An ECMO device integrated intelligent anti-collision monitoring system

Abstract

This invention discloses an integrated intelligent collision avoidance monitoring system for ECMO devices. The system includes: a multi-source environmental perception module, which collects the motion posture data of the ECMO device, the dynamic distance between the ECMO device and each surrounding obstacle, and the three-dimensional contour of each surrounding obstacle; an intelligent analysis and decision-making module, which fuses the motion posture data, all dynamic distances, and three-dimensional contours and inputs them into a collision avoidance risk assessment model to obtain a dynamic safety boundary. For each obstacle, the corresponding dynamic distance and three-dimensional contour are compared with the dynamic safety boundary to generate an individual collision risk level for each obstacle, and the highest level is selected as the comprehensive collision risk level signal; and a multi-modal early warning execution module, which triggers a target early warning combination based on the comprehensive collision risk level signal. This invention effectively avoids equipment damage and life safety risks caused by collisions, improving its reliability in clinical environments.

Description

Technical Field

[0001] This invention relates to the field of medical device safety monitoring technology, and in particular to an integrated intelligent anti-collision monitoring system for ECMO devices. Background Technology

[0002] Extracorporeal membrane oxygenation (ECMO) devices play a crucial role in the treatment of patients with cardiopulmonary failure, but their safety in complex clinical environments remains a concern. Particularly during device operation and patient transport, the large size and numerous tubing of ECMO devices make them susceptible to collisions with other medical equipment or personnel, leading to tubing twisting, detachment, or even device damage, seriously threatening patient lives. While existing technologies offer some protective measures, the lack of an effective integrated intelligent monitoring system prevents real-time and accurate detection of environmental changes and timely warnings. This makes it difficult for medical staff to fully control the device's safety status during dynamic operations, necessitating innovative solutions to improve its collision resistance and reliability.

[0003] Therefore, there is an urgent need to provide a technical solution to address the above problems. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an integrated intelligent anti-collision monitoring system for ECMO devices. The technical solution of this system is as follows: It includes: a multi-source environmental perception module, an intelligent analysis and decision-making module, and a multi-modal early warning execution module; The multi-source environmental perception module is used to: collect in real time the motion posture data of the ECMO device, the dynamic distance between the ECMO device and each surrounding obstacle, and the three-dimensional contour of each surrounding obstacle; The intelligent analysis and decision-making module is used to: fuse the motion posture data, all dynamic distances and all three-dimensional contours, generate fused data and input it into the collision risk assessment model trained based on machine learning algorithms, obtain a dynamic safety boundary that represents its own motion state and environmental characteristics, and compare the corresponding dynamic distance and three-dimensional contour with the dynamic safety boundary for each obstacle to generate an individual collision risk level for each obstacle, and select the highest level from all individual collision risk levels as the comprehensive collision risk level signal; The multimodal early warning execution module is used to: trigger a target early warning combination based on the comprehensive collision risk level signal. The target early warning combination includes: LED light strip color change set on the ECMO device body, directional voice alarm, and tactile vibration feedback integrated on the mobile chassis of the ECMO device.

[0005] In one alternative approach, the multi-source environment sensing module is specifically used for: The dynamic distance between the ECMO device and each obstacle is measured by emitting a laser beam through a laser ranging unit and receiving reflected signals from each obstacle. The depth image of the surrounding environment of the ECMO device is obtained by the depth vision unit, and the three-dimensional contour of each obstacle is parsed based on the depth image. The acceleration and angular velocity of the ECMO device in space are detected by an inertial measurement unit, and the motion attitude data of the ECMO device are calculated by integral calculation.

[0006] In one alternative approach, the intelligent analysis and decision-making module is specifically used for: All dynamic distances and all three-dimensional contours are aligned, and based on the motion posture data, all dynamic distances and all three-dimensional contours that have undergone data alignment are transformed to the unified coordinate system of the ECMO device itself. Under the unified coordinate system, all dynamic distances and all three-dimensional contours are correlated to construct an environmental situation map containing the location information and three-dimensional contours of each obstacle. The environmental situation map and the motion posture data are combined to form the fused data.

[0007] In one alternative approach, the intelligent analysis and decision-making module is specifically used for: Extract the overall distribution density features of obstacles around the ECMO device and the maximum contour complexity features among all obstacles from the environmental situation map; Extract the real-time motion speed and real-time motion direction features of the ECMO device from the motion posture data; The overall distribution density feature, the maximum contour complexity feature, the real-time motion speed feature, and the real-time motion direction feature are input into the collision avoidance risk assessment model for processing, and the motion state category of the ECMO device in the current environment is output. Based on the motion state category, select the corresponding basic boundary parameters from a plurality of preset basic safety boundary templates; By combining the real-time motion speed characteristics and the overall distribution density characteristics, the basic boundary parameters are dynamically adjusted to calculate the dynamic safety boundary.

[0008] In one alternative approach, the intelligent analysis and decision-making module is specifically used for: Through the feature fusion layer in the collision avoidance risk assessment model, the input overall distribution density feature, maximum contour complexity feature, real-time motion speed feature, and real-time motion direction feature are normalized and state decision values ​​are calculated. The state decision value is compared with multiple preset state category thresholds, and the motion state category is determined based on the comparison results.

[0009] In one alternative approach, the formula for calculating the state decision value is: S = α·V norm + β·D norm +γ·C norm + δ·Dir std S represents the state decision value, V norm D represents the real-time motion velocity characteristics after normalization. norm C represents the overall distribution density characteristics after normalization. norm Dir represents the maximum contour complexity feature after normalization. std The standardization process represents the real-time motion direction feature, where α, β, γ, and δ represent the weight coefficients corresponding to each feature, and α+β+γ+δ=1.

