Methods, equipment, and media for predicting the weaning risk of mechanically ventilated patients based on EIT

Abstract

This invention belongs to the field of intelligent healthcare, specifically relating to a method, device, and medium for predicting the weaning risk of mechanically ventilated patients based on EIT (Extra-Intensive Lung Inclination) testing. The method includes: S101: acquiring pulmonary EIT monitoring data of mechanically ventilated patients over a period of time; S102: dividing the lungs into multiple regions and extracting EIT parameters based on the pulmonary EIT monitoring data, wherein the EIT parameters include a dynamic lung tilt index, which is the degree of variation in gas redistribution in each region during spontaneous breathing tests; S103: inputting the EIT parameters into a classifier for classification, and determining the patient's weaning outcome based on the classifier's output. The EIT parameters extracted in this application can effectively predict the weaning outcome of mechanically ventilated patients, helping clinicians choose a better weaning time.

Description

Technical Field

[0001] This invention relates to the field of intelligent healthcare, and more specifically, to a method, device, medium, and program product for predicting the weaning risk of mechanically ventilated patients based on EIT. Background Technology

[0002] Mechanical ventilation (MV) is a widely used life support intervention for critically ill patients. However, prolonged MV can lead to increased mortality, higher medical costs, and various complications. The transition from MV to spontaneous breathing (known as weaning) represents a critical but challenging phase in patient care. Despite decades of research, planned extubation failure rates remain high, ranging from 10% to 20%, and are associated with increased mortality and length of hospital stay.

[0003] Traditional weaning predictors, such as the shallow and rapid breathing index (RSBI) and maximum inspiratory pressure (MIP), have limited accuracy in predicting weaning outcomes. A systematic review of 66 studies found that commonly used parameters showed only moderate discrimination, with receiver operating characteristic (AUC) values ​​typically between 0.70 and 0.80. This limitation may be partly attributed to their focus on global respiratory parameters, failing to capture subtle variations in regional ventilatory heterogeneity and breathing patterns. Electrical impedance tomography (EIT) has emerged as a promising tool for non-invasive, real-time monitoring of regional lung ventilation. Recent studies have shown that EIT-derived parameters can detect early signs of respiratory muscle fatigue and ventilatory heterogeneity during spontaneous breathing trials (SBTs). However, the optimal integration of these novel measurements with traditional clinical parameters used for weaning prediction remains unclear.

[0004] Predicting weaning outcomes in mechanically ventilated patients remains challenging. Electrical impedance tomography (EIT) provides real-time information on local lung ventilation, but its value in weaning prediction has not been fully established. Summary of the Invention

[0005] In view of the above problems, the present invention provides a method for predicting the weaning risk of mechanically ventilated patients based on EIT, and develops and validates a comprehensive prediction model that combines EIT-derived parameters with clinical variables for predicting weaning outcomes.

[0006] This application (first aspect) discloses a method for predicting the weaning risk of mechanically ventilated patients based on EIT, the method comprising:

[0007] S101: Acquire lung EIT monitoring data of mechanically ventilated patients over a period of time;

[0008] S102: Divide the lungs into multiple regions and extract EIT parameters based on the lung EIT monitoring data. The EIT parameters include the dynamic lung tilt index, which is the degree of variation in gas redistribution in each region during the spontaneous breathing test.

[0009] S103: Input the EIT parameters into the classifier for classification, and determine the patient's off-boarding outcome based on the classifier's output.

[0010] Furthermore, the dynamic lung tilt index is the sum of the ratios of the standard deviation to the mean of gas redistribution in each region during the spontaneous breathing test;

[0011] Optionally, the dynamic lung tilt index is calculated as follows:

[0012]

[0013] Where N represents the total number of regions, and when the lungs are divided into 4 quadrants, N = 4; i ∈ [1, 4]; sd(Vtpendelluft i ) represents the standard deviation of gas redistribution in the i-th region, mean(Vtpendelluft) i ) represents the average amount of gas redistribution in the i-th region;

[0014] Optionally, the lungs can be divided into four quadrants.

[0015] Furthermore, the EIT parameter also includes: lung tilt ratio, which is the ratio of gas redistribution to tidal volume during the early inspiratory phase in each region;

[0016] Optionally, the lung tilt ratio is expressed as:

[0017]

[0018] Where PR represents the tilt ratio, Vtpendelluft i VT represents the gas redistribution amount in region i, and VT represents the total tidal volume of the lungs.

