Imaging of intracranial aneurysms in animals: a systematic review of modalities

Intracranial aneurysm (IA) animal models are paramount to study IA pathophysiology and to test new endovascular treatments. A number of in vivo imaging modalities are available to characterize IAs at different stages of development in these animal models. This review describes existing in vivo imaging techniques used so far to visualize IAs in animal models. We systematically searched for studies containing in vivo imaging of induced IAs in animal models in PubMed and SPIE Digital library databases between 1 January 1945 and 13 July 2022. A total of 170 studies were retrieved and reviewed in detail, and information on the IA animal model, the objective of the study, and the imaging modality used was collected. A variety of methods to surgically construct or endogenously induce IAs in animals were identified, and 88% of the reviewed studies used surgical methods. The large majority of IA imaging in animals was performed for 4 reasons: basic research for IA models, testing of new IA treatment modalities, research on IA in vivo imaging of IAs, and research on IA pathophysiology. Six different imaging techniques were identified: conventional catheter angiography, computed tomography angiography, magnetic resonance angiography, hemodynamic imaging, optical coherence tomography, and fluorescence imaging. This review presents and discusses the advantages and disadvantages of all in vivo IA imaging techniques used in animal models to help future IA studies finding the most appropriate IA imaging modality and animal model to answer their research question.


Introduction
Intracranial aneurysm (IA) is an arterial disease resulting in abnormal enlargement of the vessel lumen.IAs generally form at bifurcations of intracranial arteries in the circle of Willis, an arterial network that supplies blood to the brain [1,2].IAs affect 2 to 5% of the population and are mostly asymptomatic [3,4].However, unstable IAs can rupture causing subarachnoid hemorrhage (SAH) that is fatal in 25-50% of cases [5].Moreover, 35% of the patients that survive SAH suffer from long-term sequelae such as physical or cognitive disabilities impairing their quality of life [6].To reduce this risk, the decision to secure the aneurysm by surgical clipping, endovascular coiling, and/or flow diversion could be taken [5].In vivo imaging techniques are of paramount importance in the management of IAs.Indeed, high-resolution imaging is needed to assess precisely size and morphology of the IA (presence of blebs, lobules, rough aspect), which is necessary to evaluate the risk of rupture of an IA at its discovery and during follow-up imaging [7,8].IAs of large diameter or displaying irregular vessel walls have been linked to an increased risk of rupture.These features are commonly used in the different IA risk scoring systems, even if the biological processes leading to IA rupture is still unknown [9][10][11].Longitudinal studies to understand the pathophysiology underlying the relation between these morphological features and the increased risk of rupture 56 Page 2 of 29 require research on IA animal models.Such models also allow for the testing of new endovascular treatments.
In this context, various IA animal models have been established.The first induced IAs in animals were surgically constructed to mimic IAs in human cerebral arteries [12].Already in 1954, German and Black [13] induced IAs by grafting a vein-pouch into the common carotid artery (CCA) of dogs.Thereafter, different venous pouch models allowing for the formation of aneurysms with adaptable sizes, at different locations, and with various shapes have been used in several species [12].Altes et al. [14] developed a rabbit model of IAs located at the origin of the right CCA using elastase incubation and ligation of the right CCA.Later, endogenous IA animal models mimicking the human disease helped to better understand IA pathogenesis and development [15].Different risk factors for the formation of IAs have been identified and can be used in animals to endogenously induce IAs.Hemodynamic stress in combination with other vascular risk factors, such as hypertension, is known to be involved in the initiation of IA formation [16].One component of hemodynamic stress is wall shear stress (WSS), which is the drag force exerted by blood flow onto the endothelium, the innermost layer of the vessel wall.High and low WSS seem both involved in the progression or growth of IAs [2].Connective tissue disorders like the Ehlers-Danlos syndrome and the Marfan syndrome also put patients more at risk to develop IAs [1].Thus, chemical compounds that weaken connective tissue extracellular matrix components such as β-aminopropionitrile (BAPN) and elastase are used in animals to favor IA formation [15].These endogenous IA animal models mimic arterial wall modifications that characterize the human disease, such as loss of the internal elastic lamina, loss of endothelial and smooth muscle cells, as well as inflammatory cell infiltration [16].
To study the size and morphology at different stages of IA development in these surgical and endogenous animal models, in vivo IA imaging is necessary.Like in human, angiography is used in animals to image the IA and, more particularly, to assess the patency of induced IAs over time or after endovascular treatment [17].The same imaging modalities have been used in human and animal models; however, IA imaging in animal models is challenging, as many induced IAs are smaller than human IAs.[15].Digital subtraction angiography (DSA) remains the gold standard imaging technique, but progress in imaging technologies increased image resolution, and many different imaging modalities are currently available to image IAs in vivo in animals [18].We performed a systematic review to list all imaging techniques used in IA animal models to help researchers finding the most appropriate IA imaging modality to answer their research question.In this review, we first describe briefly all existing in vivo imaging techniques used for IA visualization in animal studies.Then, we critically compare the different imaging modalities and discuss their advantages and disadvantages.

