Article Text

Download PDFPDF

Original research
Voxel based analysis of recurrence dynamics in intracranial aneurysms after coiling
  1. Marielle Ernst1,
  2. Anika Buchholz2,
  3. Romain Bourcier3,
  4. Hubert Desal3,
  5. Pierre-Yves Le Floch3,
  6. Markus Möhlenbruch4,
  7. Martin Bendszus4,
  8. Jens Fiehler1
  1. 1Department of Diagnostic and Interventional Neuroradiology, University Medical Centre Hamburg-Eppendorf, Hamburg, Germany
  2. 2Department of Medical Biometry and Epidemiology, University Medical Centre Hamburg-Eppendorf, Hamburg, Germany
  3. 3Service de Neuroradiologie Diagnostique et Interventionnelle, Centre Hospitalier Universitaire de Nantes, Nantes, France
  4. 4Department of Neuroradiology, Heidelberg University Hospital, Heidelberg, Germany
  1. Correspondence to Dr Marielle Ernst, University Medical Centre, Hamburg-Eppendorf Haus Ost 22 (O 22), Martinistr 5220246, Hamburg, Germany; m.ernst{at}


Background Understanding aneurysm growth is critical for the appropriate follow-up of patients after coil embolization and the need for retreatment. The purpose of the study was to stratify the growth dynamics of aneurysm recurrences after coiling by volumetric analysis and to determine predictive factors for aneurysm recurrences.

Methods Source images of follow-up three-dimensional time of flight MR angiography (ToF-MRA) scans were compared with the first post-interventional ToF-MRA scan and analyzed for residual flow after co-registration using ANALYZE-software. In the event of incomplete occlusion, the residual volume was segmented and calculated. Growth dynamic was determined for each aneurysm after embolization.

Results We analyzed 326 patients with 345 aneurysms from two centers. Each case had at least two ToF-MRA examinations after endovascular therapy. The mean observation interval was 59 months. Volumetric analysis of 1139 follow-up MRAs revealed that 218/345 aneurysms (63.2%) showed complete occlusion on initial follow-up imaging, and of these 95.0% remained stable. A steady increase in intra-aneurysmal flow was observed in 83/345 (24.1%). Less frequent observations were a steep increase (21/345; 6.1%) and a decrease (27/345; 7.8%). Independent predictors of increasing residual flow were greatest aneurysm diameter, total coil length, and incomplete occlusion.

Conclusions Volumetric analysis of registered three-dimensional ToF-MRA follow-up datasets allows the detection of different growth patterns with high precision, avoids the low inter-rater reliability, and represents a promising approach for future studies that include analysis of more complex predictors of residual flow. In cases of aneurysm recurrence after coiling, the major pattern seems to be a steady increase in intra-aneurysmal flow over several months.

  • aneurysm
  • coil
  • magnetic resonance angiography

Statistics from


It is still a matter of debate how often and for how long a coiled aneurysm of an individual patient needs to be followed and which imaging method is the most adequate. Understanding the dynamics of aneurysm growth and defining reliable parameters that help to offer the best follow-up care of patients after coil embolization is desirable. A longitudinal study of aneurysm growth demonstrated the feasibility of assessing aneurysm volume by magnetic resonance angiography (MRA) with computational 3D coregistration compared to visual assessment.1 Recently, aneurysm recurrence volumetry revealed promising results for determining the efficacy of therapy and retreatment decisions.2–4

The purpose of our study was to assess growth dynamics of aneurysm recurrences quantitatively by volumetric analysis of co-registered three-dimensional time of flight (ToF) MRA follow-up datasets and to determine factors predicting the occurrence and dynamics of residual flow. We hypothesized that there are different patterns of steep and steady increases and decreases in residual/recurrent aneurysms, that large recurrences are always early, and are predictable by location and size of the treated aneurysm. Such different behaviors might reflect different biological subgroups with different therapeutic implications.


