Evaluation of flow diverters for cerebral aneurysm therapy: recommendations for imaging analyses in clinical studies, endorsed by ESMINT, ESNR, OCIN, SILAN, SNIS, and WFITN

Abstract

Background

Multiple studies and meta-analyses have described the technical and clinical outcomes in large cohorts of aneurysm patients treated with flow diverters (FDs). Variations in evaluation methodology complicate making comparisons among studies, hinder understanding of the device behavior, and pose an obstacle in the assessment of further advances in FD therapy.

Methods

A multidisciplinary panel of neurointerventionalists, imaging experts, and neuroradiologists convened with the goal of establishing consensus recommendations for the standardization of image analyses in FD studies.

Results

A standardized methodology is proposed for evaluating and reporting radiological outcomes of FD treatment of intracranial aneurysms. The recommendations include general imaging considerations for clinical studies and evaluations of longitudinal changes, such as neointimal lining and stenosis. They cover standards for classification of aneurysm location, morphology, measurements, as well as the assessment of aneurysm occlusion, wall apposition, and neck coverage. These reporting standards further define four specific braid deformation patterns: foreshortening, fish-mouthing, braid bump deformation, and braid collapse, collectively termed ‘F2B2’.

Conclusions

When widely applied, standardization of methods of measuring and reporting outcomes will help to harmonize the assessment of treatment outcomes in clinical studies, help facilitate communication of results among specialists, and help enable research and development to focus on specific aspects of FD techniques and technology.

Introduction

Emerging nearly two decades ago, the concept of flow diverting stents (flow diverters, FDs) has revolutionized the therapeutic landscape for intracranial aneurysms.1,3 The Pipeline Embolization Device (PED, originally Chestnut Medical Technologies, Menlo Park, CA, USA) was the first commercially available FD. It received the Conformité Européene (CE) mark in June 2008 and entered the US market after receiving United States Food and Drug Administration (FDA) approval in April 2011.4,7 While FD technology underwent an iterative development, multiple review papers and meta-analyses described the technical and clinical outcomes in large patient cohorts.8,13

Numerous new FDs with differing construction and configuration have entered the market, frequently with the focus on simplified deployment. Based on recent observations,^{14 15} some interventionalists hypothesized that newer-generation FDs develop15,17 braid changes in follow-up images more frequently than expected from previous experience.^{14 15} However, actual data supporting this claim are limited. A recent meta-analysis¹² analyzed the FD literature on device braid collapse, braid narrowing, and braid bump deformation. None of the papers reporting on these conditions provided any definition. In contrast, more than a third of the studies analyzing the convergence of the proximal or distal end of the FD with a fish-mouth appearance (‘fish-mouthing’) provided a definition. The resulting data suggested a heterogeneity among the studies that might be explained by heterogeneity of definitions.

Assessing the performance of FDs in general and braid stability in particular in clinical trials requires clear definitions of the analyzed variables and clearly defined¹²operational criteria for evaluating these variables. Such standards allow more valid comparisons between different device generations and inform device development. Unfortunately, these standards are still very incompletely defined.¹²

To create such specifications, a multidisciplinary panel of neurointerventionalists, imaging experts, and neuroradiologists with extensive experience in neuroimaging and neuroendovascular therapy convened with the goal of drafting consensus recommendations for the standardization of image analyses in FD studies. Whereas general clinical and radiological standards for studies of endovascular repair of saccular intracranial cerebral aneurysms have been published previously,¹⁸ the panel aimed to focus primarily on specific aspects of the radiological evaluation of FD therapy.

We propose a standardized methodology for evaluating and reporting radiological outcomes of FD treatment of intracranial aneurysms that is endorsed by multiple international societies. The primary focus of this article is the image evaluation for clinical studies by imaging core laboratories. However, multiple variables are site-evaluated as part of patient enrollment in clinical studies before being reassessed by an imaging core laboratory. Image evaluation should be based on the same definitions to avoid unnecessary discrepancies. We hope the standards as defined in this article will support future research, as well as undergoing much scrutiny and further development.

