Extrapolation of a non-linear autoregressive model of pulmonary mechanics

Estimating the incidence of spontaneous breathing effort of mechanically ventilated patients using a non-linear auto regressive (NARX) model

“…The nonlinear autoregressive model with exogenous inputs (NARX) is chosen as the model since the structure is represented by a simple linear difference equation. It predicts one time series given past values of the same time series, the feedback input, and another time series, called the external or exogenous time series [ [21] , [22] ]. The model contains an error term because knowledge of other terms will not enable the current value of the time series to be predicted exactly.…”

Virtual patients for mechanical ventilation in the intensive care unit

“…Statistical models efficiently provide best mathematical combinations for predictive relationships, but offer little understanding of the underlying mechanics [24][25][26][27] , even if prediction accuracy is acceptable [ 26 , 27 ]. A mechanical model offers more explicit meaning, physically and physiologically, and is thus more suitable for virtual patient models [25] .…”

Prediction and estimation of pulmonary response and elastance evolution for volume-controlled and pressure-controlled ventilation

“…In the last two decades, several complex models have been proposed and can effectively capture a large range of nonlinear pulmonary dynamics [34,[44][45][46][47][48][49][50][51]. However, their complexity means they suffer poor or non-identifiability [34,52], or are too complex to identify or apply at the bedside [53][54][55][56][57][58][59], thus limiting or eliminating their potential for clinical application. Far simpler black box models can be created, but require large amounts of data to train and may lack the ability to capture or describe all physiological features in various situations [58,60,61].…”

“…However, their complexity means they suffer poor or non-identifiability [34,52], or are too complex to identify or apply at the bedside [53][54][55][56][57][58][59], thus limiting or eliminating their potential for clinical application. Far simpler black box models can be created, but require large amounts of data to train and may lack the ability to capture or describe all physiological features in various situations [58,60,61]. In addition, physiological relevance is important because it supports clinical confidence and use and provides further insight to clinical end-users [52,62], but such black-box models cannot offer physiological relevance.…”

“…Currently, while several models can identify data [9,11,44,[69][70][71], the authors are aware of only one approach able to accurately predict outcomes from changes in MV care for resulted airway pressure for VCV and tidal volume for PCV [8,19,72]. These approaches use physiologically relevant basis functions to define respiratory mechanics over all possible pressures and volumes seen in MV, which is uncommon [35,58]. However, while they predict well, the basis functions proposed are independent of known PEEP levels and changes, creating complexity in understanding and implementation, especially for PCV predictions.…”