Digital twinborn prediction method for supporting protective ventilation of lung of patient by ECMO

Abstract

The invention relates to the technical field of medical prediction, in particular to a digital twinborn prediction method for patient lung protective ventilation supported by ECMO, which comprises the following steps: extracting flow velocity change and jump points under the support of ECMO, constructing a ventilation fluctuation trend map, identifying fluctuation sections by combining compliance and pressure value ratio, delimiting lung partitions and analyzing oxygenation ratio trend. And generating an oxygenation capability classification graph, sequencing ventilation path effects, and outputting a pulmonary protective ventilation digital twinborn prediction scheme. According to the method, dynamic response mapping of ventilation behaviors is established on the basis of timing sequence changes of breathing flow velocity jump characteristics and compliance pressure value ratios, the lung regional ventilation state is accurately recognized, lung function zoning division is achieved by combining oxygenation ratio waveform classification, ventilation path utility sorting is completed through oxygenation descending characteristics under path coverage, and the lung function evaluation accuracy is improved. And a path state prediction sequence is constructed by fusing the change relationship between the tensile stress and the ventilation pressure, so that the dynamic adaptability of the ventilation strategy to the lung mechanics and oxygenation state and the region matching precision are improved.

Description

Technical Field

[0001] This invention relates to the field of medical prediction technology, and in particular to a digital twin prediction method for ECMO-supported lung-protective ventilation in patients. Background Technology

[0002] Medical predictive technology refers to an interdisciplinary research area that uses the collection, processing, and modeling of multi-source health data, along with key digital technologies such as machine learning and artificial intelligence, to predict trends and assess risks in the health status of individuals or groups. Core aspects of this technology include dynamic modeling of disease development, decision support analysis for treatment pathways, optimization of clinical resource allocation, and development of individualized intervention strategies. It encompasses sub-fields such as chronic disease prediction, early warning of severe illness progression, postoperative recovery prediction, and medical behavior simulation. With the development of intelligent medical systems, medical prediction has been widely applied in areas such as assisted diagnosis, efficacy evaluation, and treatment plan optimization, and is gradually evolving into a comprehensive medical support technology system integrating medicine, data science, and engineering modeling. Traditional methods for predicting lung-protective ventilation in ECMO-supported patients involve using static data such as respiratory mechanics parameters, blood gas analysis indicators, and some vital signs to predict the evolution of lung function during extracorporeal membrane oxygenation support through clinical experience or single mathematical modeling. This helps in developing ventilation strategies. Commonly used methods include regression analysis based on historical data, manually set combinations of static ventilation parameters, and linear decision-making path construction derived from expert experience. These methods rely on clinical rules and low-dimensional feature fitting.

[0003] Existing technologies primarily rely on a single modeling approach, depending on static inputs with limited dimensions to construct predictive models. They lack the ability to continuously acquire and structurally express information about dynamic respiratory processes. In situations where patients experience frequent state fluctuations and complex coupling relationships among indicators during ventilation, it is difficult to effectively capture key nodes in the evolution of lung function. The delayed response of ventilation strategies can easily lead to increased local lung tissue burden or decreased gas exchange efficiency. Adjustments to ventilation parameters are mostly based on empirical paths and mean analysis, lacking a mechanism for identifying the heterogeneous characteristics of different lung regions. This makes it difficult to ensure the accuracy of the distribution of regional ventilation needs, and in complex pathological conditions, it can easily lead to one-sided ventilation strategies and an imbalance in resource allocation. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a digital twin prediction method for ECMO-supported lung-protective ventilation in patients.

[0005] To achieve the above objectives, the present invention employs the following technical solution: a digital twin prediction method for ECMO-supported lung-protective ventilation in patients, comprising the following steps:

[0006] S1: Under ECMO support, record inspiratory and expiratory flow rate data during continuous respiratory cycles; obtain instantaneous change values ​​by dividing the flow rate difference between adjacent time points by the time interval, and further calculate the difference to obtain jump time points, form a fluctuation trend curve according to the time series, and generate a ventilation change jump trend map.

[0007] S2: Based on the time segments in the ventilation change jump trend map, extract the lung compliance and airway pressure trajectory within the corresponding cycle; calculate the rate of change of the compliance to pressure ratio at each jump point, compare the difference between consecutive points, filter the fluctuating segments, and generate the synchronous ventilation-compliance fluctuation mapping segment.

[0008] S3: Within the synchronous ventilation-compliance fluctuation mapping segment, lung zones are delineated, arterial blood oxygen concentration changes and total ventilation are extracted, the ratio trend of blood oxygen increase to ventilation is calculated, the waveform categories of the trend curves are compared, and a classification chart of the oxygenation capacity trend of each zone is generated.

[0009] S4: Call the structural partitions in the partition oxygenation capacity trend classification diagram, filter the ventilation paths covering the partitions, extract the oxygenation fluctuation decline rate and duration, calculate the product of the two for intensity sorting, and form a ventilation path oxygenation effectiveness sorting list.

[0010] As a further aspect of the present invention, the ventilation change jump trend map includes a jump time node sequence, gas flow rate change amplitude identifier, and periodic ventilation fluctuation curve; the synchronous ventilation-compliance fluctuation mapping segment includes ventilation and compliance corresponding time windows, ratio change rate feature set, and fluctuation difference significant segment; the regional oxygenation capacity trend classification map includes structural partition labels, oxygenation ratio change trend, and waveform category classification; and the ventilation path oxygenation utility ranking list includes ventilation path number, oxygenation decline intensity index, and utility priority ranking.

[0011] As a further aspect of the present invention, the fluctuation trend curve is a dynamic trajectory graph reflecting the amplitude and direction of ventilation changes, obtained by connecting the gas flow rate change values ​​and jump point sequences at adjacent time points in a continuous respiratory cycle in chronological order.

[0012] The difference magnitude refers to the absolute value of the difference between the rate of change of compliance and pressure ratio between two consecutive time points, which measures the magnitude of the fluctuation between jump points.

[0013] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0014] S101: Acquire the inspiratory and expiratory gas flow rate records during continuous respiratory cycles under ECMO support. Based on the flow rate value and time interval between every two adjacent time points, calculate the difference and divide it by the time interval to obtain the instantaneous flow rate change rate corresponding to the time point, and obtain the instantaneous flow rate change value sequence.

[0015] S102: Call the instantaneous flow rate change value sequence, calculate the difference between adjacent points, and extract the time points where the absolute value exceeds the respiratory fluctuation threshold. Combine them in chronological order to form a set of jump time points and obtain the gas flow rate jump time point sequence.