[0010] In one alternative approach, the intelligent analysis and decision-making module is specifically used for: Based on the real-time motion velocity characteristics and the overall distribution density characteristics, calculate the dynamic scaling factor of the basic boundary parameters; The dynamic safety boundary is obtained by scaling the basic boundary parameters according to the dynamic scaling factor.

[0011] In one alternative approach, the formula for calculating the dynamic safety boundary is: B dynamic = B base × (1 +k v × V norm + k d × D norm ); B dynamic B represents the dynamic security boundary. base Denotes the basic boundary parameter, k v k represents the speed adjustment coefficient. d This represents the density adjustment factor.

[0012] In one alternative approach, the intelligent analysis and decision-making module is specifically used for: For any obstacle, based on the three-dimensional contour of the obstacle, calculate the minimum spatial distance between each point on the surface of the obstacle and the dynamic safety boundary, add the dynamic distance corresponding to the obstacle to the minimum spatial distance to obtain the comprehensive proximity of the obstacle, and repeat the process until the comprehensive proximity of each obstacle is obtained; The overall proximity of each obstacle is compared with multiple preset risk level thresholds. Based on the comparison results, the individual collision risk level of each obstacle is determined, and the highest level among all the individual collision risk levels of all obstacles is selected as the overall collision risk level signal.

[0013] In one alternative approach, the multimodal early warning execution module is specifically used for: The comprehensive collision risk level signal is mapped to a corresponding risk level category, which includes three levels: low risk, medium risk and high risk. According to the risk level category, the target warning combination corresponding to the risk level category is called from the preset warning strategy library. The target warning combination defines the display color and flashing frequency of the LED light strip, the broadcast text and volume level of the directional voice alarm, and the coordinated activation scheme of the vibration mode and intensity level of the tactile vibration feedback. The corresponding control command set is generated based on the target warning combination and is synchronously sent to the drive circuits of the LED light strip, the directional voice alarm and the tactile vibration feedback device. By executing the control command set, the LED light strip is synchronously driven to display according to the configuration, the directional voice alarm is driven to broadcast according to the configuration, and the tactile vibration feedback device is driven to generate tactile warnings according to the configuration.

[0014] The technical solution of this invention collects data from multiple sources and assesses collision risks through intelligent decision-making, combined with real-time alerts from multimodal early warning systems. This makes the operation of ECMO devices in complex environments safer and more reliable, effectively avoiding equipment damage and life safety risks caused by collisions, and significantly improving their reliability in clinical settings.

[0015] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0017] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a schematic diagram of an embodiment of the integrated intelligent anti-collision monitoring system for ECMO equipment according to the present invention. Detailed Implementation

[0018] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0019] Figure 1 A schematic diagram of an embodiment of an integrated intelligent collision avoidance monitoring system for ECMO devices provided by the present invention is shown. Figure 1 As shown, the system includes: a multi-source environmental perception module 110, an intelligent analysis and decision-making module 120, and a multi-modal early warning execution module 130; The multi-source environmental perception module 110 is used to: collect in real time the motion posture data of the ECMO device, the dynamic distance between the ECMO device and each surrounding obstacle, and the three-dimensional contour of each surrounding obstacle.

[0020] ECMO equipment refers to extracorporeal membrane oxygenation (ECMO) devices used to provide extracorporeal circulation and gas exchange during cardiopulmonary surgery; for example, mobile ECMO devices providing life support for patients with cardiopulmonary failure in the ICU. Motion posture data refers to the spatial motion parameters of the device collected by an inertial measurement unit (IMU); for example, triaxial acceleration and angular velocity data generated by the ECMO device during transport. Obstacles refer to objects around the ECMO device that may collide with it; for example, IV stands, hospital beds, or other medical equipment in an ICU ward. Dynamic distance refers to the real-time changing relative distance between the ECMO device and obstacles; for example, the real-time distance of 1.2 meters maintained between the device and a wall during movement. 3D contour refers to the three-dimensional geometry of the obstacle in space; for example, a 3D point cloud model reconstructed from a cylindrical oxygen cylinder using a depth vision unit.

[0021] The intelligent analysis and decision-making module 120 is used to: fuse the motion posture data, all dynamic distances and all three-dimensional contours, generate fused data and input it into the collision avoidance risk assessment model trained based on machine learning algorithms, obtain a dynamic safety boundary that represents its own motion state and environmental characteristics, and compare the corresponding dynamic distance and three-dimensional contour with the dynamic safety boundary for each obstacle to generate an individual collision risk level for each obstacle, and select the highest level from all individual collision risk levels as the comprehensive collision risk level signal.

[0022] The fused data refers to the comprehensive environmental information resulting from the coordinate unification and correlation processing of multi-source sensing data; for example, a fused dataset containing obstacle positions, outlines, and equipment motion postures. The collision avoidance risk assessment model refers to a risk prediction model trained based on machine learning algorithms; for example, a neural network model trained using historical collision data. The device's own motion state refers to its current movement characteristics; for example, the device moving northeast at a speed of 0.5 m / s. Environmental characteristics refer to the distribution and morphological characteristics of surrounding obstacles; for example, a densely packed group of medical devices combined with irregularly shaped instruments. The dynamic safety boundary refers to the safety protection range that adaptively adjusts with the motion state; for example, a 1.5-meter safety radius that automatically expands when the device moves at high speed. The individual collision risk level refers to the risk level of a single obstacle assessed independently; for example, a nearby IV stand is classified as high-risk. The comprehensive collision risk level signal refers to the highest risk level among all obstacles; for example, a high-risk signal output when multiple obstacles present high risks.