Patient selection

We identified 388 consecutive patients with 407 intracranial aneurysms that had at least two ToF-MRA follow-up scans with identical imaging parameters between January 2000 and February 2014 from two centers (Centre Hospitalier Universitaire de Nantes, France and Heidelberg University Hospital, Germany). Twenty-five aneurysms were excluded because of early retreatment before the second follow-up, 16 aneurysms were excluded because the follow-up scan was of inadequate quality not allowing evaluation of aneurysm occlusion, and 21 aneurysms were excluded because of missing data. Overall, 1139 follow-up MRAs of 326 patients with 345 aneurysms were analyzed. In patients that were retreated during follow-up (n=15), residual volume was assessed on all of the last follow-up scans before retreatment.

The study was approved by the local ethics committee (Ethik-Kommission Ärztekammer Hamburg WF-040/13) and the requirement for written informed consent was waived.

MRA follow-up protocol

MR imaging examinations in Nantes were performed on a 1.5 T scanner (Magnetom Sonata; Siemens, Erlangen, Germany) eight channel coil, using a three axial slab sequence, TR=37, TE=4.69, and a flip angle of 25° with an FOV of 240 mm and an imaging matrix of 512×192×192 (0.9×0.5×0.8 mm3 voxels). In Heidelberg, images were acquired during the clinical workup using a 3 T scanner (Magnetom Tim Trio or Verio (identical technical parameters); Siemens Healthcare) with a 12 channel head matrix coil. ToF-MRA was acquired using a three-dimensional sequence with the following parameters: TR=26, TE=4.72, flip angle 15°, FOV 200×150 mm, slice thickness 0.6 mm, and pixel spacing 0.3.

Image analysis

Image analysis was performed in an independent core laboratory blinded to all patient information and unaware of treatment methods and treatment results, or other imaging examinations. For the quantitative analysis of reperfusion in embolized aneurysms, source images of each follow-up ToF-MRA scan were compared with the first post-interventional ToF-MRA scan and analyzed for residual flow after co-registration using ANALYZE-software (see figure 1). In the case of incomplete occlusion, recurrence volume was segmented slice by slice and calculated. Growth dynamic was determined for each aneurysm after embolization. Each aneurysm was assigned to one or more of the following arbitrary categories:

Figure 1

Co-registration of three-dimensional time of flight MR angiography follow-up datasets for volumetric analysis.

  • steep increase, defined as an increase >0.05 µL per day

  • steady increase, defined as an increase of 0.01-0.05 µL per day

  • steep decrease, defined as a decrease >0.05 µL per day

  • steady decrease, defined as a decrease of 0.01-0.05 µL per day

  • stable, defined as an increase or decrease < 0.01 µL per day from first to last MRA

  • increase or decrease of >10 µL

  • increase later than 48 months after treatment

  • a phase with decrease and later increase.

Statistical analysis

The influence of predictors on aneurysm growth (dichotomized to increase yes/no) was analyzed by multiple logistic regression. Potential predictors considered for model building were age, sex, greatest diameter of aneurysm, complete aneurysm occlusion on initial follow-up scan, multiple aneurysms, localization, subarachnoid hemorrhage, total length of coils, treatment with balloon, treatment with stent, and the number of coils. With regard to location, aneurysms were assigned to one of six subgroups for the regression analysis: anterior cerebral artery (ACA), anterior communicating artery, middle cerebral artery (MCA), posterior communicating arteries and internal carotid artery (ICA), basilar tip, and posterior (vertebral artery, posterior inferior cerebellar artery, superior cerebellar artery, and basilar artery). Influential covariates were selected by backward elimination, starting with the ‘full’ model containing all of the above mentioned potential predictors. Covariates were removed in an iterative fashion if they were not significant (at level 0.05; Wald p value) and not a confounder (defined as a ≥15% change in any of the remaining coefficients as compared to the ‘full’ model). Predicted marginal probabilities for increase in residual flow (as mean over aneurysms) are given with 95% CI. Statistical analysis was performed using STATA V.13 (STATA Corp., Texas, USA).


