PURE SPECT PHANTOM

User Manual

www.PureImagingPhantoms.com

Durham, N 405
2
3
Contents
Introduction ............................................................................................................................5
Use of the Benchmark, Standard and Elliptical SPECT Phantoms .......................................6
Use of cylinder without inserts ..............................................................................................6
Placement of Solid Spheres ...................................................................................................8
How much activity do I use? (also see Appendix I) ..............................................................8
How is the phantom mounted in the scanner? .......................................................................9
Slice thickness: thin vs. thick .................................................................................................9
Draining the phantom ............................................................................................................10
What if the phantom leaks?....................................................................................................10
Proper care .............................................................................................................................11
Appendix I: Filling and securing Phantom ............................................................................13
Appendix II: Example protocol and phantom specifications .................................................14
Sensitivity measurements.......................................................................................................15
Line spread measurements .....................................................................................................16
Image statistical (count density) noise ...................................................................................19
Image contrast - example using sphere ..................................................................................21
Appendix III: Sample images ................................................................................................23
SPECT phantom ..................................................................................................................24
Deluxe cold rods ..................................................................................................................25
Standard cold rods ...............................................................................................................26
Benchmark cold rods ...........................................................................................................27
Deluxe cold rods .................................................................................................................28
Bibliography ..........................................................................................................................29
General ................................................................................................................................29
Physics .................................................................................................................................29
Image artifacts and compensation techniques .....................................................................30
Warranty ................................................................................................................................31

4
 SPECT Phantom User Manual

Introduction

It is possible from a single scan of the SPECT Phantom to evaluate many system
performance characteristics, including:
 System single-slice volume sensitivity
 System total volume sensitivity
 Effect of regional variations in intrinsic system response using uniform cylindrical
portion
 Accuracy of attenuation compensation algorithms
 Variation in spatial resolution within the field-of-view using the multi-sized rod insert
 Lesion detectability using the multi-sized spheres
 Effect of finite spatial resolution and Compton scattering on image quality
 Image contrast, %-rms noise and signal-to-noise (S/N) measurements
 Line spread measurements in air and in water, without having to remove the cover plate
It is our philosophy that the reported variability of SPECT results reinforces the need
for standardized, routine quality control in ECT, even to a greater degree than in

5
SPECT Phantom User Manual

conventional nuclear medicine imaging. Emission CT is quite unforgiving of “short cuts”

and compromises in operator performance. In research and clinical settings, artifacts and
limitations in system capability must be recognized in order to understand “where your
system stands” relative to others. We feel that SPECT system evaluations and comparisons
are not easily or accurately performed using patient data due to patient variability and the
impossibility of knowing the “truth” in a patient. For example, in a SPECT liver scan how
would one determine if their system has 10, 15, or 20 mm (FWHM) resolution? We believe
this is not easily possible. A phantom such as the SPECT phantom provides a means for
accurately assessing what your ECT system is capable of and how well it is functioning.
This information can then be meaningfully transferred to the case of patient imaging. What
might appear to be a 10 mm cold lesion in a patient should be readily simulated and
demonstrated in an appropriate phantom study as a way of gaining confidence in the
interpretation and presentation of patient data. As a general rule, if you can’t see it in an
image of the phantom, you’re not likely to see it in a patient scan.

Use of the Benchmark, Standard and Elliptical SPECT

Phantoms
Within the basic SPECT Phantom cylinder, inserts may be used in a number of
possible combinations. These include:
 No inserts, in order to use the entire length of the cylinder as a uniform section.
 Cold rods alone, with uniform section in remainder of cylinder.
 Spheres alone, in uniform concentration background activity.
 Capillary line sources (user supplied) alone in air, or in “cold” (water) background.
 Cold rods, spheres and uniform activity section.
This latter configuration is the most popular for routine evaluation of camera-based
SPECT systems.