[0023] The multimodal early warning execution module 130 is used to: trigger a target early warning combination based on the comprehensive collision risk level signal. The target early warning combination includes: LED light strip color change set on the ECMO device body, directional voice alarm, and tactile vibration feedback integrated on the mobile chassis of the ECMO device.

[0024] Among these, "targeted early warning combination" refers to a collaborative scheme of multimodal early warning methods; for example, a combination of flashing red LEDs, high-frequency voice alarms, and strong vibrations. "LED light strip color changing" refers to conveying risk information through color changes; for example, green for low risk and red for high risk. "Directional voice alarm" refers to voice prompts with directional recognition; for example, a speaker at the front of the equipment playing a "obstacle ahead" warning tone. "ECMO equipment mobile chassis" refers to the mechanical structure at the bottom of the equipment that supports movement; for example, a base with casters. "Tactile vibration feedback" refers to conveying early warning information through vibration; for example, the equipment handle vibrating at different intensities according to the risk level.

[0025] The technical solution of this embodiment uses multi-source sensing to collect data and intelligent decision-making to assess collision risks, combined with multi-modal early warning and real-time reminders, to make the operation of ECMO equipment in complex environments safer and more reliable, effectively avoiding equipment damage and life safety risks caused by collisions, and significantly improving its reliability in clinical settings.

[0026] In one alternative approach, the multi-source environment sensing module is specifically used for: The dynamic distance between the ECMO device and each obstacle is measured by emitting a laser beam through a laser ranging unit and receiving reflected signals from each obstacle.

[0027] Here, a laser ranging unit refers to a sensor that measures distance based on laser phase difference; for example, a laser ranging sensor installed at the four corners of the device. A reflected signal refers to the light signal reflected back from an obstacle encountered by the laser beam; for example, a beam of light reflected from a wall received by the laser ranging unit.

[0028] Specifically, a modulated laser beam is emitted toward each obstacle by a laser ranging unit and the reflected signal from each obstacle is received. The phase difference between the emitted laser beam and the reflected signal is compared, and the dynamic distance between the ECMO device and each obstacle is calculated based on the phase difference.

[0029] The depth vision unit acquires depth images of the environment surrounding the ECMO device, and the three-dimensional contours of each obstacle are analyzed based on the depth images.

[0030] In this context, a depth vision unit refers to a visual sensor that acquires depth information about the environment; for example, the Intel RealSense depth camera. A depth image refers to image data that contains distance information about objects; for example, a depth map where each pixel records its distance value.

[0031] Specifically, depth image data of the surrounding environment of the ECMO device is acquired through a depth vision unit. The depth image contains distance information corresponding to each pixel. Obstacle segmentation processing is performed on the depth image to identify each independent obstacle region. Based on the depth data of each obstacle region, a 3D point cloud model of each obstacle is generated through a 3D reconstruction algorithm. The 3D contour of each obstacle is extracted from the 3D point cloud model.

[0032] The acceleration and angular velocity of the ECMO device in space are detected by an inertial measurement unit, and the motion attitude data of the ECMO device are calculated by integral calculation.

[0033] Inertial measurement units (IMUs) refer to a combination of sensors that detect motion parameters; for example, the MPU6050 six-axis motion processing sensor. Acceleration refers to the rate of change of the equipment's velocity; for example, 2 m / s² when the equipment starts up. 2 Acceleration value. Angular velocity refers to the rate of change of the rotation angle of the equipment; for example, the angular velocity value of 90° / s detected when the equipment turns. Integral operation refers to the mathematical method of calculating displacement from acceleration; for example, the displacement is obtained by performing a second integral on the acceleration data.

[0034] Specifically, the ECMO device's triaxial acceleration and triaxial angular velocity in three-dimensional space are detected in real time; the instantaneous velocity of the ECMO device in three-dimensional space is obtained by performing a first integration operation on the triaxial acceleration data; the attitude angle of the ECMO device in three-dimensional space is obtained by performing a first integration operation on the triaxial angular velocity data; the displacement of the ECMO device in three-dimensional space is obtained by performing a second integration operation on the instantaneous velocity; the motion attitude data of the ECMO device is calculated through coordinate transformation based on the attitude angle and displacement; and the Kalman filter algorithm is used to perform noise cancellation and error compensation processing on the motion attitude data.

[0035] Among the above-mentioned optional methods, the accuracy of environmental perception can be further improved by using laser ranging, depth vision and inertial measurement to work together to accurately acquire dynamic distance and three-dimensional contours, thereby enhancing the accuracy of motion attitude data calculation.

[0036] In one alternative embodiment, the intelligent analysis and decision-making module 120 is specifically used for: All dynamic distances and all three-dimensional contours are aligned, and based on the motion posture data, all the aligned dynamic distances and all three-dimensional contours are transformed to the unified coordinate system of the ECMO device itself.

[0037] The unified coordinate system refers to a coordinate system established with the equipment as the center; for example, a three-dimensional rectangular coordinate system with the equipment's centroid as the origin.

[0038] Specifically, a unified timestamp is assigned to all dynamic distances and all 3D contours to achieve time synchronization; a unified coordinate system for the ECMO device itself is established based on the device position and orientation parameters in the motion posture data; the dynamic distance of each obstacle is transformed from the laser ranging unit coordinate system to the unified coordinate system; the 3D contour of each obstacle is transformed from the depth vision unit coordinate system to the unified coordinate system; and spatial registration is performed on the transformed dynamic distance data and 3D contour data to ensure that the distance information and contour information of each obstacle correspond consistently in the unified coordinate system.

