CN-121686565-B - Cardiopulmonary resuscitation pressing action gesture feature AI identification and deviation correction method

CN121686565BCN 121686565 BCN121686565 BCN 121686565BCN-121686565-B

Abstract

The invention provides a cardiopulmonary resuscitation pressing action gesture feature AI identification and deviation correction method, which relates to the technical field of medical emergency treatment and comprises the steps of performing spatial reconstruction on trunk movement track data and chest pressure dynamic data when a rescuer executes pressing action, extracting trunk centroid movement rules and angular velocity change trends, calculating gesture stability components, generating a steady-state support area, constructing a pressure response state space for the chest pressure dynamic data, establishing a pressure elastic deformation field, generating an optimal pressure application path, mapping the steady-state support area and the optimal pressure application path to an elastic potential energy field, optimizing a support structure based on dynamic balance constraint, generating a real-time gesture adjustment instruction sequence, and adjusting the standing distance, the support angle and the force application direction of the rescuer according to the instruction sequence to finish deviation correction. The invention can accurately identify the pressing action deviation in real time, intelligently generate the personalized correction instruction and improve the success rate of cardiopulmonary resuscitation.

Inventors

XIA HUIHUI
YAN SHUMIN
ZHAO BEI
DU CHAOSHENG
YIN ZHAO
FENG XUEYAO
LI TING
GAO NING
HuangFu Jiaxin
CHEN SHANSHAN

Assignees

中国人民解放军总医院第九医学中心

Dates

Publication Date: 20260512
Application Date: 20251208

Claims (10)

