunit extracts the target centroid coordinate sequence from the three-dimensional spatial map in chronological order; the dynamic time warping subunit uses a standard dynamic time warping algorithm to nonlinearly align the two trajectories; the correlation calculation subunit calculates the covariance matrix of the aligned trajectories in three spatial dimensions and normalizes it to obtain the trajectory correlation coefficient.

[0019] Preferably, the intelligent binding decision unit further includes a conflict arbitration subunit. When a binding conflict is detected, the conflict arbitration subunit queries the device metadata to determine whether it supports multi-patient mode. If it does not support it, it sorts patients according to their triage level scores and retains only the binding relationship of the patient with the highest score. If it supports it, it further calculates the trajectory stability index of each candidate binding relationship. The trajectory stability index is the reciprocal of the standard deviation of the trajectory speed, and the binding relationship with the highest stability index is retained.

[0020] Preferably, the system further includes a data synchronization interface unit, used to transmit the generated structured emergency event records to the electronic medical record system, equipment asset management platform and clinical data center through the hospital's internal network protocol, so as to achieve data consistency maintenance among multiple systems.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0022] This invention integrates visual perception, 3D spatial mapping, semantic understanding, and spatiotemporal behavior analysis technologies to construct an automatic patient binding mechanism for emergency medical devices without human intervention. Compared to traditional methods relying on barcode scanning or manual input, this invention completely eliminates binding omissions, mismatches, or delays caused by busy operations, significantly improving the accuracy and timeliness of information collection at emergency sites. The invention employs dual criteria of spatial proximity and trajectory correlation to effectively distinguish between the device's "physical proximity" and "actual usage" states, avoiding misbinding; a conflict arbitration mechanism ensures the rigor of binding logic in complex multi-patient scenarios. Furthermore, this invention achieves real-time structured association between device operating status, vital sign data streams, and patient identity, providing a high-quality and complete data foundation for subsequent clinical decision support, device usage efficiency analysis, and medical quality traceability, thus promoting the development of emergency medicine towards intelligence and automation. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention;

[0024] Figure 2 This is a schematic diagram of the core principle framework of the emergency medical device-patient automatic binding mechanism based on visual perception and spatial computing in this invention;

[0025] Figure 3 This is a logical flow diagram of the three-dimensional scene mapping and target localization stages in this invention;

[0026] Figure 4 This is a logical flowchart of the spatial behavior analysis and intelligent binding decision-making stage in this invention;

[0027] Figure 5This is a schematic diagram of the multi-level interaction relationship and data flow between the visual perception unit, the 3D scene mapping and target localization unit, and the vital signs data receiving unit in this invention;

[0028] Figure 6 This is a schematic diagram of the logical framework of the internal conflict arbitration mechanism of the intelligent binding decision unit in this invention; Detailed Implementation

[0029] This invention provides a method and system for collecting and managing information on clinical emergency medical equipment. Its core lies in automatically constructing a three-dimensional spatial map containing the patient and emergency medical equipment through visual perception units deployed at the emergency scene, and achieving seamless and precise binding between the equipment and the patient based on spatial proximity and motion trajectory correlation. The following will be illustrated in conjunction with the appendix... Figures 1 to 6 The specific implementation methods of each component of the system are described in detail to ensure the full disclosure and feasibility of the technical solution.

[0030] The method begins with step S1: acquiring a continuous sequence of video frames containing the patient and emergency equipment using at least one visual sensing unit deployed at the emergency scene. The visual sensing unit is either a binocular camera built into augmented reality glasses worn by medical personnel, or a multi-view wide-angle camera array fixedly mounted on the ceiling or wall of the emergency room. The visual sensing unit has a frame rate of at least 30 frames per second, a resolution of at least 1.92 million pixels, and infrared illumination to adapt to low-light environments. In actual deployment, if augmented reality glasses are used, their binocular cameras simultaneously acquire images from both eyes for subsequent depth estimation; if a fixed camera array is used, the timestamps of each camera are strictly synchronized to ensure that the multi-view images are aligned in the time dimension. All acquired video frame sequences are accurately timestamped and transmitted in real-time to the backend processing server via wired or wireless means. The acquisition of video frame sequences continues, covering the entire emergency operation process, from the patient's entry into the emergency area until the completion of initial treatment.

