Gamma Cameras and SPECT Flashcards
The resolution of a collimator is expressed as the
FWHM of a line source
Collimators generally involve a tradeoff between
spatial resolution
sensitivity
parallel-hole collimators may be classified as
high sensitivity (low resolution)
general-purpose
high resolution (low-sensitivity)
Typical collimator for Tc-99m and Tl-201
low energy
Typical collimator for Ga-67, In-111
Medium energy
Typical collimator for I-131 and F-18 (non-PET)
High-energy
Typical count range for planar image
500,000 to 1 M
SPECT images are typically reconstructed using this matrix
128 x 128
For brain SPECT, typical collimators may include
hybrid convergin and parallel hole
Typical bit depth in terms of gray levels
8 bit, 256 gray levels
Gamma cameras typically have this number of PMTs
37, 61, 91 PMTs
A typical window for Tc-99m imaging
20%
Efficiency of the detector crystal increases with
thickness
Inclusion of scattered photons in a gamma camera acquisition causes
decrease in spatial resolution
Primary cause of limited intrinsic resolution in a gamma camera
statistical fluctuation in the distribution of light photons among photmultiplier tubes from one scintillation to the next
As gamma or x-ray energy decreases, intrinsic resolution
decreases do to larger relative statistical fluctuations in the light photons per scintillation event.
As a rule of thumb, intrinsic resolution in a gamma camera and energy are related as
1 / √E
(similar to relative uncertainty in count statistics)
Thicker detectors result in
increased sensitivity
decreased intrinsic resolution
The system resolution is largely determined by the
collimator resolution
Efficiency for 140 keV with NaI(Tl)
70% to 90% for 6.4mm and 12.7 mm thick crystals
Efficiency of NaI(Tl) for 500 keV photons
less than 20%
The energy resolution improves according to photon energy as
1 / √E
due to relatively decreased statistical fluctuations at higher energy
For Tc-99m, a lower energy threshold of 130 keV should remove photons that have been scattered through angles greater than
45 degrees, but only at about 50% due to the blurring of the spectrum (for scattered and unscattered photons)
As count rates increase, the accuracy of dead time corrections
decreases
True or False
In NM imaging studies,most are not impaced by dead time losses
True, with the exception of high-counting rate applications such as first-pass cardiac studies
In a NM image with high count rates from two sources, the resulting image may look like
An undesirable effect of pulse pile-up in terms of the image
image distortion
Modern gamma cameras employ this approach for pile-up correction
pulse-tail extrapolation
Which happens between the PMTs and over the PMTs?
Pincushion occurs over the PMT due to nonlinearities in sensitivity as you approach the center. Barrel distortion occurs between PMTs
Cause of artifacts
nonlinearities: pincushion and barrel effects seen with line pattern and homogeneity test
Two causes of gamma camera nonuniformities
- Differences in pulse-height spectrum for different PM tubes (tube response can be adjusted with careful tuning)
- position-dependent collection efficiency of scintillator light
Edge packing, due to reflection of scintillator light from the sides of the detector crystal back towards the PM tubes is mainly masked by
Useful Field of View
UFOV
For higher energy photons, interactions occur closer to the PMTs, resulting in a more narrow light spread and a __________ of detector nonuniformity
worsening
An intrinsic flood image is used to apply
spatially varying energy corrections
spatially varying energy corrections include
- spatially varying PHA settings
- normalized intensity correction (count skimming) : certain percentage of counts in hot spots are thrown out {older approach}
- use of lead sheet with uniformly spaced array of holes (~1mm diam, ~4 mm separation) at known locations, placed without collimator - lookup table corrects for offsets in x and y {mfg}
In newer cameras, tuning is performed this way
LED attached to each PMT is used to produce a light signal, PMTs are then automatically adjusted.
