In practice, can strains be computed from
time-resolved images of a model cerebral aneurysm over a cardiac cycle?
Thomas David Butterfield
Submitted in accordance with the requirements
for
the degree of MSc Medical Physics
University of Leeds
Academic Unit of Medical Physics
August 2007
The candidate confirms that the work submitted is his own and that appropriate credit has been given where reference has been made to the work of others.
Key Messages
This study developed a pulsating cerebral aneurysm model suitable for geometrical
studies.
This study investigated the feasibility of computing strain in a pulsating cerebral aneurysm model.
Abstract
Objectives. To construct a geometrically representative,
pulsating cerebral aneurysm model and assess the feasibility of measuring
the surface strain with clinical imaging facilities and image registration
techniques.
Methods. A saccular aneurysm model, constructed from latex rubber, was inflated using a pulsatile flow to a peak diameter of 25mm with a 10% deflation over a simulated cardiac cycle at 60bpm. Sequences of time-resolved images were obtained with angiography and magnetic resonance imaging (MRI). For each sequence of images, a frame was selected at diastole and registered to a frame at systole. The resulting map was used to warp a finite element surface mesh of the model and compute the strains.
Results. Physiologically-representative pulsatile inflations were found to be achievable and highly repeatable but were subject to long settle times. Angiography images possessed excellent geometrical accuracy and allowed for a high quality registration but were limited to 2D projections. Registration of 3D MRI images was possible after segmentation, despite the presence of artefacts. For both data sets, the magnitudes of the resolved strains were comparable to those expected (mean of 0.1). However, the distribution of strain over the surface was more varied than that recorded in a previous in vitro study conducted under simple static inflation pressures.
Conclusions. Computation of strain data from a physiologically-representative pulsating rubber aneurysm model is feasible using standard imaging techniques. Further work is required to validate or otherwise the resolved strains, investigate improved imaging techniques and find the minimum size limit for which this procedure remains feasible.
Keywords. Cerebral aneurysm, strain, pulsatility imaging, image registration.
Introduction
Background
The term cerebral aneurysm refers to an abnormal dilation of the blood vessel wall. This may occur anywhere within the cerebral circulation but is most commonly located within the circle of Willis. Although size and shape vary, most have a saclike pouch shape (saccular) and range in diameter from ~1mm to over 25mm (MacDonald, 1999).
Increased utilization of high resolution, non-invasive imaging modalities for investigation of head injuries has led to a large increase in the incidental discovery of asymptomatic un-ruptured cerebral aneurysms. An average estimate of the prevalence of cerebral aneurysms in the general population is around 5% (3.6-6% (White and Wardlaw, 2003), 0.2-9.9% (Wiebers et al., 2003)). The rupture rate, although low (1-2% (White and Wardlaw, 2003)), combined with the high mortality rates (~50% (MacDonald, 1999)) associated with subarachnoid haemorrhage (SAH), usually provides sufficient justification for clinical attention.
The optimum management strategy for patients with un-ruptured aneurysms is open to wide debate. A large international study of un-ruptured intracranial aneurysms (ISUIA, (1998)) concluded that aneurysm size and past history of SAH were highly influential factors for evaluating the risk/reward ratio for surgical intervention. Further extension of the analysis to obtain a consensus for individual patient risk is difficult to achieve due to fragmentation of knowledge and absence of solid supporting data.
@neurIST is a large, European, multidisciplinary project that aims to “…transform the management of cerebral aneurysms by providing new insight, personalised risk assessments and methods for the design of improved medical devices and treatment protocols” (Frangi, 2006). As an active partner in @neurIST, part of our ongoing research programme concerns an assessment of the feasibility of measuring strain in aneurysm walls from clinical images. It is well known that aneurysms are subject to stress caused by pulsatile blood flow (Kyriacou and Humphrey, 1996). Reduction of haemodynamic stress (and hence the strain) is the basis of the effective modern treatment approach of endovascular coiling (Molyneux et al., 2002). Thus, it is hypothesised that strain measurement could be used as a risk factor in patient management.