[0019] Furthermore, the EIT parameter also includes: a timing asynchrony index, which is expressed as:

[0020]

[0021] Where N represents the four regions divided, Δt i,j ΔV represents the time delay between regions. i,j It represents the difference in gas redistribution within a respiratory cycle between regions;

[0022] Optional, Δt i,j Represented as:

[0023]

[0024] in, represents the time-domain ventilation signal of the i-th region, and argmax represents the weighted average of the peak times of each respiratory cycle within the entire monitoring period;

[0025] Optional, ΔV i,j Represented as:

[0026] ΔV i,j =Vtpendelluft i -Vtpendelluft j

[0027] Vtpendelluft i Vtpendelluft j These represent the gas redistribution amounts in region i and region j, respectively.

[0028] Furthermore, the EIT parameters also include: regional ventilation distribution;

[0029] Optionally, the EIT parameter may also include: ventilation nonuniformity.

[0030] Furthermore, clinical parameters are acquired simultaneously. EIT parameters and traditional clinical parameters are input into a classifier for classification. The patient's off-boarding outcome is determined based on the classifier's output.

[0031] Optionally, the clinical parameters include the shallow and rapid breathing index;

[0032] Optionally, the clinical parameters include maximum inspiratory pressure;

[0033] Optionally, the clinical parameters include RSBI;

[0034] Optionally, the time period is 5 minutes;

[0035] Optionally, the offline result can be either offline failure or offline success.

[0036] Furthermore, the classifier includes one or more of the following: logistic regression, random forest, support vector machine, XGboost, decision tree, and extreme learning machine.

[0037] The second aspect of this application discloses a system for predicting the weaning risk of mechanically ventilated patients based on EIT, comprising:

[0038] Acquisition module 201: Used to acquire lung EIT monitoring data of mechanically ventilated patients over a period of time;

[0039] Feature extraction module 202: used to divide the lung into multiple regions and extract EIT parameters based on the lung EIT monitoring data. The EIT parameters include dynamic lung tilt index, which is the degree of variation of gas redistribution in each region during the spontaneous breathing test.

[0040] Prediction module 203: Used to input the EIT parameters into a classifier for classification, and to determine the patient's off-boarding outcome based on the classifier's output.

[0041] A third aspect of this application discloses a computer device, the device comprising: a memory and a processor; the memory being used to store program instructions; the processor being used to invoke the program instructions, which, when executed, are used to perform the steps of the method described above.

[0042] The fourth aspect of this application discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0043] The fifth aspect of this application discloses a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method.

[0044] This application has the following beneficial effects:

[0045] (1) The EIT parameters extracted in this application can also effectively predict the weaning outcomes of mechanically ventilated patients, helping clinicians to select a better time for weaning.

[0046] (2) The integration of EIT parameters with clinical variables significantly improved the prediction of weaning outcomes, and this approach may help optimize the timing of mechanical ventilation release. Attached Figure Description

[0047] 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.

[0048] Figure 1 This is a schematic diagram of the method flow provided in the first aspect of the present invention;

[0049] Figure 2 This is a schematic diagram of a program product provided in the second aspect of the present invention;

[0050] Figure 3 This is a schematic diagram of a computer device provided in an embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram of the architecture of an exemplary computing device provided in an embodiment of the present invention;

[0052] Figure 5 This is a schematic diagram of the storage medium provided in an embodiment of the present invention;

[0053] Figure 6 This is a flowchart illustrating a method for predicting the weaning risk of mechanically ventilated patients based on EIT, provided in an embodiment of the present invention.

[0054] Figure 7 This is a schematic diagram illustrating the effectiveness analysis of DPI prediction for offline risk provided by an embodiment of the present invention;

[0055] Figure 8 This is a schematic diagram illustrating one of the innovative advantages of this application provided by an embodiment of the present invention. Detailed Implementation

[0056] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0057] In some of the processes described in the specification, claims, and accompanying drawings of this invention, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as S101, S102, etc., are merely used to distinguish different operations and do not represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0058] 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.

[0059] Figure 1 This is a schematic flowchart of a method for predicting the weaning risk of mechanically ventilated patients based on EIT assessment, provided by an embodiment of the present invention. Specifically, the method includes the following steps:

[0060] S101: Acquire lung EIT monitoring data of mechanically ventilated patients over a period of time;

[0061] S102: Divide the lungs into multiple regions and extract EIT parameters based on the lung EIT monitoring data. The EIT parameters include the dynamic lung tilt index, which is the degree of variation in gas redistribution in each region during the spontaneous breathing test.