A reasonable amount of septal penetration is
about 5%
Upper limit for low energy collimators
about 150 keV
upper limit for medium energy collimators
about 400 keV
System resolution combines resolution for each part as
Intrinsic resolution may be determined by
linearity test pattern with point source at 5 x UFOV
acquire with matrix so that min pixel size is less thatn 1/10 expected resolution (usually < .35 mm)
Typical intrinsic spatial resolution of a gamma camera
2.5 to 3.5 mm
System resolution with a line source
2, 1 mm line sources spaced about 5 cm apart, 10 cm from collimator surface, several million counts
Typical system (extrinsic) spatial resolution for Tc-99m
8 to 14 mm
Integral uniformity
100 % * (max - min)(max + min)
pixel counts
Differential uniformity
Over 5 consecutive pixels
100% x (high - low)/(high + low)
Typical integral uniformity in the UFOV and CFOV
2% to 4%
Typical energy resolution for Tc-99m
8% to 11%
Typical sensitivity for gamma camer
2 to 3 x 10-4 cps/Bq
or .02% to .03%
System sensitivity can be measured using
flask with small amount of Tc-99 (10 ml or so) to minimize selb-absorption
NaI(Tl) has a thin aluminum casing to
protect it from moisture
Most general purpose cameras have a crystal thickness of
about 9.5 mm
This image is
gamma camera head with the cover removed showing PMTs above sinctillator
To form an image, gamma cameras use
absorptive collimation rather than a lens like a traditional camera uses
For a parallel hole camera, the image and the object size as a function of distance are
the same
For a pinhole collimator, the image and object size are related by
I/O = f/b
f = height of collimator
b = distance from collimator to object
In a pinhole collimator, the lead cone is about this high
20 to 25 cm
For a diverging collimator, image and object size are related in this way
I/O = (f-t)/(f+b)
f = focal length (collimator to point)
b = distance collimator to object
t = thickness of collimator
Approaches to spatial resolution testing may include
bar phantoms
line sources (line spread function)
The FWHM of a line spread function used in system resolution is
approximately 1.4-2 x width of the smallest resolvable bar pattern
the most detailed specification of spatial resolution is provided by the
modulation transfer function
The equation for the modulation transfer function
MTF = Min / Mout
Min =(Imax - Imin)/(Imax + Imin)
Mout =(Omax - Omin)/(Omax + Omin)
I and O are intensities, you would plot as a function of varying spatial frequencies
Contrast of a lesion is defined as
Cl = (Rl - Ro)/Ro
Ro is the counting rate outside the lesion
the percent contrast is
showing that a high background count rate relative to count rate outside the lesion will decrease percent contrast
Background counts can results from
septal penetration
scattered radiation
inadequately shielded sources
As the energy window is increased, spatial resolution
decreases
statistical noise or quantum mottle can be reduced by
summing measurements, unlike structured noise
The Rose criterion
An object’s CNR must exceed 3 to 5 to be detectible (shown in multiple research studies based on observers viewing lesions or objects in noise generated contrast patterns)
Lesion contrast to noise ratio is defined as
In a sinogram, each row represents
a projection at a specific angle
A break in the sinogram may represent
patient movement
The effect of the ramp filter is to
enhance high spatial frequencies and suppress low spatial frequencies to eliminate 1/r blurring
A rounded shape ramp filter will help supress unwanted high spatial frequency amplification (i.e., Hann Filter)
According to sampling theory, in order for the highest spatial frequency to be recovered, the data must be sampled at
twice this frequency, otherwise aliasing will occur.
As a rule of thumb, the sampling requirement for an imaging detector is
sampling distance <= FWHM / 3
Spoke-like artifacts occur around small high entensity objects when
the number of angles is insufficient
In a SPECT system, the minimum number of views required for a linear sampling distance, Δr is
The minimum number of angular views should be acquired over a __________ degree arc
180
This is an example of
image distortion due to insufficient sampling
Sinogram showing parts of object are not included in all projections
missing / damaged projection element
In SPECT, to keep the same SNR while increasing the resolution by a factor of 2…
the total number of counts must increase by 8 (inverse cube of pixel size)
Partial volume effect
The sum of all pixels containing the object reflect the radioactivity, but the intensity for a given pixel may be reduced due to averaging
spillover in SPECT occurs when
an area of low activity is surrounded by higher activity regions - reducing contrast and apparent size
The recovery coefficient in SPECT is
ratio of apparent concentration to true concentration, usually plotted as a function of cylinder diameter/FWHM (x) and recovery coefficient (y). It goes to 1 as the cylinder diameter / FWHM increases to > 2
In SPECT, transaxial resolution refers to
in-plane resolution
In SPECT, axial resolution refers to
component that is perpendicular to the slice along the axis of rotation (z direction) - sometimes referred to as slice thickness
For SPECT, resolution degrades as you move
away from the collimator
A typical resolution for SPECT is
12 mm FWHM
A major benefit of SPECT
ability to resolve underlying structures
The most frequent use of SPECT is for
myocardial perfusion for assessing coronary artery disease and heart muscle damage following infarction