Norbert J. Pelc, Sc.D.Norbert J. Pelc, Sc.D.
Departments of Radiologyand BioengineeringDepartments of Radiologyand Bioengineering
Stanford UniversityStanford University
Dual Energy CT:Dual Energy CT:PhysicsPrincip lesPhysicsPrinciples
DisclosuresDisclosures
Uri Shreter andRobert Senzig, GE Healthcare
ThomasFlohr and BernhardSchmidt, SiemensMedical Solutions
Researchsupport: NIH grant EB006837, GEHealthcare,theLucasfoundation
This lectureincludes off-labeluseof CT scanners
MotivationMotivation
“ Two pictures are takenof the sameslice,one at 100kV and theother at 140kV... areasof high atomicnumberscanbeenhanced... Testscarriedout to datehaveshownthatiodine (Z=53) canbereadilydifferentiatedfrom calcium (Z=20)”.
G.N. Hounsfield, BJR46, 1016-22,1973.
• Materialspecificity• Improvedtissuecharacterization
OUTLINEOUTLINE
• Physical principlesof multi-energy x-raymeasurements
• Dual energy CTprocessing
obtaining dual-energy measurements
scanning
• Strengthsandlimitations
unknown amountsofunknown amounts oftwo known materialstwo known materials
energyE1
I1
mb (g/cm2)
I01
water
bone10 10020 50 200
corticalbonewater
photonenergy(keV)
mas
sat
tenu
atio
nco
effi
cien
t(cm
2 /g)
0.1
1.0
10.0
100.0
I1 = I01e-(µ/ρ)w1mw + (µ/ρ)b1mb
I2 = I02e-(µ/ρ)w2mw + (µ/ρ)b2mb
I2
E2I02
E2E1
solve for mw and/or mb
unknown amountsofunknown amountsoftwo known materialstwo known materials
energyE1
I1
I01
water
bone10 10020 50 200
corticalbonewater
0.1
1.0
10.0
100.0
I2
E2I02
E2E1
mb = A{ ln(I01/I1) - (µw1/µw2)(ln(I02/I2)}
subtractwater
scale for lostbone signal
makes the water contributionat E2 match that at E1
••
photonenergy(keV)
mas
sat
tenu
atio
nco
effi
cien
t(cm
2 /g)
mb (g/cm2)
dual energy projection methodsdual energy projection methods
DEXA
Singleenergy “Boneminic tissue” imageBone image
DualEnergyRadiography
• 2 energies 2 materials
material analysiswithmaterial analysis withabsorptiometryabsorptiometry
• canwe generalizethis? N energiesfor Nmaterials?
• limitation: two stronginteraction mechanismsCompton scatteringandphotoelectric absorption
Eachhas~ sameenergydependencefor all elements
Basis material decompositionBasis material decomposition
0.1
1
10
100
1000
0 20 40 60 80 100 120 140
OCaCuCa'
Cu
.61*O + .04*Cu
Ca
O
Basismaterial decompositionBasis material decomposition
M gramsof Ca.04 M gramsof Cu
.61 M gramsof O
I
I0 I0
I
=
Indistinguishableat anyx-rayenergyabove their K-edge
(K-edgemethodsare not coveredhere)
• Barringa K-edge:µ(E) = a*Compton(E)+ b*Photoelectric(E)
2 fundamentalparameters characterize material behavior
electrondensity, effectiveatomicnumber
basismaterial decompositionbasismaterial decomposition basismaterial decompositionbasis material decomposition• Barringa K-edge:
µ(E) = a*Compton(E)+ b*Photoelectric(E)
2 fundamentalparameters determinematerial behavior
electrondensity, effectiveatomicnumber
• anymaterialcanbemodeled asa weightedsumof twoothermaterials
µ(E) = α* µi(E) + β* µj(E)
basis material decomposition
commonbases:aluminumandplastic
• in anyprojection measurement,we canonly isolatetwomaterials
OUTLINEOUTLINE
• Physical principles of multi-energyx-raymeasurements
• Dual energy CTprocessing
obtainingdual-energy measurements
scanning
• Strengthsandlimitations
• material selective imagesbasismaterialimagesmaterialspecificor cancelled images“ virtual non-contrast” scansweightedsubtractionsof thetwo energiesprocessingamplifiesnoise
• monoenergetic imagesweightedsumsof thetwo energiescanbehigh SNR,depending on the weighting
DualDual--energy processingenergy processing
DualDual--energy processingenergy processing
• reconstructimagesin thenormal manner,and combineHU imageseasyto implement
• combineprojectiondataprior toreconstructionsomewhat moredifficult
requiresaligned projections
enables“ exact” beamhardening correction
nonlinearcombination
PrePre--reconstr uction processingreconstruction processing
material selective images
monoenergetic images
adaptedfrom Lehmannet al: MedPhys8, 659-67, 1981.