[0039] Under the unified coordinate system, all dynamic distances and all three-dimensional contours are correlated to construct an environmental situation map containing the location information and three-dimensional contours of each obstacle.

[0040] Among them, the environmental situation map refers to a digital map that reflects the overall situation of the surrounding environment; for example, a two-dimensional plan map that marks the location and outline of all obstacles.

[0041] Specifically, under the unified coordinate system, the dynamic distance of each obstacle is matched and associated with its corresponding three-dimensional contour to ensure that the distance information and contour information of each obstacle correspond one-to-one; based on the dynamic distance and the ECMO device's own position parameters, the three-dimensional coordinate position of each obstacle in the unified coordinate system is calculated; the three-dimensional coordinate position and three-dimensional contour of each obstacle are integrated into a digital map to construct an environmental situation map containing the position information and three-dimensional contour of each obstacle.

[0042] The environmental situation map and the motion posture data are combined to form the fused data.

[0043] Specifically, the environmental situation map and motion attitude data are timestamped to ensure data synchronization; the real-time position coordinates and motion direction parameters of the device in the motion attitude data are overlaid on the environmental situation map; the device velocity and acceleration characteristics in the motion attitude data are associated and integrated with the obstacle position and contour information in the environmental situation map; and a unified data structure containing environmental situation information and device motion state information is formed as the fused data.

[0044] Among the above-mentioned optional methods, data integration and environmental modeling are further optimized. After data alignment, the data is converted to a unified coordinate system to construct an environmental situation map containing position and contour, and coordinate motion attitude data to improve the completeness of situational awareness.

[0045] In one alternative embodiment, the intelligent analysis and decision-making module 120 is specifically used for: Extract the overall distribution density features of obstacles around the ECMO device and the maximum contour complexity features of all obstacles from the environmental situation map.

[0046] The overall distribution density feature refers to the density of obstacles within a unit area; for example, a density value of 3 obstacles per square meter. The maximum contour complexity feature refers to the most complex shape feature among all obstacles; for example, the contour complexity of a surgical cart with multiple protrusions.

[0047] Specifically, based on the three-dimensional coordinate positions of all obstacles in the environmental situation map, the number of obstacles per unit volume space is calculated as the overall distribution density feature; the three-dimensional contour of each obstacle in the environmental situation map is analyzed, and the ratio of the surface area to the volume of each obstacle's three-dimensional contour is calculated as the contour complexity of that obstacle; the maximum value among all obstacle contour complexity values ​​is selected as the maximum contour complexity feature among all obstacles.

[0048] The real-time motion speed and real-time motion direction features of the ECMO device are extracted from the motion posture data.

[0049] The real-time motion speed characteristic refers to the current moving speed of the device; for example, a moving speed value of 0.8 meters per second. The real-time motion direction characteristic refers to the current moving orientation of the device; for example, moving along the positive X-axis of the coordinate system.

[0050] Specifically, instantaneous velocity vectors are obtained from motion posture data; the magnitude of the instantaneous velocity vectors is calculated as real-time motion velocity features; and the angle between the projection of the instantaneous velocity vectors onto the XY plane in a unified coordinate system and the positive X-axis is calculated as real-time motion direction features.

[0051] The overall distribution density feature, the maximum contour complexity feature, the real-time motion speed feature, and the real-time motion direction feature are input into the collision avoidance risk assessment model for processing, and the motion state category of the ECMO device in the current environment is output.

[0052] Among them, the motion state category refers to the classification result of the equipment's motion mode; for example, categories such as smooth movement, sharp turn, or emergency braking.

[0053] Specifically, the feature fusion layer of the collision avoidance risk assessment model is used to normalize the input features (overall distribution density features, maximum contour complexity features, real-time motion speed features, and real-time motion direction features) and calculate the state decision value; the state decision value is compared with multiple preset state category thresholds; and the motion state category of the ECMO device in the current environment is determined based on the comparison results.

[0054] Based on the motion state category, select the corresponding basic boundary parameters from a set of preset basic safety boundary templates.

[0055] The basic safety boundary template refers to the preset safety boundary reference parameters; for example, a basic safety radius of 1 meter in a static state. The basic boundary parameters refer to the reference values ​​of the safety boundary; for example, a basic safety distance of 1.2 meters.

[0056] Specifically, based on the motion state category, a matching search is performed among multiple preset basic safety boundary templates to find the basic safety boundary template corresponding to the motion state category, and the basic boundary parameters are obtained from the basic safety boundary template.

[0057] By combining the real-time motion speed characteristics and the overall distribution density characteristics, the basic boundary parameters are dynamically adjusted to calculate the dynamic safety boundary.

[0058] Specifically, based on real-time motion speed characteristics and overall distribution density characteristics, a dynamic scaling factor is calculated by combining predefined speed adjustment coefficients and density adjustment coefficients; the dynamic safety boundary is obtained by multiplying the basic boundary parameters by the dynamic scaling factor.

[0059] Among the above-mentioned optional methods, the intelligence of collision avoidance monitoring is further enhanced by selecting basic boundary parameters according to the type of motion state, and dynamically adjusting them by integrating real-time motion speed and obstacle density characteristics to improve boundary adaptability.

[0060] In one alternative embodiment, the intelligent analysis and decision-making module 120 is specifically used for: The feature fusion layer in the collision avoidance risk assessment model normalizes the input overall distribution density feature, maximum contour complexity feature, real-time motion speed feature, and real-time motion direction feature, and calculates the state decision value.

[0061] In this context, the feature fusion layer refers to the layer in a neural network that combines features; for example, the feature weighting fusion layer in a fully connected neural network. The state decision value refers to the numerical value used to classify motion states; for example, a classification score of 0.75 calculated using a formula.

