Development of a ‘Digital-Twin’ model for Cerebrospinal Fluid Dynamics to use in real-time monitoring of patients after severe Traumatic Brain Injury

About the project

About the Project PhD only Summary Severe traumatic brain injury (TBI) produces complex and rapidly evolving disturbances in cerebral blood flow, intracranial pressure and cerebrospinal fluid (CSF) dynamics. Modern intensive care monitoring provides continuous measurements of many of these processes, but translating these data into a coherent picture of the underlying physiological state of the patient remains a major challenge. Project Aims This project aims to develop a patient-specific digital twin of cerebral haemodynamics and CSF dynamics that can track these physiological changes in real time. Rather than relying purely on data-driven prediction, the project will combine mathematical models of brain physiology with modern machine-learning and physics-informed modelling techniques. The goal is to exploit known relationships between arterial pressure, cerebral blood flow, intracranial pressure, cerebral compliance and CSF circulation, while allowing the model parameters to adapt continuously to an individual patient. The student will begin by evaluating and synthesising existing mathematical models of cerebral haemodynamics and CSF dynamics into a parsimonious model suitable for continuous bedside application. They will then investigate computational approaches for solving both the forward and, importantly, the inverse problem: estimating otherwise unobservable physiological parameters from routinely measured signals. Methods may include physics-informed neural networks (PINNs), neural differential equations, operator-learning approaches and related hybrid mechanistic/data-driven techniques. Model development and validation will make use of the extensive high-resolution physiological datasets collected by the Brain Physics Laboratory over more than two decades. These contain continuous recordings from patients with severe TBI, including arterial blood pressure, ECG, oxygen saturation, intracranial pressure, cerebral blood-flow velocity, cerebral oxygenation and other physiological variables. The combination of rich clinical data with physiological constraints provides an unusual opportunity to develop models that are both interpretable and capable of generalising beyond the data on which they were trained. A major practical component of the project will be translation of the resulting methods into a Python-based real-time processing module for ICM+, the Brain Physics Laboratory's clinical research neuromonitoring platform. This will allow the digital-twin model to be evaluated on continuously streamed bedside data and, ultimately, to explore how changes in pathology or treatment might alter the future physiological state of an individual patient. The longer-term ambition is to move from conventional monitoring of individual physiological variables towards a dynamic, personalised representation of the patient's cerebral physiology, capable of tracking disease progression, identifying changes in physiological reserve and potentially supporting treatment optimisation. Suitable background: This project would particularly suit graduates in Biomedical Engineering, Physics, Applied Mathematics, Electrical/Electronic Engineering, Signal Processing, Computer Science or a related quantitative discipline. Experience in Python, numerical modelling, machine learning, differential equations or physiological signal processing would be advantageous, but applicants would not be expected to have expertise in all of these areas at the outset. Contact: ps10011@cam.ac.uk Website: Author: Peter Smielewski - Cambridge Neuroscience How to Apply; If you are interested in this project, please go to the University pages and apply via the online portal; PhD https://www.postgraduate.study.cam.ac.uk/courses/directory/cvcnpdpcn/apply

Research areas