[0016] S103: Based on the gas flow rate jump time sequence, extract the corresponding instantaneous change values ​​and arrange them by time, connect them to form a periodic fluctuation trend curve, and obtain the ventilation change jump trend map.

[0017] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0018] S201: Based on the jump time segments marked in the ventilation change jump trend map, extract the corresponding lung compliance records and airway pressure change trajectories within the segments, and synchronously extract them according to the jump time points to obtain the jump-corresponding compliance and pressure trajectory groups.

[0019] S202: Call the jump-corresponding compliance and pressure trajectory group, calculate the ratio of lung compliance value to airway pressure value at each jump time point, calculate the ratio change rate, compare the difference between the change rates and screen out the time period with the difference greater than the compliance change benchmark value, and obtain the ventilation compliance mutation interval sequence.

[0020] S203: Based on the positional relationship between the ventilation compliance mutation interval sequence and the jump time segment, establish the time period mapping between the two one by one, organize them into a unified structure, and obtain the synchronous ventilation-compliance fluctuation mapping segment.

[0021] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0022] S301: Based on the time window corresponding to the synchronous ventilation-compliance fluctuation mapping segment, delineate multiple partition spaces in the lung structure, extract the arterial blood oxygen concentration records and total ventilation gas values ​​of the partitions within the corresponding time period, establish a paired data set for each time period, and obtain the oxygen concentration and gas volume data set for each partition time period.

[0023] S302: Call the oxygen concentration and gas volume data group of the partition time period, divide the change in blood oxygen concentration in each time period by the total amount of ventilation gas to generate the corresponding ratio, and then sort the ratio in each partition in chronological order to extract the ratio change trend of continuous time points in the partition and obtain the partition oxygenation efficiency change trend sequence.

[0024] S303: Based on the curve shape of the ratio change trend in the sequence of oxygenation efficiency changes in the partition, the waveform category of the partition curve is determined, and the classification results are classified, labeled and integrated to obtain a partition oxygenation capacity trend classification map.

[0025] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0026] S401: Call the structural partition corresponding to the oxygenation fluctuation category in the partition oxygenation capacity trend classification diagram, filter the set of ventilation paths that can be completely covered under each partition, and record all spatial segments associated with each path to obtain the partition mapping set of covered ventilation paths.

[0027] S402: Based on each ventilation path partition in the covered ventilation path partition mapping set, extract the rate of decrease in blood oxygen concentration and the duration of the corresponding time period, perform numerical multiplication of the two indicators, calculate the oxygenation fluctuation intensity value, and obtain the ventilation path oxygenation intensity numerical sequence.

[0028] S403: Based on the numerical sequence of oxygenation intensity of the ventilation path, sort all paths by numerical value, and combine the sorting results with the path numbers to obtain a sorted list of oxygenation effectiveness of ventilation paths.

[0029] As a further aspect of the present invention, the method further includes:

[0030] S5: Based on the oxygenation efficiency ranking list of the ventilation pathway, extract the difference between the maximum and minimum values ​​of alveolar wall tensile stress within the cycle, combine the ventilation pressure and the fluctuation amplitude of tensile stress to judge the difference, record the alveolar stress matching degree and ventilation coupling trend, and generate a digital twin prediction scheme for lung-protective ventilation.

[0031] As a further aspect of the present invention, the digital twin prediction scheme for lung-protective ventilation includes a path state change sequence, alveolar stress matching score, and ventilation-stress coupling trend index.

[0032] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0033] S501: Based on the ventilation path ranked highest in the oxygenation efficiency ranking list, extract the maximum and minimum values ​​of alveolar wall tensile stress in the covered area during each ventilation cycle, and perform difference calculation on the maximum and minimum values ​​to obtain the stress variation range of the path area.

[0034] S502: Call the stress variation amplitude of the path area, extract the maximum and minimum values ​​of the periodic ventilation pressure in the corresponding path and calculate the pressure fluctuation amplitude, then perform difference judgment on the stress variation amplitude and the ventilation pressure amplitude to obtain the degree of deviation between the stress variation trend and the ventilation state under each path, and obtain the alveolar stress ventilation coupling deviation.

[0035] S503: Based on the alveolar stress ventilation coupling deviation, the ventilation paths are sorted by difference. The sorting results are combined with the path status to construct a structural template of the coupling state change law, and a digital twin prediction scheme for lung-protective ventilation is obtained.

[0036] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0037] In this invention, a dynamic response mapping of ventilation behavior is established based on the temporal variation of respiratory flow rate jump characteristics and compliance pressure ratio, accurately identifying regional ventilation states of the lungs. Lung function zoning is achieved by combining oxygenation ratio waveform classification, ventilation path utility ranking is completed by oxygenation decline characteristics under path coverage, and path state prediction sequence is constructed by integrating the relationship between tensile stress and ventilation pressure changes, thereby improving the dynamic adaptability of ventilation strategies to lung mechanics and oxygenation state and the accuracy of regional matching. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a schematic diagram of the steps of the present invention;

[0040] Figure 2 This is a flowchart of steps S1 of the present invention;

[0041] Figure 3 This is a flowchart of steps S2 of the present invention;

[0042] Figure 4 This is a flowchart of steps S3 of the present invention;

[0043] Figure 5 This is a flowchart of step S4 of the present invention;

[0044] Figure 6 This is a flowchart of step S5 of the present invention. Detailed Implementation

[0045] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0046] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0047] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0048] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0049] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0050] Please see Figure 1 A digital twin prediction method for ECMO-supported lung-protective ventilation in patients includes the following steps:

[0051] S1: Acquire the inspiratory and expiratory gas flow rate records during continuous respiratory cycles under ECMO support. Divide the flow rate difference between two adjacent time points by the interval to obtain the instantaneous flow rate change value. Then perform difference calculation on the change values ​​of continuous points and extract the jump time points. Combine the jump points in time order to form a periodic fluctuation trend curve and generate a ventilation change jump trend map.

[0052] S2: Based on the jump time segments marked in the ventilation change jump trend map, extract the lung compliance record and airway pressure change trajectory within the corresponding cycle. Calculate the ratio change rate of compliance to pressure value for each jump time point. Then compare the difference amplitude between consecutive points of the ratio change rate and screen out the time periods with prominent amplitude fluctuations to establish the mapping relationship between time intervals and ventilation fluctuations, generating the synchronous ventilation-compliance fluctuation mapping segment.

[0053] S3: Based on the time window corresponding to the synchronous ventilation-compliance fluctuation mapping segment, delineate the spatial division of the lung structure, extract the change value of arterial blood oxygen concentration in the division and the total amount of ventilation gas in the corresponding time, calculate the trend of the ratio change of blood oxygen increase and ventilation gas volume respectively, and compare the waveform category classification between the ratio trend curves to generate a classification map of the trend of oxygenation capacity of the division.