Previous Research
Previous research has focussed on vessel wall stress distributions using both physical (Isoda et al., 2006) and computational models (Burleson et al., 1995). Unfortunately due to current insurmountable difficulties regarding the collection of patient-specific wall material properties (Ma et al., 2007), it is unlikely that these techniques can be extended to form a reliable strain measurement method. Experiments have been performed on excised aneurysms from both animals and humans post mortem (Boecher-Schwarz et al., 2000, Seshaiyer et al., 2001). However, the envisaged application must be capable of measuring strain data in vivo to assess risk. Direct strain measurements using optical methods such as laser speckle (Kirkpatrick and Cipolla, 2000) obviously cannot be used deep within the cranium.
An approach under investigation is the measurement of strain from clinical images using image registration. Hartwell and Naylor (2006) demonstrated the feasibility of this method in vitro using rubber balloon model aneurysms of a variety of diameters (6, 10, 15 & 25 mm). The models were imaged with computed tomography (CT) and magnetic resonance imaging (MRI) before and after further inflation, to simulate the change in diameter expected in vivo under pulsatile blood flow. The images obtained were segmented and image registration employed to find the mapping from simulated diastole to systole. A finite element mesh of the surface of the diastolic image was constructed and warped with the registration mapping. Finite element comparison of the pre and post warped mesh allowed the principal and shear strains to be calculated for each element. The results were presented as a 3D graphical representation of the mesh with the faces coloured according to the computed strain.
Hughes (2007) attempted to validate the results of Hartwell and Naylor (2006) by measuring the strains on a model aneurysm surface using an optical camera rig and by marking dots on the models surface. The study concluded that the strain values calculated through image registration were of the correct magnitude, however, the spatial distributions did not correlate. More research is needed in this area.
Changes in aneurysm size due to pulsatile flow have been investigated by Meyer et al (1993). They used phase-contrast magnetic resonance angiography (MRA) to measure the change in diameter of ten un-ruptured and five ruptured cerebral aneurysms of various sizes (2mm to 21mm in diameter). Amongst their limited data sets they showed that percentage diameter changes vary considerably between patients from 0% up to 13%.
There is a dearth of supplementary data from other sources to quantify the range of dilations experienced by aneurysms in vivo. In a recent study, Oubel et al (2007) used digital subtraction angiography (DSA) and image registration to determine pulsatile wall displacements. Their data showed changes of no more than 0.5mm (~1%) in diameter in a very limited study of only 4 patients. There is need for a more complete picture of pulsation displacements as this data is not available from the literature. Hartwell and Naylor (2006) implemented dilations of both 10% in diameter and 10% in volume for their experiments.
Dynamic Imaging for Strain Calculations
If the strain measurement method employed by Hartwell and Naylor (2006) is to be used clinically it is necessary to collect the raw data using clinical scanners under real-time pulsatile blood flow. DSA is considered the gold standard for the investigation of cerebral aneurysms (Papke and Brassel, 2006). However, there is considerable interest in the use of other imaging methods such as computed tomography angiography (CTA), MRA and ultrasound, of which, CTA is generally considered the most robust for aneurysms (Pavone et al., 1990, White et al., 2000). Due to the harmful effects of ionising radiation, the use of CT based modalities for crucial regular monitoring of patients is undesirable, therefore, MR would be preferable.
This study advances the work of Hartwell and Naylor (2006) by replacing their static aneurysm models with dynamic models. A feasibility assessment of imaging strain in such a model with clinical digital angiography and 3T MR scanning facilities was undertaken.
Methods
Model Design
Two sizes of model aneurysms were specified and realised. A large model, indicative of a giant aneurysm (25mm±2mm nominal diameter), was chosen for primary study due to known limitations in MR scanner spatial resolution. A second, smaller model, (13mm±2mm nominal diameter) was specified for study if imaging of the larger model proved feasible. A 10% reduction in diameter from systole to diastole was specified for both models.
