Prediction of Thin-Walled Areas of Unruptured Cerebral ... · BioMedResearchInternational...

Research ArticlePrediction of Thin-Walled Areas ofUnruptured Cerebral Aneurysms through Comparison ofNormalized Hemodynamic Parameters andIntraoperative Images

Kwang-Chun Cho 1 Ji Hun Choi2 Je Hoon Oh 2 and Yong Bae Kim 3

1Department of Neurosurgery College of Medicine Catholic Kwandong University International St Maryrsquos HospitalSimgok-ro 100gil 25 Seo-gu Incheon 22711 Republic of Korea2Department of Mechanical Engineering Hanyang University 55 Hanyangdaehak-ro Sangrok-gu AnsanGyeonggi-do 15588 Republic of Korea3Department of Neurosurgery College of Medicine Yonsei University Gangnam Severance Hospital 211 Eonju-roGangnam-gu Seoul 06273 Republic of Korea

Correspondence should be addressed to Je Hoon Oh jehoonhanyangackr and Yong Bae Kim ybkim69yuhsac

Received 14 June 2018 Accepted 2 September 2018 Published 20 September 2018

Academic Editor Carl Muroi

Copyright copy 2018 Kwang-Chun Cho et al This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited

Object Rupture of a cerebral aneurysm occurs mainly in a thin-walled area (TWA) Prediction of TWAs would help to assess therisk of rupture and select appropriate treatment strategy There are several limitations of current prediction techniques for TWAsTo predict TWAsmore accuratelyHP should be normalized tominimize the influence of analysis conditions and the effectivenessof normalized combined hemodynamic parameters (CHPs) should be investigated with help of the quantitative color analysis ofintraoperative imagesMethods A total of 21 unruptured cerebral aneurysms in 19 patients were analyzed A normalized CHP wasnewly suggested as a weighted average of normalized wall shear stress (WSS) and normalized oscillatory shear index (OSI) Delta Efrom International Commission on Illumination was used to more objectively quantify color differences in intraoperative imagesResults CFD analysis results indicated that WSS and OSI were more predictive of TWAs than pressure (Plt001 P=187 P=970respectively) these two parameters were selected to define the normalized CHP The normalized CHP became more statisticallysignificant (Plt001) as theweighting factor of normalizedWSS increased and that of normalizedOSI decreased Locationswith highCHP values corresponded well to those with high Delta E values (Plt001) Predicted TWAs based on the normalized CHP showeda relatively good agreement with intraoperative images (17 in 21 cases 810) Conclusion 100 weighting on the normalizedWSS produced the most statistically significant result The normalization scheme for WSS and OSI suggested in this work wasvalidated using quantitative color analyses rather than subjective judgments of intraoperative images and it might be clinicallyuseful for predicting TWAs of unruptured cerebral aneurysms The normalization scheme would also be integrated into furtherfluid-structure interaction analysis for more reliable estimation of the risk of aneurysm rupture

1 Introduction

Spontaneous subarachnoid hemorrhages caused by rupturedcerebral aneurysms are fairly fatal and associated with highrates of morbidity andmortality Cerebral aneurysm rupturesare likely to occur mainly in thin-walled areas (TWAs)within the aneurysms [1 2] Therefore predicting TWAs is

important for assessing the risk of aneurysm rupture andselecting the appropriate treatment

TWAs have been predicted using computational fluiddynamics (CFD) with various hemodynamic parameters(HPs) [3ndash7] Recently the reliability of TWAs predictionhas been improved by comparing CFD analysis results withintraoperative images of aneurysms [8ndash11] However there

HindawiBioMed Research InternationalVolume 2018 Article ID 3047181 9 pageshttpsdoiorg10115520183047181

2 BioMed Research International

is still on-going controversy regarding which HP most accu-rately predicts TWAs Meng et al [3] reported that low wallshear stress (WSS) and high oscillatory shear index (OSI)triggered the growth and rupture of large aneurysm pheno-types whereas highWSS was associated with the growth andrupture of small aneurysm phenotypes and medial thinningTakao et al [6] concluded that pressure loss coefficient mightbe a potential parameter to predict future rupture of unrup-tured aneurysms Cebral et al [12] analyzed a rupture riskof cerebral aneurysms considering flow pattern and impinge-ment region

There are several limitations of current TWA predictionmethods First TWAs are predicted based on only single HPfrom CFD analysis although they are usually caused by com-plex phenomena [12ndash14] Second the value of HP from CFDanalysis is greatly influenced by the inlet and outlet boundaryconditions which are not only hard to accurately obtainbut also different from each patient Third identification ofTWAs in intraoperative images is subjective and significantlydependent on the visual judgment of the clinician [9 10 15]Finally as aforementioned it is still arguable which HP is themost appropriate to predict TWAs To predict TWAs moreaccurately the influence of CFD analysis conditions shouldbe minimized by properly normalizing the HP value andcombination of normalized HP values could be introducedto enhance the prediction accuracy The quantitative digitalcolor analysis of intraoperative images would help to validatethe effectiveness of normalized combined hemodynamicparameters (CHPs)

