A novel method of fabricating dynamic, heterogenous benchtop lateral and third ventricle phantoms from MRI of hydrocephalus patients: a verification and validation study

TL;DR

This study creates accurate 3D models of human brain ventricles from MRI scans to study shunt failure in hydrocephalus patients.

Contribution

A novel method for fabricating dynamic, heterogeneous benchtop ventricle models with patient-specific geometries and compliance.

Findings

Ventricle models showed low variance in clinical values between replicates, indicating high repeatability.

Most ventricle geometries were within 5% of MRI measurements, showing strong anatomical accuracy.

A strong positive correlation (R2 = 0.9932) was found between internal pressure and volume, indicating accurate compliance.

Abstract

Investigation of the failure of shunts used to treat hydrocephalus has been a topic of research for several decades, though it is difficult to understand the role of heterogeneous patient ventriculomegaly shifts over time. In this work, we present the design, verification, and validation of a novel benchtop model of the human lateral and third ventricles.

3D models were rendered from MRI of pre-revision hydrocephalic patients (n = 6), printed in hollow High Impact Polystyrene (HIPS) molds, and injected with silicone rubber allowed to rotationally cure for three hours. Wall thickness of the ventricle models was measured via random point sampling (n = 300), assessing distribution of silicone rubber within the mold. One sample z-test was used to compare mean wall thickness to the target thickness of 1 mm. To visualize the inner volume post-hoc, expanding polyurethane foam was injected into ventricle models, creating negatives of the hollow molds. The negatives were 3D scanned and measured for frontal horn diameter, occipital horn diameter, Evan’s Index, Frontal-Occipital-Horn Ratio, and frontal horn volume. A single MRI was chosen for verification of repeatability (n = 11). Ventricular size was validated across models from patients (n = 6). Accuracy of ventricular expansion (indirect compliance assessment) was tested. For all tests, a confidence interval was set at 0.95 (α = 0.05).

The verification methods indicated that low variance of experimental clinical values were observed between ventricle model replicates. Nominal-actual comparison results consistently showed similar datapoint displacement densities between models. Validation showed that most ventricle geometries fell within 5% of MRI measurements. A strong positive correlation was observed between internal pressure and volume (R2 = 0.9932). Experimental Frontal Horn and Occipital Horn Diameter measurements compared to clinical observations showed anatomical accuracy.

This study validates and verifies a model of the human ventricular system with arbitrarily chosen heterogeneous patient ventricles of varying morphologies and volumes. Paired with a pump, this model can be used to recapitulate cerebrospinal fluid flow. We show the recreation of patient-specific clinical lateral ventricle characteristics of size, shape, and static compliance control necessary to study the influence of these parameters on shunt function. The manufacturing process has the capacity to create accurate benchtop models of the lateral and third ventricles with geometric detail that should be refined over time with additive systems accounting for cranial viscoelastic compliance and varying material properties with elastic and shear moduli more similar to brain.

The online version contains supplementary material available at 10.1186/s12987-025-00742-w.

Figures