Machine Learning Supported Application of Advanced Neuromonitoring for Individualised Guided Management of Acute Traumatic Brain Injury in Intensive Care

Summary

Traumatic brain injury is an extremely complex pathology with a highly dynamic time profile developing over the course of the very first days in a critical care unit. Secondary pathological processes occur resulting from the initial trauma, with severe and often fatal consequences. The key aspects of the patient treatment/management during this period are: prevention (if possible), detection and subsequent alleviation of those insults.

Monitoring of various metrics, like pressures, flows and electrical activities from the body and the brain provides some indications for the onset and severity of those processes. However, interpretation of those measurements as presented by the patient monitors, as well as the electronic record systems, is still rather simplistic and largely based on trivial metrics, like hourly mean values of individual measurements. In addition, critical values for those individual measurements are population based, and thus do not reflect inter-individual differences and are not adjusted over the course of the patient stay in ICU.

Project aims

The aim of the project is to develop sensitive and robust metrics based on multi- parameter, continuous neuro-monitoring that would be better suited for guiding therapeutic interventions in traumatic brain injury patients in ICU. The project will involve application of cutting-edge time series analysis methods to a large number (1300+) of data sets (at full, waveform level, temporal resolution) collected by the group from neuro ICU in Cambridge over the last decades as well as the new, prospective, data set to be collected in the neuro intensive care unit.

As part of the project, methods for detecting early onsets of adverse events will be developed and algorithms for real time calculation of dynamic management targets proposed and implemented in software, building on top of the flagship software developed by the group ICM+ (https://icmplus.neurosurg.cam.ac.uk) and extending it’s functions with Python plugins. Furthermore, development and implementation of methods for automated data curation/pre-processing based on machine learning approaches will also form part of this project to ensure high quality of the input data for the decision making support algorithms.

Contact details

Dr Peter Smielewski - ps10011@cam.ac.uk

Opportunities

This project is open to applicants who want to do a:

• PhD
• MPhil