In this study we performed CFD analyses of 21 unrup-tured cerebral aneurysms and newly suggested both normal-ized schemes and a normalized CHP as a weighted average ofnormalized WSS and normalized OSI We also investigatedthe influence of weighting factors of CHP on the predictionof TWAs In order to avoid the cliniciansrsquo subjective assess-ment when determining the TWAs from correspondingintraoperative images the quantitative digital color analyseswere conducted based on CIE Delta E The results of DeltaE were compared with those of CHPs from CFD analysesto evaluate the effectiveness of the suggested normalizationscheme and CHP for TWA prediction

2 Materials and Methods

21 Patients and Image Data Acquisition This study analyzeda total of 21 unruptured aneurysms in 19 patients (14 womenand 5 men mean age 59 years mean aneurysm size 51 plusmn115 mm) All data were obtained from patients treated viasurgical neck clipping from July 2014 to December 2016 inour institution The protocols used in this studywere ethicallyapproved by our institutional review board and patient con-sent was waived Ruptured aneurysms were excluded becauseit was difficult to obtain prior shape data and the shapewas unclear Digital subtraction angiography (DSA) imageswere generated using Artis zee biplane with syngoWorkplacesoftware (Siemens Erlangen Germany) All intraoperativeimageswere captured using anOPMIPentero surgical micro-scope (Carl Zeiss Microscopy GmbH Jena Germany) witha three-chip charge-coupled device digital video camera at

640 x 480 resolution or a mounted Canon EOS 5D digitalsingle lens reflex camera at 4368 x 2912 resolution

22 Computational Fluid Dynamics Blood vessel and aneu-rysm geometry were generated from the DSA images Bloodvessel models were modified by using shape modifica-tion software (CATIA V5-6R2012 Dassault Systems ParisFrance Meshmixer version 110544 Autodesk San RafaelCA) Based on our previous study (Figure 1(a))[16] the inletlength of CFDmodels was set to 20 times larger than the inletdiameter and the outlet length was set to twice larger thanthe outlet diameter to obtain reliable simulation resultsCFD analysis was performed using ANSYS Fluent software(version 150 ANSYS Inc Canonsburg PA) The Navier-Stokes equations were solved to predict blood flow Bloodwas modeled as an incompressible Newtonian fluid with adensity of 1055 kgm3 and a viscosity of 0004 Pasdots and theflow was assumed to be laminar Blood vessels were modeledwith the assumption of no-slip and rigid-wall conditionsPulsatile flow with a Womersley velocity profile was used asthe inlet boundary condition Womersley velocity profile foreach model was calculated from the volumetric flow rate ofa humanmeasured at the internal carotid artery [17] Pressureprofile based on pressure waveforms measured at the super-ficial temporal artery [18] was applied at each outlet Detailedinformation on CFD analysis is provided in Figures 1(b) and1(c) After the mesh independence check the number ofmeshes was determined to 45 million to 55 million

23 Normalization Scheme and CHP Individual HPs suchas WSS OSI and pressure have different physical meaningsunits and value ranges so direct comparison of their values isnot reasonable WSS is the dynamic friction force generatedby viscous fluid moving along the surface of the solid walland is defined as the product of fluid viscosity and shearingvelocity gradient at the vascular wall [19] OSI is a dimen-sionless parameter that indicates how the direction of WSSchanges at a specific location during a cardiac cycle In orderto equally compare each HPrsquos contribution the value of HPshould be normalized to have a value between 0 and 1

4th-order equation based on the ellipse formula was pro-posed to create the normalized WSS WSSnorm with a fixedrange of 0 to 1 as follows

(119882119878119878119899119900119903119898minus 1)4 + ( 119882119878119878 minus119882119878119878max119882119878119878max minus119882119878119878min

)4 = 1 (1)

where WSSmin and WSSmax are the minimum and the max-imum values of WSS in the aneurysm sac above its neck atthe diastolic time respectively IfWSS at the diastolic time issubstituted into (1) thenWSSnorm is calculated in the range of0 to 1 Note thatWSSnorm is 0 forWSSmax and 1 forWSSmin119874119878119868 is originally defined as

119874119878119868 = 12 (1 minus100381610038161003816100381610038161003816int1199050 119908119904119904119894 119889119905100381610038161003816100381610038161003816int1199050

10038161003816100381610038161199081199041199041198941003816100381610038161003816 119889119905) (2)

where119908119904119904i and t represent an instantaneous WSS vector andthe cycle duration respectively [20] OSI ranges from 0 to

BioMed Research International 3

Inlet

Normal vessel

Outlets

Aneurysm

(a)

Density 1055 kgm3

Viscosity 0004 Pamiddots

Flow Laminar (Re is 500 ~ 700)

Time step size 0014409 s

Inlet condition Womersley velocity profile (Based on volumetric flow rate)

Outlet condition Pressure profile

(b)

Systolic time

Diastolic time

Systolic time

Diastolic time

0

15

25

35

45

55

Volu

met

ric F

low

Rat

e (cm

3 s)

04 08 12 16 20Time (s)

0

5000

10000

15000

20000

Pres

sure

(Pa)

04 08 12 16 20Time (s)

(c)

Figure 1 Detailed information on CFD simulation (a) Example of CFDmodel with inlet and outlet long enough for reliable simulation (b)Fluid properties and analysis conditions (c) Flow rate profile for inlet condition and pressure profile for outlet condition