[0031] The following step, S2, is executed: A 3D spatial map of the emergency scene is constructed based on the video frame sequence. Real-time semantic segmentation and spatial localization of the patient and various emergency equipment are performed within this 3D spatial map to obtain the patient's spatial coordinates and posture information, as well as the spatial coordinates, equipment type, and operating status of each emergency equipment. This step is further subdivided into several sub-processing flows. First, feature point extraction and matching are performed on the continuous video frame sequence. The ORB feature algorithm is used for feature extraction. This algorithm has high computational efficiency while ensuring rotation and scale invariance, making it suitable for real-time processing scenarios. At least 500 stable keypoints are extracted from each frame image, and fast matching is performed using Hamming distance to generate inter-frame feature correspondences. Second, a graph-optimized visual inertial odometry framework is used to execute simultaneous localization and mapping algorithms to generate a sparse point cloud map. This framework integrates visual feature matching results with acceleration and angular velocity data provided by the inertial measurement unit, and solves for camera pose and 3D point positions through nonlinear optimization to form an initial sparse map. Each point in the sparse point cloud map is accompanied by 3D position information in the world coordinate system and an observation frame index.

[0032] After obtaining the sparse point cloud map, dense semantic reconstruction is performed. A convolutional neural network with an encoder-decoder structure is used to perform pixel-level semantic segmentation on each frame of the image. During the training phase, the network is trained using a large dataset of emergency scene images labeled with patient torso regions, head regions, and emergency equipment categories such as ventilators, defibrillators, ECG monitors, infusion pumps, and pulse oximeters. The network output is a semantic label map of the same size as the input image, where each pixel is assigned a category label. The semantic segmentation results are spatiotemporally aligned with the sparse point cloud map: for each frame of segmentation, it is projected onto the local coordinate system constructed from the sparse point cloud, and a dense 3D mesh model containing semantic labels is generated through triangulation or multi-view stereo matching algorithms. This dense 3D mesh model is the 3D spatial map, where each vertex on its surface carries semantic category information.

[0033] Based on a 3D spatial map, geometric fitting and spatial positioning are performed on the patient and emergency equipment. The patient area is fitted as an ellipsoidal geometric model, with its principal axis direction determined by the extension direction of the patient's torso in 3D space, and the centroid coordinates serving as the patient's spatial coordinates. This ellipsoidal model can be dynamically updated to adapt to changes in patient position. Each identified emergency equipment area is fitted as a cuboid bounding box, with its six faces aligned with the outer contour of the equipment body, and the geometric center coordinates serving as the equipment's spatial coordinates. Simultaneously, optical character recognition technology is used to extract the unique physical identifier and equipment type information from the equipment's nameplate or display screen. If standard icons or template patterns exist on the equipment surface, a template matching algorithm is used for cross-validation to ensure identification accuracy. The equipment's operating status is determined by analyzing its display screen content, indicator light colors and flashing frequencies, or the vibration characteristics of the equipment casing. For example, a solid green indicator light indicates standby, a flashing red light indicates a malfunction, and "VENT ON" on the display screen indicates that the ventilator is in operation. Vibration characteristics are obtained by analyzing minute displacement changes in the equipment area across consecutive frames; infusion pumps in operation typically exhibit periodic micro-vibrations.

[0034] Step S3: Acquire vital sign data streams output by each emergency medical device and establish a mapping relationship between the vital sign data streams and the corresponding device physical identifiers. Emergency medical devices have multiple built-in physiological sensors that periodically broadcast physiological parameter data packets via Bluetooth or Wi-Fi modules. Each data packet contains a timestamp, device physical identifier, parameter type, and parameter value. Parameter types include one or more of heart rate, respiratory rate, blood oxygen saturation, non-invasive blood pressure, body temperature, and ventilation volume. The vital sign data receiving unit continuously monitors the wireless channel, receives data packets from all present devices, and classifies and stores the data streams according to the device physical identifier. Each device corresponds to an independent data buffer queue, in which the received parameter records are arranged in chronological order. The mapping relationship is stored in a memory database in key-value pair form, where the key is the device physical identifier and the value is a reference to the corresponding data queue. This mapping relationship is established when a device is first identified and marked as invalid when its operating state changes to power off or offline.

[0035] Then, step S4 is executed: The spatial motion trajectories of the patient and each emergency medical device are monitored within a preset time window, and the spatial proximity index and trajectory correlation coefficient between any emergency medical device and the patient are calculated. The duration of the preset time window is 10 to 60 seconds, preferably 30 seconds. Within this window, the patient's ellipsoid centroid coordinate sequence is extracted from the 3D spatial map in chronological order to form the patient's 3D motion trajectory vector. ,in For the first The coordinates of each time point are calculated. Similarly, the coordinate sequence of the center of the bounding box of any emergency medical device is extracted to form the three-dimensional motion trajectory vector of the device. .