calibration
low energyprojections
high energyprojections
“ aluminum”projections
“ plastic”projections
“ aluminum”image
“ plastic”image
linearcombination
finalimage(s)
CT recon
canincorporateaccuratebeamhardeningcorrection
basismaterialimages
Monochromatic CT from projection-based recon
Water Aluminum Water AluminumMono-
chromaticImage based Projection based
Projection based MD reduces beam hardening
80kVp 140kVp
Potential for beam hardening streak-free images
Courtesyof Uri Shreter, GE Healthcare
Potential impact on Cardiac IQ +Perfusion
Polychromatic
Courtesyof Uri Shreter, GE Healthcare
Heart Chamber Phantom, 8.3%
L = 0, W = 350 HU
80kVp 140kVp
MonochromaticIodineWater
Monochromatic CT from projection-based recon*
Material separation
Monochromatic CT – keV
tuned
Natively eliminates beam
hardening CT # shifts
Courtesyof R. Senzig, GE Healthcare
nonlinearcombination
PrePre--reconstr uction processingreconstruction processing
material selective images(noisy)
monoenergetic images(canbelow noise)
adaptedfrom Lehmannet al: Med Phys 8, 659-67, 1981.
calibration
low energyprojections
high energyprojections
“ aluminum”projections
“ plastic”projections
“ aluminum”image
“ plastic”image
linearcombination
finalimage(s)
CT recon
higher,negativelycorrelatednoise
Two known materials in aTwo known materials in a voxelvoxel55 keV 80 keV
iodineCNR=7.9 iodineCNR=3.8
water image
SNR=37
iodineimage
SNR=3.4
optimal combination(“ mixed” image)
iodineCNR=10
waterSNR=67 waterSNR=71“water” contrast
iodinecontrast
Dual energy CTDual energy CT
HUlow
HUhigh
line of identity: “water-like” materials
-1000
-1000
effectiveZ > water
effectiveZ < water
HUlow/HUhigh
(and relatedmetrics)dependon effective Z
Principle of Dual Energy CT – Image Based Evaluat ion
Each material is characterized by its „Dual Energy Index“
x80 and x140 are the Hounsfield numbers at 80 kV and 140 kV, resp.
Dual energy CT can measure chemical compo si tion!
Material DEI
Bone 0.1148
Liver 0.0011
Lung -0.0021
Sof t Tissue -0.0052
Skin -0.0064
Protein s -0.0087
Fat -0.0194
Gall fluid -0.0200
Courtesyof B. Krauss,B. Schmidt,andTh. Flohr, SiemensMedical Solutions
atomic number and density asatomic number and density asmaterial parametersmaterial parameters
• tissuespecific ity? applicationsbeinginvestigated
kidneystonecharacterization
fat or iron in theliver
plaquecharacterization
• imagesegmentationboneor plaqueremoval
identifying ligamentsor tendons
Three known materialsThr eeknown materialsdual energy CTdual energy CT
water
iodine
HUlow
water+iodine
bone water+bone
HUhigh
identity
Kelczet al: Med Phys6, 418-25,1979.