05 and higher OSI values are more dangerous from a hemo-dynamic point of view therefore OSInorm is also defined as adouble of OSI to make its range become 0 to 1

119874119878119868119899119900119903119898= 2 times 119874119878119868 (3)

Through this process CHP can be defined as a weightedaverage of the WSSnorm and OSInorm so that each HP contri-butes proportionately to the final CHP

119862119867119875 = 1199081times (119882119878119878

119899119900119903119898) + 1199082times (119874119878119868

119899119900119903119898) (4)

1199081+ 1199082= 1 (5)

where 1199081and 119908

2are weighting factors for WSSnorm and

OSInorm respectively By defining CHP as in (4) it also variesfrom 0 to 1

24 Delta E For more objective determination of TWAsDelta E was introduced to quantify reddish areas on intraop-erative images of patientsrsquo aneurysms Delta E indicates how

much a specific color differs from a reference color in CIE119871lowast119886lowast119887lowast color coordinates CIE 119871lowast119886lowast119887lowast is an internationallystandardized colorimetric system for objective color expres-sion used in applications involving color recognition by thehuman eyeThe color coordinates are composed of three axes119871lowast represents lightness and darkness 119886lowast the degree of rednessand greenness and 119887lowast the degree of yellowness and blueness(Figure 2(a))

To obtain Delta E a region of interest was first extractedfrom an intraoperative image then RGB values of all pixelsof the extracted region were converted into 119871lowast119886lowast119887lowast valuesand Delta E was calculated based on CIEDE 2000 color-difference formula[21] (Figure 2(b)) The obtained Delta Ewas finally modified to emphasize the lightness and rednessof aneurysms as follows

Δ119864119898= 119863119890119897119905119886 119864 sdot radic 119886221198712

2+ 11988622

(6)

where 1198712and 1198862are the 119871lowast and 119886lowast values of specific area on

the normal vessel selected by the clinicians Δ119864119898is the same


-a(Green) -b

(Blue)

A(Red)

b(Yellow)

L=100(White)

L=0(Black)

Delta E

Reference color

(a)

Intraoperative image

Calculate Delta E using LabVIEW

Extract sample image

Prediction the TWAs using pixel coordinates

Sample image

Thin wall

Coordinates Delta E

(b)

Figure 2 (a) CIE Delta E in the 119871lowast119886lowast119887lowast color coordinate system (b) Procedure for calculating Delta E and predicting TWAs from an intra-operative image

as the projected Delta E along the 119886lowast axis The most reddisharea was identified with the highest value of Δ119864

119898 and it was

considered TWA This process was performed using Lab-VIEW software (version 150 National Instruments AustinTX)

25 Statistical Analysis Statistical analysis was conductedusing SPSS software (version 210 IBM Corp Armonk NY)The Shapiro-Wilk test was used to evaluate the normality ofparameters used When satisfying the normality statisticalsignificance was analyzed using t-test Otherwise the MannndashWhitneyU test was used for the statistical analysis Statisticalsignificance was assumed for values of Plt053 Results

31 Usefulness of Delta E To check the usefulness of Δ119864119898in

quantifying color differences of aneurysms in the intraopera-tive images we first selected more dangerous less dangerousand reference regions and estimated Δ119864119898 for all otherregions relative to the reference Δ119864

119898results showed that the

maximumvalue ofΔ119864119898was obtained at themost reddish area

which was clinically regarded as more dangerous region (Fig-ure 3) We also compared the values of Δ119864

119898between more

dangerous and less dangerous areas for a total of 21 casesΔ119864119898of more dangerous areas was statistically higher than

that of less dangerous ones (Plt001)32 Representative Hemodynamic Parameters It is still un-clear which HPs would accurately predict TWAs [3 6 12]Therefore we analyzed the representativeHPsWSS OSI andpressure using CFD analysis on 21 aneurysm cases After the

CFDmodels for 21 caseswere visually alignedwith the corres-ponding intraoperative images each HP was obtained fromthe CFD results at positions corresponding to more danger-ous (max Δ119864

119898) and less dangerous areas in the intraoper-

ative images WSS and OSI were more predictive HPs thanpressure based on statistical analysis (Table 1) LowWSS andhigh OSI corresponded to reddish areas in 17 and 7 casesrespectively while pressure had nothing to do with reddishareas in the images (Figure 4)

33 Weighting Factors Based on representative HPs resultswe decided to include WSS and OSI but not pressure in con-structing CHP and to give WSS more weight than OSI Toinvestigate the effect of weighting factors on CHP statisticalanalysis was performed using a range of weighting fac-tors from 05 to 1 for WSSnorm (119908

1) and from 05 to 0 for

OSInorm (1199082) For all 21 aneurysm cases two CHPs were cal-

culated from the same more dangerous (max Δ119864119898) and less

dangerous areas respectively As1199081was increased the differ-

ence between the two CHPs increased and hence P-valuedecreased (Table 2) The less dangerous region showed analmost identical CHP value regardless of the combination ofweighting factors

34 Comparison of Estimated TWAs and Real TWAs For all21 aneurysm cases regions with the highest CHP values werecompared with TWAs with max Δ119864