[0054] S4: Call the structural partitions corresponding to the oxygenation fluctuation categories in the partition oxygenation capacity trend classification chart, filter the set of ventilation paths that can cover the partitions, extract the rate of decrease and duration of oxygenation fluctuations in the partitions corresponding to the paths, and then perform a product operation on the rate of decrease and duration for intensity sorting, arrange the order of ventilation paths, and generate a ventilation path oxygenation effectiveness sorting list.

[0055] S5: Based on the area covered by the highest-ranked path in the ventilation pathway oxygenation efficiency ranking list, extract the maximum and minimum values ​​of alveolar wall tensile stress within the cycle and calculate the difference. Then, judge the difference between the ventilation pressure fluctuation amplitude and the tensile stress change amplitude, and record the alveolar stress matching degree and ventilation status coupling trend under the path. Arrange the differential path status changes in sequence to generate a digital twin prediction scheme for lung-protective ventilation.

[0056] The ventilation change jump trend map includes the jump time node sequence, gas flow rate change amplitude indicator, and periodic ventilation fluctuation curve. The synchronous ventilation-compliance fluctuation mapping segment includes the ventilation and compliance corresponding time window, ratio change rate feature set, and fluctuation difference significant segment. The regional oxygenation capacity trend classification map includes structural region label, oxygenation ratio change trend, and waveform category classification. The ventilation pathway oxygenation utility ranking list includes ventilation pathway number, oxygenation decline intensity index, and utility priority ranking. The digital twin prediction scheme for lung-protective ventilation includes pathway status change sequence, alveolar stress matching degree score, and ventilation-stress coupling trend index.

[0057] Please see Figure 2 The specific steps of S1 are as follows:

[0058] S101: Acquire the inspiratory and expiratory gas flow rate records during continuous respiratory cycles under ECMO support. Based on the flow rate value and time interval between every two adjacent time points, calculate the difference and divide it by the time interval to obtain the instantaneous flow rate change rate corresponding to the time point, and obtain the instantaneous flow rate change value sequence.

[0059] To acquire inspiratory and expiratory gas flow rate records during continuous respiratory cycles under ECMO support, a high-frequency sampling flow rate sensor is installed proximal to the patient's airway. This sensor collects data every 10 milliseconds, arranging 1000 sampling points within 10 seconds in chronological order to form a raw flow rate value sequence. For example, the sampling values ​​in the first second are 1.0, 1.1, 1.3, 1.2, 1.0, 0.9, 0.8, etc. By calculating the difference between flow rate values ​​at any two adjacent moments, differences of 0.1, 0.2, -0.1, -0.2, -0.1, and -0.1 are obtained. Then, based on... The rate of change is determined at a time interval of 10 milliseconds. By dividing each difference by 10 milliseconds, the change value per second is converted to obtain a continuous instantaneous flow rate change rate. This operation is performed one by one to form a complete change rate sequence. In a real-world scenario, if the flow rate changes rapidly from 1.5 liters per minute to 2.0 liters per minute, a difference of 0.5 is formed between the two sampling points. The calculated instantaneous change rate is 50 liters per minute per second. If the recording time is continuous for 30 seconds, a total of 3,000 flow rate values ​​and 2,999 flow rate change rate values ​​are formed, creating a complete flow rate change value sequence, which provides a data foundation for subsequent abrupt change point identification.

[0060] S102: Call the instantaneous flow rate change value sequence, calculate the difference between adjacent points, extract the time points where the absolute value exceeds the respiratory fluctuation threshold, combine them in chronological order to form a set of jump time points, and obtain the gas flow rate jump time point sequence;

[0061] The instantaneous flow velocity change value sequence obtained in the previous step is called. Each pair of adjacent values ​​is iterated through, and the difference is calculated and its absolute value is taken. For example, if the current change value sequence is 15, 20, 18, 45, 100, 110, 30, 28, the calculated differences are 5, 2, 27, 55, 10, 80, 2. If the jump threshold is set to 25, then the time points with differences greater than or equal to 25 are extracted. That is, the times corresponding to points 3, 4, and 6 are 0.03 seconds, 0.04 seconds, 0.06 seconds, etc., respectively, forming a set of jump time points. Regarding setting the jump threshold... By statistically analyzing the distribution of the rate of change difference within multiple normal respiratory cycles, the mean and standard deviation are calculated. Based on this, a threshold higher than the mean plus twice the standard deviation is set. For example, if the mean difference under normal conditions is 12 and the standard deviation is 6, then the threshold is set to 12 + 2 × 6 = 24, which is rounded up to 25. This serves as a reference limit for screening jump points. This setting avoids misidentifying small fluctuations in normal breathing as jumps. All time points that meet this threshold condition are determined as jump points. After sorting, they are arranged in chronological order to form the final jump time point sequence.

[0062] S103: Based on the sequence of gas velocity jump times, extract the corresponding instantaneous change values ​​and arrange them by time, connect them to form a periodic fluctuation trend curve, and obtain the ventilation change jump trend map.

[0063] Based on the obtained sequence of jump points, the instantaneous change value corresponding to each time point is extracted and recorded. For example, the change rate at 0.03 seconds corresponds to 27, at 0.04 seconds to 45, and at 0.06 seconds to 80. These points are arranged in chronological order to form a jump trend change sequence. During execution, the list of jump points needs to be traversed, and the data items corresponding to the time points in the original change rate sequence need to be extracted and organized into a set of paired data with time and change rate values, such as {(0.03, 27), (0.04, 45)}. (0.06, 80)}, and then connect these points to form a trend curve. This trend graph can show the relative density and order of abrupt changes in a respiratory cycle. If the jump points are concentrated in the early stage of expiration and the end of inspiration, the trend graph will show a fluctuation characteristic of high at the beginning, high at the end and flat in the middle. Abnormal breathing behavior can be identified by comparing and analyzing different cycle graphs. At the same time, this graph can also be compared with the original flow rate curve to help observe whether the jump trend appears at the junction of inspiration and expiration, thus forming a gas flow rate jump trend graph.

[0064] Please see Figure 3 The specific steps of S2 are as follows:

[0065] S201: Based on the jump time segments marked in the ventilation change jump trend map, extract the corresponding lung compliance records and airway pressure change trajectories within the segments, and synchronously extract them according to the jump time points to obtain the jump corresponding compliance and pressure trajectory groups.

