Prediction of Extubation Succcessful in an Medical ICU using Machine Learning:A Preliminary Study

<bold>Background:</bold> Weaning from mechanical ventilation in the ICU is a complex clinical problem. However, despite such protocols, the rate of extubation failure remains high. Machine Learning methods are promising opportunities to positively influence patient outcomes and enhance clinical decisions processes by using the large amount of digital information generated in the ICU ventilator output parameter. This study analyzed the usefulness and accuracy of Machine Learning in predicting extubation success. <bold>Methods:</bold> We retrospectively evaluated data from patients(10/2023-11/2023). who underwent intubation for respiratory failure and received mechanical ventilation in an MICU. Information on 15 features, including patient demographics, vital signs, and ventilator data were extracted. This was a retrospective analysis of clinical registry data. We analysed patients final extubation readiness trial before 4 hours. Machine learning methods using Random Forest and ventilator parameter: Ppeak、Pmean、PEEP、FiO2、TV and RR. Extubation failure was defined as reintubation within 48 hrs. We also analyzed important features the AUC and prediction metrics. <bold>Results:</bold> Of random 78 episodes, 10.2% (N = 8) failed extubation. The mean age were 68.0(56.0-77.0) and mechanical ventilation duration days 8.8(7.3-12.1). Ventilation duration(6 feathure engineering) as being important in predicting extubation failure which for real-time successful extubation of mechanically ventilated patients in an ICU. The RF prediction accuracy: 0.884 and AUC about 0.7. <bold>Conclusions:</bold> Machine learing predicted successful extubation of ICU patients. Although the available database still be inadequate which aid the management hesitate extubation.