BJR 2015 TheAuthors.Publishedby the BritishInstituteof
RadiologyReceived:7 January 2014Revised:29 December 2014Accepted:20
January 2015doi: 10.1259/bjr.20140040Citethis article as:ShimSS, Oh
Y-W, Kong KA, Ryu YJ, KimY, Jang DH. Pulmonary nodule size
evaluation with chest tomosynthesis and CT: a phantomstudy. Br
JRadiol2015;88:20140040.FULL PAPERPulmonary nodule size evaluation
with chest tomosynthesisand CT: a phantom study1S S SHIM,2Y-W OH,3K
A KONG,4Y J RYU,1Y KIM and1D H JANG1Department of Radiology,
Mokdong Hospital, Ewha WomansUniversity School of Medicine, Seoul,
Republicof Korea2Department of Radiology, Korea University
AnamHospitalandKoreaUniversityCollegeof Medicine,Seoul, Republic of
Korea3Clinical TrialCenter, Ewha WomansUniversity Medical Center,
Seoul, Republic of Korea4Departmentof Internal Medicine, Mokdong
Hospital,Ewha WomansUniversity School of Medicine, Seoul, Republic
of KoreaAddresscorrespondence to: Yu-Whan
OhE-mail:[email protected]: We compared digital
tomosynthesis (TOMO) andchest CT in terms of assessing the sizes of
nodules located inzones where evaluation by simple radiography is
limited.Methods: Atotal of 48 images comprising phantomnodules of
four sizes in six different locations were used.Nodule size
measurement errors for measurements usingTOMOand CT images compared
with the actual size fromeach observer were calculated. The inter-
and intra-observer repeatabilityof themeasuredvaluesandtheagreement
betweenthetwotechniqueswereassessedusing the method described by
Bland and Altman.Results: The mean measurement errors for all of
thenodules andfour observers were20.84mm[standarddeviation(SD),
0.60mm]onTOMOand20.18mm(SD,0.71 mm)onCTimages.
Themeanmeasurementerrorsfor the different observers ranged
from21.11 to 20.55mmfor TOMO andfrom20.39 to 0.08mm for
CT.Assessingthe agreement between nodule size measurements
usingTOMOandCTresultedinmeanmeasurementerrorsof20.65mm,
witha95%limit of agreement of 22.53to1.22mm for comparison of TOMO
with CT.Conclusion: Our results suggest that nodule sizes
obtainedusing TOMOand chest CT are comparable, even
fornoduleslocatedinareaswherethesizemeasurementislimited on simple
radiography.Advances in knowledge: TOMOandCTcan be
usedinterchangeably, even for nodules located in a blind areaon
simple
radiography.Solitarylungnoduledetectionhasincreasedowingtothewidespread
use of CTimaging. Nevertheless, the mostcommonly used routine
examination for lung
nodulescontinuestobechestradiography,becauseituseslowra-diationdoses,
is economical andis easytouse. Becausechest radiographic images are
two-dimensional projectionsof three-dimensional structures, early
lung cancer detectionon chest radiographs is often challenging. The
projection ofpulmonary vessels, bones and part of the mediastinum
onlungeldsoftenpartiallyorcompletelyobscuresthepul-monarynodules,
resultinginfailurebytheradiologisttodetectlungnodules.1,2Digital
tomosynthesis(TOMO)hasrecentlybeenappliedtochest imagingfor the
detectionof subtlenodules onsimpleradiography,
withpromisingresults.3,4Ithasbeenintroducedas a modality withthe
potential toprovideimages similar to CT but at a comparably reduced
cost andradiation exposure.4James et al5reported that 74% of
lungnodules$4 mmindiameterthatcanbeidentiedonCTcanalso be detected
using TOMO. Vikgrenet al6alsoreported that 92%of nodules $4
mmindiameter aredetectable using TOMO. In 2012, Johnsson et
al7comparedthe ability of TOMOand CTto detect nodule size
in20patients andfoundthat bothmethods
couldbeusedinterchangeablyforthesemeasurements. Thisresult callsfor
caution, however, because the limit of agreement (LOA)between the
