Cerebrovascular disease and stroke

Why imaging is not enough

Angiography resolves vessels down to roughly a millimetre. Perfusion is decided an order of magnitude below that, in arterioles and capillaries that no clinical scan shows. A model that stops where the image stops cannot say how much tissue is at risk.

We therefore generate the missing branches computationally. Trees are grown into the tissue volume under constraints on flow, pressure and metabolic demand, so that every region of tissue is supplied by a vessel of a plausible calibre, and the result matches perfusion measurements where those exist.

Autoregulation

Cerebral blood flow stays nearly constant as blood pressure varies and as posture changes, because small vessels adjust their diameter. We model arterial and venous networks down to the pre- and postcapillary level, couple them to a three-dimensional perfusion domain, and include both passive wall mechanics and the active regulatory response.

Across aortic pressures from 30 to 150 mmHg and supine, upright and inverted postures, the model reproduces this stability. Stiffer vessel walls suppress both the volume changes and the regulatory response, which is the mechanical signature of ageing vasculature.

Collaterals and vasodilation

When an artery is blocked, two mechanisms compensate: blood finds detour vessels, and vessels downstream widen. We generate collaterals statistically and model dilation as a smooth-muscle response to metabolic and flow stimuli. Collaterals alone recover little perfusion, and only when they are wide enough; dilation is the dominant mechanism, with collaterals in a supporting role.

Anatomy also matters. The Circle of Willis differs between individuals, and those differences change how well flow redistributes after an occlusion — a per-patient question the model can answer quantitatively.