Use of the Cylinder without Inserts

Scanning the cylinder without any inserts can provide information on the accuracy of
on several system or reconstruction parameters such as an attenuation compensation
algorithm, single slice and total volume sensitivity, and the presence of circular
reconstruction artifacts due to field uniformity problems. In addition, the percent-root-mean-
square (%-rms) noise can be measured for a given activity concentration, or equivalently for
a given total number of detected gamma events within a given slice. Filling the empty
cylinder in preparation for scanning is quite easy.
With the cylinder cover plate removed, fill the cylinder to within an inch of the top
with room temperature tap water (do not use demineralized water). Then put the cover plate
back on making sure the black rubber O-ring is properly seated in its grove in the rim of the
cylinder. The O-ring provides a watertight seal by just hand tightening the small cover plate
screws provided with the phantom. These special nylon screws can be damaged or even
broken if over-tightened. Use only enough force to create a firm and uniform contact
between the O-ring, the cover plate and the cylinder rim. At this point, the radionuclide may

6
 SPECT Phantom User Manual

be introduced into the cylinder through one of the openings (“filler-holes”) in the cover plate.
After screwing both caps (plugs) into the cover plate, agitate the cylinder to completely mix
the radionuclide and the water solution. Remove both caps, to permit air to escape, and
finish filling the cylinder with water through one of the openings. A 30 or 60 cc disposable
syringe, with a “catheter” tip, is useful for this purpose. Repeatedly withdrawing into and
expelling from the syringe will afford sufficient mixing of the extra water and radionuclide.
Be careful of contamination during final filling of the cylinder. Replace both caps and set the
phantom aside for a few minutes to allow for more uniform distribution of the radionuclide
with the extra water that was added. Be sure to observe the usual precautions when handling
radioactive substances within the laboratory: wear gloves and protective clothing, use
shielding whenever possible, and avoid unnecessary exposure to yourself and others by using
time, shielding and distance.
The above procedure describes one technique of preparing the cylinder for scanning.
An alternative approach is to pre-mix the solution in another separate container which has a
larger volume than the cylinder. You may then fill the SPECT phantom from the second
container. This method may be faster for some individuals and allows immediate use of the
phantom, since there is no waiting required for the radionuclide to completely mix. A large
(30 or 60 cc) catheter tipped syringe can be used to complete the filling once the cylinder is
nearly full and you have secured the cover plate in place.
A third method is to use a magnetic stirrer to mix the radionuclide once it has been
introduced into the water. This method can be executed by filling the cylinder with “cold”
(non-radioactive) water and sealing, except for one of the caps, leaving just enough air to be
replaced by the radionuclide. Cap the phantom and place on the magnetic stirrer. A small
bar, placed in the bottom of the phantom before filling with water, will then quickly mix the
contents. This method may not work with the cold rod insert, unless the bar is placed on top
of the insert and the phantom turned upside down and placed on the stirrer. In using this
technique keep in mind that the bar might be visible in some of the reconstructed image
slices.
We encourage that you experiment to become aware of the best method of filling the
phantom for your particular application. As you gain experience filling the phantom this task
will become more efficient and seem routine. One hint with regard to water temperature: Do
not fill the phantom with cold water and tightly seal it since this water must expand as it
warms to room temperature. This expansion can damage the phantom, especially if it is
sealed extremely tightly. Prior to filling the phantom, we have found it useful to fill a
separate large container with several liters of cold tap water and allow it to set for a few
hours until it reaches room temperature. In general it is not possible to pick the "ideal" water
temperature each time, so quite often due to the increase in pressure that is associated with
small amounts of expansion, a bit of spray may result when a filler cap is removed. To
protect yourself and nearby objects from this spray, always loosen filler caps with a towel
surrounding the cap. Note that allowing the water to equilibrate to room temperature before
sealing the phantom, or simply using warmer, room temperature water to start with can help
alleviate most of these temperature related issues.

7
SPECT Phantom User Manual

Placement of Solid Spheres

The ECT phantoms are pre-assembled prior to packaging with the exception of the
cold spheres. The set of spheres (6) are located within the cylinder, the spheres are mounted
in the cylinder by gently screwing the bottom of the spheres into threads located on the
bottom of the cylinder. Inside the cylinder when looking at the cold rod insert you will see
the top plate and six pie shaped sections of cold rods. Between each section is a row of
vacant holes. The spheres are inserted into the third vacant hole, which should line up with
the threads on the bottom of the cylinder. For most users the largest sphere is set counter-
clock wise from largest pie section. Gently screw the remainder of the spheres in the threads
(going counter-clock wise) with the next size down to the smallest spheres next to each pie
section.