modalities is wider than for the
intraobservervariabilityofeachmodality.Basedonthese studies, we
hypothesizedthat TOMOiscomparabletoCTimagingfor thedetectionof
noduleslocatedinareaswheresizemeasurement islimitedusingsimple
chest radiography because of overlapping structures.The purpose of
this study was to assess the size de-terminationof
noduleslocatedinthesezonesbyTOMOandchestCT.METHODS AND
MATERIALSThorax phantom and model nodule preparationThe thorax
phantomN1 (Kyoto Kagaku Co., Ltd,Kyoto, Japan) was constructed
fromsynthetic materials(i.e. polyurethane, epoxy resin,
calciumcarbonate). Modelnodules witha homogeneous compositionof
solid-type ure-thane foam nodules of four sizes (3, 5, 8 and 10 mm)
were used.Nodules were placed in six lung zones: the right apex,
middle ofthe right subpleural lungparenchyma, rightupper hilum,
rightlowerhilum,
rightdiaphragmaticangleoftheheart(rightcar-diophrenicangle)andtherightlowerretrohepaticlungparen-chyma(Figure1).Chest
CTAtotal of48phantomsets, comprising4nodulesizes, 6loca-tions
and2imagingmodalities, wereincludedinthepresentstudy. We used a
64-channel CT scanner (Somatom Denitionscanner; Siemens Healthcare,
Forchheim, Germany) for chestCT. The scanning parameters were as
follows: individual de-tector width, 0.625 mm; gantry rotation
time, 0.5 s; tube voltage,120 kVp; tubecurrent, 30 mA; andpitch, 1.
Axial imageswerereconstructedusing a sectionof 2 mminthickness, the
B70kernel(SiemensHealthcare) anda345-mmeldofview.Digital
tomosynthesisTOMOexaminations were performed using a
commerciallyavailable unit (Sonialvision Sare II; Shimadzu Co.,
Kyoto,Japan)withaat-panel detectorsystem. WealteredthedigitalTOMO
parameters to establish a lower radiation dose conditionthat was
suitable for chest imaging. 74 low-dose projectionimages were
acquired within 4.85 s using a tube voltage of120 kVpand0.04 mA.
The detector was xedintoposition,whereas the X-ray tube was
subjectedtovertical continuousmovement, from220 to 120, around the
standard orthogonalposteroanterior position, and image data were
acquired. A totalof 74 projection images were obtained from 1
examination andwereusedtoreconstruct 84coronal images
witha2-mmre-constructioninterval.Radiation doseFor the radiation
dose assessment in TOMO, a dosemeter(UnforsThinXIntra; Unfors
Instruments AB, Billdal, Sweden)that was attached to the centre and
surface of the chest phantom(N1)recordedtheabsorptiondose.
Theabsorbeddoseat thephantomsurface for TOMOwas 0.7 mGy, andthe
absorbeddoseatthephantomcentre was0.2 mGy.Pulmonary nodule
measurementFour radiologists with 15, 4, 3 and 1 years experience
in chestimage interpretation participated in the study. Atotal
of48images, comprisingnodulesof4sizesin6differentloca-tions
andarrangedinrandomorder bytheViewDEXsoft-ware(S odraAlvsborgs
Sjukhus, SahlgrenskaUniversityandUniversityofGothenburg,
Gothenburg, Sweden), wereused.Observers were blinded to the
location and size of the nodulesin the phantom model (Figure 2).
They measured the left-to-right diameter andrecordedthelongest
length. Theuseofzoomor enlarge tool was freelyavailable, andthe
windowFigure 1. Six locations of phantomnodules: (1) right
apex;(2)middleoftherightsubpleural lungparenchyma; (3)rightupper
hilum; (4) right lower hilum; (5) right cardiophrenicangle; and (6)
right retrohepatic lung parenchyma.Figure 2. Images of the thorax
phantom performed with chesttomosynthesis (TOMO) (a, c) and chest
CT (b, d) witha phantom nodule (arrows). (a, b) A 3-mm nodule in
the
rightapexisobservedonTOMO(measuredas2.38mm)(a)andalsonotedonCT(2.28mm)(b).