1. The cardiopulmonary resuscitation pressing action posture feature AI identification and deviation correction method is characterized by comprising the following steps of: Acquiring trunk movement track data and chest pressure dynamic data when a rescuer executes a pressing action; performing space reconstruction on the trunk movement track data, extracting a trunk centroid movement rule and an angular velocity change trend, calculating stability components of trunk postures on a vertical plane and a horizontal plane, determining an optimal force application range and a force bearing supporting point, and generating a steady-state supporting area of a rescuer; Constructing a pressure response state space for the chest pressure dynamic data, extracting dynamic characteristics of a pressure action sequence, establishing a pressure elastic deformation field, analyzing energy dissipation distribution of a conduction path, and generating an optimal pressure application path; mapping the steady-state support area and the optimal pressure application path to an elastic potential energy field, optimizing the form of the support structure based on dynamic balance constraint, and generating a real-time corrected posture adjustment instruction sequence according to fatigue and stability indexes; And executing action correction according to the gesture adjustment instruction sequence, adjusting the standing distance, the supporting angle and the force application direction of the rescuer, and finishing the pressing action deviation correction.
2. The method of claim 1, wherein spatially reconstructing the torso-motion trajectory data, extracting torso centroid motion laws and angular velocity change trends comprises: Establishing a three-dimensional coordinate system taking a trunk of a rescuer as a center, mapping trunk motion track data to the three-dimensional coordinate system to obtain three-dimensional space motion data, and carrying out motion boundary correction by combining the human trunk joint motion degree range to obtain corrected trunk motion characteristics; Decomposing the corrected trunk motion characteristics by using a quaternion rotation matrix to obtain a main direction motion component and an auxiliary direction motion component, performing projection combination on the main direction motion component and the auxiliary direction motion component to obtain a trunk centroid motion track, performing time domain analysis on the trunk centroid motion track to obtain trunk centroid speed change, and performing displacement compensation to generate a trunk centroid motion rule; and processing the motion law of the trunk centroid by utilizing a spherical linear interpolation algorithm, extracting a trunk rotation characteristic quantity to calculate a triaxial rotation angle, determining a trunk posture change sequence, and carrying out angle differentiation on the trunk posture change sequence to generate an angular velocity change trend.
3. The method of claim 1, wherein calculating the stability components of the torso pose in the vertical and horizontal planes, determining the optimal force application range and force support points, generating a steady-state support region for the rescuer comprises: Decomposing the motion rule of the trunk centroid to a vertical plane and a horizontal plane to obtain a trunk centroid plane projection track; Performing amplitude analysis on the projection track of the trunk centroid plane, extracting peak points and valley points, calculating fluctuation amplitude and time intervals of adjacent peak points and valley points to obtain centroid movement fluctuation characteristics, determining a fluctuation period mapping relation, extracting centroid movement stability characteristics, performing period segmentation on the projection characteristics of the trunk angular movement plane, extracting angular movement stability characteristics, performing weighted combination on the centroid movement stability characteristics and the angular movement stability characteristics, and generating a vertical plane stability component and a horizontal plane stability component; And (3) carrying out threshold segmentation on the vertical plane stability component and the horizontal plane stability component, extracting a time period with an optimal stability index, determining a trunk posture optimal stability interval, determining a trunk posture stability action point according to the motion characteristics in the trunk posture optimal stability interval, and constructing a polygonal envelope by taking the trunk posture stability action point as a center to generate a rescuer steady-state support area.
4. The method of claim 1, wherein constructing a pressure response state space for the chest pressure dynamic data, extracting dynamic features of a pressure application sequence, establishing a pressure elastic deformation field, analyzing an energy dissipation profile of a conduction path, generating an optimal pressure application path comprises: mapping chest pressure dynamic data to a multidimensional state space, extracting a phase space track of a pressure state vector, determining a stable interval according to a topological structure of pressure fluctuation, and carrying out nonlinear reconstruction on the pressure data to obtain a single-period pressure action sequence; Performing pressure conduction analysis on the single-period pressure action sequence, extracting pressure gradients and propagation speeds of all sampling points, determining the conduction direction of pressure waves in thoracic tissues according to the pressure gradients, calculating displacement vectors of the pressure action points along the conduction direction, performing time sequence association analysis to obtain a spatial position change sequence of the pressure action points, constructing a pressure elastic deformation field based on the spatial position change sequence, and determining the propagation path of the pressure waves in thoracic tissues according to the pressure elastic deformation field to obtain a pressure conduction region; Calculating pressure attenuation coefficients between adjacent sampling points in a pressure conduction area, determining loss characteristics of pressure wave energy in a conduction process, performing cumulative calculation to obtain energy dissipation distribution on a conduction path, constructing a conduction efficiency evaluation index based on the energy dissipation distribution, sorting and screening a plurality of conduction paths according to the conduction efficiency evaluation index, determining a minimum energy dissipation conduction path, and generating an optimal pressure application path.
5. The method of claim 4, wherein mapping the chest pressure dynamic data to a multidimensional state space, extracting a phase space trajectory of a pressure state vector, determining a stability interval based on a topology of pressure fluctuations, and performing a nonlinear reconstruction of the pressure data to obtain a single cycle pressure action sequence comprises: mapping chest pressure dynamic data to a multidimensional state space, constructing a pressure response phase matrix, and performing time delay embedding transformation on the phase matrix to generate a pressure state vector; Calculating a distance matrix between adjacent state points in the pressure state vector, converting the distance matrix into a state transition diagram, extracting a communication component of the state transition diagram, acquiring an invariant set in a state space, and mapping a phase space track of pressure change by using the invariant set to form a topological structure of pressure fluctuation; And carrying out matrix decomposition on the topological structure, extracting structural parameters of the attractors, determining a convergence domain boundary of the structural parameters, marking critical state points in the convergence domain boundary, dividing a stable interval of pressure fluctuation according to the critical state points, carrying out nonlinear reconstruction on pressure data in the stable interval, and generating a single-period pressure action sequence.
6. The method of claim 1, wherein mapping the steady-state support region and the optimal pressure application path to an elastic potential energy field, optimizing support structure morphology based on dynamic balance constraints, generating a real-time corrected pose adjustment command sequence as a function of fatigue and stability metrics comprises: Acquiring a coordinate sequence and a pressure value sequence of a supporting point of a steady-state supporting area, calculating to obtain a potential energy value sequence through elastic potential energy, constructing an elastic potential energy field by using the potential energy value sequence, and mapping the optimal pressure application path into the elastic potential energy field to obtain potential energy gradient distribution; Calculating the strain characteristics of the support points based on potential energy gradient distribution, analyzing to obtain pressure conduction efficiency, generating a support point dynamic reconstruction sequence according to the pressure conduction efficiency, combining the support point dynamic reconstruction sequence with preset human biomechanical parameters, constructing dynamic balance constraint, optimizing according to the dynamic balance constraint to obtain a support structure form, and converting the support structure form into a station coordinate parameter and a posture angle parameter; The method comprises the steps of collecting real-time action data of a rescuer, calculating and determining a muscle fatigue curve and a gesture stability index, inputting the muscle fatigue curve and the gesture stability index into dynamic balance constraint, updating to obtain support point reconstruction parameters, generating gesture correction instructions based on the support point reconstruction parameters, station coordinate parameters and gesture angle parameters, and organizing the gesture correction instructions according to time sequence to obtain a gesture adjustment instruction sequence for real-time correction.
7. The method of claim 6, wherein calculating the support point strain signature based on the potential energy gradient profile, the analysis resulting in pressure conduction efficiency comprises: Extracting a main direction vector of the potential energy gradient distribution, constructing a pressure conduction path along the main direction vector, and calculating a stress concentration coefficient of a supporting point on the pressure conduction path; Determining a local strain threshold of the supporting point according to the stress concentration coefficient, classifying strain characteristics of the supporting point exceeding the strain threshold, and calculating the mapping relation between the strain quantity and the pressure conduction efficiency of various strain characteristic supporting points; and combining the mapping relation with the spatial distribution of the supporting points to construct a pressure conduction network, and determining the overall pressure conduction efficiency through calculation of the pressure conduction network.
8. Cardiopulmonary resuscitation pressing motion profile AI identification and deviation correction system for implementing the method according to any one of the preceding claims 1-7, characterized by comprising: The data acquisition module is used for acquiring trunk movement track data and chest pressure dynamic data when the rescuer executes the pressing action; The support region optimization module is used for carrying out space reconstruction on the trunk movement track data, extracting a trunk centroid movement rule and an angular velocity change trend, calculating stability components of trunk postures on a vertical plane and a horizontal plane, determining an optimal force application range and a stress support point, and generating a steady-state support region of a rescuer; The pressure path analysis module is used for constructing a pressure response state space for the chest pressure dynamic data, extracting dynamic characteristics of a pressure action sequence, establishing a pressure elastic deformation field, analyzing energy dissipation distribution of a conduction path and generating an optimal pressure application path; the gesture correction module is used for mapping the steady-state support area and the optimal pressure application path to an elastic potential energy field, optimizing the form of the support structure based on dynamic balance constraint, and generating a gesture adjustment instruction sequence corrected in real time according to fatigue and stability indexes; And the action execution module is used for executing action correction according to the gesture adjustment instruction sequence, adjusting the standing distance, the supporting angle and the force application direction of the rescuer and finishing the pressing action deviation correction.
9. An electronic device, comprising: A processor; A memory for storing processor-executable instructions; Wherein the processor is configured to invoke the instructions stored in the memory to perform the method of any of claims 1 to 7.
10. A computer readable storage medium having stored thereon computer program instructions, which when executed by a processor, implement the method of any of claims 1 to 7.

