Mechanical loading of the ventricular wall as a spatial indicator for periventricular white matter degeneration

Abstract

Progressive white matter degeneration in periventricular and deep white matter regions appears as white matter hyperintensities (WMH) on MRI scans. To date, periventricular WMHs are often associated with vascular dysfunction. Here, we demonstrate that ventricular inflation resulting from cerebral atrophy and hemodynamic pulsation with every heartbeat leads to a mechanical loading state of periventricular tissues that significantly affects the ventricular wall. Specifically, we present a physics-based modeling approach that provides a rationale for ependymal cell involvement in periventricular WMH formation. Building on eight previously created 2D finite element brain models, we introduce novel mechanomarkers for ependymal cell loading and geometric measures that characterize lateral ventricular shape. We show that our novel mechanomarkers, such as maximum ependymal cell deformations and maximum curvature of the ventricular wall, spatially overlap with periventricular WMH locations and are sensitive predictors for WMH formation. We also explore the role of the septum pellucidum in mitigating mechanical loading of the ventricular wall by constraining the radial expansion of the lateral ventricles during loading. Our models consistently show that ependymal cells are stretched thin only in the horns of the ventricles irrespective of ventricular shape. We therefore pose that periventricular WMH etiology is strongly linked to the deterioration of the over-stretched ventricular wall resulting in CSF leakage into periventricular white matter. Subsequent secondary damage mechanisms, including vascular degeneration, exacerbate lesion formation and lead to progressive growth into deep white matter regions.

Keywords: Ependymal cells; Finite element simulations; Lateral ventricles; Tissue damage; White matter hyperintensities.

Figure 1:

A) Periventricular white matter lesions gradually deteriorate from thin linings to smooth halos and eventually spread into deep periventricular white matter tissue. B) Hypothesized corresponding stages of ependyma damage in a horn of the lateral ventricles; starting from an intact ependyma with proper barrier function, propagating to loss of barrier function and increased CSF diffusion into the white matter, ultimately leading to severe white matter and cerebrovascular damage. Ependymal cells, shown in purple, are not to scale; CSF is shown in blue and parenchyma in orange.

Figure 2:

Model generation process: A) We select the axial slice with the largest ventricular area and identify the pvWMH (red mask). B) We segment gray matter, white matter, and cerebrospinal fluid, create a 2D finite element model, and mimic hemodynamic loading by applying a normal pressure to the ventricular surface [B1] and the CSF-gray matter interface [B2]. C) Representative output of maximum principal Green Lagrange strain. D) We parametrize the lateral ventricular wall to plot the variation of mechanical output measures along its perimeter.

Figure 3:

Simulation results for all eight models during peak loading. Top row: brain tissue segmentations converted into FE models; second row: displacement magnitude at peak hemodynamic loading [mm]; third row: maximum principal strain [−]; bottom row: left hemispheres show the tangential stretch field and the right hemispheres show the normal stretch fields [−]. Tangential and normal stretch fields are similar with respect to left and right hemisphere.

Figure 4:

Representation of the brain’s A) local displacement magnitudes and corresponding displacement vectors and B) maximum principal strain field. Schematic representation of our proposed damage mechanisms that shows the C) healthy ependymal wall and D) degraded ependymal lining which causes CSF leakage into periventricular tissues. Mean mechanomarker distribution along the parameterized ventricular wall based on results from all 8 models: E) tangential stretch ${\lambda }_t$ in red and normal stretch ${\lambda }_n$ in blue; F) thinning ratio, G) curvature of the LV wall; and H) pvWMH thickness measured from MRI. The figures in bottom row I) show the superposition of the WMH mask from FLAIR and the thinning ratio. We observe a consistent spatial overlap between pvWMH location and elevated ependymal cell thinning.

Figure 5:

Surrogate models of ventricular horns with varying curvature measured by $\gamma$, i.e., the ratio of the ellipsoid’s long and short axis. We report the maximum principal Green Lagrange strain (top row), tangential stretch field ${\lambda }_t$ (middle row), and normal stretch field ${\lambda }_n$ (bottom row) of the ventricular wall. The top right graph shows peak maximum Green Lagrange strain for increasing and the bottom right graph shows maximum tangential and minimum normal stretch for increasing .

Figure 6:

Impact of the septum. The first row shows the tangential ${\lambda }_t$ and normal ${\lambda }_n$ stretch fields for model M40 without the septum (left two images) and with the septum (right two images). Second row, mean and standard deviation of the thinning ratio for models without the septum (blue) and with the septum (green). Third row, shows the difference between the thinning ratio without the septum (default) and with the septum.

Figure 7:

Model sensitivity to stiffness and pressure. A) An increase in pressure or a decrease in stiffness leads to increased peak Green Lagrange strains. B) An increase in pressure or decrease in stiffness also increases the ventricular wall (VW) section affected by strains > 0.01.