| Summary: | <p>The healthy functioning of the human brain depends on continuous and sufficient blood flow and metabolism. Severe interruptions in blood and oxygen supply can lead to energy impairment, and result in irreversible brain damage. In ischaemic stroke, a large reduction in blood supply leads to the breakdown of the blood-brain barrier (BBB) and reperfusion injury after reperfusion therapy. The increased BBB permeability disrupts the homeostasis of solutes between blood and interstitial fluid and therefore gives rise to fluid leakage from blood vessels into the interstitial space. As a result, brain injuries such as brain oedema and haemorrhagic transformation can occur.</p>
<br>
<p>The simulation of the full brain has previously been performed based on mechanical and poroelastic models to study brain mechanics. In these studies, however, the modelling of post-stroke brain damage has not been considered and it is therefore necessary to propose new models for further investigations. This thesis further develops the simulation tools in the In Silico Clinical Trials for the Treatment of Acute Ischaemic Stroke (INSIST) project.</p>
<br>
<p>I first propose a model for the simulation of osmotherapy, which is a common practice to relieve intracranial pressure (ICP) in brain oedema. The model is directly compared with clinical data for model validation. Furthermore, parameter sensitivity analysis is conducted to investigate the impact of various parameters, including shear modulus, tissue hydraulic permeability, and capillary vessel wall permeability, and to determine the effects of these physiological parameters on oedema development. Finally, different administration protocols are studied using the model and a near-optimal strategy for oedema treatment is proposed.</p>
<br>
<p>Secondly, haemorrhagic transformation after stroke commonly occurs alongside oedema. Therefore, the model is further developed to include the growth of haematomas by simulating the flow of blood through the interstitial space. The simulation results are compared with clinical imaging to obtain the blood perfusion data and then utilised to investigate the effects of high blood glucose, blood pressure, etc.</p>
<br>
<p>Thirdly, a study for the modelling of the contact mechanics of oedema in the brain is conducted. As the brain swells, the cerebrospinal fluid in the ventricle is compressed and drained to release pressure. A complex contact mechanics problem thus emerges which has not been previously solved. This model provides the first computational tool for the simulation of ventricle collapse and large brain deformation. Meanwhile, this study also focuses on the midline shift caused by brain swelling as it is a crucial criterion to determine the severity of oedema. Prediction curves of MLS-ICP relationships are proposed and compared with the clinical data.</p>
<br>
<p>Finally, an AI-assisted model for the optimisation of osmotherapy is proposed. The data produced from the in-silico model are utilised to train a deep neural network for the generation of virtual patient groups with different blood-brain barrier damage levels and age. This provides an approach for the optimisation of treatment strategy for patient stratifications.</p>
|
|---|