119898 Predictions of TWAs

using the highest CHP matched in 17 of 21 cases (81 ) andregions with relatively high CHP values corresponded well tothose with high Δ119864

119898values (Plt001 Figure 5) In the non-

correspondence cases severe atherosclerotic changes differ-ences in geometry between CFD model and intraoperative


122(Max)

108

-108

141(Max)

-125

44

243(Max)

224

51

Figure 3 Demonstration of usefulness of modified Delta E Δ119864119898in intraoperative images (Left) The blue circle is a reference region the

black and yellow circles are less dangerous and more dangerous regions respectively (Right) CalculatedΔ119864119898for selected regions relative the

reference region

Table 1 Comparison of estimated HPs (mean plusmn SD) at the more dangerous (max ΔEm) and the less dangerous regionsa

HP More dangerous region Less dangerous region P-valuePressure Pa 15755 plusmn 738 15752 plusmn 739 0970WSS Pa 0337 plusmn 0705 1759 plusmn 2044 1127E-04OSI 0073 plusmn 0073 0060 plusmn 0084 0187aHP hemodynamic parameter OSI oscillatory shear index WSS wall shear stress

image and local influence of neighboring structure were ob-served in the aneurysms (Figure 6)

4 Discussion

This study is based on the idea that the aneurysm is mainlydeveloped by hemodynamic mechanism Histological studiesof retrieved cerebral aneurysm walls have shown that a re-modeling process of the aneurysm wall results in progressivewall thinning before aneurysm rupture Although the rupturerisk is dependent on several complicated factors as well asTWAs we considered the TWA as the potential rupture re-gion by focusing on the hemodynamic mechanism

41 Delta E Analysis In previous studies TWAswere definedas reddish and transparent areas within aneurysms [9 11 22]This definition relies on subjective visual judgment by clinicalexperts [9 10 15] Kawaguchi et al [10] reported that several

clinical specialists who did not know the CFD results beforecompared intraoperative video at the same time and placeto reduce human errors due to bias or subjective judgmentSuzuki et al [9] compared CFD results with the cliniciansrsquoaverage scores on the redness of the aneurysms in intraoper-ative images to determine whether they were matched eachother However these previous studies did not fully eliminatethe limitations of subjective judgment

Delta E is usually used for color quality control in indus-try Since Delta E represents a relative difference (or distance)between two colors it is suitable for determining the mostreddish location relative to a reference location in a selectedregion of interest even when the region of interest is capturedunder a different light exposure or environment Sometimestwo different colors show the same or similar value of DeltaE relative to a reference color This is because the two colorshave almost the same distance from the reference color inthe 119871lowast119886lowast119887lowast color coordinate system If the line of Delta E isprojected along the 119886lowast axis which represents the degree of


Pressure Wall shear stress Oscillatory shear index Intraoperative image

Figure 4 Comparison between the contours of HPs and intraoperative images

Table 2 Influence of weighting factors on CHP (mean plusmn SD) at the more dangerous (max ΔEm) and the less dangerous regionsa

Weighting factor CHP P-value1199081

1199082

More dangerous region Less dangerous region05 05 0359 plusmn 0105 0122 plusmn 0124 2669E-0606 04 0403 plusmn 0108 0122 plusmn 0121 6624E-0707 03 0449 plusmn 0112 0121 plusmn 0118 1517E-0708 02 0490 plusmn 0121 0121 plusmn 0118 1053E-0709 01 0534 plusmn 0129 0121 plusmn 0120 9342E-0810 00 0579 plusmn 0140 0121 plusmn 0125 6810E-08aCHP combined hemodynamic parameter

redness the component related to redness can be obtainedfromDelta E leading to determining more reddish area Thiswas implemented using Δ11986411989842 Normalized CHP From the analysis of 21 aneurysmcases we found that low WSS and high OSI were more effec-tive parameters for predicting TWAs while pressure was notThese results are similar to those fromprevious studies Xianget al [23] conducted CFD analysis using 38 ruptured and 81unruptured aneurysms and showed that low WSS and highOSI were related to aneurysm rupture Kulcsar et al [24] re-ported from their CFD analysis that the pressure at aneurysmwalls was relatively uniform

LowWSS and high OSI cause upregulation of endothelialsurface adhesion molecules dysfunction of nitrous oxide-inducing flow and increased endothelial permeability [18 2225] These phenomena can lead to weakening of the bloodvessel wall which may result in rupture [2 26] Weakeningthe blood vessel wall is hemodynamically complex phenom-ena but little study has focused on the combination of HPsor its effect on the TWA prediction [12ndash14]

Before combining HPs normalization of each HP shouldbe done using a proper scheme A range of 0 to 1 makes iteasier to compare different HP contributions Since WSS hasboth a maximum and a minimum value in an aneurysm andlowWSS is related to TWA a normalization scheme needs tobe presented in such a way that the minimum WSS becomes1 and the maximum WSS becomes 0 In order to give moreweight on low WSS we have proposed a 4th-order equationbased on the ellipse formula as in (1) For OSI it is alreadynormalized from 0 to 05 and the value of 05 means moredangerous situation Therefore OSI is only doubled to matchthe value range from 0 to 1 Normalizing makes each HPwitha different physical unit and value range contained between 0and 1 so such normalization scheme also enables the value ofeachHP to bemuch less sensitive to inlet and outlet boundaryconditions used in CFD analysis