How Much Activity Do I Use? (See Appendix I)

Phantom studies generally require a greater amount of radionuclide activity than do
patient studies. The reason is that many clinical studies are count limited, introducing
statistical variations that may mask artifacts and leave some doubt as to the actual useful
spatial resolution. Since SPECT scans can be relatively time consuming, using an additional
amount of activity can reduce the time per projection and hence the total scan time. Because
of the variation in SPECT scanner designs (multiple or single gamma cameras, single or
multiple slice systems, and positron systems), collimation, slice thickness and other factors, it
is impossible to provide specific values for the imaging parameters. However, the following
parameter values may be used as reasonable starting points. The users are expected to modify
these parameter values to suit their specific applications.
Resolution tests with the cold rod insert require about 20 million counts per slice to
adequately demonstrate the resolving capability and the susceptibility to aliasing artifacts of
an ECT system. To obtain these counts typically requires the use of a thick (approximately 8
cm) slice encompassing the entire height of the rods. The spheres may be imaged using a
thinner (approximately 1.2 cm) slice in order to avoid degradation from partial volume
effects. The capillary line source(s) should contain at least 200-300 thousand counts-per-
source per slice. As an example, we will consider a dual camera ECT system collecting data
in a 25 minute, 360 degree scan. The slices might be 25 mm thick using a general purpose,
medium sensitivity collimator. The activity concentration (measured in microcuries (Ci)
per milliliter) will influence the final value for the number of counts detected per slice. It is
primarily affected by which inserts (if any) are in place to obtain a desired number of counts.
One way to obtain the thick slices is to scan using a coarse matrix size (i.e., a larger
mm/pixel value). A better method, in that it will be more sensitive to poor uniformity, is to
scan using a finer matrix size (128 or 256 pixels wide matrix). The reconstructed slices can
then be added together after evaluating them for small ring artifacts, to see if slice adding
will help cancel out hot and cold ring artifacts. Of course, several cold rings over several
slices, or several hot rings over several slices, will reinforce each other. If you have to pick
one specific matrix size and only perform one scan, as you would for your typical quality
control procedure, a reasonable compromise might be to perform the QC procedure with the
same matrix size as you normally would perform patient scans with.

8
 SPECT Phantom User Manual

A suggested starting point is to use approximately 20 mCi of activity (for SPECT

systems) when imaging the empty (“uniform”) cylinder, or with the rods and/or spheres in
place. This value may be doubled when changing to a High-Resolution collimator. The
capillary line sources may be filled using a radionuclide source having a concentration of
about 20-50 mCi per mL. This will result in a few milliCuries (mCi) per line source due to
the small volume of the glass capillary tubes used for the line source. Remember that the
values noted above for the minimum counts-per-slice approximately correspond to the
minimum that will provide you adequate statistics to determine: 1) information relating to
uniformity of resolution throughout the reconstruction plane, and 2) reliable data on the
system's susceptibility to artifacts. The above suggested values may be increased or
decreased as your imaging situation warrants.

How Is the Phantom Mounted in the Scanner?

At this stage with your SPECT phantom filled and inserts (if any) inside you are
ready to begin imaging. The long axis of the phantom should be accurately aligned parallel
to the axis-of-rotation of the ECT system. The axis-of-rotation is the line in space about
which the cameras rotate. When properly positioned, the flat ends of the phantom are where
the head and feet of a patient would be. In this position the face of the ECT gamma
camera(s) is next to the round side of the cylinder. If available, the SPECT Phantom may be
placed on a low attenuating table designed for the SPECT systems. A narrow table (or the
portion of the table used for brain scanning) is desirable since it reduces the radius-of-
rotation for scanning. For cylindrical phantoms having a rim (i.e., flange or lip), it has
been found convenient to place the phantom on the imaging table or bed with the rim and
cover plate extending over the edge of the table. The rim and the cover plate have an outer
diameter that is somewhat larger than the main body of the cylinder. Positioning the
phantom in this way will ensure that the collimator-to-phantom distance is minimized. In
this position, the main body of the cylinder rests squarely on the table for added stability
and the long axis of the phantom coincides with the axis-of-rotation of the SPECT system.
Carefully attach the phantom to the bed with strong tape (e.g., “duct” tape) so that it does
not fall off the table. A phantom rolling off the table could easily be destroyed or it could
severely damage collimators and/or crystals. Be certain that the phantom is not tilted up,
down or laterally with respect to the axis-of-rotation.