Patient and aneurysm characteristics

Overall, 345 aneurysms in 326 patients that had at least two ToF-MRA examinations (mean 3) after endovascular therapy were analyzed. The mean observation interval was 1806 days after treatment (range 104–4519). Fifteen (4.4%) aneurysms were retreated during the observation period (mean after 3 years). The patient and aneurysm characteristics are displayed in the online supplementary file 1. 302 (87.5%) aneurysms were located in the anterior circulation, 233 (67.5%) aneurysms were ruptured and the mean greatest diameter was 5.3 mm (SD 2.8 mm). In mean 4.5 (SD 2.6) coils were used for embolization including the following coil types: GDC Detachable Coils, Matrix2 Detachable Coils (Stryker, USA); Trufill DCS Orbit, Presidio, Micrusphere, Helipaq, Ultipaq, Deltapaq (Codman Neurovascular, Raynham, MA, USA). 56 (16.2%) aneurysms were treated with balloon and 25 (7.2%) aneurysms stent assisted, there of one aneurysm with a Low-profile Visualized Intraluminal Support device (LVIS, Microvention, Tustin, CA, USA) and 24 with Enterprise Stents (Cordis Neurovascular, Miami, FL, USA).

Supplementary file 1

Volumetric analysis

Of 345 aneurysms, 218 (63.2%) showed complete occlusion on initial follow-up imaging, and of these 95.0% remained stable (figure 2). A steady decrease was observed in 25 aneurysms (7.2%), a steep decrease in 2 aneurysms (0.6%). Sixteen aneurysms (4.6%) developed a decrease of >10 µL. A phase with decrease and later increase was observed in 22 patients (6.4%). Eighty-three aneurysms (24.1%) developed a steady increase of residual flow. Their mean residual volume on initial follow-up scan was 44.1 µL. Twenty-one aneurysms (6.1%) developed a steep increase and showed at the first follow-up scan a mean residual volume of 88.8 µL. Two of these aneurysms were totally occluded at initial follow-up scan and recurrent flow was detected at the second follow-up. Twenty-five aneurysms showed a steep increase at second follow-up, there of 7 continued to show steep increase, while 2 remained stable and 8 showed steady increase; four of them were totally occluded on initial follow-up-scan. Eleven aneurysms showed steep increase not before the third follow-up scan. Later than 48 months, an increase was observed in 53 aneurysms (15.4%), while no increase was observed for 151 (43.8%) of aneurysms (for the remaining aneurysms, follow-up was shorter than 48 months). The longest period of stable recurrent flow before steep increase was at least 287 days after endovascular treatment in a patient with a Carotid-T-aneurysm. In one patient with a MCA aneurysm a stable residual volume of 180 µL was observed for at least 904 days after endovascular treatment, then a steep increase to 200 µL was proven on the fourth follow-up scan 1064 days after treatment.The greatest residual volume observed in the follow-up period was, on average, highest in aneurysms of the basilar tip with 76.4 µL, followed by aneurysms of the carotid artery with 45.1 µL. The lowest mean maximum residual volume was observed in aneurysms of the ACA with 15.1 µL. Seventy-four aneurysms (21.4%) showed an increase of >10 µL from first to last MRA. In 38 aneurysms the residual volume was >30 µL compared to the initial follow-up scan, thereof 22 aneurysms showed a steep increase. In 18 aneurysms a recurrence volume of >50µL occurred. In 12 aneurysms a recurrence volume of >50µL was detected at 2nd follow-up, in three aneurysms at 3rd follow-up and in three aneurysms at 4th follow-up; in 5 aneurysms within one year, in 7 aneurysms two years after treatment.

Figure 2

Flow diagram showing the exclusion of patients and the distribution of growth patterns.