The model aneurysms, based on the design from Hartwell and Naylor (2006), were constructed from latex rubber modelling balloons and flexible PVC tubing (fig 1A). Modifications were made to the design for dynamic flow experiments by increasing the tubing bore from 3mm to 8mm. This avoided internal folding of the rubber and facilitated repeatable construction. For the giant models, the aneurysm neck consisted of a large rectangular hole, 8x10mm in size, cut into the tubing. This minimised the risk of rupture during experiments. For the smaller models 6mm diameter circular holes were used. The models were terminated with medical quick couplings (Altec) for straightforward mounting into the test rig.
Fig 1 – Experimental apparatus: A) model aneurysms, B) test rig, C) flow resistance tuning and D) optical calibration chequerboard (1mm2).
A test rig was constructed from a sealable 4 litre plastic container which held the aneurysm model securely in place whilst submersed in tap water (fig 1B). The model was mounted at an angle to prevent air bubbles forming in the inflated volume.
A CompuFlow 1000 MR computer controlled pump (Shelley Medical Imaging Technologies, 2007) provided pulsatile flow. The device is a positive displacement pump consisting of a cylindrical piston driven along a precise lead screw by a computer-controlled micro stepping motor. It is capable of delivering user-defined pulsatile flow waveforms. A flow rate curve was programmed into the software which traced the shape of a published velocity Doppler study (Rohren et al., 2003)) of blood flow in the internal carotid artery (fig 2). The pump software was used to scale the peak and end flows of this simple waveform to enable fine tuning of the dilation magnitudes. A cycle period of 1s (60bpm) was used for all experiments.
Fig 2 – Blood velocity profile in the internal carotid artery from a Doppler study (Rohren et al., 2003)). Green waveform shows the pulse shape used in this study.
The pump system was located outside the MR scanner room and flow was delivered to the rig via two 9m lengths of 6.3mm bore, braid reinforced PVC hose (AlteVin™), chosen to minimise system capacitance. Extra resistance in the form of a 6cm length of flexible 3mm bore tubing and a miniature clamp (fig 1c) was added to the return flow to provide a coarse tuning capability. A blood analogue of 40% glycerol to 60% distilled water (by volume) was used for the circulating fluid as recommended by the pump manufacturer.
Imaging
Optical video calibration was performed using a digital camcorder (Sony DCR-PC100E) connected to a video monitor. The camera and video monitor settings were adjusted to give the sharpest interface between the balloon and a 1mm checkerboard background card (fig 1d). The setup allowed diameter measurements to within 0.2mm.
Single plane angiography was carried out on a Philips Integris V for 6 seconds at an acquisition rate of 4 frames per second (fps). The resolution of resultant 2D images was 0.06 x 0.06mm per pixel. OmnipaqueTM (GE_Healthcare, 2007) contrast agent was mixed into the circulating fluid to achieve a iodine concentration of 12mgI/ml. The model was then submersed in water before imaging.
Dynamic 3D MRI was performed on a Philips Achieva 3.0T MRI scanner and Sense head coil (Philips medical systems). Due to small air bubbles adhering to the outer surface of the rubber when in water, MR scans were performed without submersing the model to help avoid distortion. Gating was implemented using a portable hand-held ECG generator set to 60bpm to match the pump cycle rate. The imaging sequence used was the balanced fast field echo (B-FFE) cine (Philips medical systems): a fast gradient echo technique which uses retrospective reconstruction and is designed for breath-hold imaging (echo time TE=1.56s & repetition time TR=3.1s). The temporal resolution of all scans was 20 frames per second. The 2D slice resolution of all images was 2.34 x 1.43mm interpolated to 1.17 x 1.17mm by the scanner software. Scans were performed in two orthogonal planes with a slice thickness of 5mm and repeated with a slice thickness of 4mm, yielding a total of four 3D data sets. Scan times were 35s and 59s for 5mm and 4mm slice scans respectively.
Post Processing
Matlab (The_Mathworks, 2007) was utilized to analyse angiography and MR data. Custom scripts and associated graphical user interfaces were written to facilitate dynamic and step frame viewing of the raw data. A frame was hand picked at ‘diastole’ and ‘systole‘ for each data set. The pair of images was identically cropped and enhanced as appropriate using scripts from Matlab's image processing toolbox (including median filtering to remove noise). 3D MR images were interpolated to achieve a 3D resolution of 0.6mm3 per pixel.
