Pulmonary care system for airway clearance based on ultrasound technology

Abstract

This invention discloses a lung care system for airway clearance based on ultrasound technology, belonging to the field of airway clearance technology. The system includes a data acquisition module, a risk assessment and control module, an ultrasound treatment parameter matching module, an ultrasound treatment execution module, a real-time feedback and control module, and a data storage module. This invention solves the problem that existing technologies cannot collect efficacy-related data in real time during treatment, nor can they dynamically adjust treatment plans based on efficacy data. This invention can dynamically adjust ultrasound treatment parameters according to the patient's real-time airway function status, achieving data-driven precision nursing, significantly improving airway clearance effects, reducing the risk of complications caused by sputum residue, avoiding ineffective treatment and over-intervention, shortening the treatment cycle, improving nursing efficiency, significantly reducing the risk of treatment-related adverse events, and enhancing the safety and reliability of lung care.

Description

Technical Field

[0001] This invention relates to the field of airway clearance technology, specifically to a lung care system for airway clearance based on ultrasound technology. Background Technology

[0002] Airway clearance techniques can help expel sputum or secretions from the airways, aiding in the prevention and treatment of respiratory diseases. Mechanical airway clearance, in particular, uses switching between positive and negative pressure to simulate the expiratory flow rate from the lungs, thus promoting the expulsion of secretions.

[0003] Chinese Patent Publication No. CN115721535B discloses a data processing method and related system for airway clearance, including determining an EIT gas distribution map with gas spatial distribution based on EIT data, determining first lung ventilation parameters including inspiratory volume and ventilation homogeneity index for each positive pressure level based on the EIT gas distribution map, determining positive pressure results corresponding to multiple positive pressure levels based on the first lung ventilation parameters, wherein ventilation change results are determined based on the relationship between pressure change values, volume change values ​​of inspiratory volume, and exponential change values ​​of ventilation homogeneity index for at least two positive pressure levels to determine positive pressure results corresponding to multiple positive pressure levels; determining second lung ventilation parameters for each negative pressure level based on the EIT gas distribution map, including second lung ventilation parameters of expiratory volume, expiratory volume distribution, peak cough flow rate, and effective cough volume, and determining negative pressure results corresponding to multiple negative pressure levels based on the second lung ventilation parameters.

[0004] The aforementioned patents cannot collect efficacy-related data in real time during treatment, nor can they dynamically adjust treatment plans based on efficacy data; therefore, they do not meet existing needs. In response, we propose a lung care system for airway clearance based on ultrasound technology. Summary of the Invention

[0005] The purpose of this invention is to provide a lung care system based on ultrasound technology for airway clearance. This system can dynamically adjust ultrasound treatment parameters according to the patient's real-time airway function status, achieving data-driven precision care, significantly improving airway clearance effectiveness, reducing the risk of complications caused by sputum residue, avoiding ineffective treatment and over-intervention, shortening the treatment cycle, reducing the workload of medical staff in manually adjusting treatment plans, improving nursing efficiency, reducing medical resource consumption, enabling rapid response to abnormalities, and recording abnormal data for plan optimization. This significantly reduces the risk of treatment-related adverse events, improves the safety and reliability of lung care, and solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a lung care system for airway clearance based on ultrasound technology, comprising: The data acquisition module is used to collect basic data before nursing care and real-time data during treatment. The risk assessment and control module is used to receive basic physiological safety data before nursing care, calculate the patient's treatment tolerance through a preset risk assessment algorithm, output the judgment result, and generate initial control instructions based on the judgment result. The ultrasound therapy parameter matching module is used to generate initial treatment parameters through edge computing algorithms, and dynamically adjust the ultrasound therapy parameters in combination with the initial treatment parameters and control commands. The ultrasound therapy execution module is used to perform ultrasound therapy according to the ultrasound therapy parameters output by the ultrasound therapy parameter matching module; The real-time feedback control module is used to analyze dynamic data during the treatment process in real time and generate parameter optimization control instructions and safety emergency control instructions. The data storage module is used to store treatment-related data and generate individualized lung care reports based on the stored data, including details of treatment parameters, efficacy assessment results, risk warning records, and follow-up care recommendations.

[0007] Preferably, airway function-related data are collected, including diaphragmatic mobility measured by an ultrasound sensor and cough peak flow rate measured by a flow sensor; Collect respiratory mechanics-related data, including collecting pressure changes during the patient's inspiration / expiration process using pressure sensors, and calculating the maximum inspiratory pressure and maximum expiratory pressure based on these pressure changes; Collect physiological safety-related data and lung ultrasound imaging data, and identify information on lung parenchymal lesion areas; Collect sputum output, respiratory flow rate, and changes in airway pressure, and calculate changes in airway resistance using changes in respiratory flow rate and airway pressure.

[0008] Preferably, the implementation process of the risk assessment and control module specifically includes: After receiving basic physiological safety data before nursing care and determining the patient's treatment tolerance based on a preset risk assessment algorithm, the determination results are converted into standardized control signals. Based on the signal type and risk level, control signals are precisely transmitted to the corresponding functional modules; The system receives real-time physiological safety data during treatment and reassesses the patient's risk level using a preset risk assessment algorithm. If the risk level increases, the control signals are immediately adjusted and redistributed.

[0009] Preferably, the implementation process of the ultrasound therapy parameter matching module specifically includes: An initial matching model for ultrasound treatment parameters is constructed using an edge computing algorithm. Simultaneously, multi-source basic data before nursing care is received and preprocessed to obtain multi-source data feature vectors. The preprocessed multi-source data feature vectors are input into the initial matching model of ultrasound treatment parameters to generate the initial adaptation results of treatment parameters. Receive initial control instructions and generate treatment parameter instructions based on the initial adaptation results of treatment parameters; The system receives control commands from the real-time feedback control module to dynamically optimize treatment parameters, and sends the initial treatment parameter commands and the optimized treatment parameter commands to the ultrasound treatment execution module and the data storage module.

