Aneurysm Volumetry: The Unmet Need for Validation
Article information
Intracranial aneurysms have traditionally been characterized by linear measurements—maximum diameter, neck width, and dome-to-neck ratio. These parameters remain indispensable in clinical practice and have been extensively validated as predictors of rupture risk and treatment outcome [1,2]. Yet aneurysms are 3-dimensional structures, and there is growing evidence that volumetric parameters offer clinically meaningful information that linear dimensions alone cannot provide [3].
VOLUME IN PRACTICE: PACKING DENSITY AND DEVICE SIZING
The most familiar application of aneurysm volume is in calculating coil packing density—the ratio of implanted coil volume to aneurysm sac volume. A packing density of at least 20–25% has long been considered a reasonable target to reduce recanalization, although the clinical evidence supporting this threshold has proven less robust than originally assumed [4,5]. Volume has more recently gained attention in a second context: device sizing for intrasaccular flow disruptors. After a decade of clinical experience, the Woven Endo-Bridge (WEB) has established itself as a reliable treatment for wide-neck bifurcation aneurysms, with device sizing recognized as a key determinant of outcome [6,7]. The device-to-aneurysm volume (DAV) ratio has emerged as a more informative sizing parameter than width and height alone, with higher DAV ratios associated with both successful implantation and better angiographic occlusion at follow-up [8-11].
THE LIMITS OF GEOMETRIC APPROXIMATION
Although volume-based parameters are increasingly used in clinical decision-making, their clinical interpretation still depends on the accuracy of the underlying aneurysm volume measurement. This is relevant not only to packing density, whose association with angiographic outcome has been inconsistent [4,5], but also to DAV ratio-based WEB sizing, where inaccurate sac volume may affect device selection [8,10,11]. One explanation may lie in how aneurysm volume is measured. Most clinical studies have relied on geometric formulae based on linear dimensions—tools such as AngioCalc (www.angiocalc.com) apply an ellipsoid approximation derived from 2 or 3 diameters—which assumes a reasonably regular sac shape. When the sac is lobulated, multilobulated, or has daughter sacs, geometric formulae substantially underestimate the true volume, with a direct consequence of overestimating packing density or distorting volume-based device-sizing parameters [12,13]. If the denominator is systematically wrong, the relationship between volume-based parameters and clinical outcomes may be obscured accordingly.
SOFTWARE-BASED VOLUMETRY: PROGRESS WITH CAVEATS
Three-dimensional angiographic software addresses many of these shortcomings. Voxel-based segmentation from 3D rotational angiography datasets has been shown to be significantly more accurate than geometric formulae, with mean volume errors as low as 3–4% compared to over 10% for ellipsoid-based methods [14,15]. Automated segmentation is also faster, more consistent, and better suited to irregular morphologies, and these tools are now widely available on modern angiographic workstations.
However, software-based volumetry introduces 2 main sources of error, alongside other minor variables. First, and perhaps less intuitively, threshold dependency deserves particular attention. The boundary of the contrast-filled sac is defined by an intensity threshold—either set by default or adjusted visually—and there is no way to determine which threshold best represents the sac boundary [16]. This problem may be more pronounced in small aneurysms, where even minor threshold differences affect a larger fraction of the total volume. Even when a threshold is appropriately set, the intrasaccular contrast concentration itself varies with injection conditions and pulsatile blood flow, meaning that the apparent sac boundary may shift depending on hemodynamic circumstances at the time of acquisition [15,17].
Second, the neck plane—the boundary between the sac and the parent vessel—is estimated automatically by most platforms, but the result can be ambiguous in wide-neck or fusiform aneurysms and may require operator review and adjustment. There is no consensus on where exactly this boundary should be drawn, and different placements yield meaningfully different volumes. While less consequential than the above, variation in acquisition and reconstruction parameters—such as voxel size, radiation dose, and reconstruction kernel—may also contribute to measurement variability in ways that remain poorly characterized [15,17,18].
THE MISSING STEP: VALIDATION
Taken together, these issues point to a fundamental validation problem. Many studies validate new software by comparing its output against another software or manual measurement—but if neither has been verified against a reference volume, this only establishes agreement, not accuracy. A method can be highly reproducible and still systematically deviate from the reference volume. These errors are not merely technical, because they directly affect volume-based parameters such as packing density and DAV ratio-based device sizing. This point is particularly relevant to AI-based platforms, which may improve consistency and workflow but still require validation against a reference volume before being accepted for clinical or research use [19,20].
What the field needs is rigorous validation of volumetric tools against an independently determined reference volume. This means using phantoms with precisely known cavity volumes, verified using a micropipette, gravimetric measurement, or equivalent methods, so that software-derived measurements can be directly compared with a reference volume rather than with a nominal design volume, a manufacturer-provided value, or another software output. Such phantoms should not be limited to simple, regularly shaped models, but should also include clinically relevant irregular morphologies, such as lobulated, multi-sac, and daughter-sac aneurysms.
Only with such reference models can we systematically evaluate how threshold settings, neck plane definition, intra-aneurysmal contrast variation, and other factors affect measurement accuracy [14,16]. These data are needed to develop volumetric software that is not only convenient and reproducible, but also validated across the aneurysm morphologies encountered in clinical practice.
Notes
Fund
None.
Ethics Statement
This was not a human population study; therefore, neither approval from the Institutional Review Board nor the obtainment of informed consent was required.
Conflicts of Interest
YS has been an assistant editor of the Neurointervention since 2022. No potential conflict of interest relevant to this article was reported.
Author Contribution
Concept and design: YS. Analysis and interpretation: none. Data collection: none. Writing the article: YS. Critical revision of the article: YS. Final approval of the article: YS. Statistical analysis: none. Obtained funding: none. Overall responsibility: YS.