(c, d)A5-mmnoduleinthemiddleof theright subpleural lungparenchymais
seenonTOMO (4.36mm) (c) and CT (5.00mm) (d).BJR SSShimet al2 of 8
birpublications.org/bjr Br J Radiol;88:20140040centre/width for CT
was 2750/1500 HU, whereas the windowwidthforTOMOwas30004500 HU.
Thesevaluesareclin-icallyrelevant.Allofthemeasurementswererepeatedat20-dayintervalstoassessintraobservervariation.Statistical
analysisNodulesizemeasurement errors formeasurement
onTOMOandCTimages comparedwiththeactual sizefromeachob-server were
calculated. The results are presentedas means 6standard deviations
(SDs) and 95% condence intervals (95% CI)ofthemean.
Fortheinter-andintraobserverrepeatabilityofthemeasured
values,theagreementbetweenthetwotechniqueswas assessed using the
method described by Bland and Altman.8The 95% LOA was calculated as
the mean difference 61.96 SD ofthe difference.RESULTSUponinitial
measurement, oneobserver judgedtwoof the3-mm nodules as
beingmissedon the TOMO modality;
oneobserverjudgedthreeofthe3-mmnodulesasbeingmissedontheCTmodality;
whileanotherobserver judgedoneofthese nodules as being missedonCT.
Upontheir secondevaluation, one observer judged two of the 3-mm
nodules asbeing missed on TOMOand one 3-mmnodule as
beingmissedonCT.ThenumberofmeasuredphantomnodulesandmeanrelativeerrorsforCT
andTOMOmeasurementsforeachobserverareTable 1. Number of detected
nodules, mean relative error and standard deviation (SD) for each
observer regarding measurementson tomosynthesis (TOMO) and CT
imagesObserver Observer1 Observer2 Observer 3 Observer4TOMO
imagesDetected (n) 24 24 24 22Meanmeasurement error (mm) 20.55
20.75 20.94 21.11SD(mm) 0.38 0.22 0.43 0.98CT imagesDetected (n) 24
24 23 21Meanmeasurement error (mm) 0.08 20.36 20.07 20.39SD(mm)
0.28 0.27 0.66 1.16Figure 3. Measurement error for each observer
and eachnoduleontomosynthesisimagescomparedwiththeknowndiameter of
thenodule. Dashedline(centre) representsthemeanmeasurementerror
forallnodules
andobservers.Smalldashedlines(topandbottom)representthemeanmeasure-ment
error 62 standard deviation.Figure 4. Measurement error for each
observer and eachnoduleonCTimagescomparedwiththeknowndiameterofthe
nodules. Dashed line (centre) represents the meanmeasurement error
for all nodules and observers. Small dashedlines (top and bottom)
represent the mean measurement error62 standard deviation.Full
paper: Pulmonary nodule sizeevaluation withTOMOand CT BJR3 of 8
birpublications.org/bjr Br J Radiol;88:20140040Figure 5. Plots show
(a) manual measurement data from tomosynthesis (TOMO) images for
the diameters of all nodules andobserversplottedagainsttheactual
sizeand(b)manual measurementdatafromCTimages. Intheplots,
the45lineofequalityisdrawntohelpassesstheagreement
betweenthemeasurements. Plotsillustratethat theagreement
betweenmeasurementsonTOMOandtheactual
nodulesizeshowasimilarpatternwiththeagreementbetweenmeasurementonTOMOandCT.BJR
SSShimet al4 of 8 birpublications.org/bjr Br J
Radiol;88:20140040provided in Table 1. The mean measurement error
for all of thenodules and observers was 20.84 mm (SD, 0.60; 95% CI,
20.96to 20.72 mm) on TOMO and 20.18 mm (SD, 0.71; 95%
CI,20.33to20.04 mm)onCTimages.Theabsolutemeasurement errors
foreachobserver andeachnodule on TOMO and CT images are provided in
Figures 3 and4. Onaverage, all of themanual
measurementsonbothchestTOMO and CT images underestimated the nodule
size, with theexception of one observer using CT. Figure 5 shows
the manualmeasurement data from TOMO imagesforthe diameters of
allnodules and observers plotted against the actual size and
manualmeasurement data fromCTimages. The meanmeasurementerrors for
the different observers ranged from 21.11 to20.55 mmfor
TOMOandfrom20.39to0.08for CT.