Predictors of increase in residual flow in embolized aneurysms

Multiple logistic regression analysis showed that greatest diameter of aneurysm, shorter total length of coil, and incomplete occlusion at the initial follow-up imaging were independent predictors of increase in residual flow in embolized aneurysms (see online supplementary file 1). Age, sex, subarachnoid hemorrhage, aneurysm location, multiple aneurysms, and use of a balloon were identified as confounders and thus remained in the model for adjustment purposes. Patients with complete occlusion were significantly less likely to develop increase in residual flow compared with patients with incomplete occlusion (OR=0.02; 95% CI 0.01, 0.04; p<0.001). Model based predicted marginal probability of increase in residual flow was 6.3% (95% CI 3.2%, 9.4%) and 70.6% (95% CI 62.2%, 79.0%) for patients with and without complete occlusion, respectively (figure 3, top left). Per 1 mm increase in greatest diameter, the odds of increase in residual flow increased by 50% (OR=1.50; 95% CI 1.17, 1.93; p=0.002). Thus the predicted marginal probability of increase in residual flow was, for example, 16.7% (95% CI 8.6%, 24.7%) for a diameter of 2 mm, 32.0% (95% CI 28.3%, 35.7%) for 6 mm, and 50.7% (95% CI 34.0%, 67.3%) for a diameter of 10 mm (figure 3, top right). Patients with larger total coil length of 10 cm were less likely to develop an increase in residual flow (OR=0.89 per 10 cm increase in length; 95% CI 0.82, 0.97; p=0.006).

Figure 3

Predicted marginal probabilities of increase in residual flow for the three independent predictors complete occlusion (A), greatest diameter of aneurysm (B), and total length of coil (C).


Aneurysm recurrence volumetry provides a quantitative, objective, and highly sensitive approach for the understanding of growth dynamics of intracranial aneurysms after embolization using quantitative assessment of recurrence volumes in registered three-dimensional ToF-MRA. As an important strength, our study analyzed a rare patient collective with identical imaging parameters over a long follow-up period. This approach reduced the false estimation of recurrent flow due to varying imaging parameters and allowed a precise comparison of aneurysm occlusion over time. We observed no change in 62.0%, a steady increase in 24.1%, and a steep increase in 6.1% of aneurysms. A steady or steep volume decrease (7.2% and 0.6%) was much rarer. Later than 48 months, an increase was observed in 53 aneurysms (15.4%), a phase with a decrease and a later increase was observed in 22 patients (6.4%). In previous studies evaluating the stability of anatomic occlusion of aneurysms, the results of coil occlusion were classified using a nominal classification scale.5

As the degree of occlusion after treatment is a continuous variable, the use of discrete rating scales and, in particular, dichotomized scales, leads to high imprecision, being unable to distinguishing aneurysm conditions that may have a higher risk of growth or bleeding. Rezek et al observed that unfavorable angiographic appearance was noted almost twice as frequently by an independent core laboratory compared with the treating operators.6 In contrast, volumetric analysis of co-registered three-dimensional ToF-MRA follow-up datasets in a core laboratory, as in our study, represents an objective, rater independent method. According to Schönfeld et al, this approach leads to higher reliability compared with visual evaluation.4 Thus volume quantification based on ToF-MRA image sequences allows earlier estimation of safety and efficacy of new materials compared with studies requiring more patients to reduce the variability due to visual evaluation.2 By limiting the variability in the evaluation of aneurysm occlusion grade, the continuous surrogate endpoint, recurrence volume, improves statistical power and can substantially decrease sample size in future clinical trials.7