Search strategy
We systematically searched for studies in PubMed between 1 January 1945 and 13 July 2022 containing in vivo imaging of IA in animal models.We used the combination of the following Medical Subject Headings (MeSH) terms: "intracranial aneurysm" AND "animal models" AND ("diagnostic imaging" OR "diagnostic technique, cardiovascular") and excluded reviews.A hand search in the PubMed database and in the SPIE Digital library was also performed to find studies not found with the MeSH terms cited above.Then, potentially eligible studies were screened and included or excluded from this review following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [19] (Fig. 1).Included studies were carefully examined, and information on the objective(s) of the study, the IA animal model(s), and the imaging modality(ies) used was collected (Table 1).

Eligibility criteria
Exclusion criteria were defined by the objective of this study to review only articles containing in vivo IA imaging in animal models.Articles that did comprise the following criteria were not included in this review: (1) no in vivo imaging, (2) no IA created, and (3) results previously reported.Studies in which IA imaging was performed after sacrifice of the animal were also not included as this review focuses on in vivo imaging modalities that can be used in living animals.However, animal models using extracranial aneurysms as an IA animal model were included in this review as they mimic human IAs in terms of aneurysm size and vessel diameter.

Results
A total of 178 potentially eligible studies were found in PubMed using MeSH terms and 39 after the hand search (Fig. 1).Moreover, 1 additional potentially eligible study was found on SPIE Digital library.After the screening of the title and the abstract of these 218 studies, 3 articles were excluded as they were not original studies.Full texts from 2 studies were not available and had to be excluded for this reason.Following the exclusion criteria defined above, 43 articles were withdrawn from the review, yielding 170 studies included in this review (Table 1).

IA animal models
IA animal models are required to test new endovascular devices and to better understand IA pathophysiology.However, spontaneous endogenous cerebral aneurysms are extremely rare in animals [187].Consequently, many techniques to induce IAs in various animal species were successfully established.Large animal models like swines and dogs are well-characterized IA animal models with an easy access for diagnostic and IA treatment; however, they are expensive models [188].Rabbit IA models are also well characterized and commonly used as their carotid artery size is comparable to human cerebral arteries [189].Unfortunately, rabbits have a relatively high perioperative morbidity and mortality [188].Finally, the use of small rodents, like mice and rats, allow for lower study costs, but their arteries are much smaller than human cerebral arteries, making surgery and imaging more difficult and not adapted to human endovascular techniques [188].
Different techniques to surgically construct or endogenously induce IAs in animal models are available.In the first category, IAs are surgically constructed using venous or arterial pouch grafting (Fig. 2A, right side) or using artery ligation in combination with vessel wall weakening using elastase like in the frequently used rabbit elastase model [188] (Fig. 2A, left side).In the second category of endogenously induced IA animal models, animals are exposed to known IA risk factors such as hemodynamic stress or connective tissue weakening [15] (Fig. 2B).Hemodynamic stress can be increased by unilateral or bilateral CCA ligation or by the creation of a new bifurcation between 2 arteries.Therefore, hemodynamic stress can be increased in specific extracranial arteries or in intracranial arteries of the circle of Willis.Hemodynamic stress can also be increased by inducing hypertension by ligation of one renal artery (RA), with or without high salt diet or deoxycorticosterone acetate (DOCA) pellet implantation.In addition, the vessel wall can be weakened by elastase injection or inclusion of BAPN in the food of the animals.Furthermore, genetic modifications have been occasionally used in combination with other manipulations to induce IAs [15].The vast majority of the studies used saccular IA models, and only 2% of the studies used an animal model of fusiform IAs (Table 1).
Eighty-eight percent of the reviewed studies used surgical methods (Table 1).Indeed, in most of the studies, IAs are surgically induced in large animals like rabbits or dogs that are relatively simple to image (Fig. 2C).In contrast, only 12% of the reviewed articles used methods in which IAs are endogenously formed.Actually, endogenous IA models mainly use small animals like rats (Fig. 2D) making IA imaging more difficult due to the very small size of the induced lesions [8].

