This fellowship report summarizes the author's work in the Ultrasound Research Lab over 8 weeks. Key activities included learning to use lab instruments like ultrasound scanning microscopes, analyzing fetal cardiac scans using Matlab, performing literature reviews on ultrasound bone analysis research, and assisting with a student senior design project on simulating osteoporosis in bovine bone samples. Based on this experience, the author developed a new detailed osteoporosis experiment protocol to improve on issues identified in the senior design project and provide more robust data.
ISMB2014読み会 イントロ + Deep learning of the tissue-regulated splicing code
Ultrasound Analysis of Bone and Heart Tissue
1. 2016 HOFSTRA SEAS SUMMER
FELLOWSHIP REPORT
Megan Murphy
Dr.Ghorayeb
July 22nd, 2016
2. 1
SUMMARY
Thissummer,I workedasa lab assistantinthe UltrasoundResearch Lab withDr. Ghorayeb for
eightweeks.The goal of mypositionwasto be introducedtothe instrumentationandtechniquesused
inthe UltrasoundResearchLab.Thisposition involved the followingtasks:
1) Learn howto use the instrumentsinthe Lab
2) Learn Matlab (Mathworks,R2015b, Natick,MA.)
3) Analyze fetal cardiacscans
4) Performliterature reviewsof previouslyperformedresearchexperiments
5) Assistin SeniorDesign Experiments
6) Develop aprotocol fora new osteoporosisexperiment
I have learnedhowtouse the scanningacousticmicroscope (SAM) systems,the biomechanical
testingsystem,andtheirassociatedacquisitionprograms.Myworkalsoinvolvedassistingother
bioengineeringstudentsworkingontheirSeniorDesign component. Beforeusingthe instruments
myself,Ireadaboutthe physics of ultrasoundinorderto understandhow the SAMsystems function.
Anotheraspectof my fellowshipincluded the use of Matlabtoanalyze fetal cardiacscans
providedbythe Centerof Maternal Fetal Medicine atNorthwellHealthHospital.
Moreover,numerousliterature reviewswere partof thisresearchfellowship.Forinstance,I
readand reportedonpreviousresearchworkssuchas:
1) FrequencySpecificUltrasoundAttenuationisSensitivetoTrabecularBone Structure1
by
Dr. Lin
2) UltrasonicEvaluation of Bone QualityinCadaverIlia2
byDr. Ghorayeb
3) Microarchitecture,the KeytoBone Quality3
byDr. Brandi
4) Mechanical Propertiesof aSingle CancellousBone TrabeculaeTakenfromBovine
Femur4
byDr. Enoki
5) Contact StressDistributiononthe Femoral Headof the Emu5
by Dr. Troy
In addition,Iwrote anewprotocol regardingthe osteoporosisprojectinpreparationforfuture
SeniorDesign endeavors.
3. 2
ULTRASOUND
The firstday in the lab,Dr. Ghorayebtaughtme how to use the Sonix SAM systemandsoftware.
I tookseveral practice scans of a piece of plexiglasscontainingfourmachine-drilledholes.Ilearnedhow
to adjustthe resolution,the dimensions,and locationof the transducerforthe scans.It is very
importantto take the scan at the focal length,which isthe pointwhere the lensismostfocusedand
receivesthe clearest returnsignals.Focal lengthis transducerspecific.Anotherimportantguideline isto
be verycareful withthe scan windowinwhich the datais collected(Figure 3).Dependingonthe size of
the window,the programmaybe collectingtoolittleof the scanand missingdata,or it may be
collectingtoomuch of the scan andskewingthe results.
Afterthe practice scans, I read and tookdetailednotesona packetaboutthe terminologyand
physicsbehindthe system.Learningthe terminologywasveryhelpful becauseitmade iteasier forme
to communicate withthe otherstudentsinthe lab whenperformingultrasound scansof bone samples.
Also,knowingaboutthe science made iteasierforme to decide whattype of transducertouse inthe
protocol I wrote at the endof the summer.