boneand iodinein water
Cancalculatebothiodine andbone.Requiresknownmixing propertiesand consistentwater density
mb = A{ HUlow - (Μ1,I/ Μ2,I) HUhigh}
Three known materialsThree known materialsdual energy CTdual energy CT
Reliableseparationrequires large(R1 - R2)2
where R = µhigh/µlow, dependsonmaterialand energies
Works bestfor onehigh Z andonelower Z material, andverydifferentx-rayenergies
iodine
calcium
Kelczet al: Med Phys6, 418-25,1979.
Spectral separationSpectral separation
• verycritical for SNRefficiency,separationrobustness,etc.
• implementationsdifferent kVp and/or filt ration
layereddetector
photoncountingwith energyanalysis
DualDual kVpkVp, dual filt rati on, dual filt rati on
135 kVp1.5 mm bronze
85 kVp0.1 mm erbium
• switchedfi ltrationimprovesseparation
• differentmAs helpsapportiondose
Lehmannet al: Med Phys 8, 659-67,1981.
Carmi R, NavehG, andAltman A: IEEENSSM03-367,1876-78,2005
Layered detectorLayereddetector• simultaneous dualenergysensing• relatively poor spectral separation
Energy discriminating, photonEnergy discriminating, photoncounting detectorscounting detectors
broadspectrumx-ray source
pulseshaping
pulseheight
analysis
two(or more)
energybins
High enoughcountrateis diffi cult to
achieve
Spectral separationSpectral separation
• verycritical for SNRefficiency, separationrobustness,etc.
• implementationsphotoncounting with K-edgefilterphotoncounting with energyanalysisdifferent kVp andfiltrationdifferent kVplayereddetector
betterspectral
separationanddose
efficiency
Dual energy implementationsDual energy implementations
• Sequential scansat different kVpmotion sensitivity > 50%Trot
highermotion sensitivity in helical mode
• Two sourcesat 90º on thesamegantry
SiemensDefinition80 kVp & 140kVp
Dual Source Challenge: Inconsistent scans
Coincident
Moving PhantomSimulation
Dual Source system
Moving Objec ts
Does not see movement
Courtesyof R. Senzig, GE Healthcare
Dual energy implementationsDual energy implementations
• Sequential scansat different kVpmotion sensitivity > 50%Trot
• Two sourcesat 90º on thesamegantrysomemotionsensitivity (~ 25%Trot)
• Switching kVp within a singlescan1, 2
technically challenging with rapidgantryrotation
1. Lehmann etal: Med Phys 8, 659-67,1981.2. Kalenderetal, Med Phys13, 334, 1986.
Courtesyof Uri Shreter, GE Healthcare
RapidRapid kVpkVp switchingswitchingDual energy CTDual energy CT Dual energy implementationsDual energy implementations
• Sequential scansat different kVpmotion sensitivity > 50%Trot
• Two sourcesat 90º on thesamegantrysomemotionsensitivity (~ 25%Trot ?)
• Switching kVp within a singlescan
• Energydiscriminatingdetectorslayereddetector,photoncounting
betterimmunity
tomotion
Strengths and limita tionsStrengthsand limitations
• perfect beamhardeningcorrection(pre-recon)effectivemonoenergeticimages
• extrapolationto high energiesmore accurate RTPandN/M attenuation correction
• somematerial specificity (e.g.,effectiveZ, DEI)mayprovidediseasespecificityimproved imagesegmentation
Strengths and limita tionsStrengths and limita tions
• virtual non-contrast imageperfectlyregisteredandsimultaneouslyacquiredBewareof noise propagation. Separate optimizedscans
probablyhavelower total dosefor sameIQ
• isolatecontrastmedia from calcified plaquedif ficult, especially for smallamountsof either
• molecularimaging?I don’ t think so
• “ tomochemistry”only with K-edgemethodsandrelativelylargeconcentrations
Thank You