Purpose of in vivo imaging of IAs in animal models
The large majority of IA imaging in animals was performed in the reviewed studies for 4 reasons: basic research for IA  1 and Fig. 3A).The objectives of 3 studies was classified as "other" as they could not be categorized in the above classes.The vast majority of the studies included in this review focused on testing new IA treatment options (Fig. 3A).Indeed, imaging is essential during coiling or flow diverter implantation and later to check the effectiveness of the treatment over time.
IA imaging is also instrumental in research on IA animal models as it allows for the visualization and characterization of IA morphology and IA patency surveillance over time.Moreover, in vivo imaging of IAs also helps to understand IA pathophysiology better as it allows to follow IA size and morphology from IA initiation to rupture, for instance.Thus, in vivo imaging of IAs is crucial in animal models.Actually, 17% of reviewed articles explicitly aimed at research on IA imaging itself.

Description of the IA in vivo imaging techniques
The different imaging modalities found in the reviewed articles were classified in 6 different imaging techniques: conventional catheter angiography, computed tomography angiography (CTA), magnetic resonance angiography (MRA), hemodynamic imaging, optical coherence tomography (OCT), and fluorescence imaging.Pros and cons of these imaging modalities have been highlighted (Table 2).These advantages and disadvantages are likely responsible for the specific distribution of the imaging modalities among the different animal species, with conventional catheter angiography being most frequently used in larger animals, and a preference for MRA was found in smaller animals (Fig. 3B). •

