Muscle Pressure Estimation With Artificial Intelligence During Mechanical Ventilation
- Conditions
- Respiratory Failure
- Interventions
- Device: Artificial Intelligence Estimation of Muscle Pressure during Mechanical Ventilation
- Registration Number
- NCT05820347
- Lead Sponsor
- University of Sao Paulo General Hospital
- Brief Summary
The goal of this diagnostic study is to validate estimation of inspiratory muscle pressure by an artificial intelligence algorithm compared to the gold standard, the measure from an esophageal catheter balloon, in patients under assisted mechanical ventilation. The main questions it aims to answer are:
• Are inspiratory muscle pressure estimates from an artificial intelligence algorithm accurate when compared to the direct measure from an esophageal balloon?
Participants will be monitored with an esophageal balloon and with an artificial intelligence algorithm simultaneously, with inspiratory muscle pressure estimation during assisted mechanical ventilation with decremental levels of pressure support.
- Detailed Description
This is a diagnostic study to validate estimation of inspiratory muscle pressure during assisted ventilation from an artificial intelligence algorithm integrated in a mechanical ventilator (FlexiMag, Magnamed, Brazil) compared to direct measure of muscle pressure from esophageal catheter balloon (gold standard). This is a novel non-invasive method to estimate inspiratory muscle pressure.
After obtaining informed consent, participants will be monitored simultaneously with the esophageal balloon and the artificial intelligence algorithm, with decremental levels of pressure support (20 to 2 cmH2O, in steps of 20 minutes). Esophageal balloon will be removed after completing the last pressure support step.
The investigators estimated a sample of 50 participants, considering 3 cmH2O as a clinically relevant discordance between methods and 10% of missing data. Concordance analysis and correlation analysis will be performed.
Procedures will follow a specific Standard Operating Procedures and participants inclusion data will be inserted in REDCap.
Recruitment & Eligibility
- Status
- COMPLETED
- Sex
- All
- Target Recruitment
- 50
- Patients under assisted or assist-control mechanical ventilation
- Contraindication to esophageal catheter insertion (esophageal cancer or bleeding, esophageal fistula, skull base fracture, uncontrolled coagulopathies)
- Contraindication to transient neuromuscular blockade
- Bronchopleural fistula (persistent air leak)
- Hemodynamic instability (norepinephrine > 1mcg/kg/min)
- Gestation
- Current sinus infection
- Refusal from patient's family of attending physician
- Palliative care
Study & Design
- Study Type
- INTERVENTIONAL
- Study Design
- SINGLE_GROUP
- Arm && Interventions
Group Intervention Description Artificial Intelligence Estimation of Muscle Pressure during Mechanical Ventilation Artificial Intelligence Estimation of Muscle Pressure during Mechanical Ventilation All included subjects will be monitored simultaneously with the esophageal balloon (gold standard) and with the artificial intelligence algorithm integrated in the mechanical ventilator. Electrical Impedance Tomography will be used to monitor ventilatory patterns during different degrees of spontaneous effort. First, a single intravenous bolus of neuromuscular blockade (succinylcholine 1mg/kg or rocuronium 1.2mg/kg) will be performed to measure respiratory system mechanics (compliance and resistance). In cases where rocuronium is used, a single dose of sugammadex 4mg/kg will be administered intravenously to reverse neuromuscular blockade after measuring compliance and resistance. After initiation of spontaneous breathing effort, pressure support will be titrated from 20 cmH2O to 2 cmH2O, in decremental steps during 20 minutes each. After completing titrating of pressure support, the esophageal balloon will be removed.
- Primary Outcome Measures
Name Time Method Correlation between muscle pressure amplitude estimation (in cmH2O) by artificial intelligence and esophageal balloon 4 hours Correlation, reported as R-squared and a correlation plot, between amplitude in cmH2O of muscle pressure estimation by artificial intelligence and esophageal balloon.
Concordance between muscle pressure amplitude (in cmH2O) estimation by artificial intelligence and esophageal balloon 4 hours Analysis of the bias and limits of agreement (Bland-Altman plot) between muscle pressure estimated amplitude in cmH2O from artificial intelligence and measured by esophageal balloon.
Detection of initiation time and ending time of a spontaneous breathing cycle by artificial intelligence compared with esophageal balloon 4 hours Time difference (in ms) between initiation of a spontaneous breathing cycle and ending of a spontaneous breathing cycle between artificial intelligence and esophageal balloon.
- Secondary Outcome Measures
Name Time Method Sensitivity and specificity of patient-ventilator asynchrony automated detection using the Artificial Intelligence Muscle Pressure estimator 4 hours Number of patient-ventilator asynchronies detected using artificial intelligence compared with number of asynchronies detected by experts assessing airway pressure, flow and esophageal balloon waveforms.
Trial Locations
- Locations (1)
Heart Institute, University of São Paulo
🇧🇷Sao Paulo, SP, Brazil