Dynamic Noninvasive Assessment of Intracranial Pressure, Cerebral Perfusion Pressure, and Autoregulation.

BACKGROUND: Intracranial pressure (ICP) monitoring is critical for managing traumatic brain injury, yet invasive methods carry risks and require specialized settings. Noninvasive transcranial Doppler (TCD)-based methods have been proposed to estimate ICP and cerebral perfusion pressure (CPP), although validation studies predominantly assess static agreement. Clinical monitoring depends on accurate reproduction of dynamic changes. This study assesses the ability of noninvasive methods to reproduce temporal variations in ICP, CPP, and autoregulatory indices. METHODS: Nonoverlapping 5-minute segments from 96 simultaneous ICP and TCD recordings in 38 TBI patients were analyzed. Eight blood velocity-based models were used to derive noninvasive estimates of ICP, CPP, pressure reactivity index (PRx), and mean flow index (Mx). Agreement was assessed using correlation and Bland-Altman analyses, whereas dynamic performance was evaluated using first-order differences and linear mixed-effects modeling. Discriminative ability for intracranial hypertension (ICP>22 mm Hg), low CPP (<60 mm Hg), and impaired autoregulation (PRx>0.3; Mx>0.4) was examined using receiver operating characteristic and precision-recall analyses. RESULTS: Noninvasive estimates of CPP and Mx demonstrated stronger dynamic agreement with their reference measures than noninvasive estimates of ICP and PRx. Selected models showed robust discrimination of low CPP and impaired Mx (area under the curve up to 0.96 and 0.89). Conversely, reproduction of ICP dynamics and PRx was moderate, with significant interpatient and intrapatient variability. Dynamic analyses revealed discrepancies missed by static agreement metrics. CONCLUSIONS: Evaluation of noninvasive ICP monitoring should extend beyond static accuracy. TCD-based approaches may support noninvasive monitoring of CPP and Mx, but current methods remain insufficient for reliable tracking of ICP and PRx dynamics.