Basic image evaluation considerations for FDs

General imaging considerations for clinical studies

Image data quality

Real-world clinical study data do not necessarily reach the quality standards of the latest publications from research institutions using the latest imaging technology. Focusing clinical studies on such centers would likely delay study recruitment and introduce significant bias into the study results. This recommendation is based on the collective experience from multiple studies from hundreds of centers worldwide.

Given this variability in image data quality, the role of expert readers becomes pivotal in translating these image findings into a set of standardized variables for robust comparison. Specifically, comparing observations from different angles and modalities by expert readers is a key requirement of core laboratory adjudication.

Despite standardization of image acquisition as part of most study protocols, there remains considerable heterogeneity of uploaded images regarding acquisition techniques and image quality in multicenter studies. Clear central core laboratory instructions for dealing with such image quality issues (eg, incomplete series) — both for verification of quality and image adjudication — can mitigate these effects and are important for reducing both random noise and systematic errors, thereby improving the likelihood for finding differences in device performance by more precise treatment effect estimates.¹⁹

Image readers

Image readers should have a thorough understanding of the performance of FD procedures, preferably via their own personal hands-on experience. In addition to having a thorough understanding of the standardized definitions from this recommendation, the readers should undergo dedicated training for the specific study protocol and core laboratory charter requirements and definitions. The training materials should be saved for reference. Although artificial intelligence (AI) is gaining impressive momentum in the evaluation of angiographic images,^{20 21} its application without supervision by experienced adjudicators is not expected soon.

Increasing the reading precision while achieving a high inter-rater agreement is a major reason for establishing core laboratory assessments as the standard in evaluating angiography in many interventional disciplines, including neurointerventions.²² Differences in reproducibility of different variables in angiographic image evaluations may yield significant differences in study sample size requirements.²³ The goal of an imaging core laboratory is to work towards an objective, standardized, consistent, and impartial assessment of study-related imaging.¹⁴ A reading by two independent readers is recommended, at least for the primary study endpoint (double reading). Eventual disagreements should be solved by consensus among the readers. An additional rater should be involved as tiebreaker only if such consensus cannot be achieved.

Usage of reading results

First, the analyzed variables should describe isolated, simple observations that include as few assumptions as possible. The interpretation should be a separate process because our assumptions are frequently less founded. A typical example is that ‘proximal fish-mouthing due to intimal lining’ is occasionally interpreted as ‘no real fish-mouthing’. The interpretation can be done later when the combination of ‘proximal intimal lining: yes’ is combined with ‘proximal fish-mouthing: yes’. Such analysis can be realized easily with modern data analytics methods. Such an approach decreases heterogeneity and increases precision. Second, image reading is a standardized description for the purpose of a clinical study. It is not necessarily meant to judge or qualify the individual clinical decision-making which is often way more complex. According to our experience, about half of the issues detected by the core laboratory readers are rejected by the local investigators. When comparing central core laboratory adjudication with local operator assessment, there was a 24% and 26% lower rate of complete aneurysm occlusion in the PRESAT²⁴ and the Cerecyte trials,²¹ respectively, and an up to 15% lower rate in complete reperfusion in a recent meta-analysis of thrombectomy results.^{11 22 25}

Reading infrastructure

The software for capturing the reading results should allow tracking changes after first release of reading results by the individual readers, their eventual corrections, and the specifics of the consensus process.²⁶ Screen shots with markings should be stored alongside the rating results to illustrate the reasoning und understanding of the reader in the individual case. There should be a detailed imaging core laboratory charter and additional training documents with detailed definitions and reading instructions.