A 2D mesh was extracted from the angiography images using a custom-written edge-detection script which placed vertices around the aneurysm edge. 3D meshes were extracted from MR images using threshold intensity segmentation to form a binary image and employing the Matlab function isosurface to generate the vertices and edges.
Image registration and mesh warping was performed using the Sheffield Image registration toolkit (ShIRT) (Barber, 2006).
Strain calculations were carried out using the strain solver function developed by Barber (2007) which compares the pre and post warped finite element mesh, and calculates the strains for each element. Matlab was then used to superimpose this data as a colour map on the faces of the original 3D mesh and display the result.
Results
Optical Model Calibration
Figure 3a shows the peak diameter of a newly constructed giant model over four consecutive 12min test runs with a peak flow rate of 19ml/s. This initially gave the required 25mm diameter. These tests revealed considerable creep and long settle times (over 8min).
Fig 3 – Peak systole radius of models over time under periodic flow: A) New rubber model response for 4 runs at 19ml/s peak flow. B) The same stretched rubber model response for 5 runs at 17.5ml/s peak flow.
Through a process of trial and error the nominal radius was reduced to 12.5mm by reduction of the peak flow rate to 17.5ml/s. As fig 3B shows, the creep behaviour stopped but the long settle times remained. As a result of these findings a procedure was devised whereby the rubber of newly-constructed models was pre-stretched using a high flow rate (>22ml/s) prior to using them in experiments in order to avoid the creep problems. No method of reducing the settle time was found. The phase lag from pump output to model was 150ms.
Exhaustive calibration runs revealed that, provided the system was calibrated prior to transfer to scanning facilities, the resultant model diameter would be 25mm±2mm. It was also shown that a 8min settle time before acquiring data was adequate.
A flow profile which dropped to 30% of the peak flow at end diastole was found to give the required 10% dilation for the large models. A drop to 60% was required for the smaller models.
It was observed that the whole model, including the feed tubes, moved back and forth by ~1mm in response to pulsatile flow.
Angiography
Figure 4A&B show two images taken from the single plane 24 frame angiography sequence. The images show a dilation of 10.3% in diameter from peak systole to end diastole. Fig 4C shows the resulting image registration map from diastole to systole. Fig 4D shows the circumference strain calculations. The mean strain value is 0.1 and the maximum is 0.32.
Fig 4 – Angiography results: A) and B) enhanced images at systole and diastole respectively, C) image registration map and D) edge strain distribution.
MRI
Figure 5 shows the 3D distributions of the first principal strains as calculated from four separate MR scans of the 25mm diameter model. The average first principal strains were 0.11, 0.11, 0.14 and 0.08 for data sets A, B, C and D respectively. The second principal and shear strains are not shown as the first principal, being the largest, is the most useful for assessment of the acquisition techniques.
Fig 5 – Distributions of first principal strains as obtained from MR data of a 25mm aneurysm model: Data set A), slices (5mm thick) normal to artery shown from two angles. Data set B), slices (4mm thick) normal to artery shown from two angles. Data set C), slices (5mm thick) parallel to artery. Data set D), slices (4mm thick) parallel to artery.
Discussion
Rubber Aneurysm Model
Computer controlled pumps have been used to apply pulsatile flows to model aneurysms for the study of haemodynamics by others in the field (Ahn et al., 2007, Isoda et al., 2006). This study has applied this technique to a rubber latex model and demonstrated successful simulation of a pulsating giant aneurysm. The exceptionally stable and highly repeatable inflations obtained were important for successful, gated imaging.
Conditioning of the phantom model was surprisingly simple. The creep behaviour was very easily eliminated by pre-stretching the rubber. Long settle times, although inconvenient, were not a major hindrance to the study as they were comparable with scanner setup times. The pump was not thought to contribute to the long settle times because: it is a precision device, basic calibration tests confirmed its reliability to within reasonably measurable limits, any drift would be expected to be randomly centred on a median value and tests confirmed inflations did not fluctuate in this manner. The rest of the system was made up of rigid tubing and tests with shorter lengths yielded comparable settle times. For these reasons, the rubber latex material was deemed to be the source of the long settle behaviour.