[0062] The state decision value is compared with multiple preset state category thresholds, and the motion state category is determined based on the comparison results.

[0063] Among them, the state category threshold refers to the critical value for classifying motion states; for example, a score of 0.6 or higher is classified as a high-speed motion state.

[0064] Specifically, the state decision value is compared sequentially with multiple preset state category thresholds to determine the threshold range in which the state decision value is located, and the corresponding motion state category is determined based on the mapping relationship between the threshold range and the motion state category.

[0065] In the above-mentioned optional methods, to further improve the accuracy of state decision-making, the feature fusion layer normalizes multi-dimensional features, calculates state decision values, and accurately classifies the equipment motion state categories.

[0066] In one alternative approach, the formula for calculating the state decision value is: S = α·V norm + β·D norm +γ·C norm + δ·Dir std S represents the state decision value, V norm D represents the real-time motion velocity characteristics after normalization. norm C represents the overall distribution density characteristics after normalization. norm Dir represents the maximum contour complexity feature after normalization. std The standardization process represents the real-time motion direction feature, where α, β, γ, and δ represent the weight coefficients corresponding to each feature, and α+β+γ+δ=1.

[0067] It should be noted that the calculation formula for the state decision value comprehensively evaluates the ECMO device's motion state by weighted fusion of four key features. This formula uses weight coefficients obtained through machine learning algorithm training to allocate the importance of different features. The function of the state decision value is to fuse multi-dimensional features into a single quantitative indicator, providing a basis for classifying motion states. In the formula, S represents the state decision value, which ranges from 0 to 1; V... norm This represents the real-time motion velocity characteristic after normalization, with values ​​ranging from 0 to 1; D norm This represents the overall distribution density characteristic after normalization, with values ​​ranging from 0 to 1; C norm Dir represents the maximum contour complexity feature after normalization, with a value ranging from 0 to 1. std The value of α represents the real-time motion direction feature after standardization, ranging from -1 to 1; α represents the weight coefficient of the real-time motion velocity feature, ranging from 0 to 1, with a default value of 0.3; β represents the weight coefficient of the overall distribution density feature, ranging from 0 to 1, with a default value of 0.3; γ represents the weight coefficient of the maximum contour complexity feature, ranging from 0 to 1, with a default value of 0.2; δ represents the weight coefficient of the real-time motion direction feature, ranging from 0 to 1, with a default value of 0.2; and the sum of the four weight coefficients is always 1.

[0068] Among the above-mentioned optional methods, collision avoidance risks are further accurately assessed. The state decision value is calculated by formula and compared with the threshold to accurately classify the motion state and optimize the real-time risk assessment.

[0069] In one alternative embodiment, the intelligent analysis and decision-making module 120 is specifically used for: Based on the real-time motion velocity characteristics and the overall distribution density characteristics, the dynamic scaling factor of the basic boundary parameters is calculated.

[0070] The dynamic scaling factor refers to the adjustment coefficient of the safety boundary; for example, a scaling factor of 1.2 times calculated based on velocity and density.

[0071] Specifically, the speed adjustment component is calculated based on the real-time motion speed characteristics and the predefined speed adjustment coefficient, and the density adjustment component is calculated based on the overall distribution density characteristics and the predefined density adjustment coefficient. The speed adjustment component and the density adjustment component are added together and then one is added to obtain the dynamic scaling factor.

[0072] The dynamic safety boundary is obtained by scaling the basic boundary parameters according to the dynamic scaling factor.

[0073] In the above-mentioned optional methods, the safety boundary is further adaptively adjusted by calculating the dynamic scaling factor based on the real-time movement speed and obstacle density characteristics, and flexibly adjusting the basic boundary parameters.

[0074] In one alternative approach, the formula for calculating the dynamic safety boundary is: B dynamic = B base × (1 +k v × V norm + k d × D norm ); B dynamic B represents the dynamic security boundary. base Denotes the basic boundary parameter, k v k represents the speed adjustment coefficient. d This represents the density adjustment factor.

[0075] The velocity adjustment coefficient refers to the weight of velocity's influence on the safety boundary; for example, the 0.3 weight value of the velocity feature in the scaling formula. The density adjustment coefficient refers to the weight of density's influence on the safety boundary; for example, the 0.2 weight value of the density feature in the scaling formula.

[0076] It should be noted that the calculation formula for the dynamic safety boundary is based on the fundamental boundary parameters, and achieves dynamic adaptive adjustment of the safety boundary by introducing velocity and density adjustment components. The functional role of the dynamic safety boundary is to adjust the safety protection range in real time according to the equipment's motion state and environmental characteristics, improving the accuracy and adaptability of collision protection. In the formula, B... dynamic B represents the dynamic safety boundary, and its value ranges from 1 to 2 times the basic boundary parameter; base This represents the basic boundary parameter, with a value range of 0.5 meters to 2 meters. The default value is set to 1.0 meter for stationary state and 1.2 meters for moving state, depending on the motion state category; k v This represents the speed adjustment factor, which ranges from 0 to 0.5, with a default value of 0.3; k d This represents the density adjustment coefficient, which ranges from 0 to 0.5, with a default value of 0.2. This formula achieves dynamic scaling of boundary parameters through linear combination, ensuring automatic expansion of the safety boundary in high-speed motion or high-density environments.

[0077] Among the above-mentioned optional methods, to further ensure the safety of equipment operation, the dynamic safety boundary calculation formula optimizes the basic boundary, adapts to different environments and motion states, and reduces the risk of collision.