Evaluating longitudinal changes: general considerations

We recommend that follow-up vascular imaging be performed at 6 and 12 months. Further angiographic follow-ups are recommended at the discretion of the treating centers at yearly intervals. A digital subtraction angiography (DSA) at 12 months should represent the major milestone for comparing aneurysm occlusion and potential braid changes to the procedural result. Finishing the imaging follow-up can be considered if the results are perfect (complete occlusion without stenosis or braid deformation).²⁷

The most complete assessment of aneurysm healing — including the aneurysm shape beyond the contrasted vascular lumen — involves performing both a DSA and a magnetic resonance angiography (MRA) at approximately the same time point. The filling of aneurysm and vessel lumen can be assessed more easily using a MRA with contrast (figure 1). Other modalities are acceptable, per local standard. If more than one imaging modality is available from the same date (eg, DSA and MRA), a single modality should be selected as leading modality while the other(s) can be taken into consideration when interpreting the leading modality.

Figure 1. Flow diverter wall apposition can be evaluated in three-dimensional time-of-flight magnetic resonance angiography after contrast media application (0.3 x 0.6 x 0.6 mm voxel size). The quality of the evaluation needs to be recorded in parallel.

The modality of preference is first DSA, then CT angiography (CTA), then MRA. Regarding CTA, flat-panel detector CTA (FP-CTA) with intra-arterial contrast agent administration is preferred over standard CTA, due to the superior spatial resolution of the cone beam technology and improved depiction of even small stents. MRA is superior to standard CTA in cases with additional coil treatment.

In case a certain modality is not available or cannot be evaluated, the next modality in line of priority should be selected instead. Depending on the scope of the study, the comparator of a follow-up should be either the immediate post-procedural result or the previously scheduled visit, not an unscheduled visit. The quality of any longitudinal comparison can be adversely impacted by an inconsistent set of imaging modalities and protocols, with varying quality, and even projections. To account for that, the quality of the comparison should be evaluated separately from the result. We recommend a simple, three-tier grading system (table 1). Other quality criteria such as patient movement or poor contrast bolus are indirectly reflected in this grading system also as they lead to poorer comparability. The grading should be explained with specific examples during the reader training.

Table 1. Grading of image comparability between different time points.

Comparability	Description	Comment
Good	Similar projections/techniques, good images, free from relevant artifacts	Highest quality comparisons, choice for questions with highest precision requirements
Sufficient	Deviating projections/techniques but still usable for change assessment, no substantial artifacts in any of the visits	Standard quality comparisons, can be used
Insufficient	Image projections/techniques do not allow for change assessment, substantial artifacts from any source in any of the visits, also related to contrast application	Discard comparisons as ‘cannot determine’

Evaluating longitudinal changes: neointimal lining and stenosis

Some degree of intimal overgrowth is necessary for therapy success. Its characterization will become more relevant with FD surface modification and healing processes. The term ‘intimal hyperplasia’ implies an unwanted overshoot in intima growth and is reported in more than 50% of cases after FD deployment.²⁸ We suggest avoiding the qualification of ‘hyperplasia’ and to use the term ‘neointimal lining’ instead. The degree of resulting stenosis is captured independently. Both conditions can be combined in a later data analysis.

We suggest evaluating the existence of neointimal lining on unsubtracted angiograms with a simple score (yes, no, or indeterminate) independently for each FD at three points (proximal, middle, and distal). Neointimal lining can be evaluated in DSA, FP-CTA, and CTA images. An evaluation of neointimal lining in MRA is not possible in most cases and should be avoided. Stenosis is the apparent luminal narrowing of the vessel as it is opacified with contrast dye, compared with the expected diameter at that section. The degree of luminal narrowing is not necessarily related to the diameter of the visible FD braid in the affected section. The comparison is accomplished by selecting the nearest normal vessel segment as a reference (D^normal) and the narrowest part as stenosis maximum (D^stenosis). At pre-procedure, post-procedure, and follow-up visits, parent artery percent diameter stenosis should be determined by measurement of the ratio of the narrowest relevant arterial segment, as compared with an adjacent normal arterial segment (generally within 5 mm proximal to, within, or within 5 mm distal to the FD). The reference should be selected in line with the criteria as used in the Warfarin-Aspirin Symptomatic Intracranial Disease (WASID) study (figure 2).²⁹

Figure 2. Example for stenosis measurement in a flow diverter. In this case, the diameter of the proximal portion of the artery at its widest, parallel, non-tortuous normal segment (D^normal = 4.1 mm) was used as a reference (first choice) and compared with the narrowest lumen (D^stenosis = 2.4 mm). The resulting stenosis is 42% ((D^normal–D^stenosis)/D^normal)).