Global tube movement was initially eliminated by bracing the back of the model with a rigid plastic strip. However, correction of this movement was simple to achieve in post-processing, therefore no bracing was used. It also proved a useful exercise as real blood vessels also behave in this way (Oubel et al., 2007).
It should be noted that rubber is not considered to be an ideal model for elastic tissue behaviour (Haslach and Humphrey, 2004). For this reason, use of such a model is limited to geometrical studies only as pressures and flows are unlikely to be comparable to those expected in vivo.
Angiography
The sequence of angiography images was usable despite low contrast. The outline of the aneurysm was clear and very smooth. Singling out of a image at diastole and systole was trivial. Observable geometrical deformation from pulsatile flow was unmistakable in the enhanced image pair (fig 4A&B). The diameters measured from the images and estimated percentage changes concurred with optical video calibration tests. Correction of global tube movement was straightforward, and segmentation and image registration algorithms performed well.
An average strain value of 0.1 is in line with that expected from a 10% deflation and also corresponds with the strains calculated by Hartwell and Naylor (2006) for a static model of comparable size and dilation. The strain distribution around the edge of the model revealed two clear ‘hot spots’. This distribution does not match the very even distribution obtained by Hartwell and Naylor (2006) and is contrary to the smooth distribution expected from an inflated rubber membrane (Mott et al., 2003). Several explanations for this distribution can be offered. First, inaccuracies in image registration and map generation could have been caused by the intensity gradient from the centre the aneurysm image or artefacts deposited by the noise removal filter. Second, unsatisfactory correction of global model movement may have distorted the true inflation map. Thirdly, the rubber model did indeed stretch in this manner and the calculations are correct. This study was not concerned with validation of the calculated strains, therefore, optical validation is required to support or otherwise the accuracy of these strain results.
Unfortunately due to the low frame rate (4fps) it is not possible to prove if the two images picked from the sequence were absolute diastole and systole. Increase in frame rate was within the capability of the x-ray device (up to 50fps), but was not feasible at the time of acquisition due to disk space constraints.
Multiaxial strains are required to have any potential clinical application in order that surface distribution can be observed. Resolving such strains requires 3D reconstructions of the model. Ideally x-ray studies would have been performed using CTA so that 3D data could be obtained. It is likely that CT would have provided high quality raw data as has been demonstrated by others (Goddard et al., 2005, Ishida et al., 2005, Yaghmai et al., 2007). However, the required facilities and expertise were not available within the project time frame so 2D single plane angiography was performed as a proof of concept.
Due to the poor contrast it would not have been feasible to attempt imaging of the smaller model. The use of higher contrast agent concentrations would have provided sharp enough images but this was considered to be an uneconomic experiment due to the inherent limit on the usefulness of 2D images.
MRI
The strain maps resolved from MR data suggest that imaging of strain in 25mm diameter models is feasible using common clinical MR acquisition sequences. Despite geometrical distortion at the extremities (a consequence of low slice resolution), the strain values obtained in the mid regions of data sets A and B (fig 5A&B) are comparable with magnitudes measured by Hartwell and Naylor (2006). The 4mm sliced scans provide more geometrically accurate shapes than the 5mm scans, but they appeared more prone to image artefacts.
The pattern of increased strain near the neck and lower strain at the fundus in data set A (fig 5A left) is in tentative agreement with the findings of Hughes (2007), who implemented an optical strain measurement technique of a similar static model. This observation is indicative, but due to the poor spatial resolution and distortion in the other data sets no firm conclusions can be drawn from it.
Post processing of MR data was somewhat more involved than for angiography data. The raw data possessed fuzzy edges and inhomogeneities due to turbulent flow. Despite this, segmentation still performed well, resulting in well-defined edges which were suitable for image registration after smoothing and interpolation had been applied.
Measurement of the nominal diameter from the images was not possible. The limited intelligibility of the raw images coupled with geometric distortion of the tubing made obtaining a reference measurement impossible (Optical video calibration showed the nominal diameter was 25mm±2mm). For the same reason, correction of global tube movement was not attempted.