[0078] In one alternative embodiment, the intelligent analysis and decision-making module 120 is specifically used for: For any obstacle, based on the three-dimensional contour of the obstacle, calculate the minimum spatial distance between each point on the surface of the obstacle and the dynamic safety boundary. Add the dynamic distance corresponding to the obstacle to the minimum spatial distance to obtain the comprehensive proximity of the obstacle. Repeat this process until the comprehensive proximity of each obstacle is obtained.

[0079] Minimum spatial distance refers to the shortest distance between the obstacle surface and the safety boundary; for example, the shortest gap of 0.2 meters between the obstacle outline and the safety boundary. Comprehensive proximity refers to a combined risk indicator that integrates distance and outline; for example, a proximity of 1.5 meters obtained by adding the dynamic distance and the minimum spatial distance.

[0080] Specifically, for any obstacle, based on the obstacle's three-dimensional contour data, the Euclidean distances between all points on the obstacle's surface and the dynamic safety boundary are calculated, and the minimum value is selected as the minimum spatial distance. The dynamic distance corresponding to the obstacle is added to the minimum spatial distance to obtain the obstacle's overall proximity. The above calculation process is repeated for each obstacle until the overall proximity of all obstacles is obtained.

[0081] The overall proximity of each obstacle is compared with multiple preset risk level thresholds. Based on the comparison results, the individual collision risk level of each obstacle is determined, and the highest level among all the individual collision risk levels of all obstacles is selected as the overall collision risk level signal.

[0082] Among them, the risk level threshold refers to the critical value for classifying risk levels; for example, the classification standard is that within 1 meter is high risk and 1-2 meters is medium risk. The comparison result refers to the conclusion of the comparison between the risk parameter and the threshold; for example, the judgment result that the overall proximity of 0.8 meters is less than the high-risk threshold of 1 meter.

[0083] Specifically, for each obstacle, its overall proximity is compared sequentially with multiple preset risk level thresholds to determine the risk level range in which the overall proximity falls. Based on the mapping relationship between the risk level range and the individual collision risk level, the individual collision risk level of the obstacle is determined. The above process is repeated until the individual collision risk level of each obstacle is determined. The highest level among the individual collision risk levels of all obstacles is selected as the overall collision risk level signal.

[0084] Among the above-mentioned optional methods, the individual collision risk level is further precisely classified, the comprehensive proximity of obstacles is calculated and compared with the risk threshold, so as to ensure the equipment's anti-collision capability in a multi-obstacle environment.

[0085] In one alternative embodiment, the multimodal early warning execution module 130 is specifically used for: The comprehensive collision risk level signal is mapped to a corresponding risk level category, which includes three levels: low risk, medium risk, and high risk.

[0086] Among them, risk level category refers to the classification of risk level; for example, three levels: low risk, medium risk, and high risk.

[0087] Specifically, the comprehensive collision risk level signal is matched with a preset risk level mapping table to map the comprehensive collision risk level signal to the corresponding risk level category, which includes three levels: low risk, medium risk, and high risk.

[0088] Based on the risk level category, a target warning combination corresponding to the risk level category is called from a preset warning strategy library. The target warning combination defines a coordinated activation scheme for the display color and flashing frequency of the LED light strip, the broadcast text and volume level of the directional voice alarm, and the vibration mode and intensity level of the tactile vibration feedback.

[0089] The preset early warning strategy library refers to a database that stores early warning plans; for example, an SQL database containing early warning plans corresponding to each level of risk.

[0090] Specifically, based on the risk level category, a preset warning strategy library is queried, and the configuration parameters of the target warning combination corresponding to the risk level category are read. The parameters configured for the target warning combination include the display color and flashing frequency parameters of the LED light strip, the broadcast text and volume level parameters of the directional voice alarm, and the vibration mode and intensity level parameters of the tactile vibration feedback, thereby forming a collaborative activation scheme.

[0091] The corresponding control command set is generated based on the target warning combination and is synchronously sent to the drive circuits of the LED light strip, the directional voice alarm and the tactile vibration feedback device. By executing the control command set, the LED light strip is synchronously driven to display according to the configuration, the directional voice alarm is driven to broadcast according to the configuration, and the tactile vibration feedback device is driven to generate tactile warnings according to the configuration.

[0092] Among them, the control instruction set refers to the set of instructions that drive the early warning device; for example, a group of instructions that simultaneously controls the LED color, voice content, and vibration intensity.

[0093] Specifically, based on the parameters of the target warning combination, control commands are generated for controlling the display color and flashing frequency of the LED light strip, the text and volume level of the directional voice alarm, and the vibration mode and intensity level of the tactile vibration feedback device. These commands are combined into a control command set. The control command set is synchronously sent to the driving circuits of the LED light strip, the directional voice alarm, and the tactile vibration feedback device. The driving circuits execute the control command set, synchronously driving the LED light strip to display according to the configured display color and flashing frequency, driving the directional voice alarm to broadcast according to the configured text and volume level, and driving the tactile vibration feedback device to generate tactile warnings according to the configured vibration mode and intensity level.

[0094] Among the above-mentioned optional methods, the early warning effectiveness can be further improved by mapping the risk level signal to the risk category, calling the corresponding early warning combination, and multimodal collaborative feedback to improve the risk perception and response speed of medical staff.

[0095] It should be noted that the collision avoidance risk assessment model construction process comprises four main parts: an input layer, a feature preprocessing layer, a feature fusion layer, and a decision output layer. The input layer receives input data in four dimensions: overall distribution density features, maximum contour complexity features, real-time motion velocity features, and real-time motion direction features. The feature preprocessing layer normalizes the input features individually. Specifically, the real-time motion velocity features, overall distribution density features, and maximum contour complexity features are processed using a minimum-maximum normalization method to a range of 0 to 1, while the real-time motion direction features are processed using a standardization method to a range of -1 to 1. The feature fusion layer fuses the preprocessed features into a state decision value using a weighted summation method. This layer contains four trainable parameter weight coefficients, each corresponding to one of the four input features. The decision output layer compares the state decision value with multiple preset state category thresholds and outputs the motion state category of the ECMO device in the current environment.

