gms | German Medical Science

Objective: Since the primary brain injuries are merely irreversible, the main focus of neurocritical care is to prevent secondary neuronal death. To achieve this goal, an array of different monitoring techniques for the assessment of intracranial pressure (ICP), oxygenation status and metabolism was developed. However, the resulting high volume of datasets exceeds the ability of the neurointensivist to integrate these data into an adequate treatment algorithm. The goal of our study was to develop a pattern recognition tool, which will allow the real-time detection of brain swelling and impairment of autoregulation, based on the prediction of a mathematical model.

Method: To identify slow homeostatic positive or negative correlations between ABP and ICP, we utilized coherence and power spectra calculations with the multi-taper method (MTM) to determine the coherence between segments of two time series that were synchronously recorded in a cohort of 30 patients treated for either subarachnoid hemorrhage or traumatic brain injury. The data sampling rate was 0.2Hz. Additionally, we calculate the mean Hilbert phase difference between these segments. A correlation is called negative (scn) and indicative for brain swelling without loss of autoregulation if the mean Hilbert phase difference (mhpd) is higher or equal to 130 degree. Conversely if the mhpd is lower or equal 50 degree, the correlation is defined as called positive (scp) and therefore suggestive for an impaired autoregulation combined with severe brain swelling. Finally, a bedside monitoring device has been programmed to interpret the raw data derived from multimodal brain monitoring in a real time fashion.

Results: In accordance to the mathematical model, time segments of scn were correlated with changes of the intracranial compartment as illustrated by corresponding CT scans. In addition, the proportion of time which was detected as scp, significantly correlated with patients outcome (p=0.0001). All patients showing scp time segments longer than 10% of the entire observation time showed fatal outcome. After optimization of the analysis parameter sets, a real time monitoring device allowing to detect scn and scp has been developed and will be used in prospective trials focusing on treatment optimization utilizing this approach.

Conclusions: Our data demonstrate that an integrated real time analysis tool to process brain multimodal monitoring data may be useful for the early detection and management of pathophysiological events.