Mechanistic multiphysics modeling reveals how blood pulsation drives CSF flow, pressure, and brain deformation under physiological and injection conditions

TL;DR

This work develops a mechanistic multiphysics model of the CNS that couples cerebrospinal fluid flow with poroelastic brain/spinal tissue and spinal dura elasticity in a fully closed CNS geometry. The CSF dynamics are driven by pulsatile cardiac inflow, while intrathecal injections are represented by a volumetric source and a pressure-dependent absorption sink, enabling predictive simulations of ICP changes and CSF transport. The model reproduces physiological features such as craniocaudal decay and phase shifts in spinal CSF flow, pressure amplification along the spine, and spinal tissue displacements, and it provides quantitative predictions of ICP elevation and recovery during IT injections. This framework offers a versatile platform for investigating CSF flow mechanisms, optimizing IT delivery, and extending to additional physiological drivers or pathologies, with potential for real-time surrogate modeling in the future.

Abstract

Intrathecal (IT) injection is an effective way to deliver drugs to the brain bypassing the blood-brain barrier. To evaluate and optimize IT drug delivery, it is necessary to understand the cerebrospinal fluid (CSF) dynamics in the central nervous system (CNS). In combination with experimental measurements, computational modeling plays an important role in reconstructing CSF flow in the CNS. Existing models have provided valuable insights into the CSF dynamics; however, most neglect the effects of tissue mechanics, focus on partial geometries, or rely on measured CSF flow rates under specific conditions, leaving full-CNS CSF flow field predictions across different physiological states underexplored. Here, we propose a comprehensive multiphysics computational model of the CNS with three key features: (1) it is implemented on a fully closed geometry of CNS; (2) it includes the interaction between CSF and poroelastic tissue as well as the compliant spinal dura mater; (3) it has potential for predictive simulations because it only needs data on cardiac blood pulsation into the brain. Our simulations under physiological conditions demonstrate that our model accurately reconstructs the CSF pulsation and captures both the craniocaudal attenuation and phase shift of CSF flow along the spinal subarachnoid space (SAS). When applied to the simulation of IT drug delivery, our model successfully captures the intracranial pressure (ICP) elevation during injection and subsequent recovery after injections. The proposed multiphysics model provides a unified and extensible framework that allows parametric studies of CSF flow dynamics and optimization of IT injections, serving as a strong foundation for integration of additional physiological mechanisms.

Figures (6)

• Figure 1: (a) Illustration of the central nervous system. (b) Left: Blood-brain interface. Blood flows in vessels surrounded by porous brain tissue filled with interstitial fluid. Middle: Net blood inflow rate into the brain fitted from Baledent2014-blood-inflow. Right: Simplification of the driver of the CSF pulsation. The expansion during systole and contraction during diastole of blood vessels is simplified as ISF production during systole and absorption during diastole, respectively. (c) Computational domain used in this study. The domain can be partitioned anatomically or into different subdomains of materials: CSF-filled spaces ($\Omega^{f}$), including ventricular system ($\Omega^{f,v}$), cranial SAS ($\Omega^{f,c}$), and spinal SAS ($\Omega^{f,s}$); porous tissue ($\Omega^{p}$), including brain parenchyma ($\Omega^{p,c}$) and spinal cord ($\Omega^{p,s}$); an elastic membrane representing the spinal dura mater ($\Gamma^d$), the caudal end of which is marked by $\Gamma^e$.
• Figure 2: (a) Instant volumetric flow rate of net blood inflow and CSF craniocaual flow across the foramen magnum. (b) Cumulative blood inflow into the brain and craniocaual flow across the foramen magnum. (c) Left: CSF caudocranial flow rate at C2, T1, T5, T10, and L3 level. Right: CSF caudocranial waveforms along the spinal SAS. We mark the time instants of the peak craniocaudal flow with white stars and fit them with a linear curve. The estimated slope of the curve, representing the pulse wave speed, is 6.47 m/s.
• Figure 3: (a) CSF pressure at C2, T1, T5, T10, L3 levels in one cardiac cycle. (b) CSF pressure profile in spinal SAS in one cardiac cycle. (c) CSF pressure relative to foramen magnum at C2, T1, T5, T10, L3 levels in one cardiac cycle. (d) CSF pressure relative to foramen magnum in spinal SAS in one cardiac cycle.
• Figure 4: (a) Caudocranial displacement of spinal cord at C2, C3, C5, C6, T5, L1 levels. (b) Caudocranial velocity of the spinal cord at C2, C3, C5, C6, T5, L1 levels. (c) Caudocranial displacement of the spinal cord in spinal SAS. The region below the bottom axis is neglected because the spinal cord is truncated there in our model.
• Figure 5: (a) Comparison of stroke volume along the spinal SAS between our simulations and MRI measurements Coenen2019-spinal-SAS-experiment. The in vivo measured stroke volumes, obtained at C3, T1, T3, T6, T8, T10 and T12 levels, are scaled so that the value at C3 matches the simulated one, facilitating comparison of spatial trends along the spinal SAS. (b) Maximum percentage area change of different levels of cross-section in the spinal SAS.