As listed in Table 2 the most significant difference be-tweenmore dangerous and less dangerous areas was observedwhen 119908

1= 1 and 119908

2= 0 which means that WSSnorm only

produced the best results However it should be noted thatfor two out of 21 casesOSInorm showed better correlationwith


141

122

207

272

243

Figure 5 Representative correspondence cases (Left) TWAs predicted using normalized CHP were compared with (Right) reddish areas ofintraoperative images identified using Δ119864

119898 The red circle is a TWA predicted using the normalized CHP

max Δ119864119898than WSSnorm In addition almost identical CHP

values regardless of weighting factors for the less dangerousregion means that WSSnorm and OSInorm have a similar valuefor that region due to the normalization effect

43 Comparison of Normalized CHPs and Real TWAs TWAspredicted using the highest CHP agreed with regions of maxΔ119864119898in 17 of 21 caseswhile four cases did notmatch this trend

DSA is an imaging technique that only shows images of bloodvessel lumen so DSA image does not provide informationon the thickness of blood vessel wall or any geometricalchange inside the wall [27] 3D aneurysm model obtained

from DSA is not perfectly identical to the actual aneurysmshape observed in an intraoperative image If an aneurysmhas thick-walled areas due to atherosclerotic changes thedifference in shape becomes more severe In some cases theactual TWAs might be hidden or buried in the surroundingstructures so they could not be observed in the intraoperativeimages As a result a mismatch could occur in these cases

44 Limitations One of limitations of this study is that theboundaryconditions inCFDanalysiswere not patient-specificAlthough the influence of the boundary conditions is reducedby using the normalization scheme for HPs application of


189(Max)

215(Max)

249(Max)

Figure 6 Representative noncorrespondence cases Rows 1-2 severe atherosclerotic changes row 3 influence of surrounding structure orrow 4 geometry differenc observed in the aneurysm

patient-specific boundary conditions to CFD analysis is moredesirable Second the number of cases investigated is rela-tively small Additional data would allow for more accuraterelationship between the normalized CHP and TWAsThirdCFD analysis itself has technical limitations Since bloodvessels were assumed to be a non-slip and rigid wall forsimplicity it is difficult to determinewhether rupture of aneu-rysm occurs from CFD analysis only The elastic walls ofblood vessels should be included in the analysis to investigatethe rupture of aneurysms which can be implemented byfluid-structure interaction (FSI) technique Combination ofFSI simulation and TWApredictionmethod suggested in thiswork would help to estimate the risk of individual aneurysmruptures This work is under way and will be addressed infuture work

5 Conclusions

We have suggested new normalized schemes for WSS andOSI as well as the combination of the two as a weighted

average of WSSnorm and OSInorm which enable to predictTWAs of unruptured cerebral aneurysms fromCFD analysisAlthough there are several limitations this study mightprovide a clinically potential way to predict vulnerability ofcerebral aneurysm It is expected that further FSI analysis thatincorporates TWA prediction method proposed in this workwould help to diagnose the future risk of target lesions andselect the more appropriate treatment strategy for patientswith cerebral aneurysms

Data Availability

Thedata used to support the findings of this study are includ-ed within the article

Conflicts of Interest

The authors declare that there are no conflicts of interest


Authorsrsquo Contributions

KC Cho and JH Choi contributed equally as co-first au-thors JH Oh and YB Kim contributed equally as co-corres-ponding authors

Acknowledgments

This study was supported by a faculty research grant of YonseiUniversity College of Medicine [6-2016-0076]References

[1] J Frosen A Piippo A Paetau et al ldquoRemodeling of saccularcerebral artery aneurysm wall is associated with rupture histo-logical analysis of 24 unruptured and 42 ruptured casesrdquo Strokevol 35 no 10 pp 2287ndash2293 2004

[2] K Kataoka M Taneda T Asai A Kinoshita M Ito and RKuroda ldquoStructural fragility and inflammatory response of rup-tured cerebral aneurysms a comparative study between rup-tured and unruptured cerebral aneurysmsrdquo Stroke vol 30 no7 pp 1396ndash1401 1999

[3] H Meng VM Tution J Xiang and A Siddiqui ldquoHighWSS orLow WSS Complex interaction of hemodynamics with intra-cranial aneurysm initiation growth and rupture toward a uni-fying hypothesisrdquo American Journal of Neuroradiology 2013

[4] J R Cebral F Mut J Weir and C M Putman ldquoAssociation ofhemodynamic characteristics and cerebral aneurysm rupturerdquoAmerican Journal of Neuroradiology vol 32 no 2 pp 264ndash2702011

[5] Y Qian H Takao M Umezu and Y Murayama ldquoRisk analysisof unruptured aneurysms using computational fluid dynamicstechnology preliminary resultsrdquo American Journal of Neurora-diology vol 32 no 10 pp 1948ndash1955 2011

[6] H Takao Y Murayama S Otsuka et al ldquoHemodynamic differ-ences between unruptured and ruptured intracranial aneur-ysms during observationrdquo Stroke vol 43 no 5 pp 1436ndash14392012

