PET

Technical aspects

15ODetector 1 Detector 2

15N

e-β+

Evolution of PET Detectors

CTI/Siemens

15O

Constantfraction

discriminator

Constantfraction

discriminator

location “line of response”

x

y

x

y

Detector block 1 Detector block 2

15N

e-β+

PET Detectors

Most modern PET system use a different detectortechnology where a large number of scintillationcrystals are coupled to a smaller number of PMTs.

For a given crystal, the lower the cross section isthe shorter the length of the crystal has to be dueto the light transport.

NaI (Tl) BGO GSO LSO LYSO LaBr3

Density [g/ml] 3.67 7.13 6.71 7.35 7.1 5.29

1/µ [cm] 2.88 1.05 1.43 1.16 1.2 ~2Index of

Refraction 1.85 2.15 1.85 1.82 1.81 1.9

Hygroscopic Yes No No No No Yes

Rugged No Yes No Yes Yes YesPeak

Emission [nm]

410 480 430 420 420 380

Decay Constant [ns] 230 300 60 40 41 25

Light Output 100 15 35 75 75 >100Energy

Resolution 7.8 20 8.9 <9 11 7.5

Scintillator Materials

PET Detectors

In the block detector, a matrix of cuts are madeinto a solid block of scintillator material to definethe detector elements.The depth of the cuts are adjusted to direct thelight to the PMTs.The light produced in each crystal, will produce aunique combination of signals in the PMTs, whichwill allow the detector to be identified.

The more light output the more crystals can beidentified.

PET Detectors

To identify the detector elements, the two following ratios are calculated (similar to the logic used in scintillation cameras).

Below is the flood response (i.e., X and Y density distribution when exposed to a flood source of 511 keV photons) for a block detector from the ECAT EXACT HR.

Element (4,4)

Element (5,3)

A B C DXA B C DA B C DYA B C D

+ − −=

+ + +− + −

=+ + +

Standard Detector6.4 mm x 6.4 mm 64 crystals/block

144 blocks/scanner9216 crystals/scanner

3.4 mm slice width47 slices

HI-REZ Detector4.0 mm x 4.0 mm 169 crystals/block144 blocks/scanner

24336 crystals/scanner2 mm slice width

81 slices

The Technology : HiRez

Constantfraction

discriminator

Constantfraction

discriminator

12 nsAND

location

coincidence?

“line of response”

x

y

x

y

Detector block 1 Detector block 2

NaI (Tl) BGO GSO LSO LYSO LaBr3

Density [g/ml] 3.67 7.13 6.71 7.35 7.1 5.29

1/µ [cm] 2.88 1.05 1.43 1.16 1.2 ~2Index of

Refraction 1.85 2.15 1.85 1.82 1.81 1.9

Hygroscopic Yes No No No No Yes

Rugged No Yes No Yes Yes YesPeak

Emission [nm]

410 480 430 420 420 380

Decay Constant [ns] 230 300 60 40 41 25

Light Output 100 15 35 75 75 >100Energy

Resolution 7.8 20 8.9 <9 11 7.5

Scintillator Materials

Constantfraction

discriminator

Constantfraction

discriminator

12 nsAND

MCA

location

Energy spectrum

coincidence?

“line of response”

x

y

x

y

Detector block 1 Detector block 2

NaI (Tl) BGO GSO LSO LYSO LaBr3

Density [g/ml] 3.67 7.13 6.71 7.35 7.1 5.29

1/µ [cm] 2.88 1.05 1.43 1.16 1.2 ~2Index of

Refraction 1.85 2.15 1.85 1.82 1.81 1.9

Hygroscopic Yes No No No No Yes

Rugged No Yes No Yes Yes YesPeak

Emission [nm]

410 480 430 420 420 380

Decay Constant [ns] 230 300 60 40 41 25

Light Output 100 15 35 75 75 >100Energy

Resolution 7.8 20 8.9 <9 11 7.5

Scintillator Materials

Event Types

True Event Scattered Event

Random Event Multiple Event

Random Coincidences

Because of the finite width of the logic pulses that are fed into the coincidence circuit, there is a probability forrandom or accidental coincidences between unrelated events.

TrueCoinc.

Detector 1

Detector 2 Time

TrueCoinc.

SingleEvent

RandomCoinc.

SingleEventτ

Random Coincidences

If N1 and N2 are the individual average count rates ofdetector 1 and 2, respectively, then it can be shown thatthe random coincidence rate for the pair of detectors is:

NR = 2τ N1 N2

where 2τ is the coincidence window (or τ is the width ofthe singles pulses)

Random CoincidencesRandoms can be estimated by adding a second coincidencecircuit.The logic pulse from one detector is delayed beyond the timeresolution of the detector pairs.This does not require any knowledge of the coincidence timewindow, but is not as statistically accurate as the singlesmethod

Coinc.