[0010] Preferably, the generation of initial adaptation results for ultrasound treatment parameters specifically includes: When the cough peak flow rate is not as expected, the initial matching is a low-frequency ultrasound enhancement mode, the ultrasound frequency is set to the low-frequency range, and the power is increased by a preset ratio compared to the conventional mode. When the diaphragm movement is detected to be within the normal range, the initial matching is a conventional ultrasound mode, with the ultrasound frequency set to the conventional range and the power set to the conventional reference power range. When the maximum inspiratory pressure is detected to be below the normal range, the initial matching is a low-power long-duration mode, the ultrasonic power is set to a low-power ratio range of the conventional reference power, and the duration of action is extended by a preset ratio compared to the conventional mode. When lung ultrasound images detect areas of lung parenchymal lesions, an initial matching targeted focusing mode is used, setting the ultrasound energy to focus on the lesion area, and the power is dynamically adjusted to an appropriate power range based on the depth of the lesion area.

[0011] Preferably, the implementation process of the real-time feedback control module specifically includes: It receives dynamic data of the treatment process in real time, performs rapid analysis and feature extraction on the received dynamic data, and identifies abnormal data and failure to meet efficacy standards by combining preset normal data range and efficacy evaluation threshold. Based on the data analysis results, corresponding control instructions are generated according to categories. If a situation where the therapeutic effect does not meet the standard is identified, parameter optimization control instructions are generated and sent to the ultrasound treatment parameter matching module for corresponding adjustments. If a physiological safety abnormality is detected, a safety emergency control command is generated and distributed to the ultrasound treatment execution module for corresponding adjustments.

[0012] Preferably, if a situation where the therapeutic effect does not meet the target is identified, a parameter optimization and control command is generated and sent to the ultrasound treatment parameter matching module for corresponding adjustments, specifically including: When the decrease in airway resistance is not as expected, optimize the ultrasonic frequency band to the range of optimal cavitation effect. When the amount of sputum expectorated reaches the treatment target, the treatment termination procedure is automatically triggered, the treatment data is recorded, and a treatment report is generated. When the real-time cough peak flow rate is detected to increase to a safe range and the duration reaches the set duration, the system automatically switches from low-frequency ultrasound enhancement mode to regular ultrasound mode. When the real-time maximum inspiratory pressure is detected to rise to the preset safe range, the ultrasound treatment duration is automatically adjusted back to the normal mode.

[0013] Preferably, if a physiological safety abnormality is detected, a safety emergency control command is generated and distributed to the ultrasound treatment execution module for corresponding adjustments, specifically including: When arrhythmia is detected, the ultrasound power will be reduced to a safe value within a set time. If the duration of arrhythmia reaches the preset duration, the treatment will be automatically terminated. When blood oxygen saturation is detected to be below the safe value, treatment will be terminated within a set time and an alarm will be triggered. When the real-time heart rate or systolic blood pressure is detected to be outside the safe range, the ultrasound power is reduced by a preset ratio within a set time. If the physiological data does not return to normal within the set time, the treatment is terminated.

[0014] Preferably, the step of allocating data processing tasks to target nodes specifically includes: The system uses a resource monitoring daemon embedded in the system to check each candidate node in real time with a preset time granularity. It calculates and normalizes the proportion of time slices in which the CPU is in a non-idle state when processing airway function data in the most recent time window to obtain the current CPU load rate of each candidate node. At the same time, it obtains the network response delay sequence generated by probe data packets for each candidate node within a preset historical period. Calculate the statistical variance of the network response delay sequence or construct a probability density histogram of the delay distribution, and calculate the network entropy value using the Shannon entropy model. The network entropy value is used to characterize the uncertainty of the node network link when transmitting medical safety instructions. At the same time, query the system startup log or hardware running clock to obtain the continuous running time of each candidate node since the last restart. The admission priority index for each candidate node is calculated using the following dynamic resilience potential energy formula:

[0015] In this formula: Defined as the admission priority index; Defined as the normalized current CPU load rate of the candidate node; Defined as the network entropy value, whose value is positively correlated with the variance of the network response delay sequence; Defined as the continuous runtime; Defined as a preset system cold start warm-up constant; Defined as a preset load sensitivity coefficient, it is used to adjust the system's weighting of idle computing resources; Defined as a preset jitter suppression factor, it is used to achieve nonlinear compressed sensing of network instability in the calculation; The natural constant; the exponential term in the formula As a cold start suppression factor, it is used when the candidate node has just started and the continuous running time is... Less than the system cold start warm-up constant At that time, the acceptance priority index is non-linearly reduced to prevent the node from triggering an avalanche effect due to task overload in the early stage of the ultrasound treatment procedure. All candidate nodes are sorted in descending order based on the calculated admission priority index, and the node with the highest admission priority index is selected as the target node.

[0016] Preferably, the method further includes a deadlock avoidance step based on vector projection before performing the data processing task, which specifically includes: Static resource requirement analysis is performed on the pending tasks at the head of the execution queue, and all resource IDs that must be locked during the execution process are extracted as the operation object identifier set. Based on the preset resource requirement feature mapping table constructed using resource feature mapping technology, the pending tasks are converted into high-dimensional feature vectors, where the direction of the vector indicates the resource requirement type, and the numerical value of the vector in a specific dimension represents the intensity or duration weight of the resource occupation. Scan all existing tasks that are running in the current target node and have not yet released resources, convert each existing task into a corresponding high-dimensional feature vector, and stack these vectors as row vectors to construct a system state matrix representing the subsystem where resources are currently occupied. Map the high-dimensional feature vector of the task to be processed to the linear space where the system state matrix is ​​located, and calculate the orthogonal projection components of the high-dimensional feature vector on each row vector in the system state matrix; Determine whether the magnitudes of all the calculated orthogonal projection components are less than a preset safety threshold; If the magnitude of all the orthogonal projection components is less than the safety threshold, it is determined that the task to be processed is in a quasi-orthogonal state with the current system state, there is no logical resource deadlock, and an execution permission instruction is generated. If the magnitude of any of the orthogonal projection components is greater than or equal to the safety threshold, a potential resource contention risk is identified, the pending task is suspended, and a task gradient backoff mechanism is triggered: the dynamic backoff time is calculated based on the magnitude of the orthogonal projection component exceeding the safety threshold; the greater the conflict magnitude, the longer the backoff time. The determination is retried after the dynamic backoff time ends.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention upgrades data processing to real-time treatment-driven by a real-time linkage mechanism between ultrasound therapy and airway function data and multi-source data fusion. It achieves instant matching of core parameters such as cough peak flow rate and diaphragmatic movement with ultrasound therapy parameters through edge computing algorithms. Ultrasound therapy parameters can be dynamically adjusted according to the patient's real-time airway function status, realizing data-driven precision care, significantly improving airway clearance effect, and reducing the risk of complications caused by sputum residue.