Slice thickness: thin vs. thick

Image slices greater then 10 mm pixel thickness are considered thick and may be used
when the data or object structure is not expected to change from slice to slice. For example,
this would be true in the case of the Rod Insert. This increase in slice thickness is helpful in
order to increase the counts per slice in the reconstructed images. The disadvantage of using
thick slices is that structures that change rapidly from slice to slice, such as the cold spheres
or liver lesions, are blended together in the reconstructed images. This results in decreased
resolution and image contrast. It is best to use thinner (1 to 2 cm) slices for the spheres and
for the uniform section of the phantom, since thin slices are more representative of the effects
of nonuniform detector/collimator response. This is particularly true if one suspects small
localized regions of nonuniformity that might occur with separations of collimator septa or

9
SPECT Phantom User Manual

collimator dents. Thicker slices may be used with the Rod Insert and the Line Sources,
provided the phantom is mounted such that the long axis of the phantom is parallel to the
axis-of-rotation. Obtaining the scan with a thin slice thickness allows for later slice addition,
which is often useful for viewing the rod insert. For this method to work well, you must very
accurately align the phantom (in this case, the long axis of the rods) with the axis-of-rotation.
How do I know if the phantom is accurately aligned with the axis-of-rotation?
One may position the phantom and view a persistence scope or computer monitor
image to help determine proper alignment. Quickly acquired planar test images might work
better on some systems. Make sure that the top/bottom and the sides of the phantom are
“square” with the sides of the image matrix. You will need to perform this for the camera
looking up or down at the cylinder (like an anterior or posterior view), and then with the
camera looking from the “side” (like a lateral view). Repeat this another time or two until
you are satisfied with the positioning. This alignment is most critical for situations in
which you want to add many, perhaps all, of the slices containing the cold rods. Alignment
with the axis-of-rotation is not very critical for imaging the other inserts such as the spheres
or uniform activity. Line sources should be reasonably well aligned so that measurements of
resolution (FWHM or FWTM) provide you with accurate values.

Draining the phantom

There may be a considerable amount of radiotracer solution left in the phantom when
scanning is complete. Several methods may be considered for the safe disposal of the
radioactive contents. For short-lived radionuclides, the phantom may be set aside until the
activity is low enough for disposal through a hot sink. Another option is to empty the
phantom contents into another storage container. This method may allow for quicker reuse
of the phantom using a different radionuclide. Remember to wear rubber gloves and other
protective clothing when filling and draining the phantom. Use time, distance and shielding
whenever possible to minimize radiation dose to personnel.
The liquid contents of the phantom may be removed either through the uncapped
filling ports in the cover plate or by removing the cover plate itself. A small amount of the
contents will likely be spilled if removing the cover plate without first draining about 100 mL
of the contents through the ports in the cover plate. A few drops can spray out of the
phantom under certain conditions. If the phantom is filled with colder than room temperature
water, pressure can form inside as the water temperature begins to rise to room temperature.
This can be avoided by using warm water to fill the phantom. As mentioned earlier, cover
the first cap to be removed with a towel to absorb any spray.

What if the Phantom Leaks?

When properly filled, sealed and maintained, the phantom should not leak. Special care
was taken in the initial design considerations in order for the phantom to maintain its
integrity under normal operating conditions. If slight leakage is noted, check the following
points:
1. Is the large O-ring properly seated in the special groove?

