Computer methods and programs in biomedicine
-
Comput Methods Programs Biomed · Aug 2009
Intelligent analysis in predicting outcome of out-of-hospital cardiac arrest.
The prognosis among patients who suffer out-of-hospital cardiac arrest is poor. Higher survival rates have been observed only in patients with ventricular fibrillation who were fortunate enough to have basic and advanced life support initiated early after cardiac arrest. ⋯ Six different supervised learning classification techniques were used and evaluated. It has been shown that machine learning methods can provide an efficient way to detect important prognostic factors upon which further emergency unit actions are based.
-
Comput Methods Programs Biomed · Aug 2009
A minimal model of lung mechanics and model-based markers for optimizing ventilator treatment in ARDS patients.
A majority of patients admitted to the Intensive Care Unit (ICU) require some form of respiratory support. In the case of Acute Respiratory Distress Syndrome (ARDS), the patient often requires full intervention from a mechanical ventilator. ARDS is also associated with mortality rate as high as 70%. ⋯ The ability to use this identified patient specific model to optimize ventilator management is demonstrated by its ability to predict the patient specific response of PEEP changes before clinically applying them. Predictions of recruited lung volume change with change in PEEP have a median absolute error of 1.87% (IQR: 0.93-4.80%; 90% CI: 0.16-11.98%) for inflation and a median of 5.76% (IQR: 2.71-10.50%; 90% CI: 0.43-17.04%) for deflation, across all data sets and PEEP values (N=34predictions). This minimal model thus provides a clinically useful and relatively simple platform for continuous patient specific monitoring of lung unit recruitment for a patient.