[0018] 2. This invention optimizes ultrasound parameters by collecting efficacy-related data such as sputum output and airway resistance changes in real time during treatment. This allows for immediate verification of treatment effectiveness and dynamic optimization of intervention plans, avoiding ineffective treatment and over-intervention, shortening the treatment cycle, reducing the workload of medical staff in manually adjusting plans, improving nursing efficiency, and reducing medical resource consumption. Before treatment, multi-source risk data screening enables treatment access control, identifying treatment functions for high-risk patients and indicating risks. Treatment parameters are limited for borderline risk groups. During treatment, continuous monitoring of physiological data such as ECG and blood oxygenation allows for rapid response when abnormalities occur, and abnormal data is recorded for plan optimization, significantly reducing the risk of treatment-related adverse events and improving the safety and reliability of pulmonary care. Attached Figure Description

[0019] Figure 1 This is a block diagram of the lung care system for airway clearance based on ultrasound technology according to the present invention; Figure 2 This is a flowchart of the lung care system for airway clearance based on ultrasound technology according to the present invention. Figure 3 This is a graph showing the initial adaptation results of the ultrasound treatment parameters according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] To address the limitations of current technologies in collecting real-time efficacy-related data during treatment and in dynamically adjusting treatment plans based on efficacy data, please refer to [link to relevant documentation]. Figures 1-3 This embodiment provides the following technical solution: Ultrasound-based pulmonary care systems for airway clearance include: The ultrasound therapy parameter matching module is used to generate initial treatment parameters through edge computing algorithms, and dynamically adjust the ultrasound therapy parameters in combination with the initial treatment parameters and control commands. The data acquisition module is used to collect basic data before nursing care and real-time data during treatment. The risk assessment and control module is used to receive basic physiological safety data before nursing care, calculate the patient's treatment tolerance through a preset risk assessment algorithm, output three judgment results: high risk, borderline risk, and low risk, and generate initial control instructions based on the judgment results. The ultrasound therapy execution module is used to perform ultrasound therapy according to the ultrasound therapy parameters output by the ultrasound therapy parameter matching module; The real-time feedback control module is used to analyze dynamic data (airway function, efficacy, physiological safety) during the treatment process in real time, and generate parameter optimization control instructions and safety emergency control instructions. The data storage module is used to store treatment-related data, including all data related to this treatment (basic data, dynamic data, parameter data, and efficacy data) and supports offline viewing. It receives uploaded real-time data and implements encrypted data storage and access management. At the same time, it generates individualized lung care reports based on the stored data, including treatment parameter details, efficacy assessment results, risk warning records, and follow-up nursing suggestions.

[0022] The implementation process of the data acquisition module specifically includes: Collect airway function-related data, including measuring diaphragmatic movement using an ultrasound sensor and measuring peak cough flow rate using a flow sensor; Collect respiratory mechanics-related data, including collecting pressure changes during the patient's inspiration / expiration process using pressure sensors, and calculating the maximum inspiratory pressure and maximum expiratory pressure based on these pressure changes; Collect physiological safety-related data and lung ultrasound imaging data and identify lung parenchymal lesion areas. Physiological safety-related data include electrocardiogram data and heart rate collected through the electrocardiogram monitoring unit, blood pressure collected through the blood pressure monitoring unit, intracranial pressure collected through the intracranial pressure monitoring unit, and blood oxygen saturation collected through the blood oxygen monitoring unit. Collect sputum output, respiratory flow rate, and changes in airway pressure, and calculate changes in airway resistance using changes in respiratory flow rate and airway pressure.

[0023] The implementation process of the risk assessment and control module specifically includes: The system receives basic physiological safety data (heart rate, blood pressure, intracranial pressure) before nursing care. After determining the patient's treatment tolerance based on a preset risk assessment algorithm, it converts the three judgment results of high risk, critical risk, and low risk into standardized control signals. Among them, high risk corresponds to two types of signals: locked treatment signal and risk warning signal; critical risk corresponds to parameter limit signal; and low risk corresponds to routine treatment preparation signal, ensuring that the signal type accurately matches the receiving requirements of subsequent modules. The control signals are precisely distributed according to the signal type and risk level. The specific distribution logic is as follows: when the risk level is high, a lock treatment signal is sent to the ultrasound treatment execution module; when the risk level is critical, a parameter limit signal is sent to the ultrasound treatment parameter matching module; when the risk level is low, a routine treatment preparation signal is sent to the ultrasound treatment parameter matching module. The system receives real-time physiological safety dynamic data (such as real-time heart rate and real-time blood pressure) transmitted from the real-time feedback control module during the treatment process, and reassesses the patient's risk level through a preset risk assessment algorithm. If the risk level increases (such as low risk turning into borderline risk, or borderline risk turning into high risk), the control signal is immediately adjusted and redistributed. For example, when the heart rate of a low-risk patient rises to >120 beats / min, a new parameter limit signal is automatically generated and sent to the ultrasound treatment parameter matching module to ensure safety control throughout the treatment process.

[0024] The preset risk assessment algorithm adopts a weighted scoring model, specifically: weight coefficients are set for heart rate, blood pressure, and intracranial pressure (heart rate weight 0.4, blood pressure weight 0.3, intracranial pressure weight 0.3). Scores are assigned based on the degree to which each parameter deviates from the normal range (normal range is heart rate 60-100 beats / min, systolic blood pressure 90-140 mmHg, intracranial pressure 5-15 mmHg). The greater the deviation, the higher the score. A total score ≥ 60 points is considered high-risk, a total score of 40-59 points is considered borderline risk, and a total score < 40 points is considered low-risk.