[0062] S103: Input the EIT parameters into the classifier for classification, and determine the patient's off-boarding outcome based on the classifier's output.

[0063] Methods: In this multicenter retrospective study of 426 mechanically ventilated patients, we developed and validated a predictive model that combined EIT-derived parameters with clinical variables. Patients were randomly assigned to a training group (n=298) and a test group (n=128). The primary outcome was failure to wean off mechanical ventilation within 48 hours of extubation.

[0064] Conclusion: Integrating EIT-derived local ventilation parameters with clinical variables significantly improved the prediction of weaning outcomes. This approach may help optimize the timing of mechanical ventilation release.

[0065] This application is based on the following fundamental research:

[0066] I. Research Methods

[0067] 1.1 Data Collection

[0068] This multicenter retrospective cohort study was conducted from January 2020 to December 2024 at three tertiary hospitals in China (Ruijin Hospital, Peking Union Medical College Hospital, and Fujian Provincial Hospital). The entire study protocol was approved by the hospital ethics committees. Due to the retrospective nature of the study, informed consent was waived. The study followed the Enhanced Observational Epidemiology Reporting (STROBE) guidelines. The study protocol was approved by the institutional review committees of all participating centers. Adult patients (≥18 years of age) who received mechanical ventilation for more than 24 hours and met the standard weaning criteria were eligible for inclusion. We excluded patients who were unplanned extubated, had tracheostomies, were unresponsive, or had insufficient data quality. Clinical variables were extracted from electronic medical records according to standardized procedures.

[0069] The primary outcome was failure to wean off the ventilator, defined as failure of the spontaneous breathing test (SBT) or reintubation within 48 hours of extubation. Secondary outcomes included hospital mortality, length of ICU stay, and ventilator-free days up to day 28.

[0070] All EIT measurements were performed using a PulmoVista 500 ( (Medical, Lübeck, Germany) The procedure was carried out according to established protocols. Validated algorithms were used to calculate the EIT parameters.

[0071] 1.2 Feature Extraction

[0072] The lungs were divided into four regions of interest (ROI), and the temporal process of regional impedance changes was analyzed to identify and quantify the amount of gas redistribution between regions.

[0073] (1) The redistribution amount is standardized to the total tidal volume to calculate the Pendelluft ratio (PR). The Pendelluft ratio is calculated as the percentage of regional gas redistribution (Vtpendelluft) to tidal volume (VT).

[0074] Vtpendelluft identified areas for recording early inspiration using EIT recordings by analyzing asynchronous filling and emptying patterns between different lungs.

[0075] (2) This application identified lung tilt from EIT data and used the lung tilt index to characterize this phenomenon:

[0076] ① Definition of pendelluft: Pendelluft refers to the asynchrony of airflow between different lung regions, reflecting abnormal respiratory muscle function and uneven regional ventilation distribution. This is an important physiological factor affecting the successful weaning of mechanically ventilated patients.

[0077] ② Calculation method of PR index:

[0078] Electrical impedance tomography (EIT) allows for regional monitoring and analysis of the lungs. Specifically, we divide the lungs into four regions of interest (ROIs) and calculate the gas redistribution (Vtpendelluft) in each region during the early inspiratory phase. Then, we define the tilt ratio:

[0079]

[0080] Among them, PR i Vtpendelluft represents the tilt ratio of the i-th region. i VT represents the amount of gas redistribution in the current area;

[0081] When PR > 3%, it indicates a significant lung tilt.

[0082] Since the lungs are divided into four regions (ROIs), the above calculation yields the PR values ​​for the four regions. If any PR > 3%, it indicates a significant lung tilt.

[0083] In some embodiments, PR is taken as the maximum tilt ratio among the four regions:

[0084]

[0085] Gas redistribution volume reflects the volume (in mL) of gas redistributed from the dependent region to the dependent region during positive pressure ventilation.

[0086] The calculation steps are as follows: by analyzing the impedance-time curve from the end of expiration to the end of inspiration, identify the specific time period of gas redistribution (usually the middle and late stages of inspiration), integrate the impedance change of the dependent area (such as the dorsal lung region) within this time period, and then multiply it by the calibration coefficient (calibrated by the known tidal volume).