Conventional catheter angiography
Catheter angiography is the conventional angiography method that uses X-ray and injection of a nonionic iodinated contrast agent, usually through an arterial catheter, to image blood vessels [190].In 1954, German and Black [13] were the first to image a constructed IA in the CCA of dogs, using 2D catheter angiography.Until 1993, catheter angiography was the only available modality to image IAs in animal models (Fig. 3C).Imaging improvements continued [25] and led Wakhloo et al. [36] to use DSA to image IAs in CCA of dogs in 1994.DSA is a 2D catheter angiography that uses subtraction of precontrast images to obtain an image of the vasculature only [191].From then on, DSA became the gold standard to detect IAs in human [192] and in animal models [18].Actually, 84% of the reviewed studies used conventional catheter angiography (Table 1).Indeed, as mentioned before, the most frequent purpose to use in vivo imaging of IAs in animal studies is to test new endovascular treatment modalities, and 97% of these studies use catheter angiography.Conventional catheter angiography allows the correct assessment of IA occlusion following IA endovascular treatment.DSA in animal models is mainly performed through the femoral artery.However, it can also be performed through the central artery of the ear in rabbits [78] or through veins (e.g., ear vein) [18,48,50,65,71,73,91].Intra-venous DSA (i.v.DSA) is less invasive and allows for repeated imaging compared to intra-arterial DSA (i.a.DSA), which requires exposure, cutdown, and ligation of arteries making serial imaging sessions more difficult and risky [50,71].Doerfler et al. [18] showed that i.v.DSA is as precise as i.a.DSA to efficiently predict IA size and geometry in the rabbit elastase IA model.However, other studies revealed that i.v.DSA underestimates the IA dimension [112] or is insufficient to assess aneurysm occlusion after flow diverter treatment [166] compared to i.a.DSA.The authors postulated that this difference results from a decreased contrast agent aneurysm filling due to contrast agent dilution in the bloodstream and/or decreased velocity during i.v.DSA.Moreover, it should be noted that, despite its decreased invasiveness, i.v.DSA did not replace i.a.DSA in clinic because of the decreased contrast density and the vessel overlapping visualization due to the simultaneous imaging of veins and arteries [193].
3D rotational DSA (3D-DSA) images cerebral vessels from all viewpoints, thereby increasing precision for 3D geometry assessment of arteries [194].This allows for visualization of vascular modeling to create precise vascular devices [66,133].It is assumed that 3D-DSA offers the greatest resolution compared to other imaging techniques [112].However, it has been shown in human studies that DSA displays a better resolution of small vessels, which allows a greater sensitivity in small IAs than with 3D-DSA [192].
As it provides an very high resolution allowing the visualization of small cerebral vessels (< 1 mm), e.g., the anterior choroidal artery [195], DSA continues to be the gold standard [7,18].However, DSA remains an invasive method using radiation and injection of contrast agent through a catheter and is associated with a complication rate of 0.04-0.30% in humans [194].Therefore, less invasive imaging methods represent a safe alternative in human and animal models [18].• Computed tomography angiography CTA uses X-ray and contrast agent injection through a venous catheter to image the vasculature and allows for a 60% decrease of ionizing radiation making CTA less invasive than conventional catheter angiography [7].This imaging modality uses the rotation of a CT scanner in combination with a motile patient table that allows a continuous 2D or 3D image acquisition at a higher speed than conventional catheter angiography [196,197].The first use of CTA in an IA animal model was described in 2004, and only 10 studies included in this review used CTA (Table 1 and Fig. 3C).Human studies revealed that CTA has an insufficient resolution in small IAs (diameter < 3 mm) [198].However, several studies showed that CTA is as efficient as DSA to detect IAs in animal models [18,79,80,102,142].
Different types of detectors can be used with CT: multislice detectors (MS-CTA) and flat-detectors (FD-CTA), which have been introduced later and use a smaller detector element size.Struffert et al. [102] showed that CT with both detectors allowed for measurement of similar IA dimensions in the rabbit elastase model, but images seem to be better delineated using FD-CTA due to a higher spatial resolution.
A major limitation of CTA in human is the presence of artifacts when clips, stent, or coils are used [192].Yet, Dudeck et al. [79] did not observe such effects in CCA aneurysms constructed in swine and coiled with a liquid embolic agent.Moreover, Ott et al. • Magnetic resonance angiography Magnetic resonance imaging allows for less invasive 2D and 3D angiography using powerful magnetic fields without ionizing radiation nor iodine-based contrast agent injection [199].This less invasive imaging modality makes MRA an optimal instrument for serial imaging in long-term studies.Indeed, safe serial imaging in long-term studies using MRA was performed in dogs [113], mouse [143,156,174], rats [96,136,137,181], and rabbit elastase IA models [18,169].Several studies in human suggest that MRA could be considered equivalent to DSA to detect IAs [7].However, it has also been shown that the resolution was insufficient to detect small IAs (< 3 mm) [7].Sixteen percent of the animal studies included in this review used MRA (Table 1), and this imaging modality was more often used in the last 10-15 years (Fig. 3C).In 1996, Kirse et al. [38] were the first to image IAs in a surgical rat model using MRA.In this study, however, DSA showed a better resolution than MRA.Different MRA methods are available: time of flight MRA (TOF-MRA), phase contrast MRA (PC-MRA), or contrast-enhanced MRA after venous contrast agent injection (CE-MRA) [200].Krings et al. [56] imaged IAs in 5 rabbits using the elastase model with the 3 afore-mentioned MRA methods.They observed that, in contrast to CE-MRA, TOF-MRA and PC-MRA were not sufficient to detect all constructed IAs and that the gold standard DSA detected all IAs.The authors postulated that this can be explained by the induced turbulent blood flow which results in signal loss in TOF-MRA and PC-MRA.This effect is overcome in CE-MRA, which uses contrast agents and images vasculature regardless of blood flow.Another study confirmed that CE-MRA is as good as DSA to detect IAs in animal models [59].
Paramagnetic objects, such as coil, may disturb the magnetic field and therefore create artifacts on MRA images.Therefore, Spilberg et al. [113] evaluated the signal overestimation, i.e., the created artifact, of CCA IAs in dogs using CE-MRA during 28 weeks after coiling.They described a gradual decay of the signal overestimation until 4 weeks post-surgery to reach a 25% decrease.Moreover, as for CTA, Dudeck et al. [79] did not report any artifact when imaging swine CCA IAs coiled with a liquid embolic agent.
MRA lacks resolution with endogenous IAs especially in small animals like rats or mice.Therefore, few studies using endogenous IA animal models used MRA (Table 1).In 2015, using MRA, Makino et al. [143] were able to detect a large aneurysm induced in a mouse cerebral artery after elastase injection in the cerebrospinal fluid.However, because of their small size, most endogenous IAs are impossible to image with MRA despite huge improvement in MRI technology with the development of 7 T or even 9.4 T MRI.Thus, most studies use MRA not to image IAs, but rather to image vascular remodeling in the circle of Willis or to determine intra-arterial hemodynamics using computational fluid dynamics (CFD) analyses [146,156,168].
Furthermore, as IAs with thin walls are associated with an increased risk of rupture, MRA has also been used to measure IA wall thickness in the rabbit venous pouch IA model [139].Unfortunately, it appeared that 3 T MRI overestimates the wall thickness and that a better resolution is needed to study differences of < 0.4 mm in wall thickness.MRI can also be used to study vessel wall enhancement (VWE) after contrast agent injection in IA animal models.Indeed, in a rabbit elastase model, VWE was observed and correlated positively with the number of inflammatory cells [169].Moreover, molecular imaging can be performed in animal models using MRA.Thus, IA wall inflammation was imaged in the rabbit elastase model after lipopolysaccharide injection using an MRI contrast agent targeting myeloperoxidases [101,138].In addition, Shimizu et al. [181] imaged ferumoxytol contrast agent accumulation in a rat IA wall with macrophage infiltration.Ferumoxytol is considered a true blood pool contrast agent and in addition can leak through permeable endothelium and is taken up by macrophages [201].With targeted MRI contrast agent, it is thus possible to visualize an excess of inflammatory cells, but whether the method is sensitive enough for other potential IA instability markers remains to be proven.In this respect, it is interesting to mention that Zhang et al. [183] used a nanoplatform (zinc and iron oxide nanoparticles targeting the platelets) to target thrombus in the rabbit elastase IA model.