[7] M Shojima M Oshima K Takagi et al ldquoMagnitude and roleof wall shear stress on cerebral aneurysm computational fluiddynamic study of 20 middle cerebral artery aneurysmsrdquo Strokevol 35 no 11 pp 2500ndash2505 2004

[8] SOmodaka S-I Sugiyama T Inoue et al ldquoLocal hemodynam-ics at the rupture point of cerebral aneurysms determined bycomputational fluid dynamics analysisrdquo Cerebrovascular Dis-ease vol 34 no 2 pp 121ndash129 2012

[9] T Suzuki H Takao T Suzuki et al ldquoDetermining the Presenceof Thin-Walled Regions at High-Pressure Areas in UnrupturedCerebral Aneurysms byUsingComputational FluidDynamicsrdquoNeurosurgery vol 79 no 4 pp 589ndash595 2016

[10] T Kawaguchi S Nishimura M Kanamori et al ldquoDistinctiveflow pattern of wall shear stress and oscillatory shear indexsimilarity and dissimilarity in ruptured and unruptured cere-bral aneurysm blebsrdquo Journal of Neurosurgery vol 117 no 4 pp774ndash780 2012

[11] L M Kadasi W C Dent and A M Malek ldquoColocalization ofthin-walled dome regions with low hemodynamic wall shearstress in unruptured cerebral aneurysms clinical articlerdquo Jour-nal of Neurosurgery vol 119 no 1 pp 172ndash179 2013

[12] J R Cebral M A Castro J E Burgess R S Pergolizzi MJ Sheridan and C M Putman ldquoCharacterization of cerebralaneurysms for assessing risk of rupture by using patient-specific

computational hemodynamics modelsrdquo American Journal ofNeuroradiology vol 26 no 10 pp 2550ndash2559 2005

[13] C O Willis K N T Kayembe M Sasahara and F HazamaldquoCerebral aneurysms and variations in therdquo Stroke vol 15 no5 pp 846ndash850 1984

[14] P Berg G Janiga O Beuing M Neugebauer and DTheveninldquoHemodynamics in Multiple Intracranial AneurysmsThe Roleof Shear Related to Rupturerdquo International Journal of BioscienceBiochemistry and Bioinformatics pp 177ndash181 2013

[15] J Song J E Park H R Kim and Y S Shin ldquoObservation ofcerebral aneurysm wall thickness using intraoperative micro-scopy clinical and morphological analysis of translucent aneu-rysmrdquo Neurological Sciences vol 36 no 6 pp 907ndash912 2015

[16] Y Hua J H Oh and Y B Kim ldquoInfluence of parent arterysegmentation and boundary conditions on hemodynamic char-acteristics of intracranial aneurysmsrdquo Yonsei Medical Journalvol 56 no 5 pp 1328ndash1337 2015

[17] K Kono T Fujimoto A Shintani and T Terada ldquoHemody-namic characteristics at the rupture site of cerebral aneurysmsA case studyrdquo Neurosurgery vol 71 no 6 pp E1202ndashE12082012

[18] P Reymond F Merenda F Perren D Rufenacht and NStergiopulos ldquoValidation of a one-dimensional model of thesystemic arterial treerdquo American Journal of Physiology-HeartandCirculatory Physiology vol 297 no 1 ppH208ndashH222 2009

[19] A M Malek S L Alper and S Izumo ldquoHemodynamic shearstress and its role in atherosclerosisrdquoThe Journal of the AmericanMedical Association vol 282 no 21 pp 2035ndash2042 1999

[20] X He and D N Ku ldquoPulsatile flow in the human left coronaryartery bifurcation average conditionsrdquo Journal of BiomechanicalEngineering vol 118 no 1 pp 74ndash82 1996

[21] G Sharma W Wu and E N Dalal ldquoThe CIEDE2000 color-difference formula implementation notes supplementary testdata and mathematical observationsrdquo Color Research amp Appli-cation vol 30 no 1 pp 21ndash30 2005

[22] LM KadasiW C Dent and AMMalek ldquoCerebral aneurysmwall thickness analysis using intraoperative microscopy effectof size and gender on thin translucent regionsrdquo Journal ofNeuroInterventional Surgery vol 5 no 3 pp 201ndash206 2013

[23] J Xiang S K Natarajan M Tremmel et al ldquoHemodynamic-morphologic discriminants for intracranial aneurysm rupturerdquoStroke vol 42 no 1 pp 144ndash152 2011

[24] Z Kulcsar A Ugron M Marosfoi Z Berentei G Paal and ISzikora ldquoHemodynamics of cerebral aneurysm initiation therole of wall shear stress and spatial wall shear stress gradientrdquoAmerican Journal of Neuroradiology vol 32 no 3 pp 587ndash5942011

[25] A M Nixon M Gunel and B E Sumpio ldquoThe critical role ofhemodynamics in the development of cerebral vascular diseaseA reviewrdquo Journal of Neurosurgery vol 112 no 6 pp 1240ndash12532010

[26] L Pentimalli A Modesti A Vignati et al ldquoRole of apoptosisin intracranial aneurysm rupturerdquo Journal of Neurosurgery vol101 no 6 pp 1018ndash1025 2004