[0087] Vtpendelluft=∑(ΔZ dependent (t)×Calibration Factor)

[0088] Where ΔZ represents the local impedance change.

[0089] Tidal volume (VT) refers to the total amount of gas (in mL) that enters or exits the lungs during a single respiratory cycle. EIT estimates tidal volume by integrating impedance changes across the entire lung region.

[0090] Spatial integration of impedance changes across all lung pixels, combined with calibrated data from a ventilator-synchronized flow sensor, establishes a linear relationship (calibration coefficient) between ΔZ and volume.

[0091] VT EIT =∑(ΔZ) global(t) ×Calibration Factor).

[0092] (3) Dynamic Pendellft Index (DPI)

[0093] This index is used to assess the weaning risk of mechanically ventilated patients. For the first time, it quantifies the gas redistribution characteristics between lung regions during a spontaneous breathing test (SBT) from a dynamic perspective.

[0094] Specifically, the method for obtaining and calculating DPI is as follows: First, the lungs are divided into four quadrants (ROIs 1-4) using electrical impedance tomography (EIT), and continuous monitoring is performed during the spontaneous breathing test. Every 5 minutes is a time window, and the variability of gas redistribution within each quadrant is calculated. The DPI is calculated by multiplying the sum of the ratios of the standard deviations to the mean values ​​of gas redistribution in each quadrant by 100%.

[0095]

[0096] Where N represents the total number of regions, and when the lungs are divided into 4 quadrants, N = 4; i ∈ [1, 4]; sd(Vtpendelluft i ) represents the standard deviation of gas redistribution in the i-th region, mean(Vtpendelluft) i ) represents the average amount of gas redistribution in the i-th region;

[0097] In some embodiments, the EIT data collection uses 32*32 pixels, each with a corresponding value. The standard deviation and mean of the gas redistribution amount are calculated based on the distribution of the pixels.

[0098] This calculation method takes into account the differences in gas redistribution between lung regions and the relative transformations within the tidal cycle.

[0099] (4) Temporal Asynchrony Index (TDI) new )

[0100]

[0101] Where N represents the four regions divided, Δt i,j ΔV represents the time delay between regions. i,j This indicates the difference in gas redistribution between regions within a respiratory cycle:

[0102] ΔV i,j =Vtpendelluft i -Vtpendelluft j

[0103] Regional time delay is expressed as:

[0104]

[0105] in, represents the time-domain ventilation signal of the i-th region, and argmax represents the weighted average of the peak times of each respiratory cycle within the entire monitoring period;

[0106] In some embodiments, the lungs are divided into four quadrants / regions, and the average phase difference between adjacent regions is calculated to obtain the ventilation cycle difference of the four regions:

[0107]

[0108] Calculated TDI new The predicted independent AUC was 0.77, significantly higher than the baseline TDI's 0.70 (p<0.01).

[0109] (5) Composite Oscillator Index (CSI) (integrating bidirectional flow + amplitude heterogeneity)

[0110] formula:

[0111]

[0112] Independent AUC: 0.81, significantly different during the SBT process (p<0.005).

[0113] (6) Enhance phase variation Introducing dynamic weight processing:

[0114]

[0115] in, Δφ(t) represents the phase difference over time, with an independent AUC of 0.79, which is 6% higher than the original phase variability. T represents the time of one ventilation cycle.

[0116] 1.3 Prediction Model Training

[0117] Sample size calculations were based on previous research, ensuring at least 10 events for each predictor variable. The dataset was randomly split into training (70%) and test (30%) sets, stratified by outcome status. Multiple imputation was used to handle missing data, following current recommendations. Feature selection was performed using LASSO regression with minimum absolute shrinkage and selection operator (LASSO) and 10-fold cross-validation. Several machine learning models, including logistic regression, random forest, gradient boosting machine, and deep neural networks, were developed and compared using scikit-learn (version 0.24.2) and TensorFlow (version 2.4.1). Model performance was evaluated using Area Under the Receiver Operating Characteristic (AUROC), calibration plots, decision curve analysis, net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Internal validation was performed using bootstrap resampling for 1000 iterations. Furthermore, sensitivity analyses were conducted to assess the robustness of our findings across different subgroups and clinical settings.