ThemeanmeasurementerrorswithanLOAforeachobserverandsizeareprovidedinTable2.The
intra- and interobserver 95% LOA for each imaging type areprovided
in Table 3. The intraobserver 95% LOA for the diametermeasurements
calculated using the mean of the two measurementsas a reference
ranged from 20.42 to 0.50mm for the least variableobserver to 21.85
to 2.31 mm for the most variable observer usingtheTOMOmodality. For
CTmeasurements, the intraobserverTable 2. Mean measurement error
and limit of agreement (LOA) for real size of the nodules by each
observer and sizeModality/measurementTOMO CTMeanmeasurementerror
(mm)LowerLOA(mm)UpperLOA(mm)Meanmeasurementerror
(mm)LowerLOA(mm)UpperLOA(mm)Observer1 20.55 21.29 0.19 0.08 20.46
0.632 20.75 21.19 20.32 20.36 20.89 0.173 20.94 21.78 20.09 20.07
21.37 1.234 21.11 23.02 0.81 20.39 22.66 1.88Size(mm)3 20.63 22.11
0.86 20.72 22.87 1.445 20.64 21.21 20.07 20.00 20.82 0.818 21.14
21.72 20.56 20.05 20.83 0.7410 20.95 22.42 0.52 0.03 20.80
0.85TOMO, tomosynthesis.Table 3. Intra- and interobserver mean
measurement error by observerModality/measurementTOMO
CTMeanmeasurementerror
(mm)LowerLOA(mm)UpperLOA(mm)Meanmeasurementerror
(mm)LowerLOA(mm)UpperLOA(mm)Intraobserver1 0.04 20.42 0.50 20.08
20.75 0.582 20.20 20.68 0.28 20.43 21.01 0.163 0.06 20.44 0.57
20.14 21.30 1.034 0.23 21.85 2.31 20.01 22.47 2.45Interobserver1and
2 0.20 20.31 0.71 0.44 20.34 0.221and 3 0.38 20.06 0.83 0.15 21.22
1.531and 4 0.55 21.49 2.59 0.48 21.72 2.672and 3 0.18 20.46 0.83
20.29 21.58 1.002and 4 0.35 21.52 2.22 0.03 22.12 2.193and 4 0.17
22.02 2.36 0.32 21.49 2.13LOA, limit of agreement; TOMO,
tomosynthesis.Full paper: Pulmonary nodule sizeevaluation
withTOMOand CT BJR5 of 8 birpublications.org/bjr Br J
Radiol;88:2014004095%LOA ranged from20.75 to 0.58 mmand 22.47
to2.45mm for the least and most variable observer,
respectively.Theinterobserver95%LOA
concerningthediameterestimatesfor eachpossible pair of radiologists
rangedfrom20.06 to0.83
mmfortheleastvariablepairofradiologiststo22.02to2.36 mmfor the most
variable pair of radiologists using theTOMO modality. For
measurements using the CT modality, theinterobserver
95%LOArangedfrom20.34to0.22 mmandfrom 22.12 to 2.19 mm for the
least and most variable
observerpairs,respectively.AssessingtheagreementbetweenthenodulesizemeasurementsusingTOMOandCTbythedifferencebe-tweentheobservers
meandiametermeasurement
forthetwomodalitiesresultedinmeanmeasurementerrorsof20.65 mm(95%CI,
20.91 to 20.40) with a 95%LOAof 22.53 to1.22 mm forthe comparison
of TOMO and CT.