• First choice involves measuring the diameter of the proximal part of the parent artery (at its widest, non-tortuous normal segment).

• If Choice #1 is diseased, the second choice involves measuring the diameter of the distal portion of the parent artery (at its widest, non-tortuous normal segment).

• If Choice #2 is also diseased, the third choice involves measuring the most distal normal segment of the feeding artery (non-tortuous).

The reasons for in-stent stenosis (ISS) are lumen loss (‘true ISS’) and negative remodeling with fully expanded mesh around a stenotic lumen.³⁰ Reversible ISS may be related to thrombus underneath endothelization which cannot be differentiated reliably by routine imaging.

Based on our experience with core laboratories in FD studies, the degree of stenosis in MRA can be determined in approximately half of the cases, dependent on the FD type, the acquisition method, the patient anatomy, and the relationship to the direction of the magnetic field. We define any stenosis of greater than 25% as ‘relevant,’ and a stenosis of greater than 50% as ‘significant.’ To optimize the use of available MRA imaging data, a qualitative assessment for significant stenosis should be considered (yes, no, or indeterminate).

Only a thrombus within the inner FD lumen is considered a thrombus on the device. A clot within the aneurysm is not counted (as this is a therapeutic goal). Reporting should be based on the final angiographic series or run.

Evaluating aneurysms and their occlusion

Aneurysm location

Aneurysms are named according to the vessel of neck origin or closest branch vessel origin. There are several systems for grouping aneurysm locations.831,34 We recommend the application of the Aneurysm Reporting Terminology for aneurysm location¹⁸ and in case of ICA aneurysms the Bouthillier classification.³¹ If several segments are affected, we recommend using the most proximal one (‘starting point’) to classify the aneurysm. For selection of the aneurysm side in anterior communicating or basilar aneurysms we recommend selecting a side only if the aneurysm neck is located predominantly on one side (clearly >2/3). Otherwise, the midline should be chosen.

Aneurysm morphology

Side branches and branches originating from the aneurysm as well as relation of the aneurysm to its parent artery determine treatment strategy and likelihood of success in FD therapy. Differentiating side branch aneurysms from bifurcation aneurysms is typically based on the typical arterial anatomy (eg, anterior cerebral artery (ACA)-middle cerebral artery (MCA), MCA-M2, basilar artery-posterior cerebral artery (PCA)). Although the respective labeling is useful as anatomical nomenclature, it is not as helpful as a predictor for a successful outcome. Different anatomies and arterial diameters (eg, of posterior communicating (PCom)) create different flow conditions, leading to different treatment strategies with different pre-treatment probabilities for aneurysm occlusion. Therefore, aneurysms with the same anatomical label (eg, an aneurysm originating from PCom) represent a heterogeneous cluster of different flow conditions.

We propose a pragmatic method for reporting aneurysm morphology, considering the relationship between parent arteries and branches, as shown in figure 3. The aneurysm is considered a sidewall aneurysm when a single parent artery has a gap between a neighboring artery of any diameter and the aneurysm, or when the artery originating at the aneurysm base has a diameter of less than 30% of the parent artery. A bifurcation aneurysm is defined as an aneurysm originating from an arterial bifurcation in which the smaller artery has the diameter of a least 20% of the larger one. A branch originating from the aneurysm is described only when the contour in any projection suggests aneurysm wall instead of artery wall. This approach may lead to classifying a paraophthalmic aneurysm in some cases as a bifurcation aneurysm, instead of a sidewall aneurysm, which may create problems when comparisons with older studies are required. However, the data can still easily be matched by using the location as an additional variable.