[0025] The implementation process of the ultrasound therapy parameter matching module specifically includes: An initial matching model for ultrasound treatment parameters was constructed using an edge computing algorithm. Simultaneously, multi-source baseline data from pre-nursing care was received and preprocessed to obtain multi-source data feature vectors. The initial matching model for ultrasound treatment parameters was based on the clinical diagnosis and treatment guidelines for airway clearance nursing and was generated by training with a large amount of clinical sample data. It covers a baseline library of treatment parameters corresponding to different airway function states, respiratory mechanics characteristics, and lung lesions. The preprocessing included data format standardization and conversion, abnormal data removal, and data feature extraction (extracting features such as lesion area and depth from lung ultrasound images, and extracting core features such as CPF and diaphragmatic mobility from airway function data). The preprocessed multi-source data feature vector is input into the initial matching model of ultrasound treatment parameters to generate the initial adaptation result of ultrasound treatment parameters. The input layer of the edge computing algorithm is the multi-source data feature vector, the hidden layer is a 3-layer fully connected layer, the deep fusion of features is achieved through the activation function, and the output layer is the combination of ultrasound treatment parameters. Receive initial control instructions and generate treatment parameter instructions based on the initial adaptation results of treatment parameters; The system receives control commands from the real-time feedback control module to dynamically optimize treatment parameters, and sends the initial treatment parameter commands and the optimized treatment parameter commands to the ultrasound treatment execution module and the data storage module.

[0026] Generate initial adaptation results for ultrasound therapy parameters, including ultrasound frequency, power, and duration of action, specifically: When the cough peak flow rate is not as expected, the initial matching is a low-frequency ultrasound enhancement mode, the ultrasound frequency is set to the low-frequency range, and the power is increased by a preset ratio compared to the conventional mode. When the diaphragm movement is detected to be within the normal range, the initial matching is a conventional ultrasound mode, with the ultrasound frequency set to the conventional range and the power set to the conventional reference power range. When the maximum inspiratory pressure is detected to be below the normal range, the initial matching is a low-power long-duration mode, the ultrasonic power is set to a low-power ratio range of the conventional reference power, and the duration of action is extended by a preset ratio compared to the conventional mode. When lung ultrasound images detect areas of lung parenchymal lesions, an initial matching targeted focusing mode is used, setting the ultrasound energy to focus on the lesion area, and the power is dynamically adjusted to an appropriate power range based on the depth of the lesion area.

[0027] The implementation process of the real-time feedback control module specifically includes: The system receives dynamic data during the treatment process in real time. This dynamic data includes: dynamic airway function data (real-time peak cough flow rate, real-time diaphragmatic movement), efficacy-related data (sputum expectoration volume, changes in airway resistance), and dynamic physiological safety data (real-time ECG data, blood oxygen saturation, real-time heart rate, and real-time systolic blood pressure). The abnormal physiological safety data identification logic is as follows: the presence of arrhythmias such as ventricular premature beats, atrial premature beats, or tachycardia in the ECG data; blood oxygen saturation <90%; real-time heart rate >120 beats / min; or real-time systolic blood pressure >160 mmHg. The efficacy data non-compliance identification logic is as follows: airway resistance decrease <5% (compared to the baseline value at the start of treatment). The system rapidly analyzes and extracts features from the received dynamic data in real time, and identifies abnormal data and cases where the efficacy is not met by combining the preset normal data range and efficacy evaluation threshold. Based on the data analysis results, corresponding control instructions are generated according to categories. If a situation where the therapeutic effect does not meet the target is identified, a parameter optimization control instruction is generated and sent to the ultrasound treatment parameter matching module for corresponding adjustments. The instruction content includes specific optimization parameters such as the target ultrasound frequency band and the power adjustment range. If a physiological safety abnormality is detected, a safety emergency control command is generated and distributed to the ultrasound treatment execution module for corresponding adjustments. The command content is divided into three categories according to the severity of the abnormality: power reduction, power dropping to the safety threshold, and immediate termination of treatment, with corresponding response time requirements.

[0028] If a failure to meet therapeutic standards is detected, a parameter optimization and control command is generated and sent to the ultrasound treatment parameter matching module for corresponding adjustments, including: When the decrease in airway resistance is not as expected, the ultrasonic frequency band is optimized to the optimal range of cavitation effect. The optimal range of cavitation effect is calibrated by preset experimental data. When the amount of sputum expectorated reaches the treatment target, the treatment termination procedure is automatically triggered, the treatment data is recorded, and a treatment report is generated. When the real-time cough peak flow rate is detected to increase to a safe range and the duration reaches the set duration, the system automatically switches from low-frequency ultrasound enhancement mode to regular ultrasound mode. When the real-time maximum inspiratory pressure is detected to rise to the preset safe range, the ultrasound treatment duration is automatically adjusted back to the normal mode.

[0029] If a physiological safety abnormality is detected, a safety emergency control command is generated and distributed to the ultrasound treatment execution module for corresponding adjustments, specifically including: When arrhythmia is detected, the ultrasound power will be reduced to a safe value within a set time. If the duration of arrhythmia reaches the preset duration, the treatment will be automatically terminated. When blood oxygen saturation is detected to be below the safe value, treatment will be terminated within a set time and an alarm will be triggered. When the real-time heart rate or systolic blood pressure is detected to be outside the safe range, the ultrasound power is reduced by a preset ratio within a set time. If the physiological data does not return to normal within the set time, the treatment is terminated.

[0030] The implementation process of the ultrasound therapy execution module specifically includes: An ultrasonic generator generates ultrasonic signals of corresponding frequency and power, and an ultrasonic transducer converts the electrical signals into ultrasonic mechanical waves that act on the patient's lung airway area. The ultrasound energy focusing adjustment unit adjusts the focusing position of the ultrasound energy according to the lesion area information in the lung imaging data to achieve targeted treatment; wherein, the ultrasound transducer supports frequency band adjustment of 38-55kHz and power adjustment range of 0.1-0.5W / cm².

[0031] Working principle: When using the ultrasound-based lung care system for airway clearance of this invention, according to... Figure 1 , Figure 2 and Figure 3 This includes the following steps: S1: Collect multi-source basic data, including airway function, respiratory mechanics, physiological safety and basic lung imaging data; S2: Based on basic physiological safety data, risk prediction and treatment access determination are made, and locked treatment, parameter restriction and routine treatment preparation operations are executed according to high, critical and low risk results respectively. S3: Based on basic data of airway function, respiratory mechanics and lung imaging, a model is built through edge computing algorithm to achieve individualized initial adaptation of ultrasound treatment parameters; S4: Start ultrasound treatment and simultaneously activate real-time data acquisition to continuously collect dynamic data on airway function, efficacy-related data, and dynamic data on physiological safety during the treatment process; S5: Dynamically regulate the treatment process based on real-time data acquisition, including optimizing treatment parameters based on efficacy data and performing optimized regulation based on abnormal physiological safety conditions; S6: After treatment is terminated, integrate and store various data, and generate an individualized lung care report based on the stored data, which includes details of treatment parameters, efficacy assessment results, risk warning records, and follow-up care recommendations.

