Abstract
Magnetic Resonance Imaging (MRI) visible perivascular spaces (PVS) have been associated with age, decline in cognitive abilities, interrupted sleep, and markers of small vessel disease. Therefore, several computational methods have been developed for their assessment from brain MRI. But the limits of validity of these methods under various spatial resolutions, and the accuracy in detecting and measuring the dimensions of these structures have not been established. We use a digital reference object (DRO) previously developed for this purpose, to construct an in-silico phantom for answering these questions; and validate it using a physical phantom. Our in-silico and physical phantoms use cylinders of different sizes as models for PVS. Using both phantoms, we also evaluate the influence of the “PVS” orientation on the accuracy of the diameter measured, different sets of parameters for two vesselness filters that have been used for enhancing tubular structures, namely Frangi and RORPO filters, and the influence of the vesselness filter per-se in the accuracy of the measurements. Our experiments indicate that PVS measurements in MRI are only a proxy of their true dimensions, as the boundaries of their representation are consistently overestimated. The success in the use of the Frangi filter for this task relies on a careful tuning of several parameters. The combination of parameters α=0.5, β=0.5 and c=500 proved to yield the best results. RORPO, on the contrary, does not have these requirements, and allows detecting smaller cylinders in their entirety more consistently in the ideal scenarios tested. The segmentation of the cylinders using the Frangi filter seems to be best suited for voxel sizes equal or larger than 0.4 mm-isotropic and cylinders larger than 1 mm diameter and 2 mm length. “PVS” orientation did not influence their measures for image data with isotropic voxel size. Further evaluation of the emerging deep-learning methods is still required, and these results should be tested in “real” world data across several diseases.
Introduction
Perivascular spaces (PVS) are fluid-filled channels surrounding the brain's vasculature, serving as a primary waste-clearance system. While they are a normal part of brain anatomy, their enlargement is a recognized biomarker for small vessel disease, cognitive decline, and sleep disorders. Detecting and measuring these structures is challenging because they are typically 1–3 mm in diameter, which is close to the resolution limits of standard clinical MRI (often ~1 mm isotropic). This leads to significant partial volume effects, where a single voxel contains a mix of PVS fluid and surrounding brain tissue, complicating automated segmentation and measurement. This study aims to establish the limits of validity for computational tools used to assess PVS across different spatial resolutions.
Methods
The researchers employed a dual-phantom validation strategy to establish a ground truth for PVS measurements:
- In-silico Phantom (DRO): A Digital Reference Object was constructed using mathematical models of cylinders with varying diameters (0.2–3 mm), lengths, and orientations to simulate PVS in a controlled digital environment.
- Physical Phantom: A Perspex plate with precision-drilled holes (0.2–3 mm) was filled with water and scanned using a high-powered 7T MRI at multiple resolutions (0.3 mm to 0.5 mm isotropic).
Two common "vesselness" filters were evaluated for their ability to enhance and detect these tubular structures:
- Frangi Filter: A classic Hessian-based method that analyzes local image intensity curvature.
- RORPO Filter: A morphological method (Ranking Orientation Responses of Path Operators) that identifies shapes based on path alignment without requiring extensive manual tuning.
Results and Discussion
The experiments revealed several critical insights into the accuracy of PVS quantification:
The Overestimation Effect
A primary finding is that MRI-based measurements of PVS are consistently larger than their true physical dimensions. This overestimation is largely due to partial volume effects at the boundaries of the structures. Consequently, MRI measurements should be viewed as a proxy for PVS burden rather than an absolute physical measurement.
Filter Performance and Sensitivity
The RORPO filter demonstrated superior sensitivity, detecting cylinders as narrow as 0.4 mm at a 1 mm resolution. In contrast, the Frangi filter required a diameter of at least 1.0 mm to reliably detect the structures. Furthermore, the Frangi filter was found to be highly sensitive to parameter tuning; the optimal configuration was identified as α=0.5, β=0.5, c=500.
Impact of Resolution and Orientation
Measurement accuracy was not significantly affected by the orientation of the "PVS" when using isotropic voxels. However, the Frangi filter's effectiveness dropped significantly for voxel sizes smaller than 0.4 mm or for cylinders smaller than 1 mm in diameter and 2 mm in length.
Conclusion
This study highlights the inherent limitations of current MRI-based PVS measurement techniques. While automated tools are valuable, researchers must account for the consistent overestimation of PVS dimensions and the sensitivity of detection filters to both resolution and parameter settings. Future research should leverage these phantom benchmarks to evaluate emerging deep-learning segmentation methods and validate these findings in clinical populations with diverse neurological conditions.
Summary updated