Using Machine Learning and Biomarkers for Early Detection of Delayed Cerebral Ischemia

Overview

The overall goal of this project is to determine if machine learning and analysis of neurospecific biomarkers can enable early detection of upcoming or ongoing cerebral ischaemia in patients suffering from subarachnoid haemorrhage with altered consciousness due to cerebral injury or sedation. Analyses of heart rate variability, electroencephalgraphy,nearinfrared spectroscopy, cerebral autoregulation, and brain injury specific biomarkers in blood and cerebrospinal fluid will be performed.

A new and promising approach to detect ongoing cerebral ischemia might be the detection of neurospecific biomarkers in blood. A biomarker for cerebral ischaemia, similar to troponin T and troponin I for detecting cardiac ischaemia, would be precious; however, such a biomarker for cerebral ischaemia is currently lacking. (9) There are several interesting neurospecific biomarkers for this purpose, such as Glial fibrillary acidic protein (GFAP), neuron-specific enolase (NSE), total tau, S-100, and neurofilament light chains (NFL). At this point, we do not have enough knowledge about levels of neurospecific biomarkers in blood and cerebrospinal fluid during delayed cerebral ischemia after subarachnoid hemorrhage. The sampling of neurospecific biomarkers have a dual purpose, the first is to investigate if we can detect ongoing cerebral ischemia with these biomarkers, and the second purpose is to compare levels of biomarkers to outcome in mortality and morbidity determined by the Glasgow Coma Scale Extended at 1-year, 3-years and 5-years after admission. Machine learning algorithms for predicting outcomes after delayed cerebral ischemia using a combination of clinical and imaging data have emerged. Nevertheless, prediction of delayed cerebral ischemia does not prevent it; to prevent delayed cerebral ischemia, an easily applied, cheap and reliable monitoring system that can warn physicians of the imminent risk of cerebral ischemia needs to be developed, making it possible to intervene. The overall goal of this project is to develop methods that enable the detection of upcoming or ongoing cerebral ischaemia in patients with subarachnoid haemorrhage Our primary aims are: * To develop a machine learning-based model that can identify patterns in signals obtained from HRV, NIRS, and EEG monitoring, which are consistent with upcoming cerebral ischemia and provide a warning about this to attending physicians. * To define the specificity and time relation of neurospecific biomarkers in blood and cerebrospinal fluid in patients with subarachnoid haemorrhage with and without delayed cerebral ischemia to evaluate if any of these biomarkers can be used as an indicator for ongoing cerebral ischemia. * To assess the prognostic value of changes in physiological and neurospecific biomarkers changes during the acute phase after subarachnoid hemorrhage on long-term outcome.