[0032] In one embodiment, the step of assigning data processing tasks to target nodes specifically includes: The system uses a resource monitoring daemon embedded in the system to check each candidate node in real time with a preset time granularity. It calculates and normalizes the proportion of time slices in which the CPU is in a non-idle state when processing airway function data in the most recent time window to obtain the current CPU load rate of each candidate node. At the same time, it obtains the network response delay sequence generated by probe data packets for each candidate node within a preset historical period. Calculate the statistical variance of the network response delay sequence or construct a probability density histogram of the delay distribution, and calculate the network entropy value using the Shannon entropy model. The network entropy value is used to characterize the uncertainty of the node network link when transmitting medical safety instructions. At the same time, query the system startup log or hardware running clock to obtain the continuous running time of each candidate node since the last restart. The admission priority index for each candidate node is calculated using the following dynamic resilience potential energy formula:

[0033] In this formula: Defined as the acceptance priority index, with a dimension of 1; Defined as the normalized current CPU load rate of the candidate node, with a value ranging from 0 to 1; Defined as the network entropy value, whose value is positively correlated with the variance of the network response delay sequence; Defined as the continuous runtime, in seconds; Defined as a preset system cold start warm-up constant, in seconds; Defined as a preset load sensitivity coefficient, it is used to adjust the system's weighting of idle computing resources; Defined as a preset jitter suppression factor, it is used to achieve nonlinear compressed sensing of network instability in the calculation; The natural constant; the exponential term in the formula As a cold start suppression factor, it is used when the candidate node has just started and the continuous running time is... Less than the system cold start warm-up constant At that time, the acceptance priority index is non-linearly reduced to prevent the node from triggering an avalanche effect due to task overload in the early stage of the ultrasound treatment procedure. All candidate nodes are sorted in descending order based on the calculated admission priority index, and the node with the highest admission priority index is selected as the target node.

[0034] The working principle and beneficial effects of the above technical solution are as follows: This system establishes a comprehensive node physical state awareness model. In the system's hardware architecture, the data acquisition module is connected to each candidate computing node via a high-speed local area network or industrial fieldbus, forming a tightly coupled distributed computing cluster. To select the most suitable target node from these candidate nodes to undertake high-priority lung care computing tasks at any given moment, the system first needs to perform a real-time acquisition step of multi-source basic data.

[0035] In this step, the high-performance resource monitoring daemon embedded in the system checks each candidate node with a millisecond-level time granularity. Specifically, the system obtains the current CPU load rate of each candidate node in real time, denoted as... This parameter is not obtained by simply reading the instantaneous value from the operating system. Instead, it is obtained through a kernel-level performance counter interface, which calculates the proportion of CPU time slices in a non-idle state within the most recent time window (e.g., 500 milliseconds) and normalizes it to a dimensionless floating-point number between 0 and 1. The closer the value is to 1, the busier the node's computing core is, the more strained its internal instruction pipeline and cache resources are, and the higher the risk of blocking due to accepting new tasks. Conversely, the closer the value is to 0, the more idle and ready the node is. Simultaneously, the system also acquires a network response latency sequence within a preset historical period (e.g., the past 30 seconds). This sequence is a set of round-trip times (RTTs) generated by a series of probe packets (such as ICMP Echo messages or application-layer heartbeat commands) with high-precision nanosecond timestamps sent by the master node to candidate nodes. This sequence not only records the absolute latency of each communication but also fully preserves the dynamic characteristics of network transmission quality evolution over time through a discrete distribution on the timeline, providing raw data support for subsequent network stability analysis.

[0036] After obtaining the aforementioned basic physical parameters, the system enters the feature extraction and calculation stage. This embodiment introduces network entropy (denoted as...). This physical quantity characterizes the uncertainty of the network state of candidate nodes. In complex hospital electromagnetic environments or wireless network coverage areas, network links often experience invisible electromagnetic interference or channel congestion, leading to severe jitter in data transmission delays. Traditional average delay metrics mask this jitter risk, while network entropy... This allows for precise identification of the potential problem. In the specific calculations, the system first calculates the statistical variance based on the network response delay sequence, or constructs a probability density histogram of the delay distribution, and then uses the Shannon entropy model to calculate the network entropy value. Network entropy The magnitude of the entropy is strictly positively correlated with the variance and disorder of the network response delay sequence. If the absolute value of a node's network delay is low but fluctuates wildly and is extremely unstable, its calculated entropy value will increase significantly, indicating that the link is in a poor state and is not suitable for transmitting ultrasonic control commands with extremely high real-time requirements. Conversely, if the network delay is extremely stable, even if the absolute value is slightly large, its entropy value will remain at a low level, indicating that the link has high predictability.

[0037] Furthermore, to address the common cold start deception issue in distributed systems during node restarts or service resets, the system also precisely obtains the continuous runtime of each candidate node since its last restart by querying system startup logs or hardware runtime clocks, denoted as... The unit is seconds. A newly started node often has its hot data cache in memory, database connection pool, and Java Virtual Machine (JVM) Just-In-Time (JIT) compiler not yet fully initialized or optimized. At this time, although its CPU load is extremely low (because it hasn't started processing tasks yet), its actual instruction throughput is extremely fragile. If a large number of tasks are distributed to it based on low load metrics, it can easily cause the node to become instantly overloaded due to resource initialization blocking, or even trigger a secondary crash (i.e., a cascading failure). Therefore, continuous runtime... It is a key time-dimensional parameter for assessing the true robustness of nodes.