[27] A W Leber A Knez F Von Ziegler et al ldquoQuantification ofobstructive and nonobstructive coronary lesions by 64-slicecomputed tomography A comparative study with quantitativecoronary angiography and intravascular ultrasoundrdquo Journal ofthe American College of Cardiology vol 46 no 1 pp 147ndash1542005


is still on-going controversy regarding which HP most accu-rately predicts TWAs Meng et al [3] reported that low wallshear stress (WSS) and high oscillatory shear index (OSI)triggered the growth and rupture of large aneurysm pheno-types whereas highWSS was associated with the growth andrupture of small aneurysm phenotypes and medial thinningTakao et al [6] concluded that pressure loss coefficient mightbe a potential parameter to predict future rupture of unrup-tured aneurysms Cebral et al [12] analyzed a rupture riskof cerebral aneurysms considering flow pattern and impinge-ment region

There are several limitations of current TWA predictionmethods First TWAs are predicted based on only single HPfrom CFD analysis although they are usually caused by com-plex phenomena [12ndash14] Second the value of HP from CFDanalysis is greatly influenced by the inlet and outlet boundaryconditions which are not only hard to accurately obtainbut also different from each patient Third identification ofTWAs in intraoperative images is subjective and significantlydependent on the visual judgment of the clinician [9 10 15]Finally as aforementioned it is still arguable which HP is themost appropriate to predict TWAs To predict TWAs moreaccurately the influence of CFD analysis conditions shouldbe minimized by properly normalizing the HP value andcombination of normalized HP values could be introducedto enhance the prediction accuracy The quantitative digitalcolor analysis of intraoperative images would help to validatethe effectiveness of normalized combined hemodynamicparameters (CHPs)

In this study we performed CFD analyses of 21 unrup-tured cerebral aneurysms and newly suggested both normal-ized schemes and a normalized CHP as a weighted average ofnormalized WSS and normalized OSI We also investigatedthe influence of weighting factors of CHP on the predictionof TWAs In order to avoid the cliniciansrsquo subjective assess-ment when determining the TWAs from correspondingintraoperative images the quantitative digital color analyseswere conducted based on CIE Delta E The results of DeltaE were compared with those of CHPs from CFD analysesto evaluate the effectiveness of the suggested normalizationscheme and CHP for TWA prediction

2 Materials and Methods

21 Patients and Image Data Acquisition This study analyzeda total of 21 unruptured aneurysms in 19 patients (14 womenand 5 men mean age 59 years mean aneurysm size 51 plusmn115 mm) All data were obtained from patients treated viasurgical neck clipping from July 2014 to December 2016 inour institution The protocols used in this studywere ethicallyapproved by our institutional review board and patient con-sent was waived Ruptured aneurysms were excluded becauseit was difficult to obtain prior shape data and the shapewas unclear Digital subtraction angiography (DSA) imageswere generated using Artis zee biplane with syngoWorkplacesoftware (Siemens Erlangen Germany) All intraoperativeimageswere captured using anOPMIPentero surgical micro-scope (Carl Zeiss Microscopy GmbH Jena Germany) witha three-chip charge-coupled device digital video camera at

640 x 480 resolution or a mounted Canon EOS 5D digitalsingle lens reflex camera at 4368 x 2912 resolution

22 Computational Fluid Dynamics Blood vessel and aneu-rysm geometry were generated from the DSA images Bloodvessel models were modified by using shape modifica-tion software (CATIA V5-6R2012 Dassault Systems ParisFrance Meshmixer version 110544 Autodesk San RafaelCA) Based on our previous study (Figure 1(a))[16] the inletlength of CFDmodels was set to 20 times larger than the inletdiameter and the outlet length was set to twice larger thanthe outlet diameter to obtain reliable simulation resultsCFD analysis was performed using ANSYS Fluent software(version 150 ANSYS Inc Canonsburg PA) The Navier-Stokes equations were solved to predict blood flow Bloodwas modeled as an incompressible Newtonian fluid with adensity of 1055 kgm3 and a viscosity of 0004 Pasdots and theflow was assumed to be laminar Blood vessels were modeledwith the assumption of no-slip and rigid-wall conditionsPulsatile flow with a Womersley velocity profile was used asthe inlet boundary condition Womersley velocity profile foreach model was calculated from the volumetric flow rate ofa humanmeasured at the internal carotid artery [17] Pressureprofile based on pressure waveforms measured at the super-ficial temporal artery [18] was applied at each outlet Detailedinformation on CFD analysis is provided in Figures 1(b) and1(c) After the mesh independence check the number ofmeshes was determined to 45 million to 55 million

23 Normalization Scheme and CHP Individual HPs suchas WSS OSI and pressure have different physical meaningsunits and value ranges so direct comparison of their values isnot reasonable WSS is the dynamic friction force generatedby viscous fluid moving along the surface of the solid walland is defined as the product of fluid viscosity and shearingvelocity gradient at the vascular wall [19] OSI is a dimen-sionless parameter that indicates how the direction of WSSchanges at a specific location during a cardiac cycle In orderto equally compare each HPrsquos contribution the value of HPshould be normalized to have a value between 0 and 1