While workingwith studentsontheirSeniorDesign component,I usedboth low-frequencyand
highfrequency ultrasonicSAMsystems inthe lab (Figure 1 andFigure 2). The systems are verysimilar
and relativelyeasytouse.Boththe OKOSand SONIX machinesare verydelicate andare connected to
computerseachsetup withOKOS acquisition software whichdisplays the A-scanwave forms,scan
specifications, transducerheightadjustmentarrows,andthe resultsafterthe scan.
Figure 1: Low Frequency Ultrasonic Scanning Acoustic Microscope system.
4. 3
A)
B) C)
Figure 2: High Frequency Ultrasonic Scanning Acoustic Microscope system.
Figure 3: Outputof ODIS acquisition system. A) Wave during the A-scan. B) Arrows
used to adjust the position of the transducer on the machine before the scan. C)
Measurements used to adjust the size, area, and resolution of the scan.
5. 4
MATLAB
I spentthe firstweekreacquaintingmyself with Matlab. Whenanalyzingmedical scans, medical
ultrasonicimagesare analyzedusing aMatlabprogram developedbyDr.Ghorayeb.The program
performstexture analysisanddetermines the homogeneity leveloveraspecificareaof the scan. Dr.
Ghorayebwantedme to be able to understand the functionandpurpose of eachline of the code soI
wouldbe able to design a similarcode forfuture experimentsif necessary.Iwentthroughthe file and
figuredoutwhatresultedfromeachline of the code.
I spentmanyhourstryingto adjustthe program to automatically inputthe datafound inMatlab
intoan excel spreadsheet.The waythe file worksiswhenthe file is run,awindow popsupwhichasks
the userto choose a picture to analyze.Afterapicture isselected,aseparate windowwith the chosen
picture opens.Byclickingtwopointsonthe picture,the user createsa regionof interest (ROI) forthe
program to collectdatafrom.Then,the program calculates the level of homogeneityin the ROI and
displaysthe maximum,minimum, andaverage homogeneityof the region.
Currently,whendataisfound,the datafor the average isautomaticallyputinthe computer’s
clipboard (Figure 4) andthat data can be inserted intoaseparate excel file bypasting.Thisisa
convenientfeature whenonly the average isrequired:however, itissometimes helpfultolookatall
data valuescollected. Iwantedtoadjustthe code so the maximum, minimum, andaverage datawould
be automaticallysaved inrespective columns inaspecificexcelfile.Ialsowantedthe programto be
able to shiftdowna row whena newpicture ischosensothe new data wouldbe inputinitsownrow.I
couldnot finda code that wouldkeepthe programfromcontinuously writingthe newestdatacollected
intothe same row, erasingthe previouslycollected data.
Figure 4: File excerpt that copies the average data value directly to the clipboard.
6. 5
FETAL CARDIOLOGY
As part of a jointcollaborationbetweenthe Maternal Fetal Medicine CenteratNorthwell and
Dr. Ghorayeb,I got involvedinthe analysisof fetal cardiacultrasound scanstosee whetheracorrelation
existedbetweenthe tissuehomogeneityandthe heartconditionsfoundin the fetuses. Priorresearch
has showedinotherorgansthat levelsof homogeneityorheterogeneity were anindicationof certain
conditions.These levelscanalsoshowif the organis well developedornot.Forexample,the more
heterogeneous afetal lungis,the more developeditis.We were curioustosee if there wouldbe a
similartrendinthe fetal heart.
Figure 5: Cardiac scan with ROIs on the septum (top) and left ventricle (bottom).
7. 6
Whenrunningthroughthe scans, I took separate homogeneityreadingsforthe septumandthe
lowerleftventricularwall ineachimage andrecordedall valuesinanexcel file.We came torealize that
it wasdifficulttochoose agood ROI to use to analyze the scans.Since each scan was ina different
orientationandslightlyblurry, itwasveryhardto distinguish the partsof the heart.Fetal heartsare very
small andit isunderstandable thatitishardto get a clear scan,but clearerscanswitha more uniform
orientationwouldhave beenmore useful.Italsowouldhave beenbetter if everyhearthadthe same
condition.Eachscan we receivedandanalyzedhadadifferentcondition,sothere wasnowayto
correlate a specificconditionwithatrendinhomogeneity.We askedformore scansto findmore
information,but we neverreceived them.
Figure 6: Data found after scans analyzed with Matlab.