A limitation in these results is the extent of geometric distortion and artefacts. These range from the spurious masses most evident below the main aneurysm in data set B (fig 5B), to distortion affecting the strain, such as the large area of zero strain in fig A right. The strains in the protruding lobes of data sets C and D (fig 5C&D) are of the correct magnitude. However, the tubing parallel to the slice direction caused a band of distortion through the images which caused the mesh directly above the tubing to be over-warped.
It is likely that imaging in air contributed to artefacts. The decision to exclude the surrounding water was based on several factors including: concerns over obtaining enough contrast and the formation of small air bubbles on the external face of the model when in water. Air bubbles would have almost certainly introduced unacceptable artefacts as was experienced by Hartwell and Naylor (2006). Formulation of a method to deoxygenate the surrounding water and use of an MR contrast agent in the circulating fluid would have helped obtain cleaner results.
The B-FFE imaging sequence used was not the most effective for obtaining geometrically accurate data due to rapid acquisition rates (30-40s). Various scans were attempted with longer (~5mins) spin-echo based sequences. However, these did not yield usable data as the pump and ECG simulator were subject to phase disruption over time due to lack of synchronicity. A method using a ECG signal triggered from the computer controlled pump instead of a separate device would have eliminated this phase drift and allowed higher resolution scans to be performed. Alternative MR techniques such as time-of-flight, black blood or contrast enhanced MRA warrant investigation for possible use in this context as they have been used to obtain geometrical data of aneurysms in other studies (Faries et al., 2003, Nael et al., 2006, Papke and Brassel, 2006).
Imaging of the smaller model was not attempted with MRI. Limitations in resolution and artefacts need to be addressed before scaling down as these will certainly be exacerbated with smaller models. A lower size limit than can be usefully imaged with the current scanner generation will need to be quantified. Furthermore, many aneurysms exhibit minute blisters (commonly referred to as blebs) on their surface which are thought to be prime rupture sites (Challa and Han, 2007). It is vital that any strain measurement system has sufficient resolution to resolve such entities.
Future Context
The future possibility of measuring strain through image registration techniques appears promising despite current practical constraints. Potential applications are not limited to the cranium. For example, similar risk assessment data is desirable in the management of aortic aneurysms (Fillinger, 2007). Much work is still required but the ever improving state of available technology can only help to break through the practical barriers for those who pursue this in the future.
Conclusions
Saccular aneurysm models can be constructed from rubber latex modelling balloons
and inflated by applying pulsatile flow to simulate the geometric dilation
of an aneurysm. The phantom that was developed provided accurate and repeatable
inflations, making it highly fit for purpose.
Angiography images possessed excellent geometrical accuracy which allowed for a high quality 2D registration. Data collection using an x-ray based 3D imaging modality such as CTA would yield more valuable results.
3D MR images were subject to distortion and artefacts. It is likely that longer acquisition sequences would yield more accurate results, but this would require specialist hardware to enable phase-locked gating. Exploration of alternative MR imaging techniques may bear more favourable results. The lower limit of aneurysm size for which usable data can be obtained with MR is unknown.
Resolved strain magnitudes from both data sets were comparable with those expected. The distributions were more uneven that those obtained from a previous static model study. The work done supports the hypothesis that extraction of usable strain data from pulsating aneurysms is likely to be feasible. Further work is required to validate or otherwise the resolved strains.
Acknowledgments
Pat Lawford and Rod Hose, my supervisors, for their guidance, support and
expertise.
Jon Hughes for his support and countless helpful conversations.
Andrew Narracott for his support and advice.
Steven Wood for his help in obtaining angiography data.
Iain Wilkinson for his time and expertise in obtaining MR data.
David Barber for his provision of image registration software and strain solver functions.
References
AHN, S., SHIN, D., TATESHIMA, S., TANISHITA, K., VINUELA, F. & SINHA, S. (2007) Fluid-induced wall shear stress in anthropomorphic brain aneurysm models: MR phase-contrast study at 3 T. J Magn Reson Imaging, 25, 1120-30.
ALTEC (2007) CPC Medical Quick Couplings. Altec Products Limited (http://www.altecweb.com).