BlandAltmanplotsillustratingtheagreementareprovidedinFigure6.The
95%LOAof the diameter measurements for
individualobserversusingTOMOandCTimagesrangedfrom21.16to0.37 mm for
the least variable observer to 23.47 to 2.04 mm forthe most
variable observer. The 95% LOA for all of the observersand sizes
regarding the measurement of the nodule diameter onTOMO and CT are
provided in Table 4. BlandAltman plots ofthe agreement between
measurements on TOMOand CTimages for the most and least experienced
radiologists are
showninFigure7.DISCUSSIONWeevaluatedTOMO-derivedparameters,
includingnoduledi-ameter, repeatabilityof
TOMOobservationsandtheextent ofagreement
betweenTOMOdataandthoseacquiredbyCT. Arecent study found that the
repeatability values of manualmeasurementsmadebyTOMOandCT
werecomparable.7Inthe cited study, clinical nodules that were
considered adequatelysegmented were investigated. However, missed
nodules onsimpleradiographyareusuallyat theapices, lungbases or
incentral locations adjacent tovessels. It is difcult
toperformadequate segmentation(not toget confusedby the
adjacentstructuresinsideorbesidethelesion)ofsuchnodules.9Several
TOMOphantomstudies have reported limited
un-derestimationoftheactualsize,whereasstudiesusingCThavefound
underestimated or overestimated nodule sizes.7,10,11In
thepresentstudy,bothmodalitiesunderestimatedthediametersofphantomnodules.
TOMOappearedtoslightlyunderestimatethenodulesize(meanmeasurement
error, 20.84 mm); how-ever, suchanerrorisclinicallyrelevant
becausenoduleman-agement is based on an absolute nodule size
threshold. Inaddition, wefoundthatCT
slightlyunderestimatedthenodulesize(meanmeasurementerror, 20.18
mm).The extent of nodule size underestimation by TOMOwassomewhat
more marked than that affected by nodule de-lineation.12,13Such
artefacts were mostly investigated
inTOMOimagesforthebreasts.14,15Although wemeasuredthenodule
diametersfromleft to right(thus,notinthedirectionof thescan),
darkerareas createdahaloaroundthenodule.This halowas not
includedinmanual measurements of thenodule diameter, a factor that
affectedthe results. Amorerelevant issue is the possible effect of
in-plane artefacts onclinical images. Recently, underestimationof
nodulesizeas-sociated with manual measurement of clinical nodules
onTOMOhasbeenreported.7Svahnetal16foundthattheextentofin-planeartefactsvariedlinearly
with the spherical diameter and relative contrast ofFigure 6. A
BlandAltman plot illustrating the agreementbetween two modalities.
Dashed line (centre), mean difference[20.65mm (95% confidence
interval, 0.91 to 0.40)]. Dashedlines(topandbottom),
upperandlowerlimitsofagreement.TOMO, tomosynthesis.Table 4. Mean
measurement error at tomosynthesis (TOMO) incomparison with CT by
observer and sizeTOMOcompared withCTMeanmeasurementerror
(mm)LowerLOA(mm)UpperLOA(mm)Observer1 20.64 21.45 0.172 20.40 21.16
0.373 20.87 22.69 0.964 20.72 23.47 2.04Size(mm)3 0.09 22.46 2.645
20.64 21.63 0.358 21.09 22.07 20.1210 20.98 22.60 0.64LOA, limit of
agreement.BJR SSShimet al6 of 8 birpublications.org/bjr Br J
Radiol;88:20140040nodules. Their nding that the tendency towards
size
un-derestimationonTOMOimagesescalatedwithanincreaseinnodulediameterseemstobeconsistentwithourresults.
