Assessing the prediction potential of an in silico computer model of intracranial pressure dynamics

Abstract

Objective: Traumatic brain injury (TBI) frequently results in poor outcome, suggesting that new approaches are needed. We hypothesized that a patient-specific in silico computer model of ICP dynamics may predict the ICP response to therapy. Design: In silico model analysis of prospective data. Stetting: 16-bed pediatric intensive care unit in a tertiary care academic hospital. Patients: 9 subjects with severe TBI undergoing ICP monitoring (7M/2F, age range 3-17 years). Interventions: Random changes in head-of-bed (HOB)(0 0 , 10 0 , 20 0 , 30 0 , 40 0) elevation and respiratory rate (to achieve a ∆ETCO2 = ±3-4 mmHg) daily as long as an ICP monitor was in place. Methods and Main Outcome Measures: A 6-compartment dynamic ICP model was constructed based on data from the interventions and session-specific model parameters were estimated. The accuracy of session-specific model-calculated ICP was compared to the accuracy ICP calculated using non-session specific parameter values. To assess the prediction accuracy of the model, two analyses were performed: a "within" session analysis of Segment A for model estimation and Segment B for prediction, and a "between" session analysis to predict later session ICP using parameters from > 1 earlier sessions. A mean absolute error to mean absolute deviation ratio (MAE/MAD) of < 1 was considered favorable. Results: For non-session specific parameters, MAE/MAD was <1 in 2/24 (8%) sessions. For session-specific parameter values MAE/MAD was <1 in 21/24 (88%) sessions, and <0.5 in 9/24 (38%) sessions. Sessions with low (<12 mm Hg) (N=8; 33%) or high (>18 mm Hg) (N=6; 25%) ICP had lower error than moderate ICP (12-18 mmHg) (N=10; 42%). MAE/MAD was <1 for 6/22 (27%) for within-session predictions and 3/31 (10%) for between-session predictions. 3 Conclusions: The protocol for collecting physiologic data in subjects with severe TBI was feasible and without undue risk. The in silico ICP model with session-specific parameters accurately reproduced actual ICP response to changes in HOB and RR. We demonstrated modest success at predicting future ICP within a session and to a lesser extent between sessions.

Predicting Intracranial Pressure and Brain Tissue Oxygen Crises in Patients With Severe Traumatic Brain Injury

“…Several works have been published attempting to either predict the onset of ICP crisis events or to forecast future ICP values themselves with limited success (17)(18)(19). The closest work to this study is that of Güiza et al (8).…”

Functional relationship of arterial blood pressure, central venous pressure and intracranial pressure in the early phase after subarachnoid hemorrhage

“…Furthermore, on average increasing CVP led to increase in ICP, which is reported previously in the literature as occurring after a loss of compliance (22,23). Nonlinear models of ICP dynamics have previously been proposed which specifically model the logarithmic relationship between intracranial volume and pressure (24,25) and also incorporate the effect of cerebral autoregulation through linear transfer functions (26)(27)(28)(29)(30). However, parameter estimation in nonlinear systems is much more involved than estimating linear system parameters because of the nonconvexity of the loss function.…”

Use of models in identification and prediction of physiology in critically ill surgical patients

“…More sophisticated analyses using longer time interval data and other entropy calculations in the ECG tracing have subsequently been reported as predictive. In the area of traumatic brain injury, Wakeland et al 40 have reported on the feasibility of an in silico model of intracranial pressure dynamics. Together these represent examples of the use of advanced monitoring technology to provide in silico determination of physiological signals and predictions.…”