Description

Cardiopulmonary resuscitation pressing action gesture feature AI identification and deviation correction method Technical Field The invention relates to the technical field of medical emergency treatment, in particular to a cardiopulmonary resuscitation pressing action posture feature AI identification and deviation correction method. Background The quality of cardiopulmonary resuscitation pressing action directly influences survival rate and nerve function recovery effect of the patient, and because the pressing action involves a complex human biomechanics mechanism, a rescuer often has difficulty in accurately mastering the correct posture and force application mode in the execution process. With the development of artificial intelligence technology and sensing technology, the use of intelligent recognition method to monitor and guide the pressing action in real time has become an important research direction for improving the quality of cardiopulmonary resuscitation. The traditional training mode mainly depends on manual demonstration and feedback, lacks quantitative analysis on action details, and is difficult to realize personalized accurate guidance. In recent years, although some monitoring devices based on sensors are presented, there is still a technical bottleneck in terms of comprehensive analysis and real-time correction of motion gestures. The prior art has obvious limitations in the aspects of cardiopulmonary resuscitation pressing action monitoring and guiding, usually focuses on monitoring single indexes such as pressing depth and frequency, lacks systematic analysis on the overall trunk motion state of a rescuer, cannot identify the problem of low force transmission efficiency caused by unreasonable gestures, and can not comprehensively consider the fatigue state, support stability and dynamic balance relation among the pressure application efficiency of the rescuer because of improper support point selection or body gravity center deviation, meanwhile, the fatigue accumulation and damage risk of the rescuer are increased, the analysis on chest pressure response characteristics in the prior art stays on the surface layer, a dynamic deformation model in the pressure application process cannot be established, the conduction path and energy dissipation condition of force in chest tissues cannot be accurately estimated, the optimal force application direction and action point cannot be found by the rescuer, the condition of uneven pressure distribution or insufficient effective compression depth can possibly occur, and the traditional correction guiding method lacks instantaneity and systematicness, the provided adjustment suggestion is often too general to be converted into specific action correction parameters, and accurate personalized guiding cannot be realized, and the continuous pressing quality improvement effect is influenced. Disclosure of Invention The embodiment of the invention provides a cardiopulmonary resuscitation pressing action gesture feature AI identification and deviation correction method, which can solve the problems in the prior art. In a first aspect of the embodiment of the present invention, a cardiopulmonary resuscitation pressing motion gesture feature AI identification and deviation correction method is provided, including: Acquiring trunk movement track data and chest pressure dynamic data when a rescuer executes a pressing action; performing space reconstruction on the trunk movement track data, extracting a trunk centroid movement rule and an angular velocity change trend, calculating stability components of trunk postures on a vertical plane and a horizontal plane, determining an optimal force application range and a force bearing supporting point, and generating a steady-state supporting area of a rescuer; Constructing a pressure response state space for the chest pressure dynamic data, extracting dynamic characteristics of a pressure action sequence, establishing a pressure elastic deformation field, analyzing energy dissipation distribution of a conduction path, and generating an optimal pressure application path; mapping the steady-state support area and the optimal pressure application path to an elastic potential energy field, optimizing the form of the support structure based on dynamic balance constraint, and generating a real-time corrected posture adjustment instruction sequence according to fatigue and stability indexes; And executing action correction according to the gesture adjustment instruction sequence, adjusting the standing distance, the supporting angle and the force application direction of the rescuer, and finishing the pressing action deviation correction. In an optional embodiment, performing spatial reconstruction on the trunk motion trajectory data, and extracting a trunk centroid motion rule and an angular velocity change trend includes: Establishing a three-dimensional coordinate system taking a trunk of a rescuer as a center, mappin