AI-powered brain pressure monitoring could predict critical episodes in trauma patients
Researchers found that an artificial intelligence metric tracking pressure wave patterns in the brain can forecast dangerous ICP spikes hours in advance in traumatic brain injury patients. The discovery could improve intensive care management and patient outcomes by enabling earlier intervention—a significant advance for hospitals treating severe head trauma.
Originaltitel: Intracranial compliance monitoring using pulse shape index in traumatic brain injury: relation to cerebral physiology and clinical outcome
BACKGROUND: The intracranial pressure (ICP) pulse waveform reflects intracranial compliance. The pulse shape index (PSI), an artificial intelligence (AI)-based metric ranging from 1 (normal) to 4 (disturbed), quantifies morphological waveform pathologies. Early findings in smaller cohorts indicate that elevated PSI is associated with mass lesions, aging, higher ICP, and worse outcome. This study examined how PSI relates to other markers of intracranial compliance, the risk of ICP crisis, and clinical outcome in a large traumatic brain injury (TBI) cohort. METHODS: This retrospective study included 321 TBI patients with ≥ 12 h of ICP/PSI monitoring. PSI was analysed in relation to ICP, ICP pulse amplitude (AmpICP), the moving correlation coefficient between mean ICP and ICP amplitude (RAP index), and arterial blood pressure (ABP) using generalised additive models (GAMs). Linear mixed-effects models assessed whether PSI during preceding hours predicted later ICP elevations (e.g., ICP > 20/22 mmHg). The prognostic value of PSI was tested using univariable Mann-Whitney U test and multivariable (after adjustment for age, Glasgow Coma Scale, ICP, CPP, and PRx) logistic regression for mortality (Glasgow Outcome Scale = 1) and favourable outcome (Glasgow Outcome Scale = 4-5). RESULTS: ~30%). PSI was lower in survivors and in patients with favourable outcomes. However, in multivariable logistic regressions, PSI was not independently associated with mortality or favourable outcome. CONCLUSIONS: PSI was linked to established indicators of intracranial compliance and provided early warning of impending intracranial hypertension. While PSI showed univariable associations with outcome, these did not remain in multivariable models. Retrospectively, PSI does not appear to have independent prognostic value, but rather a complementary value. Prospectively, it may serve as a dynamic marker of deteriorating intracranial compliance and an early signal for future ICP crisis.