• Hemodynamic imaging
Hemodynamic imaging is a functional method that measures active changes in hemodynamic parameters, which is fundamental in studies using IA animal models as disturbed blood flow patterns control IA pathophysiology.For instance, PC-MRA is a hemodynamic imaging modality that allows for the quantitative measurement of blood flow velocity [200] and has been used in IA animal models [56,109,113,181].Doppler ultrasonography, which is another modality that measures the velocity of flow [202], was the first imaging technique enabling the study of intraaneurysmal hemodynamics in such animal models.In 1993, Hashimoto [31] used this modality to measure blood flow velocity in a rabbit venous pouch IA model.Since then, this technique has been used in a number of IA studies to study blood flow velocity in cerebral arteries or within the IA and to check IA patency over time or after treatments [41,42,44,91,154,176,178].
Other hemodynamic imaging methods like CFD started to be used in animal models of IAs.CFD uses vessel geometry obtained with high-resolution 3D imaging to numerically simulate complex vascular hemodynamics [203].Already in 2007, Kadirvel et al. [88] simulated hemodynamic forces (e.g., WSS), using CFD from 3D-DSA images in the elastase rabbit model.Interestingly, they found a correlation between altered WSS and markers of vascular remodeling.In IA animal models, CFD has been simulated from 3D-DSA [88, 98, 106, 109, 110, 123, 131-134, 144, 146] and 3D-MRA [109,156,181].In the principle, CFD can also be generated from 3D-CT, but no studies using IA animal models were found.CFD simulations in IA animal models allowed for a better understanding of IA pathophysiology and participated Page 19 of 29 56 in research for new IA treatments.Indeed, Cebral et al. [131,134] used CFD to study hemodynamic patterns after flow diversion treatment.CFD analyses were also used to study hemodynamics in induced IAs [109,144] and even confirmed that IA hemodynamics are similar in human IAs and elastase-induced IAs in rabbits [106].Moreover, several studies highlighted correlations between dynamic changes in hemodynamics and vascular remodeling [88,91,98,110,132,133,156].
High-resolution 3D images are needed to generate CFD, which may be difficult to obtain in small animals.Therefore, it could be an option to use vascular casts created after the sacrifice of animals to re-create a precise 3D arterial geometry and therefore a precise CFD, as shown by Tutino et al., for instance [146].
• Optical coherence tomography OCT is a high-resolution and less invasive optical imaging technique that uses light produced by a vascular probe (e.g., linear scanning probes or MEMS-based probes [204]) to obtain high-resolution tomography of tissues like eyes or blood vessels [205].The near-infrared light reflects on tissue and the depth in which this reflection occurred is calculated using the delays of the back-reflected wave [205].Indeed, OCT uses an interferometer composed of a sample arm and a reference arm to measure the interference granting a high-resolution imaging modality [204].In 2005, Thorell et al. [67] used bench-TOP OCT on ex vivo dog-coiled surgical CCA aneurysms.They could easily identify the IA neck and coil pattern and obtained a good correlation between OCT images and histological findings.OCT was then used in vivo in induced IAs in dogs, rabbits, and rats (Table 1).OCT is mainly used for the evaluation of endovascular devices [161-163, 171, 173], i.e., IA recanalization following an incomplete coil occlusion, flow diverter malposition, or neointimal hyperplasia, which are important limitations of these treatment modalities.However, OCT does not allow the visualization of the IA form and size.Interestingly, Liu et al. [165] were able to observe internal and external elastic lamina disruption using OCT in elastase-induced IAs in rabbits.Moreover, Fries et al. [177] found OCT more sensitive as it allowed the detection of 18 residual aneurysms after flow diverter implantation in the rabbit elastase model as compared to DSA, which allowed the detection of 12 residual aneurysms only.More recently, Vardar et al. [182] showed the potential of high-frequency OCT (HF-OCT) in the rabbit elastase model to assess the correct IA occlusion after endovascular treatment as well as during follow-up imaging.