8. 7
LITERATURE REVIEW
Duringmy fellowshipIwasgiventworesearchpapersaboutthe effectsof osteoporosisin
trabecularbone.The firstresearchpaperI receivedwasbyDr.Lin et al.1
, titled“FrequencySpecific
UltrasoundAttenuationisSensitive toTrabecularBone Structure,”andthe secondpaperwasby Dr.
Ghorayeb et al. 2
, titled“UltrasonicEvaluationof Bone QualityinCadaverIlia.”
Currently,the mostpopularwaytomeasure bone propertiesisdual energyx-ray
absorptiometry(DXA),butDXA usesionizingradiation andthere are limitationsonhow much
informationthe x-rayscancan reveal.Anothersystemthatmeasuresbone propertiesisbroadband
ultrasoundattenuation(BUA),whichisnoninvasive,butitisnotverysensitivebecause itcanonlywork
ina frequency range between300and 700 kHz.Dr. Lin’sstudyinvestigatedthe possibilityof using
anotherformof ultrasoundcalled frequencymodulatedultrasoundattenuation(FMA) toevaluate
trabecularstructural properties.FMA canfunctionina much widerrange thanBUA, so for this
experimentfourdifferentfrequencybandsinthe range between300 KHzand 1.9 MHz. Dr. Linused
throughtransmissionforthe ultrasonicscans,andhe was searchingtosee if there wasa correlation
withthe effectivenessof the bandrange andthe orientationof the bone.Itwasfoundthat there wasa
connectionwiththe orientationandfrequencybandinthe anterio-posterial orientation,butinthe end
it wasconcludedthatFMA shouldbe usedtosupplementBUA findingsbecause FMA relies onthe
linearityof the attenuationfrequencyrelationship.
Dr. Ghorayeb’sstudyalsolookedintothe possibilitiesof usingquantitative ultrasound(QUS)
toolsto measure andmonitorbone density.Inthisexperiment,informationaboutthe iliasampleswas
gatheredwithultrasound,the same informationwasfoundwithDXA,andthenthe resultswere
compared.Itwas foundthat the information gatheredwithQUS hadan error of 3.5% whencompared
to the DXA measurements,butitwasproportional tothe trends inpercentbone lossandbone mineral
density.The conclusionforthisexperimentwasthatultrasoundmeasurementsof the bone properties
wouldbe veryhelpful withthe monitoringof progressionorregressionof osteoporosisinalivingbeing.
AfterI readbothpapers,I searched forsimilarstudies online.The thirdpaperIread,
“Microarchitecture,the KeytoBone Quality,”byDr. Brandi et al. 3
, whichstressedthe importance of
knowledge of the microarchitecture in the trabecularbone.Osteoporosisweakensandbreaksdownthe
trabecularbone connections,andasthe disease progresses,the bone becomeshollower,whichmakes
9. 8
it harderfor the bone to resistfracture.Dr.Brandi concluded thatthe way to bestdiagnose andtreat
osteoporosisisbymeasuring the spatial distributionof the microarchitectureandthe bone mass.She
mentionedboth DXA andCT scanning asa meansto provide measurementsof bone massandactual
imagesof the bone’sactual internal architecture.
AnotherpaperIread that focusedon microarchitecture was “Mechanical Propertiesof aSingle
CancellousBone Trabeculae TakenfromBovineFemur,”byDr.Enoki et al. 4
,whichfocused onboth
cancellousandtrabecularbone andelasticityindifferent anatomicalorientations.Cancellousand
trabecularsampleswere obtainedfromthe bovinefemur,andall testswere performedin each
anatomical orientation.Compressiontestswere performedonthe cancellousbone samplesandthree
pointbendingtestswere performedonthe trabecularsamples. The testsleadtothe conclusions that
trabecularbone structure hasinfluence overthe mechanical propertiesof cancellousbone,and that
orientationaffectsthe elasticityinbothtypesof bone.
Finally,I reada paperslightlyoutsideof whatIspentthe summerresearchingcalled, “Contact
StressDistributiononthe Femoral Headof the Emu,”by Dr. Troy et al. 5
whichfocusedonthe loadingat
the hipand the femoral headof the emu.The study wastestingemufemursto see if the emu wasa
viable candidate forsimulationsof osteonecrosis inhumanhips.The datafoundthatthe emuwould in
fact be ideal forsuch simulations.