[0038] Based on the above data, the current CPU load rate Network entropy and continuous runtime This embodiment proposes a comprehensive scoring algorithm based on dynamic resilience potential, which uses a formula to calculate the admission priority index of each candidate node. This formula is a digital mapping of the carrying capacity of computing nodes in the physical world. The specific expression of the formula is: The following provides a detailed breakdown and explanation of the physical meaning, parameter selection, and technical effects of each mathematical term in the formula: First, regarding the loading potential term in the numerator of the formula. The core of this project lies in... This represents the idle computing power space of a node. Defined as a preset load sensitivity coefficient, it is a dimensionless constant greater than 0. In practical applications, the load sensitivity coefficient... The value is typically set between 1.5 and 3.0. Its technical function is to adjust the system's demand for computing resources. When the system is handling computationally intensive tasks (such as real-time 3D rendering of lung ultrasound images), the load sensitivity coefficient is increased. The value can amplify the weight of idle computing power in the final score, so that nodes with extremely idle CPUs can get significant bonuses, thereby guiding tasks to the nodes with the most abundant computing power and ensuring the smooth execution of complex algorithms.

[0039] Secondly, regarding the network damping term in the denominator of the formula... This term introduces the natural logarithm function. To process network entropy values. Defined as a preset jitter suppression factor, typically ranging from 0.5 to 1.0. This design achieves nonlinear compressed sensing of network instability. (The last part, "in network entropy value," appears to be an error and doesn't translate directly.) Within a relatively small (relatively stable network) interval, the slope of the logarithmic function is large, indicating that the system is highly sensitive to even minor network fluctuations. Any slight instability will lead to an increase in the denominator and a decrease in the score. However, when the network entropy is already very high (extreme network congestion or severe interference), the slope of the logarithmic function gradually flattens out to prevent the denominator from expanding infinitely and causing the score to drop directly to zero. This preserves the minimum theoretical availability of the node under extremely harsh conditions at the algorithm level. This design perfectly matches the complex and ever-changing reality of the medical field network environment, avoiding the risk of system paralysis caused by an overly stringent algorithm that eliminates all nodes.

[0040] Finally, regarding the multiplier term in the formula, namely the cold start inhibition factor... .in, It is a natural constant. Defined as a preset system cold start warm-up constant, in seconds. The specific value depends on the node's hardware characteristics and software environment. For example, for server nodes running complex middleware, the system cold start warm-up constant... It can be set to 60 seconds, representing the typical time required for the entire system to warm up. From a mathematical perspective, the continuous runtime when a node is first started... The exponent term approaches 0. Approaching 1, leading to the entire inhibitory factor It approaches 0. This means that no matter how idle the node's CPU is or how stable the network is, its final acceptance priority index will remain constant. All of these will be constrained to a very low level by this factor. This forces the task scheduler to bypass the node during its initial operation. However, as continuous runtime increases... The shift exceeds the system cold start warm-up constant. The inhibition factor will rise smoothly along an exponential curve and eventually approach 1, indicating that the node has passed the vulnerable period and can carry services at full speed. This continuous control mechanism based on a time function enables smooth transition management of the node throughout its entire lifecycle, from startup to maturity.

[0041] After calculating the admission priority index of all candidate nodes Next, the system executes a sorting and optimization step. The system sorts all candidate nodes in descending order of their indices and selects the node with the highest index as the target node. Subsequently, the data acquisition module sends the packaged task data to the target node. This process is fully automatic and dynamically refreshed, ensuring that each task allocation is the optimal solution based on the current physical state of the system, thereby greatly improving the overall throughput and anti-interference capability of the lung care system.

[0042] In one embodiment, the method further includes a deadlock avoidance step based on vector projection before performing the data processing task, the step specifically including: Static resource requirement analysis is performed on the pending tasks at the head of the execution queue, and all resource IDs that must be locked during the execution process are extracted as the operation object identifier set. Based on the preset resource requirement feature mapping table constructed using resource feature mapping technology, the pending tasks are converted into high-dimensional feature vectors, where the direction of the vector indicates the resource requirement type, and the numerical value of the vector in a specific dimension represents the intensity or duration weight of the resource occupation. Scan all existing tasks that are running in the current target node and have not yet released resources, convert each existing task into a corresponding high-dimensional feature vector, and stack these vectors as row vectors to construct a system state matrix representing the subsystem where resources are currently occupied. Map the high-dimensional feature vector of the task to be processed to the linear space where the system state matrix is ​​located, and calculate the orthogonal projection components of the high-dimensional feature vector on each row vector in the system state matrix; Determine whether the magnitudes of all the calculated orthogonal projection components are less than a preset safety threshold; If the magnitude of all the orthogonal projection components is less than the safety threshold, it is determined that the task to be processed is in a quasi-orthogonal state with the current system state, there is no logical resource deadlock, and an execution permission instruction is generated. If the magnitude of any of the orthogonal projection components is greater than or equal to the safety threshold, a potential resource contention risk is identified, the pending task is suspended, and a task gradient backoff mechanism is triggered: the dynamic backoff time is calculated based on the magnitude of the orthogonal projection component exceeding the safety threshold; the greater the conflict magnitude, the longer the backoff time. The determination is retried after the dynamic backoff time ends.

[0043] The working principle and beneficial effects of the above technical solution are as follows: In pulmonary care systems, multiple concurrent tasks (such as simultaneous ultrasound parameter adjustment, database record writing, and real-time waveform rendering) often compete for limited system resources (such as memory buffers, hardware device handles, and database write locks). Traditional deadlock detection algorithms (such as loop detection in resource allocation graphs) have high computational complexity and often exhibit lag, making them difficult to meet the requirements of real-time control. Therefore, this invention proposes a pre-prediction method based on high-dimensional linear spatial geometric projection.

[0044] The implementation of this method first involves the vectorized feature extraction of the task. When a new task arrives at the head of the execution queue, the system first performs a static resource requirement analysis, extracting its set of operation object identifiers, i.e., a list of all resource IDs that the task must lock during execution. For mathematical operations, the system has a built-in pre-defined resource requirement feature mapping table. This table uses resource feature mapping technology to construct a high-dimensional linear vector space. In this space, each type of contestable key resource in the system (e.g., ultrasound probe A, ultrasound probe B, memory block region M, database table T, etc.) is mapped to a unique basis vector dimension. Based on this mapping table, the system numerically superimposes the dimensions corresponding to all resources required by the task, thereby transforming the abstract task logic into a unique high-dimensional feature vector. The direction of this vector geometrically precisely represents the task's resource requirement; the magnitude of the vector in a certain dimension represents the intensity or duration weight of the resource occupancy.