[0066] Based on the jump time segments marked in the ventilation change jump trend chart, the start and end times of each jump time segment marked in the trend chart are first obtained. For example, if the chart shows jump segments as 5 seconds to 6.2 seconds, 9 seconds to 9.7 seconds, 12 seconds to 13.5 seconds, etc., the start and end sampling numbers of these time segments in the original record need to be accurately located during data extraction. Using a sampling interval of 10 milliseconds, the sampling numbers for 5 seconds to 6.2 seconds correspond to 500 to 620, totaling 120 data points. Then, continuous compliance values ​​within this time segment are extracted from the lung compliance recording curve, for example, 38.4, 38.1, 37.8, 3... 7.4, 36.9, 36.1, etc., and simultaneously extract the data sequence within the corresponding sampling number segment from the airway pressure change trajectory, such as 14.8, 15.0, 15.2, 15.5, 15.7, 15.9, etc., to ensure that each set of compliance values ​​and airway pressure values ​​strictly correspond at the time point, forming a synchronous data pair. For each jump segment, the above synchronous extraction action is repeated to form multiple trajectory combinations of compliance and pressure values. Each combination can be represented as a continuous parameter sampling sequence under a certain jump segment, and the time segment number is marked so that ratio change and segment mapping processing can be performed in subsequent steps, and finally the corresponding compliance and pressure value trajectory group of the jump is obtained.

[0067] S202: Call the jump-corresponding compliance and pressure trajectory group, calculate the ratio of lung compliance value to airway pressure value at each jump time point, calculate the ratio change rate, compare the difference between the change rates and screen out the time period with the difference greater than the compliance change baseline value to obtain the ventilation compliance mutation interval sequence.

[0068] The specific formula for calculating the ratio change rate is as follows:

[0069]

[0070] Where, ΔR i C represents the rate of change of the ratio between the jump time point i and the adjacent i+δ time point. i C represents the lung compliance value at time i. i+δ P represents the lung compliance value at the next adjacent time step. i P represents the airway pressure at time i. i+δ The value represents the airway pressure at the next moment; C represents the arithmetic mean of all lung compliance values ​​within the jump time period; P represents the arithmetic mean of all airway pressure values ​​within the jump time period; α is the compliance change amplification adjustment coefficient, used to weight the contribution of the absolute compliance change term; β is the airway pressure change inhibition coefficient, used to correct the amplification effect of the airway pressure change comparison value difference; ε is a positive real constant to prevent the denominator from being zero; and δ is the sampling interval step size of the ratio change at the jump time point.

[0071] Parameter definition and acquisition process description:

[0072] C i Lung compliance value at time i, in mL / cmH2O, was obtained by sampling every 0.1 seconds using a pulmonary function instrument. In this example, it was recorded as 43.5 mL / cmH2O at the 3rd second.

[0073] C i+δ : The compliance value at time i+δ, where δ=1 represents adjacent time points, the sampling interval is 0.1 seconds, and the corresponding value is 41.8mL / cmH2O;

[0074] P i : The airway pressure at time i, in cmH2O, is obtained by a pressure sensor and is 15.1 cmH2O.

[0075] P i+δ The airway pressure at time i+δ corresponds to 14.7 cmH2O.

[0076] C: The average compliance value within the transition period. Let this period be from the 3rd second to the 4th second, totaling 10 points, with compliance values ​​of {43.5, 43.2, 42.8, 42.6, 42.3, 42.0, 41.8, 41.6, 41.4, 41.2}. Calculate the average:

[0077]

[0078] P: Mean airway pressure within the same segment, with values ​​{15.1, 15.0, 14.9, 14.8, 14.7, 14.6, 14.7, 14.6, 14.5, 14.4}, average value:

[0079]

[0080] α: Adjustment coefficient for the increase in compliance change. The reference sets the range to 0.3-0.8. In this example, it is set to 0.6 based on the fluctuation range of the previous parameters. Since the compliance change value is about 1.7, it belongs to the medium amplitude change.

[0081] β: Airway pressure change suppression coefficient, set to 0.2, with low weighting to handle pressure difference effects with fluctuations less than 0.6;

[0082] ε: The constant is set to 0.001 to prevent the denominator from being zero;

[0083] δ: 1 indicates adjacent sampling points.

[0084] Input the data and perform the calculation:

[0085] molecular:

[0086]

[0087] Denominator:

[0088]

[0089] Final calculation results:

[0090]

[0091] Table 1. Sampling parameters for lung compliance and airway pressure during the transition period.

[0092] Sampling time i <![CDATA[C i (mL / cmH2O)]]> <![CDATA[P i (cmH2O)]]> <![CDATA[C i+δ ]]> <![CDATA[P i+δ ]]> 3 seconds 43.5 15.1 41.8 14.7

[0093] As shown in Table 1, the compliance and airway pressure data at the continuous jump point at the 3rd second are recorded. This data is used as input for the calculation items in this section of the formula.

[0094] Explanation of calculation results:

[0095] The results indicate that the rate of change of the ratio between the current time points is 0.722. According to the baseline amplitude of compliance change set in this embodiment, which is 0.5 (this baseline value is calculated from the normal respiratory fluctuation, the maximum ΔR in the normal segment is 0.38, the standard deviation is 0.06, so the baseline is 0.38 + 2 × 0.06 = 0.50), this value has exceeded the baseline, so it is judged to be a mutation interval and marked as entering the ventilation compliance mutation sequence.

[0096] The advantage of the formula is that by introducing two weighted adjustments, α×ΔC and β×ΔP, it can effectively control the interference of airway pressure changes when capturing the compliance-dominated trend of change. Furthermore, by integrating with the normalized baseline value, it improves the accuracy of mutation identification, thereby achieving fine-grained and sensitive detection in lung stress modeling.

[0097] S203: Based on the positional relationship between the ventilation compliance mutation interval sequence and the jump time segment, establish the time period mapping between the two one by one, organize them into a unified structure, and obtain the synchronous ventilation-compliance fluctuation mapping segment;

[0098] Based on the positional relationship between the ventilation compliance mutation interval sequence and the jump time segment, the coverage and overlap between the two are determined segment by segment. For each mutation segment, such as a time range of 5.2 seconds to 5.3 seconds, the start and end times of the corresponding jump segment are matched. If the jump segment time is 5.0 seconds to 6.2 seconds, it is determined whether its start and end times are contained within or whether there is an overlap. For example, if the mutation segment is completely inside the jump segment, the correspondence between the mutation segment and the jump segment is directly established. If the mutation segment is 6.1 seconds to 6.4 seconds, and only a part of it overlaps with the jump segment of 5.0 seconds to 6.2 seconds, it is determined to be a partial cross mapping. When mapping, the start and end times of the cross and the coverage ratio are marked. The same matching process is then performed on all mutation and jump segments to form a structured mapping table. Each record clearly indicates the mutation segment number, start and end time, corresponding jump segment number, jump segment time, and crossover type (complete mapping or partial crossover). For example, the first record is mutation segment T1 (5.2s-5.3s) → jump segment J1 (5.0s-6.2s), type: complete mapping; the second record is T2 (6.1s-6.4s) → J1 (5.0s-6.2s), type: partial crossover. This table is used for subsequent correlation analysis and stage division to finally obtain the synchronous ventilation-compliance fluctuation mapping segment.