Figure 3. Relation of parent arteries and branches in the vicinity of the aneurysm neck. (A) Sidewall aneurysm (anatomic): originating from a single parent artery. (B) Sidewall aneurysm (hemodynamic): the aneurysm neck touches a second artery with a diameter of less than 30% of the other artery. (C) Bifurcation aneurysm: originating from an actual pre-existing arterial bifurcation. The second artery has a diameter of at least 30% of the major artery. (D) An artery originating from the harmonic shape of the aneurysm sac.

Aneurysm shape

Based on their hypothesized pathogenesis, aneurysms are typically separated into the overlapping shape subgroups of saccular, fusiform, and dissecting.

Saccular (‘berry-shaped’) aneurysms are the most common type of intracranial aneurysm, typically arising from bifurcations with a clearly defined neck.

Aneurysms shaped like a spindle (‘fusiform’) are thought to have a dissecting origin in 73%, atherosclerotic origin in 18%, and collagen disease origin in 9% of cases.³⁵ Therefore, terms such as ‘fusiform’ have little meaning per se.¹⁸ Nevertheless, it appears useful to apply a pragmatic classification of aneurysm morphology. We recommend using the term ‘fusiform’ for aneurysms that are spindle-shaped.

In contrast, the label ‘dissecting aneurysm’ should be used for false, irregularly shaped aneurysms, frequently with a stenosis or evidence of intimal injury or underlying vasculopathy. The term ‘blister’ should be reserved for specific carotid lesions characterized by a half-dome-shaped appearance.

Aneurysm measurements

There are several methods available for using automated and semi-automated measurements to determine the diameter and volume of an aneurysm and its parent artery. However, a correcting step or manual adjustment is required in most cases. For clinical studies, manual measurement by experienced readers is still considered the standard method of reference (figure 4).

Figure 4. Critical measurements of aneurysm dimensions for saccular and fusiform/dissecting aneurysms. Height, defined as the maximum linear distance between the proximal margin of the aneurysm and the distal margin of the aneurysm dome, measured perpendicular to the parent artery longitudinal axis (where applicable). Width, defined as the maximum linear distance between the lateral margins of the aneurysm dome, measured parallel to the parent artery longitudinal axis (where applicable, A). In fusiform and dissecting aneurysms (B), height and neck are defined as the widest diameter of the contrasted aneurysm lumen. Neck, defined as the maximum linear distance between the lateral margins of the aneurysm neck, measured parallel to the parent artery longitudinal axis (where applicable). In fusiform and dissecting aneurysms (B), neck and height are equally defined as the widest diameter of the contrasted aneurysm lumen.

All assessments of maximum distances are constrained by the projections provided to the central core laboratory. Sizes should be measured in millimeters (or to the nearest tenth of a millimeter) and are ideally recorded from two sources such as MRA or CTA and DSA (ideally 3D-DSA with maximum intensity projections).

Reliable measurements in 2D-DSA images by the core laboratory require calibration factors from angiographic images that are frequently unavailable. Using external calibration methods, such as coins affixed to the scalp, may be unrealistic for most studies.²¹ Using a guiding diameter as reference can be an option also if the size information is available to the core laboratory. Any existing daughter sacs should be included in the measurements.

However, the most reliable measurement of the parent artery or aneurysm dimensions require the availability of 3D raw data, which are rarely available for central core laboratory analysis in clinical studies.

Evaluation of aneurysm occlusion

The goal of aneurysm treatment is its complete exclusion from the circulation by occlusion. The most widely used scoring system is the Raymond–Roy Scale (RR),³⁶ which divides the residual static filling of endovascularly treated saccular aneurysms into three categories: complete occlusion, residual neck, and residual aneurysm. It should be part of a standard reading for the FD studies, even if it neglects the different conditions and therapeutic goals in the FD treatment of bifurcation aneurysms and aneurysms with an artery integrated into the aneurysm sac (figure 3). A neck remnant in a coiled aneurysm has a different quality than residual filling after FD therapy.