[0045] Meanwhile, the system kernel scans in real time all running tasks on the current target node that have not yet released resources. Using the same method, the system converts each running task into a corresponding high-dimensional feature vector. Then, the system stacks these running task vectors as row vectors to construct a real-time system state matrix. This matrix forms the subspace where current resources are occupied. Any new task falling into this subspace means it will compete for existing resources.

[0046] The deadlock detection logic is based on orthogonal projection calculation. The system maps the high-dimensional feature vector of the task to be processed to the linear space of the system state matrix, and calculates the orthogonal projection component of the vector on each row vector (i.e., each existing task) in the matrix. According to the principles of vector geometry: if two vectors are orthogonal (with an angle of 90 degrees), their dot product is 0, and the magnitude of the projection component is 0. This means that the resources required by the two tasks do not overlap at all, they do not interfere with each other, and they can be perfectly parallelized. If the two vectors are in the same direction (with a small angle), the magnitude of their projection component will be significantly greater than 0. This means that the two tasks need to compete for the same critical resources and there is a fierce competition relationship. The larger the projection magnitude, the higher the overlap of resource requirements, and the greater the probability of deadlock or blocking.

[0047] The system presets a safety threshold, which is an empirical constant set based on the system's resource redundancy and concurrent processing capabilities. The system iterates through the magnitudes of all calculated orthogonal projection components and compares them to this safety threshold. If the result shows that the magnitudes of all calculated orthogonal projection components are less than the safety threshold, this proves that the task vector to be processed maintains a sufficient safe angle with all current old task vectors and is in a quasi-orthogonal state. Based on this, the system determines that there is no logical resource deadlock between the task to be processed and the current system state. An execution permission instruction is then generated, and the task is immediately pushed into the CPU pipeline for execution. This determination method does not require traversing a complex resource allocation graph; it only requires a few vector dot product operations, which can be completed in microseconds using the SIMD (Single Instruction Multiple Data) instruction set of modern processors, making it extremely fast.

[0048] If the judgment result shows that the magnitude of any orthogonal projection component is greater than or equal to the safety threshold, this geometrically means that the new task and an old task are too close in the direction of resource demand. Based on this, the system determines that there is a potential risk of resource contention. At this point, to prevent the system from freezing, the system suspends the issuance of the pending tasks and triggers a task gradient backoff mechanism. This mechanism calculates a dynamic backoff time based on the extent to which the projection magnitude exceeds the threshold. If the magnitude just exceeds the threshold, the conflict is minor, and the backoff time may only be a few milliseconds; if the magnitude far exceeds the threshold, the conflict is severe, and the backoff time may be as long as hundreds of milliseconds. This gradient backoff based on the degree of conflict avoids deadlock while maximizing task throughput efficiency, achieving flexible scheduling of the system.

[0049] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0050] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A lung care system for airway clearance based on ultrasound technology, characterized in that, include: The data acquisition module is used to collect basic data before nursing care and real-time data during treatment. The risk assessment and control module is used to receive basic physiological safety data before nursing care, calculate the patient's treatment tolerance through a preset risk assessment algorithm, output the judgment result, and generate initial control instructions based on the judgment result. The ultrasound therapy parameter matching module is used to generate initial treatment parameters through edge computing algorithms, and dynamically adjust the ultrasound therapy parameters in combination with the initial treatment parameters and control commands. The ultrasound therapy execution module is used to perform ultrasound therapy according to the ultrasound therapy parameters output by the ultrasound therapy parameter matching module; The real-time feedback control module is used to analyze dynamic data during the treatment process in real time and generate parameter optimization control instructions and safety emergency control instructions. The data storage module is used to store treatment-related data and generate individualized lung care reports based on the stored data, including details of treatment parameters, efficacy assessment results, risk warning records, and follow-up care recommendations.

2. The lung care system for airway clearance based on ultrasound technology according to claim 1, characterized in that, The implementation process of the data acquisition module specifically includes: Collect airway function-related data, including measuring diaphragmatic movement using an ultrasound sensor and measuring peak cough flow rate using a flow sensor; Collect respiratory mechanics-related data, including collecting pressure changes during the patient's inspiration / expiration process using pressure sensors, and calculating the maximum inspiratory pressure and maximum expiratory pressure based on these pressure changes; Collect physiological safety-related data and lung ultrasound imaging data, and identify information on lung parenchymal lesion areas; Collect sputum output, respiratory flow rate, and changes in airway pressure. Calculate changes in airway resistance using changes in respiratory flow rate and airway pressure.

3. The pulmonary care system for airway clearance based on ultrasound technology according to claim 1, characterized in that, The implementation process of the risk assessment and control module specifically includes: After receiving basic physiological safety data before nursing care and determining the patient's treatment tolerance based on a preset risk assessment algorithm, the determination results are converted into standardized control signals. Based on the signal type and risk level, control signals are precisely transmitted to the corresponding functional modules; The system receives real-time physiological safety data during treatment and reassesses the patient's risk level using a preset risk assessment algorithm. If the risk level increases, the control signals are immediately adjusted and redistributed.

4. The pulmonary care system for airway clearance based on ultrasound technology according to claim 1, characterized in that, The implementation process of the ultrasound therapy parameter matching module specifically includes: An initial matching model for ultrasound treatment parameters is constructed using an edge computing algorithm. Simultaneously, multi-source basic data before nursing care is received and preprocessed to obtain multi-source data feature vectors. The preprocessed multi-source data feature vectors are input into the initial matching model of ultrasound treatment parameters to generate the initial adaptation results of treatment parameters. Receive initial control instructions and generate treatment parameter instructions based on the initial adaptation results of treatment parameters; The system receives control commands from the real-time feedback control module to dynamically optimize treatment parameters, and sends the initial treatment parameter commands and the optimized treatment parameter commands to the ultrasound treatment execution module and the data storage module.