ALTEVIN™ Braid Reinforced PVC Hose. Altec Products Limited (http://www.altecweb.com).
BARBER, D. C. (2006) Sheffield Image Registration Toolkit: ShIRT. Medical Physics, University of Sheffield, Royal Hallamshire Hospital, Sheffield.
BARBER, D. C. (2007) Strain solver Matlab functions. Medical Physics, University of Sheffield, Royal Hallamshire Hospital, Sheffield.
BOECHER-SCHWARZ, H. G., RINGEL, K., KOPACZ, L., HEIMANN, A. & KEMPSKI, O. (2000) Ex vivo study of the physical effect of coils on pressure and flow dynamics in experimental aneurysms. AJNR Am J Neuroradiol, 21, 1532-6.
BURLESON, A. C., STROTHER, C. M. & TURITTO, V. T. (1995) Computer modeling of intracranial saccular and lateral aneurysms for the study of their hemodynamics. Neurosurgery, 37, 774-82; discussion 782-4.
CHALLA, V. & HAN, H. C. (2007) Spatial variations in wall thickness, material stiffness and initial shape affect wall stress and shape of intracranial aneurysms. Neurol Res.
FARIES, P. L., AGARWAL, G., LOOKSTEIN, R., BERNHEIM, J. W., CAYNE, N. S., CADOT, H., GOLDMAN, J., KENT, K. C., HOLLIER, L. H. & MARIN, M. L. (2003) Use of cine magnetic resonance angiography in quantifying aneurysm pulsatility associated with endoleak. J Vasc Surg, 38, 652-6.
FILLINGER, M. (2007) Who Should We Operate On and How Do We Decide: Predicting Rupture and Survival in Patients with Aortic Aneurysm. Seminars in Vascular Surgery, 20, 121-127.
FRANGI, A. (2006) @neurIST: Integrated biomedical informatics for the management of cerebral aneurysms. Fabra.
GE_HEALTHCARE (2007) OmnipaqueTM. http://www.omnipaque.com.
GODDARD, A. J., TAN, G. & BECKER, J. (2005) Computed tomography angiography for the detection and characterization of intra-cranial aneurysms: current status. Clin Radiol, 60, 1221-36.
HARTWELL, J. & NAYLOR, S. (2006) The feasibility of imaging strain in intracranial aneurysms. MSc Dissertations, Universtiy of Sheffield.
HASLACH, H. W. & HUMPHREY, J. D. (2004) Dynamics of biological soft tissue and rubber: internally pressurized spherical membranes surrounded by a fluid. International Journal of Non-Linear Mechanics, 39, 399-420.
HUGHES, J. (2007) Are static strains measured in phantom aneurysms in 3T MRI comaprable to those measured optically? , University of Sheffield, MSc Project.
ISHIDA, F., OGAWA, H., SIMIZU, T., KOJIMA, T. & TAKI, W. (2005) Visualizing the dynamics of cerebral aneurysms with four-dimensional computed tomographic angiography. Neurosurgery, 57, 460-71; discussion 460-71.
ISODA, H., HIRANO, M., TAKEDA, H., KOSUGI, T., ALLEY, M. T., MARKL, M., PELC, N. J. & SAKAHARA, H. (2006) Visualization of hemodynamics in a silicon aneurysm model using time-resolved, 3D, phase-contrast MRI. AJNR Am J Neuroradiol, 27, 1119-22.
ISUIA (1998) Unruptured intracranial aneurysms--risk of rupture and risks of surgical intervention. International Study of Unruptured Intracranial Aneurysms Investigators. N Engl J Med, 339, 1725-33.
KIRKPATRICK, S. J. & CIPOLLA, M. J. (2000) High resolution imaged laser speckle strain gauge for vascular applications. J Biomed Opt, 5, 62-71.
KYRIACOU, S. K. & HUMPHREY, J. D. (1996) Influence of size, shape and properties on the mechanics of axisymmetric saccular aneurysms. J Biomech, 29, 1015-22.
MA, B., LU, J., HARBAUGH, R. E. & RAGHAVAN, M. L. (2007) Nonlinear anisotropic stress analysis of anatomically realistic cerebral aneurysms. J Biomech Eng, 129, 88-96.