[0118] II. Results

[0119] 2.1 Patient characteristics

[0120] Of the 458 patients screened between January 2020 and December 2024, 278 met the inclusion criteria and were included in the final analysis. The most common reasons for exclusion were incomplete EIT records (n=82), tracheotomy (n=48), and poor signal quality (n=31). The median age was 64 years (IQR, 55–73), and 168 patients (60.4%) were male. The primary admission diagnoses included pneumonia (35.8%), postoperative respiratory failure (28.4%), and acute respiratory distress syndrome (ARDS) (22.6%). Baseline characteristics were comparable across the three participating centers.

[0121] Based on predefined criteria, the study population was divided into a successful weaning group (n = 186, 66.9%) and a failure group (n = 92, 33.1%). Patients in the failure group were older (median 68 years vs. 61 years, P = 0.003), had higher APACHE II scores (median 24 vs. 19, P = 0.003), and had longer durations of mechanical ventilation before SBT (median 8.5 days vs. 6.0 days, P = 0.002).

[0122] 2.2. Relationship between PR and offline risk:

[0123] Analysis of data from 278 patients revealed a significant difference in partial ventilator (PR) between the successful weaning group and the failed weaning group (2.54±0.51 vs 7.32±0.92, p<0.001). This indicates that the novel PR indicator is closely statistically correlated with the weaning risk of mechanically ventilated patients.

[0124] 2.3. The role of PR in predictive models:

[0125] When PR was incorporated into a predictive model that included clinical parameters, the model's predictive accuracy (AUC = 0.88) was significantly better than models that used PR or clinical parameters alone. This indicates that PR, as a novel predictive feature, can be combined with other indicators to significantly improve the predictive ability of mechanically ventilated patients' weaning risk.

[0126] 2.4. The predictive role of DPI

[0127] To verify the clinical value of DPI, we conducted an observational study on 30 mechanically ventilated patients. The results showed that in the successful weaning group (20 cases), the DPI value was 2.1±0.4, while in the failed weaning group (10 cases), the DPI value was significantly higher to 5.8±0.7, and the difference between the two groups was statistically significant (p<0.001).

[0128] When DPI is used alone for prediction, the area under the ROC curve (AUC) reaches 0.75, indicating that this indicator has good predictive efficacy. More importantly, the predictive accuracy is further improved when DPI is combined with other parameters. When combined with traditional clinical parameters (such as the shallow and rapid breathing index and maximum inspiratory pressure), the AUC increases to 0.82; when combined with baseline EIT parameters (such as regional ventilation distribution and ventilation unevenness), the AUC reaches 0.85; and when DPI is integrated with all available parameters, the AUC of the predictive model reaches 0.89. This shows that DPI can not only be used as a predictive indicator on its own, but also complements existing parameters well, jointly improving predictive accuracy.

[0129] Compared to traditional methods, the ensemble prediction model demonstrated superior performance. On the test set (n=42), the model's AUC was 0.89 (95% CI, 0.85–0.93), significantly higher than the two traditional offline parameters (AUC 0.76, 95% CI, 0.71–0.81, P<0.001) and the individual EIT parameter (AUC 0.82, 95% CI, 0.77–0.87, P=0.003). The model showed excellent calibration (Hosmer-Lemeshowx²=6.8, P=0.56) and consistent performance across all three centers.

[0130] All the above data have been validated through a clinical study of 30 patients, and the statistical differences are significant (Table 1) (p<0.05).

[0131] Table 1. Between-group differences in features and AUC predicting offline outcomes

[0132]

[0133] 2.4 Sensitivity analysis confirmed robust performance across different patient subgroups. The model performed particularly well in patients with longer mechanical ventilation durations (>7 days, AUC 0.91, 95% CI, 0.87–0.95) and those with higher APACHEI scores (>25, AUC 0.90, 95% CI, 0.86–0.94). Exclusion of patients meeting the borderline weaning criteria (AUC 0.88, 95% CI, 0.84–0.92) maintained stable performance in sensitivity analysis.

[0134] 2.5 Relative Contribution of Parameters

[0135] Importance analysis revealed that EIT-derived parameters significantly contributed to model performance. The Pendelluft ratio showed the highest importance score (0.82), followed by the global inhomogeneity index (0.76) and regional ventilation delay (0.71). Among the conventional parameters, RSBI (0.68) and P0.1 (0.65) exhibited moderate importance, while gas exchange parameters showed lower contributions (importance scores 0.45–0.55).