• Fluorescence angiography
The development of fluorescence microscopy in the beginning of the 1900s brought the possibility to observe emitted fluorescence after the excitation of a fluorophore in cultured cells or on slides [206].Furthermore, this imaging modality can also be applied in vivo to image cells and tissues in IA animal models for instance.Indeed, in 1993, Nakatani et al. [30] used fluorescent particles to visualize blood flow in an IA rat model.More recently, a transgenic rat line expressing a green fluorescent protein specifically in endothelial cells [164,179] was used to visualize the wall motion in IAs.Moreover, fluorescence angiography using fluorescein injection has been described to visualize blood flow and assess IA patency in rat and rabbit models [167,170].This imaging technique is not associated with increased mortality or morbidity and shows high contrast and sensitivity for a low-cost imaging modality.However, it is an invasive method as the artery and IA have to be dissected to be exposed to the light source.

• Combination of imaging modalities
A total of 44 reviewed articles (i.e., 26% of the reviewed articles) combined several imaging modalities (see Table 1).Obviously, a large portion of these studies used numerous modalities to compare different imaging techniques and research on in vivo imaging.However, other studies combined several imaging techniques to use the advantages of the different imaging modalities and acquire more information on the induced IAs.For instance, hemodynamic parameters or IA patency can be measured using Doppler ultrasonography and combined with conventional catheter angiography or MRA to image accurately the morphology and size of the IAs.Moreover, DSA being the gold standard IA imaging modality, 38 reviewed articles combined DSA with one or several other imaging techniques to visualize IA in animals.Of note, DSA is commonly used in animals during the surgical construction of IAs and in combination with CTA, MRA, or OCT after IA construction to obtain more detailed information on the morphology of the IA.