MACDONALD, D. J. (1999) Wall characteristics of saccular aneurysms from polarized light microscopy. Ontario, University of Western Ontario.
MEYER, F. B., HUSTON, J., 3RD & RIEDERER, S. S. (1993) Pulsatile increases in aneurysm size determined by cine phase-contrast MR angiography. J Neurosurg, 78, 879-83.
MOLYNEUX, A., KERR, R., STRATTON, I., SANDERCOCK, P., CLARKE, M., SHRIMPTON, J. & HOLMAN, R. (2002) International Subarachnoid Aneurysm Trial (ISAT) of neurosurgical clipping versus endovascular coiling in 2143 patients with ruptured intracranial aneurysms: a randomised trial. Lancet, 360, 1267-74.
MOTT, P. H., ROLAND, C. M. & HASSAN, S. E. (2003) Strain in an inflated rubber sheet. Rubber chemistry and technology, 76, 326.
NAEL, K., VILLABLANCA, J. P., SALEH, R., POPE, W., NAEL, A., LAUB, G. & FINN, J. P. (2006) Contrast-enhanced MR angiography at 3T in the evaluation of intracranial aneurysms: a comparison with time-of-flight MR angiography. AJNR Am J Neuroradiol, 27, 2118-21.
OUBEL, E., DE CRAENE, M., PUTMAN, C. M., CEBRAL, J. R. & FRANGI, A. F. (2007) Analysis of intracranial aneurysm wall motion and its effects on hemodynamic patterns. Proceedings of the SPIE. SPIE--The International Society for Optical Engineering.
PAPKE, K. & BRASSEL, F. (2006) Modern cross-sectional imaging in the diagnosis and follow-up of intracranial aneurysms. Eur Radiol, 16, 2051-66.
PAVONE, P., DI CESARE, E., DI RENZI, P., MARSILI, L., VENTURA, M., SPARTERA, C. & PASSARIELLO, R. (1990) Abdominal aortic aneurysm evaluation: comparison of US, CT, MRI, and angiography. Magn Reson Imaging, 8, 199-204.
ROHREN, E. M., KLIEWER, M. A., CARROLL, B. A. & HERTZBERG, B. S. (2003) A spectrum of Doppler waveforms in the carotid and vertebral arteries. AJR Am J Roentgenol, 181, 1695-704.
SESHAIYER, P., HSU, F. P. K., SHAH, A. D., KYRIACOU, S. K. & HUMPHREY, J. D. (2001) Multiaxial Mechanical Behavior of Human Saccular Aneurysms. Taylor & Francis.
SHELLEY_MEDICAL_IMAGING_TECHNOLOGIES (2007) CompuFlow Pump. Ontario, CANADA.
THE_MATHWORKS (2007) Matlab 7. Mathworks.
WHITE, P. M. & WARDLAW, J. M. (2003) Unruptured intracranial aneurysms. J Neuroradiol, 30, 336-50.
WHITE, P. M., WARDLAW, J. M. & EASTON, V. (2000) Can Noninvasive Imaging Accurately Depict Intracranial Aneurysms? A Systematic Review. Radiology, 217, 361-370.
WIEBERS, D. O., WHISNANT, J. P., HUSTON, J., 3RD, MEISSNER, I., BROWN, R. D., JR., PIEPGRAS, D. G., FORBES, G. S., THIELEN, K., NICHOLS, D., O'FALLON, W. M., PEACOCK, J., JAEGER, L., KASSELL, N. F., KONGABLE-BECKMAN, G. L. & TORNER, J. C. (2003) Unruptured intracranial aneurysms: natural history, clinical outcome, and risks of surgical and endovascular treatment. Lancet, 362, 103-10.
YAGHMAI, V., ROHANY, M., SHAIBANI, A., HUBER, M., SOUD, H., RUSSELL, E. J. & WALKER, M. T. (2007) Pulsatility Imaging of Saccular Aneurysm Model by 64-Slice CT with Dynamic Multiscan Technique. J Vasc Interv Radiol, 18, 785-8.