Thecontrast affordedbythe40acquisitionangleof ourTOMOimages may be
greater than that of the images produced in thestudybyJohnssonet
al17in2010, whousedanacquisitionangleof 30. Thiscontrast might
aggravatein-planeartefactsand underestimate the size of larger
nodules because
thecontrastinTOMOimagesincreasesasacquisitionanglerises.Despite
these limitations of TOMO, our present phantomstudy suggested that
the measurement values on CTandTOMO were comparable; the mean
difference was 20.65 mm.Evenwith larger nodules ($5 mmin diameter),
the meandifferencewaslessthan21.1 mm.Regardingrepeatability, the
interobserver 95%LOAfor
mea-suringTOMOdiametersbythepairofradiologistswhorecor-dedthemostsimilarresultswas20.06to0.83
mm, thevalueswere similar to the CT data. These values were
slightly less thanthe variation of 21.3 to 1.5 mm in the
measurement of
clinicalnodulesreportedbyJohnssonetal7in2012.Theintraobserveragreement
was of similar magnitude. It is possible that the evenshape of the
phantom nodules was associated with less variationin measurements
than in clinical nodules. We found that TOMOdata, including
measurement errors, intra- andinterobserveragreement levels and
LOAs, improved in proportion to the
yearsofexperience.ThisisalsotrueofCT
measurements.Ourstudyhadseveral limitations.
Onemajordrawbackinthemeasurement study in the phantomnodule was
that theobservers could guess the true sizes of the studied objects
duringthetwoobservations, resultinginincreasedrepeatability;
thus,we sought to make observers completely unaware of the
nodulesizeandlocationbypresentingrandomimagestakenusingei-ther
imagingmodality. Inaddition, theintrinsiclimitationofTOMO discussed
above has affected our results. Optimization ofTOMO technique,
including the plane and angle of acquisition,andconstruction
ofthephantomnodulerequirefurtherwork.A clinical nodule may vary in
shape. Thus, a limitation of
TOMOisthatthesizecanbeestimatedonlyinasingleplane,
whereasmultidetector CTis three dimensional. However, the
thoraciccoronal projection of TOMO is larger than the routine axial
imageafforded by CT; therefore, it is possible to measure the
superior toinferior nodule diameter on TOMO images more rapidly
than ispossible usingCT data.
Moreover,theentirethoraxispresentedbyTOMOwithlessradiationandashorterstudytimethanbyCT.
Therefore, TOMO should be considered for nodule detectionand follow
up for young patients (particularly young women witha risk of
breast cancer), and probable benign lesions.CONCLUSIONSOur results
suggest that the nodule sizes obtained using TOMOand chest CTare
comparable, even for nodules located in areaswhere size measurement
is limited on simple radiography,because the size difference was ,1
mm, the LOAs were
ofsimilarwidthandtherepeatabilityvaluesweresimilar. How-ever,
measurements made by TOMO tended to be smaller thanthose byCT,
andthis tendencyampliedas the nodule
di-ameterincreased,whichmaybeofconcernifTOMOandCTare to be used
interchangeably during nodule follow-up. Theseconcerns maybe
alleviatedinthe near future byupgradingTOMOto reduce the number of
artefacts encountered inclinicalpractice.Figure7.
Theplotsshowthedifferencebetweenthenodulediameter measurement on
tomosynthesis (TOMO) and CTagainst theaverageof themeasurements for
(a) themostexperienced radiologist and (b) the least
experiencedradiologist.Full paper: Pulmonary nodule sizeevaluation
withTOMOand CT BJR7 of 8 birpublications.org/bjr Br J
Radiol;88:20140040REFERENCES1. BrogdonBG,Kelsey CA,Moseley
RDJr.Factorsaffectingperceptionofpulmonarylesions.Radiol ClinNorth
Am1983; 21:63354.2. ShahPK,AustinJH,White CS,Patel P,Haramati LB,
Pearson GD, et al. Missed non-smallcell
lungcancer:radiographicndingsofpotentiallyresectablelesionsevidentonlyinretrospect.Radiology2003;
226:23541.3. Gomi T, Nakajima M, Fujiwara H, Takeda T,Saito K,
Umeda T, et al. Comparison betweenchestdigitaltomosynthesisand CT
asascreeningmethod todetectarticialpulmonarynodules: aphantom
study.BrJRadiol2012; 85:e6229.doi: 10.1259/bjr/126430984.