[0099] Please see Figure 4 The specific steps of S3 are as follows:

[0100] S301: Based on the time window corresponding to the synchronous ventilation-compliance fluctuation mapping segment, delineate multiple spatial zones in the lung structure, extract the arterial blood oxygen concentration records and total ventilation gas values ​​of the zones within the corresponding time period, establish a paired data set for each time period, and obtain the oxygen concentration and gas volume data sets for each zone time period.

[0101] Based on the time windows corresponding to the synchronized ventilation-compliance fluctuation mapping segments, the start and end times of each mapping segment are first determined. For example, segment 1 is from the 4th to the 6th second, and segment 2 is from the 10th to the 12th second. Then, according to the pulmonary anatomy, the entire lung region is divided into several spatial zones, such as the left upper lobe, left lower lobe, right upper lobe, right middle lobe, and right lower lobe. For each zone, continuous arterial oxygen concentration records are extracted within the aforementioned time periods. For example, in segment 1, the oxygen concentration data recorded every 0.5 seconds for the left upper lobe are 92%, 93%, 92%, 91%, 90%, and 89%. The total ventilation volume was recorded as 320mL, 310mL, 305mL, 300mL, 295mL, and 290mL, respectively. The blood oxygen concentration and the total ventilation volume were mapped one-to-one to construct a data set for each time period. All data were numbered with time identifier and zone label. Each data pair was the blood oxygen concentration value and the total ventilation volume value of the zone at a certain time point. The operation was repeated for all lung zones and all synchronous segments to ensure that the data of each lung zone in each time period was complete and structurally consistent. Finally, a complete data matrix was organized according to the two dimensions of time and space to form a data set of oxygen concentration and ventilation volume for each time period.

[0102] S302: Call the oxygen concentration and gas volume data group of the partition time period, divide the change in blood oxygen concentration in each time period by the total amount of ventilation gas to generate the corresponding ratio, and then sort the ratio in each partition in time order to extract the ratio change trend of continuous time points in the partition and obtain the partition oxygenation efficiency change trend sequence.

[0103] The system retrieves data sets of oxygen concentration and gas volume for different time periods. For each time point, a ratio calculation is performed. Specifically, the change in blood oxygen concentration at each time point is divided by the corresponding total ventilation volume to obtain the unit ventilation oxygenation efficiency at that time point. For example, if the blood oxygen concentration in the left upper lobe decreases from 93% to 91% between 4.5 and 5 seconds, and the total ventilation volume is 305 mL, the corresponding ratio is -2 / 305, resulting in -0.0066. The oxygenation efficiency values ​​for all time points are then sequentially arranged into a continuous sequence, such as -0.0051, -0.0066, -0.0043, -0.0078, and -0.0057, representing the trend of blood oxygen change per unit ventilation volume. This ratio is then used to calculate the oxygenation efficiency. The numerical values ​​between consecutive time points in the sequence are analyzed to identify trends, extracting states such as rising, falling, and stable. For example, a continuous decline indicates a continuous deterioration in oxygenation efficiency, while a continuous rise indicates a recovery in oxygenation efficiency. Separate trend sequences are established for different lung regions. For instance, the ratio sequence for the right middle lobe might be 0.0012, 0.0010, 0.0009, 0.0007, and 0.0006, forming a downward trend. The left lower lobe might be -0.0020, -0.0015, -0.0011, -0.0010, and -0.0010, forming a trend of first rising and then stabilizing. Finally, the ratio trend sequences of all regions are compiled into a unified dataset to obtain the regional oxygenation efficiency trend sequence.

[0104] S303: Based on the curve shape of the ratio change trend in the regional oxygenation efficiency change trend sequence, the waveform category of the regional curve is determined, and the classification results are classified, labeled and integrated to obtain a regional oxygenation capacity trend classification map.

[0105] Based on the curve shape of the ratio change trend in the oxygenation efficiency change trend sequence of each region, the waveform category of the regional curves is determined. First, based on the direction of the ratio trend change, four categories are categorized: monotonically increasing, monotonically decreasing, oscillating, and stable with no obvious trend. The trend sequence of each region is then traversed to determine whether each segment of change is continuous and consistent in trend. For example, if the ratio of a region continuously decreases from 0.0010 to -0.0030 within 5 seconds, it can be classified as monotonically decreasing. If another region increases for the first 2 seconds and then rapidly decreases, forming a peak, it is classified as oscillating. The judgment criterion is that three consecutive sets of data have the same direction as the single trend. The judgment criteria are adjusted, and the change amplitude exceeding the set judgment threshold is used as a reference for judging the effectiveness of the trend. The judgment threshold is set to 0.0005. Based on historical data, the normal oxygenation efficiency fluctuation range is generally no more than 0.0004. Only when it exceeds this value can it be judged as a significant trend. During the annotation process, the curve category of each region is marked. For example, the upper left lobe belongs to the declining type, the lower right lobe belongs to the stable type, and the lower left lobe belongs to the oscillating type. Then, the classification results of all regions are organized into a structured diagram. The horizontal axis is time, the vertical axis is lung region, and the color or icon indicates the curve classification type. Finally, a classification chart of oxygenation capacity trend of component regions is drawn.

[0106] Please see Figure 5 The specific steps of S4 are as follows:

[0107] S401: Call the structural partition corresponding to the oxygenation fluctuation category in the partition oxygenation capacity trend classification chart, filter the set of ventilation paths that can be fully covered under each partition, and record all spatial segments associated with each path to obtain the partition mapping set of covered ventilation paths.

[0108] First, all lung regions are searched according to the classification icon labels. The lung region numbers categorized into oxygenation trend types such as "monotonically decreasing," "oscillating," and "stable" are located. For example, the left lower lobe is classified as decreasing, and the right middle lobe as oscillating. Then, for each region, the anatomical ventilation pathways connected to its structure are called up one by one. Using lung spatial structure mapping data, it is determined which ventilation pathways can spatially traverse the region from beginning to end without interruption. These ventilation pathways are selected as a candidate pathway set. For example, a ventilation pathway might start in the trachea, pass through the right main bronchus, cross the right middle and right lower lobes, and extend to the alveolar end. If all its spatial segments are distributed in the right middle lobe, it is determined to be a ventilation path that "can completely cover the right middle lobe" and the path is recorded. Then, all the spatial segments passed through by the path are traversed and numbered. For example, if path P3 contains spatial segment numbers S17, S18, S19, and S20, all of which belong to the right middle lobe region, then the path is marked as "path P3 corresponding to the right middle lobe". The above operation is repeated to filter and record the path for each partition, generating a mapping structure with the structure {partition number: path set, list of spatial segment numbers under each path}, and finally obtaining the partition mapping set of the covered ventilation path.