Delay

Coinc.

Amp./PHA

Amp./PHA

Prompts

Randoms

Estimation of Random Coincidences

Singles MethodMonitor singles for each detectorStatistically accurateRequires accurate knowledge of τDead-time different between prompt coinc. & singles

Delayed Coincidence MethodSeparate coincidence circuitry/processorPoor statistical accuracy (improvement with variance

reduction)

ScatterScatter can be reduced by

better energy resolution

more collimation (2D)

True Coincidences~ ActivityGood events!

Signal-to-Noise

/ ~ TS NT

Signal-to-Noise

Random Coincidences~ Activity2

Can be accurately corrected forCorrection increases image noiseDetector material dependent

/ ~2

TS NT R+

Signal-to-Noise

Scattered Coincidences~ ActivityReduces Image ContrastRequires correctionAnalytical estimationCorrection increases image noise

/ ~2

TS NT S R+ +

Signal-to-Noise

Multiple Coincidences:~ Activity3

Never savedSource of Dead time

Noise Equivalent Count Rate - NEC

The correction for random and scattered events increases the noise in the net true events

The correction therefore reduce the effective true count rate

The NEC describes the equivalent true count rate in the absence of random and scattered events:

~ /2

TNEC S NT S kR

=+ +

0.0 100

5.0 105

1.0 106

1.5 106

2.0 106

0 0.01 0.02 0.03 0.04 0.05

TruesPromptsRandomsNEC

Cou

nt R

ate

[cps

]

Specific Activity [MBq/ml]

Spatial Resolution

The spatial resolution in PET is primarily determined by:

Detector size

Physics of positron decay

System geometry

Detector material

Spatial Resolution

For a source placed at the midpoint between two scintillation detectors with a width wd, the geometric line spread function has a triangular shape with a FWHM of wd /2.

wd

2FWHM =

wd

Spatial Resolution -Tangential

For sources located between the midpoint and the detector surface the LSF will have a trapezoidal shape with width varying from wd /2 (at the center) and wd at the detector surface.

Spatial Resolution - Radial

Radial

R=0

Axial section Tangential

012

345678

0 5 10 15 20

FWHMtangFWHMrad

FWHM

(mm

)

R (cm)

Transaxial Resolution or ECAT EXACT HR

Depth-of-Interaction

c m

e te c to r e p a r a te

Depth of Interaction Decoding Techniques

• Multiple layers of scintillators with different decay constants.

• Wong, 1985; Carrier et. al., 1987; Casey et. al. 1996• Additional light readouts on the scintillator.• McIntyre et. al., 1980; Moses et. al., 1991• Modulation of light output• Rogers et. al., 1995

UCLA School of Medicine

30 cm

40 cm

30 mm 2x15 mm 3x10 mm 4x7.5 mm

DOI Detectors

Siemens Molecular Imaging

SSPM/SiPM

UCLA School of Medicine

SSPM/SiPM

0.0

5.0 102

1.0 103

1.5 103

2.0 103

2.5 103

3.0 103

3.5 103

300 400 500 600 700 800

Cou

nts

Channel

1.37 ns FWHM

UCLA School of Medicine

Avalanche Photo Diodes (APD) vs. Photo Multiplier Tubes (PMT)

PMTs APDs

Size 10-52mm dia. 5x5mm

Gain Up to 106 Up to 200

Rise time ~1 ns ~5 ns

QE 20% 70%

Magnetically sensitive insensitive

Although the most energetic positrons can travel several mm before annihilating,only a few of these are emitted.

The average positron energy emitted is approximately 1/3-1/2 of the maximumenergy.

The total path length the positrons travel is not along a straight path. Throughinelastic interactions with electrons in the positrons path is deflected. The distancefrom the mother nucleus is therefore much shorter.

Spatial Resolution

Positron Range

From Levin & Hoffman PMB 44, 1999

Resolution loss due to the positron range

Isotope β+ Energy[MeV]

FWHM[mm]

FWTM[mm]

18F 0.64 0.10 1.09

11C 0.96 0.19 1.86

13N 1.19 0.28 2.53

15O 1.72 0.50 4.14

82mRb 3.35 2.60 13.20

Positron Range

18F635 keV

124I1.53 & 2.14 MeV

~65% ~50%

~0.3° FWHM100 cm Ø ~ 2.5 mm FWHM

15 cm Ø ~ 0.3 mm FWHM

Non-colinearity

Spatial Resolution

The measured resolution (intrinsic resolution) of the system is a convolution of the various resolution components.