[0096] The training process of the collision avoidance risk assessment model employs a supervised learning method. First, a large amount of historical operational data of ECMO equipment in a clinical environment is collected as training samples. Each training sample includes equipment motion parameters, environmental obstacle information, and corresponding actual motion state labels. Data cleaning and labeling of the training samples ensure data quality and label accuracy. Initialization of the model sets the initial values ​​of the weight coefficients, typically using random or uniform initialization. During training, the mean squared error loss function is used to calculate the difference between the model's predicted state and the actual state labels. The gradient of the loss function with respect to each weight coefficient is calculated using the backpropagation algorithm, and the weight coefficient values ​​are updated using the stochastic gradient descent optimization algorithm. An early stopping mechanism is implemented during training to prevent overfitting; training terminates when performance on the validation set no longer improves. After model training, performance evaluation is performed on an independent test set to ensure the model has good generalization ability. The final collision avoidance risk assessment model can accurately identify the motion state category of the equipment based on real-time perceived environmental features and equipment motion characteristics, providing a reliable basis for calculating dynamic safety boundaries.

[0097] To better illustrate the technical solution of this embodiment, the following complete example is used for explanation, specifically: S10: The laser ranging unit emits a laser beam towards the IV stand in the ICU ward and receives the reflected signal, measuring the dynamic distance between the ECMO device and the IV stand to be 1.2 meters. S20: The depth image of the surrounding environment of the ECMO device is acquired by the depth vision unit, and the three-dimensional outline of the infusion stand is analyzed as a cylindrical structure based on the depth image. S30: The inertial measurement unit detects the acceleration and angular velocity of the ECMO device in space in real time, and calculates the motion attitude data of the device moving northeast at a speed of 0.5 meters per second through integral calculation; S40: Align all dynamic distances and 3D contours, and convert the processed data to the ECMO device's own unified coordinate system based on motion posture data; S50: Under a unified coordinate system, all dynamic distances and three-dimensional contours are correlated to construct an environmental situation map that includes the location information of the IV stand and the three-dimensional contour. S60: Combines environmental situational data with motion posture data to form fused data; S70: The overall distribution density feature of obstacles around the ECMO device extracted from the environmental situation map is 2 obstacles per square meter, and the maximum contour complexity feature among all obstacles is 0.85; S80: The real-time motion speed feature of the ECMO device extracted from the motion posture data is 0.5 meters per second, and the real-time motion direction feature is northeast. S90: Input the overall distribution density characteristics, maximum contour complexity characteristics, real-time motion speed characteristics, and real-time motion direction characteristics into the collision avoidance risk assessment model, and output the motion state category of the ECMO device in the current environment as a steady motion state. S100: Select the corresponding basic boundary parameter of 1.2 meters from multiple preset basic safety boundary templates according to the motion state category; S110: The basic boundary parameters are dynamically adjusted by combining real-time motion speed characteristics and overall distribution density characteristics, and the dynamic safety boundary is calculated to be 1.5 meters. S120: Based on the three-dimensional contour of the infusion stand, the minimum spatial distance between each point on its surface and the dynamic safety boundary is calculated to be 0.3 meters. The dynamic distance of 1.2 meters and the minimum spatial distance of 0.3 meters are added together to obtain a comprehensive proximity of 1.5 meters. S130: By comparing the overall proximity of 1.5 meters with multiple preset risk level thresholds, the individual collision risk level of the IV stand is determined to be high risk; S140: Select the highest level from the individual collision risk levels of all obstacles as the comprehensive collision risk level signal; S150: Maps the comprehensive collision risk level signal to a high-risk level category; S160: Retrieve the corresponding target early warning combination from the preset early warning strategy library according to the high-risk level category; S170: Generates a corresponding set of control commands based on the target warning combination and sends them synchronously to the drive circuits of the LED light strip, the directional voice alarm and the tactile vibration feedback device; S180: By executing the control command set, it synchronously drives the LED light strip to display red and flash at high frequency, drives the directional voice alarm to play a warning sound indicating an obstacle ahead, and drives the tactile vibration feedback device to generate strong vibration.

[0098] Furthermore, the system provided in the above embodiments is only illustrated by the division of the above functional modules when implementing its functions. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to the actual situation to complete all or part of the functions described above.

[0099] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0100] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0101] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An integrated intelligent collision avoidance monitoring system for ECMO equipment, characterized in that, The system includes: a multi-source environmental perception module, an intelligent analysis and decision-making module, and a multi-modal early warning execution module; The multi-source environmental perception module is used to: collect in real time the motion posture data of the ECMO device, the dynamic distance between the ECMO device and each surrounding obstacle, and the three-dimensional contour of each surrounding obstacle; The intelligent analysis and decision-making module is used to: fuse the motion posture data, all dynamic distances and all three-dimensional contours, generate fused data and input it into the collision risk assessment model trained based on machine learning algorithm, obtain a dynamic safety boundary that represents its own motion state and environmental characteristics, and compare the corresponding dynamic distance and three-dimensional contour with the dynamic safety boundary for each obstacle to generate an individual collision risk level for each obstacle, and select the highest level from all individual collision risk levels as the comprehensive collision risk level signal; The multimodal early warning execution module is used to: trigger a target early warning combination based on the comprehensive collision risk level signal. The target early warning combination includes: color change of LED light strip set on the ECMO device body, directional voice alarm, and tactile vibration feedback integrated on the mobile chassis of the ECMO device.