5. The pulmonary care system for airway clearance based on ultrasound technology according to claim 4, characterized in that, The generation of initial adaptation results for ultrasound therapy parameters specifically includes: When the cough peak flow rate is not as expected, the initial matching is a low-frequency ultrasound enhancement mode, the ultrasound frequency is set to the low-frequency range, and the power is increased by a preset ratio compared to the conventional mode. When the diaphragm movement is detected to be within the normal range, the initial matching is a conventional ultrasound mode, with the ultrasound frequency set to the conventional range and the power set to the conventional reference power range. When the maximum inspiratory pressure is detected to be below the normal range, the initial matching is a low-power long-duration mode, the ultrasonic power is set to a low-power ratio range of the conventional reference power, and the duration of action is extended by a preset ratio compared to the conventional mode. When lung ultrasound images detect areas of lung parenchymal lesions, an initial matching targeted focusing mode is used, setting the ultrasound energy to focus on the lesion area, and the power is dynamically adjusted to an appropriate power range based on the depth of the lesion area.

6. The pulmonary care system for airway clearance based on ultrasound technology according to claim 1, characterized in that, The implementation process of the real-time feedback control module specifically includes: It receives dynamic data of the treatment process in real time, performs rapid analysis and feature extraction on the received dynamic data, and identifies abnormal data and failure to meet efficacy standards by combining preset normal data range and efficacy evaluation threshold. Based on the data analysis results, corresponding control instructions are generated according to categories. If a situation where the therapeutic effect does not meet the standard is identified, parameter optimization control instructions are generated and sent to the ultrasound treatment parameter matching module for corresponding adjustments. If a physiological safety abnormality is detected, a safety emergency control command is generated and distributed to the ultrasound treatment execution module for corresponding adjustments.

7. The lung care system for airway clearance based on ultrasound technology according to claim 6, characterized in that, If a situation where the therapeutic effect does not meet the target is identified, a parameter optimization and control command is generated and sent to the ultrasound treatment parameter matching module for corresponding adjustments, specifically including: When the decrease in airway resistance is not as expected, optimize the ultrasonic frequency band to the range of optimal cavitation effect. When the amount of sputum expectorated reaches the treatment target, the treatment termination procedure is automatically triggered, the treatment data is recorded, and a treatment report is generated. When the real-time cough peak flow rate is detected to increase to a safe range and the duration reaches the set duration, the system automatically switches from low-frequency ultrasound enhancement mode to regular ultrasound mode. When the real-time maximum inspiratory pressure is detected to rise to the preset safe range, the ultrasound treatment duration is automatically adjusted back to the normal mode.

8. The pulmonary care system for airway clearance based on ultrasound technology according to claim 6, characterized in that, If a physiological safety abnormality is detected, a safety emergency control command is generated and distributed to the ultrasound treatment execution module for corresponding adjustments, specifically including: When arrhythmia is detected, the ultrasound power will be reduced to a safe value within a set time. If the duration of arrhythmia reaches the preset duration, the treatment will be automatically terminated. When blood oxygen saturation is detected to be below the safe value, treatment will be terminated within a set time and an alarm will be triggered. When the real-time heart rate or systolic blood pressure is detected to be outside the safe range, the ultrasound power is reduced by a preset ratio within a set time. If the physiological data does not return to normal within the set time, the treatment is terminated.

9. The data processing method according to claim 1, characterized in that, The step of assigning data processing tasks to target nodes specifically includes: The system uses a resource monitoring daemon embedded in the system to check each candidate node in real time with a preset time granularity. It calculates and normalizes the proportion of time slices in which the CPU is in a non-idle state when processing airway function data in the most recent time window to obtain the current CPU load rate of each candidate node. At the same time, it obtains the network response delay sequence generated by probe data packets for each candidate node within a preset historical period. Calculate the statistical variance of the network response delay sequence or construct a probability density histogram of the delay distribution, and calculate the network entropy value using the Shannon entropy model. The network entropy value is used to characterize the uncertainty of the node network link when transmitting medical safety instructions. At the same time, query the system startup log or hardware running clock to obtain the continuous running time of each candidate node since the last restart. The admission priority index for each candidate node is calculated using the following dynamic resilience potential energy formula: In this formula: Defined as the admission priority index; Defined as the normalized current CPU load rate of the candidate node; Defined as the network entropy value, whose value is positively correlated with the variance of the network response delay sequence; Defined as the continuous runtime; Defined as a preset system cold start warm-up constant; Defined as a preset load sensitivity coefficient, it is used to adjust the system's weighting of idle computing resources; Defined as a preset jitter suppression factor, it is used to achieve nonlinear compressed sensing of network instability in the calculation; The natural constant; the exponential term in the formula As a cold start suppression factor, it is used when the candidate node has just started and the continuous running time is... Less than the system cold start warm-up constant At that time, the acceptance priority index is non-linearly reduced to prevent the node from triggering an avalanche effect due to task overload in the early stage of the ultrasound treatment procedure. All candidate nodes are sorted in descending order based on the calculated admission priority index, and the node with the highest admission priority index is selected as the target node.

10. The data processing method according to claim 1, characterized in that, The method further includes a deadlock avoidance step based on vector projection before performing data processing tasks, which specifically includes: Static resource requirement analysis is performed on the pending tasks at the head of the execution queue, and all resource IDs that must be locked during the execution process are extracted as the operation object identifier set. Based on the preset resource requirement feature mapping table constructed using resource feature mapping technology, the pending tasks are converted into high-dimensional feature vectors, where the direction of the vector indicates the resource requirement type, and the numerical value of the vector in a specific dimension represents the intensity or duration weight of the resource occupation. Scan all existing tasks that are running in the current target node and have not yet released resources, convert each existing task into a corresponding high-dimensional feature vector, and stack these vectors as row vectors to construct a system state matrix representing the subsystem where resources are currently occupied. Map the high-dimensional feature vector of the task to be processed to the linear space where the system state matrix is ​​located, and calculate the orthogonal projection components of the high-dimensional feature vector on each row vector in the system state matrix; Determine whether the magnitudes of all the calculated orthogonal projection components are less than a preset safety threshold; If the magnitude of all the orthogonal projection components is less than the safety threshold, it is determined that the task to be processed is in a quasi-orthogonal state with the current system state, there is no logical resource deadlock, and an execution permission instruction is generated. If the magnitude of any of the orthogonal projection components is greater than or equal to the safety threshold, a potential resource contention risk is identified, the pending task is suspended, and a task gradient backoff mechanism is triggered: the dynamic backoff time is calculated based on the magnitude of the orthogonal projection component exceeding the safety threshold; the greater the conflict magnitude, the longer the backoff time. The determination is retried after the dynamic backoff time ends.

Images