Increasing the efficiency of cerebrovascular autoregulation-guided neuroprotection for traumatic brain injury patients

Abstract

Background: Optimal cerebral perfusion pressure (CPPopt) targeted treatment of traumatic brain injury (TBI) requires 4 h or longer multimodal monitoring data accumulation to identify CPPopt for individual patients. Nevertheless, even with a long 4-h data accumulation window, the CPPopt value can be identified in about 50–60% of monitoring time. Aim: To increase the efficiency of CPPopt targeted treatment by minimizing the time of monitoring data accumulation needed to choose the appropriate individualized clinical therapy. Method: A retrospective analysis of multi-modal physiological monitoring data of 87 severe TBI patients was performed by representing cerebrovascular autoregulation (CA) indexes in relation to CPP, arterial blood pressure (ABP), and intracranial pressure (ICP) separately in order to improve the existing CPPopt identification algorithms. Machine learning (ML) based algorithms were developed for automatic identification of selected informative data segments that were used for reliable CPPopt, ABPopt, ICPopt or lower/upper limits of CA (LLCA/ULCA) identification. Results/Conclusions: The reference datasets of the informative data segments, artifact-distorted segments, and datasets of different clinical situations were used for training ML-based algorithms allowing to choose the appropriate individualized CPP, ABP or ICP management in 79% of full monitoring time of studied population data. Developed ML-based algorithms allow to identify CPPopt, (or ABPopt or LLCA/ULCA) values within 24 min (10-times shorter comparing to 4 h of existing in clinical practice data accumulation time) in the cases when informative physiological ABP/ICP variations are detected. Prospective clinical studies are needed in order to prove the efficiency of developed algorithms.