A recent meta-analysis comparing ToF-MRA and contrast enhanced (CE) MRA with DSA for follow-up of coiled aneurysms revealed a high diagnostic performance for both MRA techniques.8 Nonetheless, the quantification of small residual volumes is difficult and the sensitivity of visual evaluation of 3D ToF-MRA compared with DSA for detection of reperfusion or residual perfusion of coiled intracranial aneurysms was found to be lower for aneurysms ≤5 mm (72.2%) and ≤3 mm (63.6%).9 However, volumetric analysis of registered 3D-ToF-MRA follow-up data sets was found to be highly sensitive and significantly more reliable than visual inspection in the detection of aneurysm recurrences.4 Allowing the detection of even smaller recurrence volumes compared to visual evaluation of 3D ToF-MRA, it presents a promising approach, though the validation against a reference standard is still missing. One difficulty in the evaluation of the accuracy of MRA in detecting residual flow in comparison with DSA derives from the lack of a reference standard for DSA, such as histopathologic confirmation. Thus false positive findings seen on DSA due to artifacts can result in false negative values for MRA and a lower sensitivity. The sensitivity for detection of residual neck or residual aneurysm was found to be lower in contrast enhanced MRA compared with ToF-MRA.8 This might be due to inadequate contrast timing leading to venous enhancement and vessel overlaps that degrade image quality, preventing an accurate delineation of residual flow. Additionally, the short time window between the arterial and venous phases lowers spatial resolution.10–12 In a large prospective study, three-dimensional ToF-MRA led to better visibility of coils and fewer artifacts.13 Moreover, contrast medium administration has the risk of allergic reactions, renal damage, and patient discomfort.14 In our study the field strength was 3 T in one center and 1.5 T in the other center. Incremental improvements in the detection of aneurysm recurrences may be achieved by imaging at 3 T.15 However, up to now, there has been a lack of evidence, and in the only study comparing both techniques directly, ToF-MRA at 3 T was not observed to be significantly better.16 With regard to Magnetic resonance artefacts from coil constructs, especially Platinum–iridium alloys in coils were shown to cause significant magnetic resonance susceptibility artefacts, leading to loss of perianeurysmal anatomical data.17 Thus, MRA images of intracranial aneurysms treated with Nexus coils showed significantly more severe artifacts than those of intracranial aneurysms treated with bare platinum coils.18 Moreover, Barricade coils were reported to cause MR susceptibility artefacts making it impossible to ascertain aneurysm closure and anatomy.19 None of these coil types were used in our study. 25 aneurysms were treated stent assisted, thereof one aneurysm with an LVIS and 24 with an Enterprise stent, a self-expandable nitinol stent. The standard imaging protocol following endovascular treatment included a DSA in addition to an MRA at 6 months post coiling. If there was good correlation between the two studies and no significant artefact, further follow-up was performed with MRA alone. Otherwise DSA was continued. Nitinol does not appear to cause significant artifacts in MR images. In an evaluation of stented intracranial arteries, MRA after the placement of Nitinol stents depicted the intracranial artery lumen without distortion or signal loss, whereas stainless-steel stents caused segmental signal loss.20

Our results are remarkably similar to the findings of previous studies, based on visual rating. In our study, 218 aneurysms (63.2%) showed complete occlusion on the initial follow-up imaging. In a recent systematic review on initial occlusion and reopening, complete initial occlusion was reported in 4355 aneurysms (62.3%; 95% CI 61.2%, 63.4%).21 The International Subarachnoid Aneurysm Trial (ISAT) trial observed 66% complete angiographic occlusion.22 This important subgroup of patients with completely occluded aneurysms on the initial follow-up remained stable in 95.0% of cases in our study. This again is in accordance with a large multicenter study observing late reopening in 11 of 440 aneurysms (2.5%; 95% CI 1.0, 4.0) that showed adequate aneurysm occlusion (complete occlusion or only a small neck remnant) at the 6 month angiographic follow-up.23 Some authors suggest that patients with aneurysms adequately occluded on the 6 month angiographic follow-up do not need further imaging follow-up in the first 5–10 years after endovascular treatment as the recurrence rate is low.23-25 In our study, 12 of 218 aneurysms (5.5%) that were initially completely occluded showed reopening. In eight aneurysms, recurrence was detected at the 6 month follow-up. In the remaining four aneurysms, reopening occurred after 6 months; none of these patients was retreated. However, there are subgroups of patients that might benefit from screening beyond 5 years, as higher reopening rates were observed for partially thrombosed aneurysms,26 large and giant aneurysms,21 23 27 posterior circulation location,21 and wide neck aneurysms.21