[0136] 2.6 Clinical Implementation and Decision Support

[0137] The simplified risk score derived from the model categorized patients into low-risk (n=98, weaning failure rate 8.2%), medium-risk (n=112, 31.3%), and high-risk (n=68, 64.7%) (P<0.001, trend). Adding the EIT parameter to the conventional predictive variables resulted in significant improvements in net reclassification (NRI=0.32, 95% CI, 0.25–0.39, P<0.001) and overall discrimination (IDI=0.15, 95% CI, 0.11–0.19, P<0.001).

[0138] 2.7 Clinical Results

[0139] All 278 patients had follow-up data within 28 days after extubation. The overall 28-day mortality rate was significantly higher.

[0140] The failure to wean compared to the successful group was significantly lower (23.9% vs 8.6%, P<0.001). Patients who failed to wean also experienced longer ICU stays (median 15 days vs 9 days, P<0.001) and hospital stays (median 28 days vs 18 days, P=0.002). After adjusting for age, APACHEI score, and initial diagnosis, failure to wean remained independently associated with increased mortality (adjusted HR 2.45, 95% CI 1.78–3.36, P<0.001).

[0141] Stratification by risk category in our predictive model showed a strong correlation with clinical outcomes. High-risk patients had significantly higher incidences of adverse events, including re-intubation (38.2% vs 5.1%, P < 0.001), tracheotomy (28.4% vs 7.2%, P < 0.001), and hospital-acquired pneumonia (42.6% vs 15.3%, P < 0.001) compared to low-risk patients. The relationship between model-predicted risk and observed outcomes was consistent across all three participating centers (interaction P = 0.78).

[0142] The median time without mechanical ventilation on day 28 showed a progressively decreasing trend across risk categories: 24 days for low-risk patients (IOR 21–26), 19 days for intermediate-risk patients (IOR 15–22), and 12 days for high-risk patients (IOR 8–16) (trend, P < 0.001). Resource utilization, as measured by ICU stay and duration of mechanical ventilation, also exhibited a similar pattern. Economic analysis revealed significantly higher healthcare costs in the high-risk group (median $52,000 vs. $31,000 in the low-risk group, P < 0.001).

[0143] Subgroup analysis showed that the model had particularly strong predictive accuracy for clinical outcomes in patients with ARDS (AUC 0.91, 95% CI 0.87–0.95) and patients on mechanical ventilation for >7 days (AUC 0.90, 95% CI 0.86–0.94). The model also showed strong predictive accuracy across age groups, body mass index, and admission diagnoses (all AUC > 0.85).

[0144] Event occurrence time analysis using Kaplan-Meier curves revealed significant differences in 28-day survival (log-rank P < 0.001) and ventilator-free survival (log-rank P < 0.001) between risk groups. Cox proportional hazards modeling confirmed that, after controlling for potential confounding factors, the high-risk category was independently associated with reduced survival (adjusted HR 2.8, 95% CI 1.9–4.1).

[0145] Implementing predictive models in clinical practice was associated with improved decision-making efficiency. The median time from the start of SBT to making a clinical decision decreased from 126 minutes before implementation (IQR 98–156) to 84 minutes after implementation (IQR 62–108) (P<0.001). Furthermore, a survey of healthcare providers indicated high satisfaction with the clinical usability of the model (median score 4.2 / 5.0, IOR 3.8–4.5).

[0146] III. Discussion

[0147] Our study demonstrates that combining EIT-derived local ventilation parameters with conventional clinical variables significantly improves the prediction of weaning outcomes. Our ensemble model (AUC 0.89) exhibits superior performance compared to conventional parameters alone (AUC 0.76), highlighting the added value of incorporating local ventilation information through EIT monitoring. This improvement in predictive accuracy is particularly noteworthy given the relatively low performance of conventional weaning parameters reported in previous studies.

[0148] Comparative analysis among different modeling approaches yielded some important insights. Conventional models, including only clinical variables such as RSBI, P<0.1, and respiratory mechanics, had an AUC of 0.76 (95% CI: 0.71–0.81) on the test set, significantly lower than the 0.89 AUC achieved by the ensemble model. EIT models, including only EIT-derived local ventilation parameters (P<0.001), also outperformed conventional clinical models, with an AUC of 0.82 (95% CI: 0.77–0.87, P = 0.003). These findings highlight that local ventilation information provides additional predictive value beyond conventional indices.