4th-order equation based on the ellipse formula was pro-posed to create the normalized WSS WSSnorm with a fixedrange of 0 to 1 as follows

(119882119878119878119899119900119903119898minus 1)4 + ( 119882119878119878 minus119882119878119878max119882119878119878max minus119882119878119878min

)4 = 1 (1)

where WSSmin and WSSmax are the minimum and the max-imum values of WSS in the aneurysm sac above its neck atthe diastolic time respectively IfWSS at the diastolic time issubstituted into (1) thenWSSnorm is calculated in the range of0 to 1 Note thatWSSnorm is 0 forWSSmax and 1 forWSSmin119874119878119868 is originally defined as

119874119878119868 = 12 (1 minus100381610038161003816100381610038161003816int1199050 119908119904119904119894 119889119905100381610038161003816100381610038161003816int1199050

10038161003816100381610038161199081199041199041198941003816100381610038161003816 119889119905) (2)

where119908119904119904i and t represent an instantaneous WSS vector andthe cycle duration respectively [20] OSI ranges from 0 to


Inlet

Normal vessel

Outlets

Aneurysm

(a)

Density 1055 kgm3






(b)

Systolic time

Diastolic time

Systolic time

Diastolic time

0

15

25

35

45

55

Volu

met

ric F

low

Rat

e (cm

3 s)

04 08 12 16 20Time (s)

0

5000

10000

15000

20000

Pres

sure

(Pa)

04 08 12 16 20Time (s)

(c)



119874119878119868119899119900119903119898= 2 times 119874119878119868 (3)


119862119867119875 = 1199081times (119882119878119878

119899119900119903119898) + 1199082times (119874119878119868

119899119900119903119898) (4)

1199081+ 1199082= 1 (5)

where 1199081and 119908






Δ119864119898= 119863119890119897119905119886 119864 sdot radic 119886221198712

2+ 11988622

(6)




-a(Green) -b

(Blue)

A(Red)

b(Yellow)

L=100(White)

L=0(Black)

Delta E

Reference color

(a)





Sample image

Thin wall

Coordinates Delta E

(b)



119898 and it was








119898between more


















122(Max)

108

-108

141(Max)

-125

44

243(Max)

224

51



reference region




4 Discussion










1199082






1= 1 and 119908




141

122

207

272

243










189(Max)

215(Max)

249(Max)



5 Conclusions



Data Availability







Acknowledgments































Inlet

Normal vessel

Outlets

Aneurysm

(a)

Density 1055 kgm3






(b)

Systolic time

Diastolic time

Systolic time

Diastolic time

0

15

25

35

45

55

Volu

met

ric F

low

Rat

e (cm

3 s)

04 08 12 16 20Time (s)

0

5000

10000

15000

20000

Pres

sure

(Pa)

04 08 12 16 20Time (s)

(c)



119874119878119868119899119900119903119898= 2 times 119874119878119868 (3)


119862119867119875 = 1199081times (119882119878119878

119899119900119903119898) + 1199082times (119874119878119868

119899119900119903119898) (4)

1199081+ 1199082= 1 (5)

where 1199081and 119908






Δ119864119898= 119863119890119897119905119886 119864 sdot radic 119886221198712

2+ 11988622

(6)




-a(Green) -b

(Blue)

A(Red)

b(Yellow)

L=100(White)

L=0(Black)

Delta E

Reference color

(a)





Sample image

Thin wall

Coordinates Delta E

(b)



119898 and it was








119898between more


















122(Max)

108

-108

141(Max)

-125

44

243(Max)

224

51



reference region




4 Discussion










1199082






1= 1 and 119908




141

122

207

272

243










189(Max)

215(Max)

249(Max)



5 Conclusions



Data Availability







Acknowledgments































-a(Green) -b

(Blue)

A(Red)

b(Yellow)

L=100(White)

L=0(Black)

Delta E

Reference color

(a)





Sample image

Thin wall

Coordinates Delta E

(b)



119898 and it was








119898between more


















122(Max)

108

-108

141(Max)

-125

44

243(Max)

224

51



reference region




4 Discussion










1199082






1= 1 and 119908




141

122

207

272

243










189(Max)

215(Max)

249(Max)



5 Conclusions



Data Availability







Acknowledgments































122(Max)

108

-108

141(Max)

-125

44

243(Max)

224

51



reference region




4 Discussion










1199082






1= 1 and 119908




141

122

207

272

243










189(Max)

215(Max)

249(Max)



5 Conclusions



Data Availability







Acknowledgments



































1199082






1= 1 and 119908




141

122

207

272

243










189(Max)

215(Max)

249(Max)



5 Conclusions



Data Availability







Acknowledgments































141

122

207

272

243










189(Max)

215(Max)

249(Max)



5 Conclusions



Data Availability







Acknowledgments































189(Max)

215(Max)

249(Max)



5 Conclusions



Data Availability







Acknowledgments

































Acknowledgments






























Prediction of Thin-Walled Areas of Unruptured Cerebral ... · BioMedResearchInternational...

Documents

Transcript of Prediction of Thin-Walled Areas of Unruptured Cerebral ... · BioMedResearchInternational...