2. The integrated intelligent anti-collision monitoring system for ECMO equipment according to claim 1, characterized in that, The multi-source environment sensing module is specifically used for: The dynamic distance between the ECMO device and each obstacle is measured by emitting a laser beam through a laser ranging unit and receiving reflected signals from each obstacle. The depth image of the surrounding environment of the ECMO device is obtained by the depth vision unit, and the three-dimensional contour of each obstacle is parsed based on the depth image. The acceleration and angular velocity of the ECMO device in space are detected by an inertial measurement unit, and the motion attitude data of the ECMO device are calculated by integral calculation.

3. The integrated intelligent anti-collision monitoring system for ECMO equipment according to claim 1, characterized in that, The intelligent analysis and decision-making module is specifically used for: All dynamic distances and all three-dimensional contours are aligned, and based on the motion posture data, all dynamic distances and all three-dimensional contours that have undergone data alignment are transformed to the unified coordinate system of the ECMO device itself. Under the unified coordinate system, all dynamic distances and all three-dimensional contours are correlated to construct an environmental situation map containing the location information and three-dimensional contours of each obstacle. The environmental situation map and the motion posture data are combined to form the fused data.

4. The integrated intelligent anti-collision monitoring system for ECMO equipment according to claim 3, characterized in that, The intelligent analysis and decision-making module is specifically used for: Extract the overall distribution density features of obstacles around the ECMO device and the maximum contour complexity features among all obstacles from the environmental situation map; Extract the real-time motion speed and real-time motion direction features of the ECMO device from the motion posture data; The overall distribution density feature, the maximum contour complexity feature, the real-time motion speed feature, and the real-time motion direction feature are input into the collision avoidance risk assessment model for processing, and the motion state category of the ECMO device in the current environment is output. Based on the motion state category, select the corresponding basic boundary parameters from a plurality of preset basic safety boundary templates; By combining the real-time motion speed characteristics and the overall distribution density characteristics, the basic boundary parameters are dynamically adjusted to calculate the dynamic safety boundary.

5. The integrated intelligent anti-collision monitoring system for ECMO equipment according to claim 4, characterized in that, The intelligent analysis and decision-making module is specifically used for: Through the feature fusion layer in the collision avoidance risk assessment model, the input overall distribution density feature, maximum contour complexity feature, real-time motion speed feature, and real-time motion direction feature are normalized and state decision values ​​are calculated. The state decision value is compared with multiple preset state category thresholds, and the motion state category is determined based on the comparison results.

6. The integrated intelligent anti-collision monitoring system for ECMO equipment according to claim 5, characterized in that, The formula for calculating the state decision value is: S = α·V norm + β·D norm + γ·C norm + δ·Dir std S represents the state decision value, V norm D represents the real-time motion velocity characteristics after normalization. norm C represents the overall distribution density characteristics after normalization. norm Dir represents the maximum contour complexity feature after normalization. std The standardization process represents the real-time motion direction feature, where α, β, γ, and δ represent the weight coefficients corresponding to each feature, and α+β+γ+δ=1.

7. The integrated intelligent anti-collision monitoring system for ECMO equipment according to claim 6, characterized in that, The intelligent analysis and decision-making module is specifically used for: Based on the real-time motion velocity characteristics and the overall distribution density characteristics, calculate the dynamic scaling factor of the basic boundary parameters; The dynamic safety boundary is obtained by scaling the basic boundary parameters according to the dynamic scaling factor.

8. The integrated intelligent anti-collision monitoring system for ECMO equipment according to claim 7, characterized in that, The formula for calculating the dynamic safety boundary is: B dynamic = B base × (1 + k v × V norm + k d × D norm );B dynamic B represents the dynamic security boundary. base Denotes the basic boundary parameter, k v k represents the speed adjustment coefficient. d This represents the density adjustment factor.

9. The integrated intelligent anti-collision monitoring system for ECMO equipment according to claim 8, characterized in that, The intelligent analysis and decision-making module is specifically used for: For any obstacle, based on the three-dimensional contour of the obstacle, calculate the minimum spatial distance between each point on the surface of the obstacle and the dynamic safety boundary, add the dynamic distance corresponding to the obstacle to the minimum spatial distance to obtain the comprehensive proximity of the obstacle, and repeat the process until the comprehensive proximity of each obstacle is obtained; The overall proximity of each obstacle is compared with multiple preset risk level thresholds. Based on the comparison results, the individual collision risk level of each obstacle is determined, and the highest level among all the individual collision risk levels of all obstacles is selected as the overall collision risk level signal.

10. The integrated intelligent anti-collision monitoring system for ECMO equipment according to claim 9, characterized in that, The multimodal early warning execution module is specifically used for: The comprehensive collision risk level signal is mapped to a corresponding risk level category, which includes three levels: low risk, medium risk and high risk. According to the risk level category, the target warning combination corresponding to the risk level category is called from the preset warning strategy library. The target warning combination defines the display color and flashing frequency of the LED light strip, the broadcast text and volume level of the directional voice alarm, and the coordinated activation scheme of the vibration mode and intensity level of the tactile vibration feedback. The corresponding control command set is generated based on the target warning combination and is synchronously sent to the drive circuits of the LED light strip, the directional voice alarm and the tactile vibration feedback device. By executing the control command set, the LED light strip is synchronously driven to display according to the configuration, the directional voice alarm is driven to broadcast according to the configuration, and the tactile vibration feedback device is driven to generate tactile warnings according to the configuration.

Images