[0036] The spatial proximity index is calculated using a time-weighted Euclidean distance average. Specifically, for the trajectory... and Perform time alignment so that both are at the same point in time. Each point in time has coordinate values. Calculate the coordinates for each time point. Euclidean distance between the upper device and the patient's coordinates:

[0037] ;

[0038] for Time points Above, trajectory and The three-dimensional Euclidean distance between them; , , For time points Time, trajectory Coordinates in the X, Y, and Z dimensions; , , For time points Time, trajectory Coordinates in the X, Y, and Z dimensions;

[0039] Subsequently, the distance sequence is weighted using an exponential decay function:

[0040] ;

[0041] in For the first The time decay weight corresponding to each object, For the current moment, This is the attenuation coefficient, with a value of 0.2. Spatial proximity index. Defined as weighted average distance:

[0042] ;

[0043] Weighted by time decay As weight, for Spatial distance of objects The smaller the weighted average, the closer the device is to the patient in the near future.

[0044] Calculating the trajectory correlation coefficient requires dynamic time warping and alignment of the two trajectories. Since there may be slight differences in the motion sampling frequencies of the patient and the device, directly calculating the correlation can lead to bias. A standard dynamic time warping algorithm is used to construct a cumulative distance matrix and find the alignment path that minimizes the total warping distance. After alignment, trajectory sequences of equal length are obtained. and The Pearson correlation coefficient was then calculated. :

[0045]

[0046] in For the trajectory of the patients after the completion The A coordinate vector, For the equipment trajectory after completion The A coordinate vector, and These are the mean vectors of the aligned trajectories. This is the length of the aligned trajectory sequence. This coefficient measures the consistency of the motion trend between the two trajectories, ranging from -1 to 1. A higher value indicates stronger motion synchronization.

[0047] Step S5: When the spatial proximity index is lower than a first preset threshold and the trajectory correlation coefficient is higher than a second preset threshold, it is determined that the emergency medical device is in use with the patient. The device's physical identifier, device type, operating status, and corresponding vital sign data stream are automatically bound to the patient's identity information, generating a structured emergency event record. The first preset threshold is set to 1.5 meters, and the second preset threshold is set to 0.7 meters. Patient identity information is obtained in the following ways: when the patient enters the emergency area, the QR code on their wristband is captured by the visual perception unit and decoded to obtain the patient's unique identifier; or identity information is obtained by comparing the facial recognition module with the hospital's registered photo database. The binding operation includes creating an associated record in the central database. Fields include the patient's identity identifier, device physical identifier, device type, operating status, binding start time, spatial proximity index value, trajectory correlation coefficient value, and a pointer to the corresponding vital sign data stream.

[0048] After binding is completed, the system executes a conflict resolution mechanism. If the same device is detected to meet the binding conditions with multiple patients simultaneously within a short period, an arbitration process is triggered. First, the device metadata is queried to determine if it supports multi-patient sharing mode. Dedicated devices such as ventilators and defibrillators typically do not support this, while some models of ECG monitors support multi-channel monitoring. If the device does not support multi-patient mode, patients are sorted according to their triage level score, and only the binding relationship of the patient with the highest score is retained. The triage level score is provided by the emergency triage system and is divided into Level 1 (critical), Level 2 (urgent), Level 3 (sub-urgent), and Level 4 (non-urgent). If the device supports multi-patient mode, the trajectory stability index of each candidate binding relationship is further calculated. The trajectory stability index is defined as the reciprocal of the standard deviation of the trajectory velocity. The velocity sequence is obtained by differentiating the trajectory coordinates; the smaller the standard deviation, the smoother the motion, and the higher the stability index. The system retains the binding relationship with the highest stability index, and the other candidate relationships are rejected.

[0049] The generated structured emergency event records are pushed in real time to the electronic medical record system, equipment asset management platform, and clinical data center via the hospital's internal network protocol. The pushed content includes complete binding information and a summary of the associated vital signs data stream. The electronic medical record system automatically updates equipment usage records, the equipment asset management platform records equipment usage duration and status logs, and the clinical data center is used for subsequent quality analysis and research.