[0149] The key advantage of our model lies in its ability to capture complex physiological interactions. Relative contribution analysis shows that EIT-derived parameters, particularly measurements of ventilation heterogeneity and regional distribution patterns, provide substantial incremental value compared to conventional predictors. This aligns with emerging evidence suggesting that regional ventilation heterogeneity may serve as an early indicator of respiratory muscle fatigue and impending weaning failure. The analytical platform developed in our work makes these complex measurements practically feasible in routine clinical care.

[0150] Our findings have significant clinical implications for weaning management. The strong correlation between the model's predicted risk categories and actual outcomes supports the model's clinical applicability. The progressively increasing number of adverse events across risk categories indicates that the model effectively stratifies patients beyond simple binary classification, enabling more personalized weaning strategies. This refined risk stratification can help clinicians optimize the timing of extubation attempts, potentially reducing premature extubation and unnecessary prolongation of mechanical ventilation.

[0151] The innovation of this invention is mainly reflected in the following aspects: it proposes for the first time a quantitative indicator reflecting the dynamic changes of lung tilt; it innovatively incorporates the time dimension, more comprehensively reflecting the physiological changes of patients during weaning; this indicator has good independent predictive value and can complement other parameters, significantly improving predictive accuracy. This provides a new technical solution for improving the weaning risk assessment of mechanically ventilated patients.

[0152] The above features and technical solutions have not been reported in existing literature and technologies, demonstrating high innovation. Preliminary validation using data from 30 patients has confirmed the feasibility and effectiveness of this method. This lays the foundation for subsequent clinical application and promotion.

[0153] Figure 3 This is a schematic diagram of a computer device provided in an embodiment of the present invention, such as... Figure 3As shown, the device 2000 may include: one or more processors 2010 and one or more memories 2020; wherein the memories store computer-readable code that, when run by the one or more processors, can perform the methods described above.

[0154] The processor in this embodiment can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, operations, and logic block diagrams disclosed in this embodiment. The general-purpose processor can be a microprocessor or any conventional processor, and can be based on an x86 or ARM architecture.

[0155] In general, the various exemplary embodiments of this disclosure can be implemented in hardware or dedicated circuitry, software, firmware, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device. When aspects of embodiments of this disclosure are illustrated or described as block diagrams, flowcharts, or using some other graphical representation, it will be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as non-limiting examples in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0156] For example, the method or apparatus according to embodiments of this disclosure can also be used by means of Figure 4 The architecture of the computing device 3000 shown is used for implementation. For example... Figure 4 As shown, the computing device 3000 may include a bus 3010, one or more CPUs 3020, a read-only memory (ROM) 3030, a random access memory (RAM) 3040, a communication port 3050 connected to a network, an input / output component 3060, a hard disk 3070, etc. The storage devices in the computing device 3000, such as the ROM 3030 or the hard disk 3070, may store various data or files used for processing and / or communication of the methods provided in this disclosure, as well as program instructions executed by the CPU. The computing device 3000 may also include a user interface 3080. Of course, Figure 4 The architecture shown is merely exemplary and can be omitted as needed when implementing different devices. Figure 4 One or more components in the computing device shown.

[0157] This invention also includes a computer-readable storage medium, such as... Figure 5The diagram illustrates a storage medium 4000 provided in an embodiment of the present invention. The computer storage medium 4020 stores computer-readable instructions 4010. When the computer-readable instructions 4010 are executed by a processor, the method described above according to embodiments of the present disclosure can be performed. The computer-readable storage medium in the embodiments of the present disclosure may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. Non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), Synchronous Link Dynamic Random Access Memory (SLDRAM), and Direct Memory Bus Random Access Memory (DRRAM). It should be noted that the memory used in the methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0158] This disclosure also provides a computer program product or computer program that, when executed by a processor, implements the steps of the above-described method, such as... Figure 2 As shown, the computer program product or computer program includes:

[0159] Acquisition module 201: Used to acquire lung EIT monitoring data of mechanically ventilated patients over a period of time;

[0160] Feature extraction module 202: used to divide the lung into multiple regions and extract EIT parameters based on the lung EIT monitoring data. The EIT parameters include dynamic lung tilt index, which is the degree of variation of gas redistribution in each region during the spontaneous breathing test.

[0161] Prediction module 203: Used to input the EIT parameters into a classifier for classification, and to determine the patient's off-boarding outcome based on the classifier's output.