Purpose of in vivo imaging of IAs in animal models
The large number of articles included in this review using surgical IA animal models reveals the paramount importance of in vivo IA imaging in such models.Indeed, to surgically construct and check the correct IA patency over time, in vivo imaging is essential.As size of surgically created IAs in 56 Page 20 of 29 animals is similar to the human ones, imaging techniques used in clinical settings can be employed in these large animal models.Moreover, surgical models are mainly used to test endovascular procedures, and in vivo imaging is necessary to assess treatment efficacy.
In contrast, only 12% of the reviewed articles use endogenous IA animal models.Indeed, in vivo imaging of IAs seems to be less often used in endogenous animal models as discussed in a recent review by Tutino et al. [15].They showed that only 7% of studies on endogenous IAs in animals were combined with medical imaging.This lack of use can be partially explained by the fact that in vivo imaging is not essential in these studies.Indeed, most of them aim to better understand IA pathophysiology and not to test new endovascular treatment modalities.Therefore, in vivo imaging is not essential as they can directly observe the IA samples after animal euthanasia.Most studies report using (immuno) staining to characterize IA wall changes or observe artery bulging under a binocular microscope or scanning electron microscopy of circle of Willis casts.Furthermore, the lack of in vivo imaging in studies using endogenous models can be explained by an insufficient image resolution to visualize endogenously induced IAs of small size.Indeed, spatial resolution of the commonly available modalities to image IAs in animals is limited: DSA (< 0.5 mm [207]), CT (≈1 mm [7]), and MRA (1-2 mm [208]).Yet, in vivo imaging has greatly improved, and several high-resolution imaging modalities exist: 3D-DSA (0.15 mm [209]), high-resolution CTA (0.25 mm [8]), highresolution MRA (50 µm [156]), as well as HF-OCT (10 µm [182]).Small rodents like rats and mice, which are mostly used for endogenous IA models, are often exposed to MRA and hemodynamic imaging and less frequently to conventional catheter angiography, which is mainly used in bigger animal models (Fig. 3B).As diverse IA imaging modalities are nowadays more generally available (Fig. 3C), imaging of endogenously induced IAs has become more accessible.
Despite these limitations of in vivo imaging of IAs in small animal models, there are many good reasons to perform in vivo imaging in rodents like mice or rats.Indeed, without in vivo imaging, endogenous IAs can only be studied at the sacrifice of the animal, whereas in vivo imaging permits the observation of IA size, shape, and hemodynamics at different stages during IA development.Studies monitoring IA development require endogenous animal models because surgical models do not reflect the natural IA formation and progression.Studies with follow-up imaging would lead to a better understanding of the morphological IA changes appearing before IA rupture.Such knowledge would greatly help in clinical follow-up imaging to determine whether an unruptured IA is at risk of rupture or not and whether it needs to be secured or not.Moreover, linking in vivo imaging and histology could also greatly help in this decision process.Indeed, essential changes in wall composition have been identified in ruptured human IAs when compared to unruptured IAs.Increased inflammatory cell infiltration, luminal thrombosis, and less smooth muscle cells and collagen fibers have been observed in wall of ruptured human IAs [210,211].The emergence of molecular imaging could allow for the in vivo visualization of these changes in the IA wall.So far, molecular imaging using targeted MRI contrast agents allowed for the visualization of the inflammation-associated tissue marker, myeloperoxidase [101,138], macrophage infiltration using ferumoxytol [181], and thrombus using a nanoplatform targeting the platelets [183] in animal IA models.The development of other targeted MRI contrast agents would critically help to elucidate modifications in the vessel wall during IA development and prior to rupture.During MRI, the observation of VWE, which reflects a gadolinium-based contrast agent accumulation in the aneurysm wall, has been associated with an increased risk of IA rupture in human.The pathophysiological reasoning behind the occurrence of VWE is unknown, but enhanced permeability of arterial endothelium, excessive macrophage infiltration, or presence of (leaky) vasa vasorum have been proposed as potential mechanisms [212].Studies using in vivo imaging and animal models of IA could help to elucidate this phenomenon, like the study of Wang et al. [169].
Therefore, in vivo IA imaging should be used more frequently in studies using endogenous animal models.Indeed, this would help to better understand morphological and hemodynamic changes of IAs during their evolution before rupture.Moreover, MRA molecular imaging allows for the observation of in vivo wall modifications.Thus, it is essential that in vivo imaging continues to improve to obtain images of small IAs in endogenous animal models at sufficient resolution.