If the different resolution components are assumed to be Gaussian in shape and are described by a FWHM then the combined resolution is the squared sum of the individual resolution components:

2 2 2dettotal ector positron angulationFWHM FWHM FWHM FWHM= + +

2d

22et 20% posittotal ector angulatin onroFWHM FWHM W FWHF MM= + +

Sánchez Crespo et. al. EJNM, 2004

Improved NEC Performance

Time-of-flight PET

x

R R

R - x R + x

1

2

2 12 ( ) 2

s v tR x vtR x vt

c tx v t t x

= ⋅+ =− =

∆= − ⇒ =

Det 2 Det 1

Time-of-flight PET

For ideal detectors, TOF would eliminate the need for image reconstruction, since the measurement would allow each event to be accurately positioned in space.

All detectors have a finite time resolution, or uncertainty in timing. This translates to an uncertainty in positioning.

BGO ~ 5 ns 75 cmNaI ~ 1.5 ns 22.5 cmCsF, LaBr3 ~ 0.45 ns 6.7 cmBaF2, LSO, LYSO ~ 0.3 ns 4.5 cm

Time-of-flight PET

Figure 1. Image elements contributing to a LOR, for conventional PET (left) and TOF PET (right).

Time-of-flight PET

Even with a finite time resolution, using the TOF information an improvement in signal-to-noise ratio (S/N) can be achieved:

. .2

≅ =∆ ∆TOF conv convD DSNR SNR SNRx c t

Can we see TOF improvement?

Noticeable improvement with TOF with large size phantom

1 min

non TOF TOF

3 min

6-to-1 contrast; 35-cm phantom

5 min

J. Karp, U of Penn

QUALITY CONTROL (QC)

Quality Control

Ensure operational integrity of the systemMaintain consistent and high image qualityMinimize chance for artifactsCatch potential problems earlyMaintain quantitative accuracy

PET System, storage of data

Each detector is in coincidence with a sector of opposite detectors.

The angle of the section defines the Field of View (FOV)

PET System, storage of data

All coincidence lines that are parallel at a given angle form a projectionwhich is the accumulated radioactivity in that angle (1D) of the object (2D).

In the computer we store the projection in a matrix

x-axis radioactivity profile, y-axis the different angles

Projection & Sinogram

Sinogramt

θ

Sinogram:All projections

P(θ,t)

f(x,y)

t

θ

y

x

X-rays

Projection:All ray-sums in a direction

π

PET System

In most PET scanners, a large number of scintillation detectors arearranged in a circle. Each detector is in coincidence with a number ofopposite detectors. The field-of-view (FOV) of the scanner is defined bythe width or the angle of the fan.

All coincidence lines that are parallel at a given angle form a projection inthe sinogram.

r

θ

PET System

All coincidence lines (or lines of response) for a given detector form adiagonal trace in the sinogram.

r

θ

Daily QC / Blank Scan

Normal QC Detector Failure

Detector Controller Failure Coincidence Processor Failure

The sinogram as a help to identify errors

Buchert et. al. JNM 40, 1999

Daily QC / Cylinder Scan

BeforeNormalization

AfterNormalization

Daily QC Procedure

Daily Detector CheckSources & Phantoms needed

Transmission/Rotating Rod SourcesUniform 68Ge cylinder phantom

Daily / Weekly Quantitation Scan

Scan uniform 20 cm Ø 68Ge or 18F Cylinder

Reconstruct:

All corrections applied

Standard reconstruction parameters

Visual inspection

Compare image ROI activity to calibrated activity

Always perform after:

Service

Re-tuning

Re-normalization

Quantitation

2D Cylinder (0.0299 µCi/ml)

FBP OSEM

Calculated Atten. Corr. 0.0296 -1.06% 0.0295 -1.5%

Meas. Atten. Corr. CT 0.0285 -4.7% 0.0284 -4.0%

Meas. Atten. Corr. Rods 0.0242 -19.2% 0.0239 -20.2%

Quarterly QC Procedures

Detector setup (if needed)PMT tuning

Detector setup

Coincidence timing

Normalization

Gantry alignment (for PET/CT)Always after Service

Software upgrades

Other calibrations (well counter, etc)

Jaszczak Phantom

CT PET

PET/CT Gantry Alignment

Annual QC Procedures

Perform a sub-set of the acceptance test:UniformityResolutionCount Rate TestDead Time correction

Test Phantoms