[0162] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0163] In general, the various exemplary embodiments of this disclosure can be implemented in hardware or dedicated circuitry, software, firmware, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device. When aspects of embodiments of this disclosure are illustrated or described as block diagrams, flowcharts, or using some other graphical representation, it will be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as non-limiting examples in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0164] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0165] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0166] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0167] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0168] The exemplary embodiments of this disclosure described in detail above are merely illustrative and not restrictive. Those skilled in the art will understand that various modifications and combinations can be made to these embodiments or their features without departing from the principles and spirit of this disclosure, and such modifications should fall within the scope of this disclosure.

Claims

1. A method for predicting the weaning risk of mechanically ventilated patients based on EIT, characterized in that, The method includes: S101: Acquire lung EIT monitoring data of mechanically ventilated patients over a period of time; S102: Divide the lungs into multiple regions, extract EIT parameters based on the lung EIT monitoring data, the EIT parameters include the dynamic lung tilt index, the dynamic lung tilt index is calculated by calculating the degree of variation of gas redistribution in each region during continuous monitoring of spontaneous breathing test, the dynamic lung tilt index is calculated by the sum of the ratios of the standard deviation and the mean of gas redistribution in each region during spontaneous breathing test, the specific calculation method is as follows; Where N represents the total number of regions, and when the lungs are divided into 4 quadrants, N=4; i∈[1,4]; This represents the standard deviation of gas redistribution in the i-th region. This represents the average amount of gas redistribution in the i-th region; The gas redistribution amount reflects the volume of gas redistributed from the non-dependent zone to the dependent zone during positive pressure ventilation. The calculation steps for the gas redistribution amount Vtpendelluft are as follows: by analyzing the impedance-time curve from end-expiration to end-inspiration, a specific time period for gas redistribution is identified; the impedance change in the dependent zone during this time period is integrated and then multiplied by a calibration coefficient; the specific time period is the mid-to-late inspiratory phase; the calibration coefficient is calibrated using known tidal volumes; the calibration coefficient is established by spatially integrating the impedance change of all lung pixels and calibrating it using flow sensor data synchronized with the ventilator, establishing a linear relationship between ΔZ and volume; this linear relationship is the calibration coefficient; ΔZ represents the local impedance change. S103: Input the EIT parameters into the classifier for classification, and determine the patient's off-boarding outcome based on the classifier's output.

2. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 1, characterized in that, The EIT parameter also includes: lung tilt ratio, which is the ratio of gas redistribution to tidal volume during the early inspiratory phase in each region.

3. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 2, characterized in that, The lung tilt ratio is expressed as: Where PR represents the tilt ratio. VT represents the gas redistribution amount in region i, and VT represents the total tidal volume of the lungs.

4. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 1, characterized in that, The EIT parameter also includes: a timing asynchrony index, which is expressed as: Where N represents the four regions divided, Indicates time delay between regions Late , It indicates the difference in gas redistribution within a breathing cycle between regions.

5. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 4, characterized in that, Represented as: in, This represents the time-domain ventilation signal of the i-th region. This indicates that the weighted average of the peak times of each respiratory cycle within the entire monitoring period is taken.

6. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 4, characterized in that, Represented as: , These represent the gas redistribution amounts in region i and region j, respectively.

7. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 1, characterized in that, The EIT parameters also include: regional ventilation distribution.

8. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 1, characterized in that, The EIT parameters also include: ventilation nonuniformity.

9. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 1, characterized in that, Simultaneously, clinical parameters are acquired, and EIT parameters and traditional clinical parameters are input into a classifier for classification. The patient's off-boarding outcome is determined based on the classifier's output.

10. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 9, characterized in that, The clinical parameters include the shallow and rapid breathing index.

11. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 9, characterized in that, The clinical parameters include maximum inspiratory pressure.

12. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 9, characterized in that, The clinical parameters include RSBI.

13. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 1, characterized in that, The time period is 5 minutes.

14. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 9, characterized in that, The offline result is either offline failure or offline success.

15. The method for predicting the weaning risk of mechanically ventilated patients based on EIT according to claim 1, characterized in that, The classifier includes one or more of the following: logistic regression, random forest, support vector machine, XGboost, decision tree, and extreme learning machine.

16. A computer device, characterized in that, The device includes: a memory and a processor; the memory is used to store a computer program; the processor executes the computer program to implement the steps of the method according to any one of claims 1-15.

17. A computer-readable storage medium, characterized in that, It stores a computer program thereon, which, when executed by a processor, implements the steps of the method as described in any one of claims 1-15.

18. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method described in any one of claims 1-15.

Images