[0050] At the system level, the clinical emergency equipment information acquisition and management system includes a visual perception unit, a 3D scene mapping and target localization unit, a vital signs data receiving unit, a spatial behavior analysis unit, and an intelligent binding decision-making unit. The visual perception unit is responsible for acquiring raw video. The 3D scene mapping and target localization unit includes feature extraction subunits, synchronous localization and mapping subunits, semantic segmentation subunits, and dense reconstruction subunits, collaboratively completing the conversion from images to a semantic 3D map. The vital signs data receiving unit maintains the mapping relationship between the device's physical identifier and the data stream. The spatial behavior analysis unit includes trajectory extraction subunits, dynamic time warping subunits, and correlation calculation subunits, responsible for trajectory processing and indicator calculation. The intelligent binding decision-making unit integrates threshold judgment logic and conflict arbitration subunits to ensure the uniqueness and rationality of the binding results. All units are interconnected through a high-speed internal bus, sharing a unified time reference and coordinate system to ensure data consistency.

[0051] The entire system runs on a dedicated server cluster, equipped with high-performance graphics processors to accelerate neural network inference and 3D reconstruction calculations. After startup, the system continuously monitors the emergency area, immediately adding any new patient or activated device to the monitoring scope. All processing latency is controlled within 500 milliseconds to ensure real-time binding decisions. The system also features anomaly handling mechanisms: if the visual sensing unit signal is interrupted, it automatically switches to a backup camera; if vital sign data is lost, the device status is marked as a communication anomaly and a reconnection attempt is made; if 3D map reconstruction fails, it reverts to 2D planar positioning mode, using single-view images to estimate relative position and maintain basic binding functionality.

[0052] Through the above methods and systems, this invention achieves fully automatic and high-precision binding between emergency medical equipment and patients, completely eliminating manual operation and significantly improving the reliability and efficiency of emergency scene information management.

Claims

1. A method for collecting and managing information on clinical emergency medical equipment, characterized in that, include: A continuous sequence of video frames containing the patient and emergency equipment is acquired by at least one visual sensing unit deployed at the emergency scene. A three-dimensional spatial map of the emergency scene is constructed based on the video frame sequence, and real-time semantic segmentation and spatial positioning of the patient and each emergency device are performed in the three-dimensional spatial map to obtain the patient's spatial coordinates and posture information, as well as the spatial coordinates, device type and operating status of each emergency device. Acquire vital sign data streams output by each emergency medical device and establish a mapping relationship between the vital sign data streams and the corresponding device physical identifiers; Monitor the spatial movement trajectory of the patient and each emergency medical device within a preset time window, and calculate the spatial proximity index and trajectory correlation coefficient between any emergency medical device and the patient. When the spatial proximity index is lower than the first preset threshold and the trajectory correlation coefficient is higher than the second preset threshold, it is determined that the emergency equipment is in use for the patient, and the physical identifier, equipment type, operating status and corresponding vital sign data stream of the emergency equipment are automatically bound to the patient's identity information to generate a structured emergency event record.

2. The clinical emergency equipment information collection and management method according to claim 1, characterized in that, A three-dimensional spatial map of the emergency rescue scene is constructed based on the video frame sequence. Real-time semantic segmentation and spatial positioning of the patient and various emergency rescue devices are performed within this three-dimensional spatial map to obtain the patient's spatial coordinates and posture information, as well as the spatial coordinates, device type, and operating status of each emergency rescue device, including: Feature points are extracted and matched from a continuous video frame sequence, and a sparse point cloud map is generated using a simultaneous localization and mapping algorithm. Pixel-level semantic segmentation of each frame of image is performed using a deep neural network to identify the patient's torso region, head region, and the outline of various emergency equipment. By combining sparse point cloud and semantic segmentation results, a dense 3D mesh model containing semantic labels is generated through a dense reconstruction algorithm, which serves as the 3D spatial map. The patient region in the semantic segmentation results is fitted to an ellipsoidal geometric model, and its centroid coordinates are used as the patient's spatial coordinates. Each identified emergency medical device area is fitted as a cuboid bounding box, and its geometric center coordinates are used as the spatial coordinates of the device. By using equipment appearance template matching or optical character recognition technology, the unique physical identifier and equipment type information of the equipment can be extracted from the nameplate or display screen on the equipment surface.

3. The clinical emergency equipment information collection and management method according to claim 2, characterized in that, Acquire vital sign data streams output by each emergency medical device, and establish a mapping relationship between the vital sign data streams and the corresponding device physical identifiers, including: The system receives physiological parameter data packets transmitted from the built-in sensors of the emergency medical equipment via a wireless communication module. The physiological parameter data package includes a timestamp, device physical identifier, parameter type, and parameter value; The parameter types include one or more of the following: heart rate, respiratory rate, blood oxygen saturation, non-invasive blood pressure, body temperature, and ventilation. The physiological parameter data packets are classified and stored according to the device physical identifier, and a mapping relationship between the device physical identifier and the corresponding vital sign data stream is established.