Comparison of the IA in vivo imaging techniques
In vivo imaging is important for studies using surgical and endogenous IA animal models, and many imaging techniques are now available (Fig. 3C).All imaging modalities have advantages and disadvantages (Table 1), and it is essential to choose the most appropriate modality.
All imaging modalities do not provide the same resolution, which is a first consideration.Based on human data, conventional MRA and CTA exhibit an insufficient resolution for IAs having a diameter < 3 mm [7].Therefore, these techniques are not appropriate for endogenously induced IAs in small animals which require high-resolution imaging.DSA is the gold standard technique as it displays a high resolution that can further be increased with 3D-DSA [8].The resolution of CTA seems to be increased when combined with a flat detector [102].MRA resolution Page 21 of 29 56 can be improved using a higher magnetic field (7 T or even 9.4 T), which has been shown to allow for accurate imaging of the rat circle of Willis [156,168,174].More recently, HF-OCT showed a great potential to assess appropriate treatment of IAs, thanks to a very high-spatial resolution close to 10 μm [182].CFD uses high-resolution 3D imaging to simulate the flow at every position in the IA and adjacent arteries.In comparison, Doppler ultrasonography measures the average blood flow velocity for the entire IA, which is less precise for IA studies.Human studies comparing both techniques show that WSS measured by Doppler ultrasonography is consistently smaller compared to CFD simulations [213].
The choice of the imaging modality should obviously take the invasiveness of the procedure into consideration.Indeed, when several imaging modalities allow for IA visualization at sufficient resolution for the goal of the study, the less invasive technique should be selected.DSA, despite being the gold standard technique, remains the more invasive modality.However, the use of a venous instead of an arterial catheter decreases DSA invasiveness.Fluorescence angiography is also invasive, as it requires artery dissection.CTA is a less invasive technique as the ionizing radiation is lower and as the contrast agent is injected through a venous catheter.Finally, MRA and OCT are the less invasive in vivo imaging modalities for IAs as they do not require ionizing radiation nor contrast agent injection except for CE-MRA that requires a venous contrast agent injection.Long-term and repetitive studies should obviously use the less invasive IA in vivo imaging modality.Doerfler et al. [18] showed that induced IAs in the rabbit elastase model can be serially imaged during a long-term study using i.a.DSA, i.v.DSA, CTA, and MRA.Therefore, the less invasive CTA and MRA modalities should be preferred over the more invasive techniques.
IA patency, with or without endovascular treatment, can be evaluated using different in vivo imaging modalities.DSA, CTA, MRA, and Doppler ultrasonography are routinely used, and OCT, which is a more recent high-resolution technique, is very efficient to assess IA patency accurately [182].As discussed above, DSA is the more invasive technique and should be avoided when possible.OCT being non-invasive and displaying a high resolution should be preferred.However, this imaging modality does not allow for global morphology visualization of IAs and may be combined with another imaging modality.The presence of artifacts in some imaging modalities due to endovascular treatments should also be considered.Indeed, artifacts in presence of clips, stent, or coils can be observed in CT and MRA [192].Metal artifact reduction software are available for clinical CT [214]; Yuki et al. [147] successfully decreased CT stent artifacts in a swine model of IAs.Moreover, Spilberg et al. [113] observed a decay in MRA artifact until 4 weeks post-surgery, which could be linked to IA thrombus modifications.This review did not discuss 4D imaging because only 2 reviewed studies used time-resolved 4D imaging, which combined sequentially obtained 3D images [108,109].However, such imaging techniques are known to significantly improve imaging in clinic.For instance, 4D-DSA could lead to a voxel volume of 0.008mm 3 [215].Temporal resolution is an important parameter to consider in time-resolved imaging, as it will determine the capability of the imaging modality to distinguish fast physiological temporal processes.

Limitations of the study
Despite a careful database search following the PRISMA guidelines and using precise MeSH terms and additional hand searches, this systematic review might have missed some studies using in vivo imaging in IA animal models.Therefore, we cannot exclude a slight bias in the distribution of the different IA animal models and imaging modalities.

Conclusion
In vivo imaging of IAs has tremendously improved in recent years and should be used more frequently in IA animal models.However, all imaging techniques have advantages and disadvantages, and the most appropriate imaging modality should be chosen.The imaging resolution and invasiveness should be considered with respect to the goal of the study.In particular, studies aiming to test endovascular treatment should consider ability to assess IA patency of the imaging modality and the presence of potential metal artifacts.Research to improve imaging modalities should continue, in particular in the field of molecular imaging to better understand IA physiopathology.
56 Page 22 of 29 included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this licence, visit http:// creat iveco mmons.org/ licen ses/ by/4.0/.

Fig. 2
Fig.2IA animal models can be divided in surgical and endogenous models.A Surgical IAs can be constructed using venous or arterial pouch grafting or elastase incubation in combination with right CCA ligation.B Endogenous IA models comprise models using IA risk factors such as increased hemodynamic stress, vessel wall weak-

Fig. 3
Fig. 3 Purpose, distribution among species, and evolution of the different in vivo IA imaging modalities in animal studies.A Reasons for IA in vivo imaging in animals.B Distribution of the use of imaging modalities in swines, dogs, rabbits, rats, and mice.The two studies [142] observed limited coil artifacts with FD-CTA in comparison with MS-CTA.In addition, metal reduction software used on high-resolution CT scans considerably decreases stent artifacts [147].

Table 1
Included studies in the review for in vivo imaging in IA animal models

Table 1 (
XThe 170 studies are presented chronologically, and information on the objective(s) of the study, the IA animal models and the imaging modality(ies) employed are reported Page 15 of 29 56 models, testing of new IA treatment modalities, research on in vivo imaging of IAs, and research on IA pathophysiology (Table