If patients with bifurcation aneurysms and aneurysms with an artery integrated into the aneurysm sac are included in the study population, the modified Cekirge–Saatci (mCS) classification should be used to describe the new dynamic flow-remodeling process.^{37 38} Cekirge, Saatchi, and Hanel expanded this classification using novel subgroups for evaluation of cerebral aneurysms treated with any endovascular technique. The selection of other aneurysm occlusion scores depends on the specific scope of the study and is not covered here (table 2).

Table 2. Major aneurysm occlusion scores.

Variable	Scale	Scope
Static occlusion	Raymond–Roy	Occlusion
	Original (2001)36
	Modification (2016)41
	Meyers (2009)11	Occlusion
	Cekirge–Saatci	Occlusion + Flow + Anatomy specific (expanding Raymond–Roy)
	Original (2016)37
	Modification (2023)38
Residual flow	O'Kelly–Marotta (2010)42	Focused on flow dynamics in aneurysm sac, mainly before and directly after implantation
	Kamran–Byrne (2011)43
	SMART (2012)44
Occlusion prediction	FDSS (2017)35	Occlusion + Flow + Anatomy + Clinical (outcomes-based grading scale)
	4F-FPS (2021)45	Flow + Anatomy

• Aneurysm recurrence: Aneurysm recurrence is the change in aneurysm occlusion status from complete occlusion (RR class I) to new neck remnant or residual filling of the aneurysm.

• Aneurysm occlusion change: Additionally, we recommend capturing the change in contrasted lumen (increased, decreased, different shape with similar volume, no change, indeterminate). This change can be observed without change in the RR class (but decreasing contrasted aneurysm volume).

Any occlusion scale describes the contrasted lumen of the aneurysm and potentially the adjacent arteries. However, even in cases of absence or partial absence of contrast filling, an aneurysm can remain unstable with persistent remodeling processes, intrasaccular thrombus, or changing outer shape that can be monitored by cross-sectional imaging. A respective validated scale is missing.

Aneurysm recurrence and occlusion change should be assessed only in case of good comparability (table 1).

FD deployment and braid deformation

Full coverage of the aneurysm neck and wall apposition are required to achieve the goal of redirecting the flow toward the parent artery and providing an endoluminal scaffold for vascular reconstruction. FDs are primarily constructed of cobalt chromium (eg, Pipeline, Pipeline Flex, and Pipeline Shield, Medtronic, USA and Surpass, Stryker, USA) or nitinol (eg, Silk, Balt, France or Derivo, Acandis, Germany).⁷ The materials are typically braided with multiple strands woven together to create a mesh structure. Cobalt/chromium adds stiffness and radial force and allows a better response to angioplasty when incomplete wall apposition is observed (eg, in curved arteries).

Newer generations of nitinol-based FDs have increased flexibility and improvements in navigation and deployment, but may have lower radial force, which may be associated with less wall apposition or a higher risk of kinking and braid deformation.

The best imaging modality for assessing braid deformation is FP-CTA, ideally with reconstructions and source images. If not available, unsubtracted DSA images should be assessed without contrast (for braid deformation alone) and with contrast (for assessing intimal lining).

FD wall apposition

Beyond the inherent biomechanical properties, the deployment efficacy of a braided stent or FD is contingent on the accurate sizing (diameter and length) and the specific anatomy of the proximal and distal landing zones. It is more likely for any braided device to open completely when the proximal landing zone is in a straight vessel segment. Insufficient opening in a curve is one of the most frequent problems and might require additional maneuvers or second stent implantation. Proximal, incomplete opening is frequent and can be mostly solved with re-passing the device with the delivery catheter. In rare cases, stent migration might occur proximally, even with bulging into the aneurysm. A possible explanation for this complication might be a soft FD with insufficient landing zones. The assessment of wall apposition is crucial and should be meticulously conducted in the final DSA run.