10. 9
SENIOR DESIGN ASSISTANCE
As mentionedpreviously,partof thisfellowshipwastobe involvedwithassisting astudent
workingonthe biomedical engineering SeniorDesign component.The aimof the projectwasto study
levelsof osteoporosisinducedinbovine femoral headsamples.The studyfollowedthe protocol
describedbyLin et al. 5
. Seventrabecularbone sampleswere cutoutof a bovine femurhead withthe
dimensionsof 1 cm X 1cm X .5cm. The bone marrow was cleaned off the samples withahighpressure
washer,andthe sampleswere then wrappedingauze soakedinX10phosphate buffersaline.
Sevensamples wererequired becauseeachsample wasanexample of adifferentdegree of
osteoporosis,whichwas simulatedbytimedexposuretoacid. Twosamplesservedascontrols,andfive
sampleswere exposedto1.8%formicacid at 20 minute time intervals(20min.,40 min.,60 min.,80
min.,100 min.).
Aftertwodays of beingimmersedina buffersalinesolution,we demineralized the samples
were thendemineralizedbyplacingtheminfivebeakers thatwere filled halfwaywith 1.8% formicacid
and placedona shaker.The shakerwasset to 120 rotationsperminute.Eachbeakerwaslabeledto
showhowlongeach sample wasexposedtothe acid.Afterthe firsttwentyminutespassed,we tookthe
20 min.sample off the shaker,removedthe samplefromthe beaker,andplaceditinammoniaforthirty
minutestostopthe demineralizationprocess.Everytwentyminutesafterthat,we repeatedthe same
stepsforthe 40 min.,60 min.,80 min.,and100 min.,samples.Afterthe demineralizationprocesswas
finished forall samples,we placedthe trabecularbones innew beakersfilledwithwaterandputthem
inthe fridge.
The followingdaywe performed ultrasoundscansof eachsample with a75 MHz transducer
usingthe OKOSSAM system.Afterwe took the scansof eachsample,we sentthe scan picturestoDr.
Ghorayeb.Once received,the scans were analyzedwith Matlabtofind the porosityin eachsample. The
data was close to what we expected:however, we believethe datawasbeingskewedbysomeof the
cortical bone that wasnot removedin the 80 min.and100 min.samplesbecause bothsamplesshould
have been muchmore porousthan whatthe analysisshowed.
The nextday,we foundthe wetanddry mass of each sample.The followingweek,we
performed mechanicaltestingonthe samples withanInstron biomechanical testing machine.After
adjustingthe settings of the machine, asample wasplacedona metal plate.Force wasthenappliedto
each sample until the bone broke,whichallowedustofindthe Young’sModulusforeachsample.
11. 10
PROTOCOL DEVELOPMENT
Afterassistingwiththe SeniorDesign,Iwasaskedtowrite a more detailedandspecificprotocol
of a similarexperimentwith alargersample size basedon Dr.Lin’sandDr. Ghorayeb’spapers. While
workingonthe SeniorDesign,Inoticedafew problemsthatcouldhave alteredthe endresults.
One issue dealtwithcortical bone thatwasleft onsome of the samples.Cortical bone ismuch
strongerthan trabecularbone anditwouldnot be as affectedbythe demineralizationtreatmentasthe
trabecularbone.Since the cortical layerwasstill strongduringthe ultrasound andmechanical tests,the
resultswere skewed.If there wasnocortical onthe samples, the sampleswouldhave appearedmore
porouson the ultrasoundscans,and the samples mighthave brokenfasterduringmechanical testing.
Alsothe presence of cortical bone didnotallow ultrasoundwavestofullypropagateintothe trabecular
layer,therebygivingrise toa more homogeneousreadingof the sample.
Additionally,the aciddidnotuniformlydemineralize all sides of the samplesbecause of the way
they were sittinginthe beakersduringthe demineralization process shaking.Some of the sampleswere
lyingflatagainstthe bottomor the side of the beaker,whichprotected aside fromthe acid.The
protectedside wasthen strongerandlessporousthanthe rest of the sample,whichskewed the
porosity resultsandthe mechanical testing.