In a multivariate analysis, Ferns et al found that the only independent risk factors for late reopening were aneurysm size ≥10 mm (OR 4.7; 95% CI 1.3, 16.3) and location on the basilar tip (OR 3.9; 95% CI 1.1, 14.6).23 In our analysis, independent predictors of increase in residual flow were incomplete occlusion, greatest diameter of aneurysm, and shorter total length of coil. Compared with MCA aneurysms, aneurysms in the posterior subgroup have a lower odd of increase, while aneurysms in the anterior communicating artery, ACA, internal carotid artery/posterior communicating artery, and basilar tip have higher odds. Yet, neither of these effects was significant. The lack of significance might be due to the low number of aneurysms in each subgroup.

In contrast, there are also studies which could not find any relation between recurrence and simple patient and aneurysm related characteristics, such as location23 or neck width.23 In view of the large number of risk factors and contradictory results, the real underlying causes and mechanisms of aneurysm recurrence after endovascular coil embolization might be more complex and not yet sufficiently understood. Given the lack of robust data, further volumetric analyses are needed to analyze more complex aneurysm related characteristics, such as aneurysm morphology and shear wall stress, that allow prediction of the occurrence and the dynamics of residual flow. This knowledge might offer an individual design of the most adequate follow-up care for every patient after coil embolization.

Definition and risk of growth

To date there has been no definition of how much growth or decrease is enough to term an aneurysm ‘unstable’. Our definition of steep increase as an increase higher than the mean value of >0.05 µL might differ in other study populations. Assessment of growth dynamics is influenced by the precision with which follow-up measurements can be compared with earlier imaging studies. The semiautomated assessment of residual flow allows high precision, and our definition of any measured change in residual volume as ‘unstable’ leads to the same proportion of stable and unstable aneurysms as in previous studies. The number of aneurysms with a steep increase in residual flow (n=24) or steep (n=2) or steady (n=25) decrease was too small to allow more precise analysis of the magnitude of change in residual flow. Differentiation between steep and steady change is dependent on the time point of evaluation. Thus 15 aneurysms showed a steep increase during follow-up but their growth dynamic was finally judged a steady increase with regard to the whole follow-up period. Therefore, the different follow-up length of individual patients is a further limitation of the study and might have influenced the judgement of growth dynamics. Moreover, 15 aneurysms were retreated during the follow-up period and might have changed their growth dynamics. Thus the proportion of aneurysms with a steep increase or decrease in residual flow might be higher. An important source of bias is the exclusion of early retreated aneurysms, leading to an underestimation of the rate of patients with a steep increase.


Our study confirms the stability of endovascular treatment for cerebral aneurysms. Two-thirds of treated aneurysms were completely occluded at the first follow-up imaging, and of these 95.0% remained stable in later follow-up. A quarter of the aneurysms showed a steady increase in recurrence volume during follow-up, whereas a decrease and an undulant dynamic were rare observations. Volumetric analysis of registered three-dimensional ToF-MRA follow-up datasets allows the detection of different growth patterns with high precision, avoids low inter-rater reliability, and thus represents a promising approach for future studies analyzing more complex aneurysm and patient related predictors of the occurrence and dynamics of residual flow.



  • Contributors ME designed the data collection tools, monitored the data collection for the whole trial, wrote the statistical analysis plan, cleaned and analyzed the data, and drafted and revised the paper. She is the guarantor. AB wrote the statistical analysis plan, and revised the draft paper. RB, P-YL, and MB revised the draft paper. JF initiated the collaborative project, designed the data collection tools, monitored the data collection for the whole trial, analyzed the data, and drafted and revised the paper. HD and MM designed the data collection tools, monitored the data collection for the whole trial, analyzed the data, and drafted and revised the paper.

  • Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

  • Competing interests None declared.

  • Ethics approval The study was approved by Ethik-Kommission Ärztekammer Hamburg WF-040/13.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement There are no unpublished data to share.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.