[0109] S402: Based on each ventilation path partition in the coverable ventilation path partition mapping set, extract the rate of decrease and duration of blood oxygen concentration in the corresponding time period, perform numerical multiplication of the two indicators, calculate the oxygenation fluctuation intensity value, and obtain the numerical sequence of oxygenation intensity of the ventilation path.

[0110] Formula for calculating the intensity of oxygenation fluctuations:

[0111]

[0112] Among them, S λ The oxygenation fluctuation intensity value of the ventilation path λ is expressed as a percentage per second (%·s), ΔΨ. λ Θ represents the total decrease in blood oxygen concentration along path λ, expressed as a percentage (%). λ The duration of the decrease in blood oxygen saturation is expressed in seconds (s). ζ is the local fluctuation amplification factor, a dimensionless parameter. The setpoint is adjusted from 0.1 to 1.0 based on the amplitude of blood oxygen fluctuations. Ω λ This represents the total number of blood oxygen concentration sampling points involved in the analysis along path λ. Ψ represents the blood oxygen concentration value at the μ-th sampling point in path λ, expressed as a percentage (%). λ This represents the arithmetic mean of all blood oxygen concentration sampling points along path λ, expressed as a percentage (%).

[0113] The formula consists of three parts:

[0114] Main difference ratio The average rate of change (% / s) of blood oxygen concentration along path λ within a specified time period is calculated. The difference part is calculated using absolute value to reflect the total decrease, and the denominator is the corresponding time span. The unit rate of change is constructed using standard division.

[0115] Fluctuation-weighted terms The average deviation within a sampling point is expressed by the square root, based on the mean Ψ. λ The differences are weighted and combined with the adjustment coefficient ζ to adjust the sensitivity to oxygenation fluctuations.

[0116] Product term Θ λ : Used to restore the overall intensity dimension (%·s), so that the final result is the degree of oxygenation fluctuation accumulated per unit time.

[0117] Parameter definition and assignment instructions:

[0118] ΔΨ λ The variation in blood oxygen concentration within path λ is the difference between the maximum and minimum values. This is extracted from the percentage difference between the maximum and minimum values ​​in a continuous sampling sequence. For example, if the oxygen concentration decreases from 97.8% to 91.4%, then:

[0119] ΔΨ λ =97.8 - 91.4 = 6.4;

[0120] Θ λ The time span of the blood oxygen saturation decrease is calculated from the sampling interval and the difference between sampling points. If the decrease takes 28 sampling points at a sampling frequency of 10Hz, then:

[0121]

[0122] ζ: Fluctuation amplification factor, set based on the standard deviation of the oxygenation curve and the empirical range of pathological fluctuations, with a value range of 0.2 to 0.8. If the fluctuation intensity variation range is moderate, it is set to:

[0123] ζ = 0.5;

[0124] Ω λ The number of sampling points used in the statistics is 28, which is used for fluctuation analysis of all sampling points within the falling segment.

[0125] The blood oxygen concentration value at the μth sampling point along path λ, expressed as a percentage, is shown below (unit: %):

[0126] [97.8, 97.6, 97.3, 96.8, 96.0, 95.2, 94.4, 93.5, 92.6, 91.4];

[0127] The remaining 18 points are omitted, for a total of 28 sampling points.

[0128] Ψ λ The average blood oxygen value of the sampling segment is calculated as follows:

[0129] (Example of calculation results);

[0130] Actual example:

[0131] First calculate the first term:

[0132]

[0133] Next, calculate the square root fluctuation term:

[0134] Suppose that the following 7 sampling points deviate from the mean (in %):

[0135] [3.65, 3.45, 3.15, 2.65, 1.85, 1.05, 0.25];

[0136] After completing the calculation to 28 points, the mean absolute deviation is:

[0137]

[0138] Result after square root:

[0139]

[0140] Multiply by ζ = 0.5:

[0141] 0.5 × 1.319 = 0.6595;

[0142] The values ​​within the parentheses are obtained by merging:

[0143] 2.2857 + 0.6595 = 2.9452;

[0144] Finally multiplied by Θ λ =2.8:

[0145] S λ =2.9452 × 2.8 = 8.2466;

[0146] Table 2. Parameters for blood oxygen fluctuation analysis of ventilation pathway λ

[0147]

[0148] Table 2 lists the key parameters and intermediate calculation results of oxygenation fluctuations corresponding to path λ. All values ​​are from simulation sampling data and derivation process.

[0149] Explanation of calculation results:

[0150] The results indicate that the oxygenation fluctuation intensity of ventilation path λ within 2.8 seconds is 8.25%·s, which is higher than the reference value of 6.0%·s in this implementation standard (this benchmark is defined as the maximum fluctuation of the curve in normal ventilation patients being 4.5, with a standard deviation of 0.75, and the mutation threshold being: 4.5 + 2 × 0.75 = 6.0). Therefore, this path is identified as having a moderate to severe oxygenation mutation region. This value will directly serve as an important basis for subsequent coupling bias analysis and priority adjustment of path ranking.

[0151] The innovative aspects of the formula:

[0152] The advantage of the formula is that by combining the instantaneous descent rate and the mean square deviation of local fluctuations with different structures, and introducing an adjustment coefficient ζ to correct the intensity judgment for different oxygenation waveform types, it has structural adaptability when identifying nonlinear variations, thereby completing multidimensional modeling and differential ranking of regional imbalances in a multipath ventilation system.

[0153] S403: Based on the numerical sequence of oxygenation intensity of ventilation pathways, sort all pathways by numerical value, and combine the sorting results with the pathway numbers to obtain a sorted list of oxygenation effectiveness of ventilation pathways.