Finally,whenimmersingthe bonesinX10phosphate buffersolution,the gauze wassqueezed to
remove anyexcesssolution beforewrappingthe bone.This causedthe samplestodryoutwhichmade
the samplesnotwell-preparedforthe restof the experiment.
Whenwritingoutthe protocol for the new osteoporosisexperiment,Iwasveryspecificand
detailedabouthoweverystepistobe done tomake sure that the resultswouldnotbe skewed.Inthis
newprotocol,Idecidedthatthere are to be seventrabecularsamplesfromfive separatebovine femur
headsto increase the sample size andgetmore informationoutof the experimentasawhole.Itis
expectedthatthe samplesfrom differentboneswill all reactsimilarly,butthere maybe some slight
variation.Ispecifiedthatpre-demineralizationscanswillbe takenof everysample inaspecific
orientation.Thiswill give usanideaof how porousthe samplesare before the demineralizationtests.
Afterwe performthe demineralization,we will be able tofindthe exactpercentage of porosityandthe
difference between the initial sampleandsample aftertreatment.The orientation the scanistakeninis
alsoimportant. If post-demineralizationscansare takenwiththe same orientationasthe pre-scans,we
12. 11
will hopefullybe able tosee the same “landmarks”withinthe scansasthe brightnessof the scans
decreaseswithincreasedtime inacid.
AfterI finishedwritingthe protocol,Dr.GhorayebcontactedDr.Lin at StonyBrook to see if he
wouldlike toworkwithuson this experiment. StonyBrooklabhasdifferentlabequipmentthan
Hofstra,whichwouldallowustofindmore informationfromthe experiment.If theyusedtheirdiamond
cutterto cut outthe trabecularsamplesfromthe femur heads,theywould be able tobe remove all
unwantedcortical bone. The diamondcutterwouldalsoexpose the samplestolessheatthenaregular
bandsaw. StonyBrookalso hasa throughtransmissionultrasoundmachinethatwould allow ustofind
the Young’sModulus by findingthe velocityatwhichthe ultrasoundsignal passesthroughthe samples
ineach anatomical orientation.A collaborationcouldleadtoa more accurate andinformational
experiment,whichhasthe potential tobe veryimportantandinfluential.
Figure 7: Outlined steps of the Experiment
Protocol described in this section.
13. 12
CONCLUSIONS
The SummerResearchFellowshipProgramallowedme tolearn manythingsthat I needtoknow
to be able to contribute to the work done inthe UltrasoundResearch Lab,and I am eagerto conduct
researchand workon future experimentsincludingmySeniorDesign.Ireallyappreciatedthe
opportunitytoworkwithDr. Ghorayeb,andI believethatthissummerpositionwas educational and
successful,andthatitwill lead towardsmore importantresearchinthe nearfuture.
14. 13
REFERENCES
1
Lin W., Serra-HsuF.,ChengJ.,and QinY.-X.,"FrequencyspecificultrasoundAttenuationissensitive to
Trabecularbone structure," Ultrasound in Medicine& Biology,vol.38, no.12, pp.2198–2207, Dec. 2012.
2
GhorayebS. R., RooneyD.M., "Ultrasonic evaluationof bone qualityincadaverIlia," Annalsof
Biomedical Engineering,vol.41, no.5, pp.939–951, Jan. 2013.
3
Brandi M. L., "Microarchitecture,the keytobone quality," Rheumatology,vol.48,no.suppl 4, pp.iv3–
iv8,Sep.2009.
4
Enoki S., SatoM., Tanaka K.,KatayamaT., "Mechanical Propertiesof aSingle CancellousBone
Trabeculae TakenfromBovine Femur,"InternationalJournalof Modern Physics:ConferenceSeries,vol.
06, pp.349–354, Jan. 2012.
5
Troy K. L., BrownT. D., ConzemiusM.G., "Contactstressdistributionsonthe femoral headof the emu
(Dromaiusnovaehollandiae)," Journalof Biomechanics,vol.42,no. 15, pp. 2495–2500, Nov.2009.