Wall apposition should be evaluated along the entire FD length, and separately recorded for its proximal and distal end. The completeness of wall apposition along the full FD length is evaluated in a DSA by the presence or absence of visible contrast media between the FD and the parent artery wall. Wall apposition can be evaluated using MRA, when of sufficiently high quality for follow-up (figure 1). MRA may show hyperintensive signal intensity on T1-weighted imaging representative of slow flow or thrombus between the wall of the parent artery and the FD also.

FD neck coverage

The longitudinal coverage of the aneurysm neck requires an evaluation separate from radial wall apposition. For meeting the criteria of ‘full neck coverage’ the aneurysm ostium must be covered in its entirety. The FD must touch the parent artery wall, even if only 1 mm proximal or distal to the neck.

Intentionally placing the FD into a more proximal artery without aneurysm neck coverage is a viable option to achieve indirect flow diversion.³⁹ Consequently, a missing or incomplete neck coverage according to this classification is not necessarily a poor placement result (figure 5). We suggest differentiating indirect flow diversion (intended) from a misplaced neck coverage (‘failed’, eg, due to device shortening, migration) by adding a secondary question in case of missing or incomplete neck coverage (intended, not intended, cannot determine).

Figure 5. Types of aneurysm neck coverage. (A) No neck coverage. The intention is to create an indirect flow diversion.³⁹ (B) Partial neck coverage. The flow diverter (FD) reaches the junction between the aneurysm and the parent artery on one side. (C) Full neck coverage. The FD completely covers the junction between aneurysm and parent artery. It applies only if confirmed in all available projections.

For a statistical analysis, these cases can be identified additionally by combining neck coverage with aneurysm location and morphology. The dissociation of the aneurysm morphology and the FD placement allows for a more granular and precise assessment of FD performance.

Braid changes

The underlying mechanism of action of FDs is to provide a stable scaffold for endothelial cell growth along the device struts, thereby facilitating aneurysm exclusion from the circulation.^{1 40} Therefore, one of the central questions for the radiological evaluation of FD performance is the shape and long-term stability of the FD braid. Braid changes are deviations of the radiological appearance of the FD braid from the result after initial deployment. Changes in FD braid appearance gained more attention recently, despite having been observed and reported for many years.¹² Neither a widely accepted definition nor standardization of these findings has been established.

Based on our experiences, we recommend subdividing the FD braid changes into four categories (F2B2): (fore)shortening, fish-mouthing, braid bump deformation, and braid collapse (table 3 and figure 6, figure 7). These categories should apply to both the procedural post-deployment imaging and the follow-up visits. They should be strictly applied to the FD braid, irrespective of its wall apposition and the vessel lumen. A fifth category should be considered as well (‘Other braid changes: yes, no, or indeterminate’) to capture patterns that cannot be assigned to the F2B2 categories.

Table 3. Braid deformation patterns (F2B2): foreshortening, fish-mouthing, braid bump deformation, and braid collapse.

	Type	Definition	Location	Grading	Comments
1	Foreshortening*	Movement on either or both ends of the FD towards the aneurysm neck of at least 5%	Proximal	Yes, no, cannot determine	Clearly visible compared with specific previous image, no semiquantitative estimate
			Distal
2	Fish-mouthing	Converging of focal FD end (cone shape) with more than 25% decrease of the outer FD diameter compared with expected continuation of FD shape	Proximal	None (<25%)†, 25%–50%, >50% of braid diameter, cannot determine	Involves just the proximal and/or distal tip
			Distal
3	Braid bump deformation	Braid bump or localized outward bulging of the FD mid-segment	Mid-segment (mid 50%)	Yes, no, cannot determine	Deformation exceeds proximal and/or distal braid diameter; should have dense rim
4	Braid collapse	Bottle-necked or cigar-shaped deformation diameter, decrease of >1/3 involving >1/3 of FD length	Either proximal or distal	Yes, no, cannot determine	Commonly noted, acutely requiring angioplasty or device removal; as opposed to fish-mouthing, involves a larger FD segment

^*Foreshortening is reported by comparing any follow up-image compared with the final image of the index procedure. Elongation (which accompanies negative remodeling) or foreshortening ±5% of total length.