[0154] Based on the numerical sequence of oxygenation intensity of the ventilation pathways, the oxygenation fluctuation intensity values ​​corresponding to each pathway are sorted. First, all pathways and their corresponding intensity values ​​are organized into a list of binary pairs, such as {P1: 3.2, P2: 6.4, P3: 7.0, P4: 2.5, P5: 5.0}. Then, the intensity values ​​are compared from largest to smallest and the order is rearranged to determine the size relationship of each pathway's intensity value in the entire list. For example, P3 is the largest, followed by P2, then P5, P1, and P4, forming the sorting order P3, P2, P5, P1, P4. At the same time, the sorting position number is assigned to each pathway, i.e., P3 ranks first, P2 ranks second, and so on. Each pathway records its pathway number, intensity value, and sorting position. Finally, the results are combined into a sorted list of ventilation pathway oxygenation effectiveness. The list structure is a multi-column dataset, with columns containing pathway number, intensity value, and sorting position. The sorting result is used as the row order to form the final output list.

[0155] Please see Figure 6 The specific steps of S5 are as follows:

[0156] S501: Based on the ventilation path ranked highest in the oxygenation efficiency list, extract the maximum and minimum values ​​of alveolar wall tensile stress in the covered area during each ventilation cycle, and perform difference calculation on the maximum and minimum values ​​to obtain the range of tensile stress change in the path area.

[0157] Based on the ventilation pathway ranked highest in the oxygenation efficiency ranking list, the pathway number ranked first is read first, for example, pathway number P3. This pathway covers the right middle lobe and right lower lobe. Corresponding to all alveolar segments traversed by this pathway, the maximum and minimum values ​​of tensile stress data for each segment are extracted sequentially within each ventilation cycle. The start and end times of each respiratory cycle are recorded; for example, the first cycle is from second 5 to second 6. Within this time, tensile stress sensor data is retrieved; for example, the maximum value of S17 in the right middle lobe segment is 28.6 cmH2O, and the minimum value is 6.3 cmH2O. The maximum value of the lower right segment S18 is 30.2 cmH2O, and the minimum value is 7.1 cmH2O. After taking the maximum and minimum values ​​for each spatial segment, the stress variation amplitude of each segment is calculated, i.e., the maximum value minus the minimum value. For example, the amplitude of S17 is 22.3, and the amplitude of S18 is 23.1. The average or maximum value of the stress amplitudes calculated for all spatial segments in the entire path is used as the overall stress variation amplitude of the path. In this example, the overall amplitude of path P3 can be taken as the average value of 22.7 or the maximum value of 23.1. This value is recorded as a reference benchmark for stress variation, and finally the stress variation amplitude of the path area is obtained.

[0158] S502: Call the stress variation amplitude of the path area, extract the maximum and minimum values ​​of the periodic ventilation pressure within the corresponding path and calculate the pressure fluctuation amplitude, then perform a difference judgment on the stress variation amplitude and the ventilation pressure amplitude to obtain the degree of deviation between the stress variation trend and the ventilation state under each path, and obtain the alveolar stress ventilation coupling deviation.

[0159] The method retrieves the stress variation amplitude of the path region and extracts the maximum and minimum periodic ventilation pressure values ​​for the area covered by the path within the same time period. For example, if the maximum ventilation pressure of path P3 is 22.4 cmH2O and the minimum is 5.8 cmH2O between the 5th and 6th seconds, then the ventilation pressure amplitude is 22.4 minus 5.8, which equals 16.6 cmH2O. Comparing the stress variation amplitude of 22.7 and the ventilation pressure amplitude of 16.6, a difference judgment operation is performed, i.e., 22.7 minus 16.6 equals 6.1. This difference is used as the deviation value between the stress and ventilation pressure to determine whether there is a coupling deviation. The larger the deviation value, the more significant the change in stress within the same time period. Abnormal expansion or tension extension exists within the ventilation pressure range. If the difference is less than a certain reference limit, the change in tension stress and the fluctuation in ventilation pressure can be considered to be synchronous. The reference limit is set based on statistical data from multiple ventilation pathways. For example, if the difference is concentrated between 0 and 8 in 100 pathways, with a mean of 4.2 and a standard deviation of 1.6, then the significant deviation judgment threshold is set to the mean plus one standard deviation, i.e., 5.8. The deviation value of the current pathway P3 is 6.1, which exceeds the threshold value, and it is determined that it has a strong stress-ventilation non-coupling state. In this way, the difference between tension stress and ventilation pressure is calculated for other pathways respectively to form the alveolar stress-ventilation coupling deviation value for the corresponding pathway.

[0160] S503: Based on the alveolar stress-ventilation coupling deviation, the ventilation pathways are sorted by difference. The sorting results are combined with the pathway status to construct a structural template of the coupling status change law, and a digital twin prediction scheme for lung-protective ventilation is obtained.

[0161] Based on alveolar stress-ventilation coupling bias, all pathways are sorted by difference. Each pathway's number and its coupling bias value form a dataset. For example, pathways P1 have a bias of 3.9, P2 4.7, P3 6.1, P4 2.3, and P5 5.0. These pathways are sorted from largest to smallest bias value as P3, P5, P2, P1, and P4. The sorting position is recorded and bound to the pathway number, forming a sorted list. Each record contains the pathway number, coupling bias value, and sorting rank. The sorting results are then further combined with pathway status information. The path status information includes oxygenation efficiency category, spatial segment coverage area, and tensile stress fluctuation status. Through structured data integration, a unified structural template is formed. For example, the list structure is {path number: P3, deviation value: 6.1, ranking: 1, oxygenation category: decreasing type, tensile stress status: high amplitude type, ventilation pressure amplitude: medium level}. After repeating the operation on all paths, a structural template of coupled state change law is summarized. This template has the ability to annotate multi-path features and is ultimately used to represent the digital twin prediction scheme of lung-protective ventilation.

[0162] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A digital twin prediction method for lung-protective ventilation of ECMO supported patients, characterized in that, Comprise the following steps: S1: Under the support of ECMO, record the inspiration and expiration flow rate data in the continuous respiratory cycle, obtain the instantaneous change value by the flow rate difference between adjacent time points divided by the time interval, and further calculate the difference to obtain the jump time point, form the fluctuation trend curve in time sequence, and generate the ventilation change jump trend atlas; S2: According to the time section in the ventilation change jump trend atlas, extract the lung compliance and airway pressure trajectory in the corresponding cycle, calculate the compliance and pressure value ratio change rate of each jump point, and compare the difference amplitude between the continuous points, screen the fluctuation section, and generate the synchronous ventilation-compliance fluctuation mapping section; S3: In the synchronous ventilation-compliance fluctuation mapping section, the lung subarea is divided, the arterial blood oxygen concentration change and the total ventilation are extracted, the ratio trend of the oxygen lifting amount and the ventilation amount is calculated respectively, the waveform category of the trend curve is compared, and the subarea oxygenation capacity trend classification diagram is generated; S4: Call the structure subarea in the subarea oxygenation capacity trend classification diagram, screen the ventilation path covering the subarea, extract the oxygenation fluctuation decline rate and duration, calculate the product of the two for intensity sorting, and form the ventilation path oxygenation utility sorting list.