4. The clinical emergency equipment information collection and management method according to claim 3, characterized in that, Monitor the spatial movement trajectories of the patient and various emergency medical devices within a preset time window, and calculate the spatial proximity index and trajectory correlation coefficient between any emergency medical device and the patient, including: Within a preset time window, the Euclidean distance sequence between the spatial coordinates of the device and the spatial coordinates of the patient is taken, and its time-weighted average value is calculated as the spatial proximity index. The time-weighted method uses an exponential decay function, where the distance weight increases as it approaches the current time. Extract the three-dimensional motion trajectory vector sequences of the patient and emergency equipment within a preset time window; Dynamic time warping and alignment of two trajectory vector sequences; The Pearson correlation coefficient of the aligned sequence is calculated and used as the trajectory correlation coefficient.

5. The clinical emergency equipment information collection and management method according to claim 4, characterized in that, The duration of the preset time window is 10 to 60 seconds; the first preset threshold is 1.5 meters; and the second preset threshold is 0.7 meters.

6. The clinical emergency equipment information collection and management method according to claim 5, characterized in that, After the device is successfully linked to the patient, the method further includes: The binding results are pushed to the hospital information system in real time to update the device usage records in the electronic medical records; If the same device is detected to meet the binding conditions with multiple patients simultaneously within a short period of time, a conflict verification mechanism is triggered. Arbitration is conducted based on whether the device type supports multiple patients sharing, patient priority scores, and trajectory stability indicators to retain a unique and valid binding relationship.

7. The clinical emergency equipment information collection and management method according to claim 6, characterized in that, The conflict verification mechanism includes: Query the device metadata to determine if it supports multi-patient mode; If not supported, patients will be sorted according to their triage level scores, and only the binding relationships of patients with the highest scores will be retained; If supported, the trajectory stability index of each candidate binding relationship is further calculated. The trajectory stability index is the reciprocal of the standard deviation of the trajectory velocity, and the binding relationship with the highest stability index is retained.

8. The clinical emergency equipment information collection and management method according to claim 7, characterized in that, The patient's identity information was obtained through the following methods: When a patient enters the emergency area, the QR code on their wristband is captured by the visual perception unit and decoded to obtain the patient's unique identification. Alternatively, identity information can be obtained by comparing photos in the hospital's registration database using a facial recognition module.

9. A clinical emergency equipment information collection and management system, characterized in that, include: A visual perception unit is used to acquire a continuous sequence of video frames containing the patient and emergency equipment. The three-dimensional scene mapping and target localization unit is used to construct a three-dimensional spatial map of the emergency scene based on the video frame sequence, and to perform real-time semantic segmentation and spatial localization of the patient and each emergency equipment in the three-dimensional spatial map to obtain the patient's spatial coordinates, posture information, and the spatial coordinates, equipment type and operating status of each emergency equipment. The vital signs data receiving unit is used to acquire the vital signs data streams output by each emergency medical device and establish a mapping relationship between the vital signs data streams and the corresponding device physical identifiers. The spatial behavior analysis unit is used to monitor the spatial movement trajectory of the patient and each emergency equipment within a preset time window, and to calculate the spatial proximity index and trajectory correlation coefficient between any emergency equipment and the patient. The intelligent binding decision unit is used to determine that the emergency medical device is in use with the patient when the spatial proximity index is lower than a first preset threshold and the trajectory correlation coefficient is higher than a second preset threshold. It automatically binds the physical identifier, device type, operating status and corresponding vital sign data stream of the emergency medical device with the patient's identity information to generate a structured emergency medical event record.

10. The clinical emergency equipment information collection and management system according to claim 9, characterized in that, The 3D scene mapping and target localization unit includes a feature extraction subunit, a simultaneous localization and mapping subunit, a semantic segmentation subunit, and a dense reconstruction subunit. The feature extraction subunit uses the scale-invariant feature transformation algorithm or the ORB feature algorithm to extract key points of the image. The synchronous positioning and mapping subunit adopts a graph-optimized visual inertial odometry framework; The semantic segmentation subunit adopts a convolutional neural network with an encoder-decoder structure, and its training dataset contains emergency scene images labeled with patients and various types of emergency equipment. The dense reconstruction subunit employs a multi-view stereo vision algorithm or a neural radiation field reconstruction algorithm.

Images