^†Fish-mouthing of <25% cannot be reliably detected in typical standard of practice images with their different angulations of digital subtraction angiography (DSA) acquisitions and vessel anatomies.

FD, flow diverter.

Figure 6. Implanted flow diverter in the common carotid artery of the rabbit. Microcomputed tomography (microCT) (left panel) shows malformation of the device with expansion of the proximal aspect. Cross-sectional imaging with high-frequency tomography (right panel, corresponding with the black line on the microCT) shows medial thickening with braid bump deformation; however, conserving the cross-sectional area with an elliptical configuration.

Figure 7. Flow diverter braid deformation (F2B2). Foreshortening, fish-mouthing, braid bump deformation, and braid collapse illustrated in a figure and magnified unsubtracted image without contrast (for further definitions see table 3).

For analysis of these patterns, single unsubtracted, fluoroscopic spot images are recommended as a gold standard. If not available, high-quality CTA or MRA can be used to exclude some of the conditions. Such evaluations are subject to individual expert judgement. The quality of the comparisons should be classified as recommended above.

Due to the braided design, foreshortening of the device may occur during deployment to varying degrees, depending on the biomechanical FD properties and the ratio of the diameters of the stent to the target vessel. Unintended foreshortening can be observed after FD deployment, with most negative consequences occurring when the landing zones are near the aneurysm neck. Further expansion and foreshortening at the neck might lead to FD displacement, either proximally, distally, or on both ends. Foreshortening can be observed as result of positive remodeling as a fundamental property of braided devices.

Convergence of the proximal or distal end of the FD may give it a fish-mouth appearance, and this is more often occurs in a curved segment of the vessel where a soft stent is overexpanded at the level of the aneurysm neck (for specific definition see table 3). The term ‘fish-mouthing’ should not be applied to cases in which a cone-shaped FD braid is apposed to the wall of a tapering parent artery with decreasing diameter that was present before FD deployment. This observation should be described briefly in a comment. Fish-mouthing still applies in cases in which the degree of deformation is, or becomes, more pronounced.

Potential causes of these patterns are hard to assign reliably without advanced imaging and should be documented separately (eg, intimal lining, parent artery stenosis, in-stent thrombosis). The assessments can later be merged if desired. The patterns can occur in combination with each other.

Braid collapse can be observed in the case of negative remodeling constraining the mesh of the implanted FD (FD and artery lumen with same diameter) or with neointimal lining after vascular injury or due to foreign body reaction to the metal struts (FD diameter larger than artery lumen).³⁰

Limitations

Due to the lack of any published literature, many of our recommendations are based on expert opinions only. The authors aimed to create a common language and to define standards for clinical studies in FDs where they were lacking. We hope these standards will undergo much scrutiny and, based on that, will be revised several times. Our suggestions for the reading processes and their documentation are based on few FDA documents.²⁶ In the absence of substantial ‘core laboratory research’, the other recommendations represent expert opinions only.

Another limitation is that we do not fully understand the various underlying mechanisms for braid deformation and how this relates to device design and material. These challenges impact grading and post-procedure management/imaging.

Conclusions

We propose a set of standards for evaluating and reporting radiological outcomes of FD treatment of intracranial aneurysms, including several aspects of imaging analyses that were previously not fully addressed.

The widespread adoption of these standardized methods could enhance accuracy and precision in assessing treatment outcomes. Such standardization is pivotal not only for ensuring consistent communication of results among neurointerventional specialists but also for catalyzing focused research and development of FD technologies.