2. The digital twin prediction method for lung-protective ventilation of an ECMO- supported patient according to claim 1, characterized in that, The ventilation change jump trend atlas comprises a jump time node sequence, a gas flow rate change amplitude identifier and a periodic ventilation fluctuation curve, the synchronous ventilation-compliance fluctuation mapping section comprises a ventilation and compliance corresponding time window, a ratio change rate feature set and a fluctuation difference significant section, the subarea oxygenation capacity trend classification diagram comprises a structure subarea label, an oxygenation ratio change trend and a waveform category attribution type, and the ventilation path oxygenation utility sorting list comprises a ventilation path number, an oxygenation decline intensity indicator and an utility priority sorting.

3. The digital twin prediction method for lung-protective ventilation of an ECMO- supported patient according to claim 1, characterized in that, The fluctuation trend curve is a dynamic trajectory graph reflecting the ventilation change amplitude and direction obtained by connecting the gas flow rate change values and jump point sequence in the continuous respiratory cycle in time sequence; The difference amplitude refers to the absolute value of the difference between the compliance and pressure value ratio change rates of two continuous time points, which measures the fluctuation degree between the jump points.

4. The digital twin prediction method for lung-protective ventilation of an ECMO- supported patient according to claim 1, characterized in that, The specific steps of S1 are: S101: Obtain the inspiration and expiration gas flow rate record in the continuous respiratory cycle under the support of ECMO, calculate the difference value and divide by the time interval based on the flow rate value and time interval of each two adjacent time points, obtain the instantaneous flow rate change rate corresponding to the time point, and obtain the instantaneous flow rate change value sequence; S102: Call the instantaneous flow rate change value sequence, calculate the difference between adjacent points, and extract the time points whose absolute value exceeds the respiratory fluctuation threshold, combine to form a jump time point set in time sequence, and obtain the gas flow rate jump time point sequence; S103: According to the gas flow rate jump time point sequence, extract the corresponding instantaneous change value and arrange in time, connect to form a periodic fluctuation trend curve, and obtain the ventilation change jump trend atlas.

5. The digital twin prediction method for lung-protective ventilation of an ECMO- supported patient according to claim 1, characterized in that, The specific steps of S2 are: S201: Extract the corresponding lung compliance record and airway pressure change trajectory in the segment according to the marked jump time segment in the ventilation change jump trend map, and synchronize and intercept at the jump time point to obtain a jump corresponding compliance and pressure value trajectory group; S202: Calculate the ratio of the lung compliance value and the airway pressure value at each jump time point by calling the jump corresponding compliance and pressure value trajectory group, calculate the change rate of the ratio, compare the difference amplitude between the change rates, and select the time period with an amplitude greater than the compliance change reference amplitude to obtain a ventilation compliance mutation interval sequence; S203: According to the positional relationship between the ventilation compliance mutation interval sequence and the jump time segment, establish a time period mapping between them one by one, arrange them into a unified structure, and obtain a synchronous ventilation-compliance fluctuation mapping segment.

6. The digital twin prediction method for lung protective ventilation of an ECMO supported patient according to claim 1, characterized in that, The specific steps of S3 are: S301: According to the time window corresponding to the synchronous ventilation-compliance fluctuation mapping segment, divide the partition space in the lung structure, extract the arterial blood oxygen concentration record and ventilation gas total value in the corresponding time, establish a paired data set in each time period, and obtain a partition time period oxygen concentration and gas quantity data group; S302: Divide the ventilation gas total amount by the change amount of the blood oxygen concentration in each time period by calling the partition time period oxygen concentration and gas quantity data group to generate the corresponding ratio, and then sequentially arrange the ratio in each partition according to time sequence, extract the ratio change trend of the continuous time points in the partition, and obtain a partition oxygenation efficiency change trend sequence; S303: According to the curve shape of the ratio change trend in the partition oxygenation efficiency change trend sequence, perform waveform category attribution judgment on the partition curve, classify and integrate the attribution results, and obtain a partition oxygenation capacity trend classification graph.

7. The digital twin prediction method for lung-protective ventilation of an ECMO- supported patient according to claim 1, characterized in that, The specific steps of S4 are: S401: Call the structure partition corresponding to the oxygenation fluctuation category in the partition oxygenation capacity trend classification graph, select a complete coverable ventilation path set under each partition, and record all the space segments associated with each path to obtain a coverable ventilation path partition mapping set; S402: According to each ventilation path partition in the coverable ventilation path partition mapping set, extract the blood oxygen concentration decrease rate and duration in the corresponding time period, perform numerical multiplication operation on the two indexes, calculate the oxygenation fluctuation intensity value, and obtain a ventilation path oxygenation intensity numerical sequence; S403: According to the ventilation path oxygenation intensity numerical sequence, sort all the paths according to the numerical size, and then combine and arrange the sorting results and the path numbers to obtain a ventilation path oxygenation effect sorting list.

8. The digital twin prediction method for lung protective ventilation of an ECMO supported patient according to claim 1, characterized in that, The method further comprises: S5: According to the ventilation path oxygenation effect sorting list, extract the maximum and minimum value difference of the alveolar wall stress in the period, combine the ventilation pressure and the stress fluctuation amplitude to perform difference judgment, record the alveolar stress matching degree and the ventilation coupling trend, and generate a digital twin prediction scheme of lung protective ventilation.

9. The digital twin prediction method for lung-protective ventilation of an ECMO- supported patient according to claim 8, characterized in that, The digital twin prediction scheme of lung protective ventilation includes a path state change sequence, an alveolar stress matching degree score, and a ventilation-stress coupling trend index.

10. The digital twin prediction method for lung-protective ventilation of an ECMO- supported patient according to claim 8, characterized in that, The specific steps of S5 are: S501: Extract the maximum and minimum values of the alveolar wall stress in each ventilation cycle in the covered area according to the highest ranked ventilation path in the ventilation path oxygenation utility ranking list, and perform a difference calculation on the maximum and minimum values to obtain the path area stress change amplitude; S502: Call the path area stress change amplitude, extract the maximum and minimum values of the cycle ventilation pressure in the corresponding path and calculate the pressure amplitude, and then perform a difference judgment on the stress change amplitude and the ventilation pressure amplitude to obtain the deviation between the stress change trend and the ventilation state under each path, and obtain the alveolar stress ventilation coupling deviation; S503: According to the alveolar stress ventilation coupling deviation, the ventilation paths are sorted differently, the sorting results are combined with the path state, a structure template of the coupling state change rule is constructed, and a digital twin prediction scheme of lung protective ventilation is obtained.

Images