SvalkvistA, JohnssonAA, Vikgren J,HakanssonM,Ullman G,BoijsenM,
etal.Evaluation ofanimprovedmethod ofsimulatinglungnodules inchest
tomosyn-thesis.ActaRadiol 2012;
53:87484.doi:10.1258/ar.2012.1202305.
JamesTD,McAdamsHP,SongJW,LiCM,GodfreyDJ,DeLong DM,et
al.Digitaltomosynthesis ofthechestfor
lungnoduledetection:interimsensitivityresults from
anongoingNIH-sponsoredtrial. Med Phys2008; 35:25547.6. Vikgren
J,ZachrissonS,SvalkvistA,JohnssonAA, BoijsenM,FlinckA,
etal.Comparisonofchest tomosynthesisandchest radiographyfor
detection ofpulmo-narynodules:human observer study ofclinicalcases.
Radiology2008; 249:103441.doi:10.1148/radiol.24920803047.
JohnssonAA, FagmanE,VikgrenJ,FisichellaVA, BoijsenM,FlinckA,
etal.Pulmonary nodule size evaluation with
chesttomosynthesis.Radiology2012; 265:27382.8. Bland JM, Altman DG.
Statistical methods forassessing agreement between two methods
ofclinical measurement. Lancet 1986; 1: 30710.9.
GirvinF,KoJP.Pulmonarynodules:de-tection, assessment,and CAD.AJR
AmJRoentgenol 2008; 191: 105769. doi: 10.2214/AJR.07.347210.
WilleminkMJ,Leiner T,Budde RP,deKortFP,VliegenthartR, vanOoijen
PM,etal.Systematic
errorinlungnodulevolumetry:effectofiterativereconstructionversusltered
back projection at different CTparameters. AJR AmJRoentgenol 2012;
199:12416.doi: 10.2214/AJR.12.872711. Revel MP,BisseryA,Bienvenu
M,Aycard L,LefortC, FrijaG.Aretwo-dimensionalCTmeasurements
ofsmallnoncalciedpulmo-narynodules reliable?Radiology2004;
231:4538.12.
TingbergA,ZackrissonS.Digitalmammog-raphyandtomosynthesisforbreastcancerdiagnosis.ExpertOpin
MedDiagn2011; 5:51726.doi: 10.1517/17530059.2011.61649213. HuYH,
ZhaoB,ZhaoW.Imageartifactsindigitalbreasttomosynthesis:investiga-tionoftheeffectsofsystemgeometryandreconstructionparametersusingalinearsystemapproach.MedPhys2008;35:524252.14.
SvahnT,Ruschin M,Hemdal B, Nyhlen L,Andersson
I,TimbergP,etal.Inplaneartifacts inbreasttomosynthesisquantiedwitha
novelcontrast-detail phantom.ProcSPIE2007; 6510:65104R.15.
TuckerAW,LuJ,ZhouO.Dependency ofimagequality
onsystemcongurationparameters inastationary digital
breasttomosynthesissystem. MedPhys2013; 40:031917.doi:
10.1118/1.479229616. SvahnT, Hemdal B, RuschinM,Chakraborty
DP,AnderssonI,TingbergA,etal.
Dosereductionanditsinuenceondiagnosticaccuracyandradiationriskindigitalmammography:anobserverperfor-mancestudyusingananthropomorphicbreastphantom.BrJRadiol2007;80:55762.17.
JohnssonAA, SvalkvistA, VikgrenJ,BoijsenM, FlinckA,
KheddacheS,etal.
Aphantomstudyofnodulesizeevaluationwithchesttomosynthesisandcomputedtomography.
RadiatProtDo-simetry2010;139: 1403. doi: 10.1093/rpd/ncq003BJR
SSShimet al8 of 8 birpublications.org/bjr Br J
Radiol;88:20140040