10
 SPECT Phantom User Manual

2. Is the O-ring slightly lubricated with silicone grease?

3. Are the O-rings damaged (cuts or abrasions)?
4. Is the phantom cracked or otherwise damaged?
5. Are the cover plate plugs and/or caps tight? Be sure not to over-tighten the caps.
6. Are the small O-rings (on plugs and/or caps) seated properly?
7. Are the round O-rings (on plugs and/or caps) slightly lubricated?
8. Are the plugs and/or caps “cross-threaded”?
9. Are the cover plate retaining screws tight? Be sure to not over-tighten the retaining screws.
Please contact Data Spectrum Corporation if additional problems occur.

Proper Care
1. Periodically clean the phantom using only a mild (dishwashing or other) soap
solution.
2. If the phantom is not to be used for an extended period of time, remove liquid, clean
and dry. Remove caps and leave the inside of the phantom open to air for ventilation.
3. Periodically inspect phantom for damage. If damage has occurred, repair or replace
phantom prior to further use.
4. The O-rings require lubrication and should have a liberal amount of silicone-based
grease applied every month or so depending on how often the phantom is used.
Silicone grease may be purchased in the plumbing department of a home
improvement store such as Home Depot or Lowes.
5. Handle the phantom and inserts with care. Each component is precision- and custom-
made. Avoid dropping or striking since acrylics are relatively fragile..
** WARNING ** Avoid over-tightening caps or other plastic or nylon parts. Threads
can be damaged or screws broken. Usually, hand tightening of any threaded part is adequate.
Avoid using metal screws on the cover plate. With the flangeless model never, never use
longer replacement cover plate screws than the ones that come with the original phantom.
Longer screws, particularly longer metal screws, will “bottom out” before the cover plate is
seated and potentially will crack the thin acrylic cylinder. This would void the warranty.
Also, with the flangeless model, be careful not to confuse the threaded screw holes in the
cover plate that can be used to easily remove the cover plate from the cylinder with the un-
threaded holes that are use to secure the cover plate to the cylinder. Failure to follow this
precaution may damage the screws and/or the phantom. Data Spectrum Corporation can
supply replacement components. The screws can also be lubricated with silicone-based
grease. Silicone grease can be obtained in the plumbing department of home improvement
stores such as Home Depot™ and Lowes™.
Avoid strong detergents or “glass cleaners” containing ammonia, alcohols,
solvents, chlorinated hydrocarbons, petroleum products, extreme hot or cold water, and
exposure to extreme heat and cold. Never, under any circumstances, fill the phantom
with alcohols, solvents, chlorinated hydrocarbons, or petroleum products.
A maximum of 5 parts per million of chlorine can be safely dissolved in the water
used to fill the phantom to retard biological growth. Swimming pool chlorine additives may

11
SPECT Phantom User Manual

be used, but it is important that the chlorine concentration be measured accurately with an
appropriate test kit, such as those used to test swimming pool or spa water.
We wish you the best of success with the use of your emission computed tomography
system, and hope that your Data Spectrum SPECT Phantom can play an important role in
acceptance testing and routine quality control procedures. We would be most interested in
your comments, questions, suggestions and results. Thank you for considering Data
Spectrum Corporation. For your convenience we have included in the following pages, a few
representative quantitative calculations that can be performed using the SPECT Phantom.

12
 SPECT Phantom User Manual

APPENDIX I: Filling and Securing Phantom

How to estimate the amount of activity to fill Phantom in order to obtain
approximately 20 million counts in an 8 cm-thick SPECT section of cold rods:
Assume 30 minutes (1800 sec) of active data acquisition.
Method: Take 1 mCi of Tc-99m and place it in air at a distance of 10 cm from the
parallel-hole collimator surface and measure the point-source-sensitivity in
(cnts•sec-1)•mCi-1, using a centered 15-20% energy window. Use the measured point-source-
sensitivity (for your specific collimators/system) and the following table to estimate the
amount of activity necessary to obtain 22 million counts in an 80 mm-thick slice containing
the rods. This is the total activity that is to be used in filling the SPECT Phantom.

Point Source Approx. Scan

Sensitivity Time Activity*
Collimator Type (cnts•sec-1)•mCi-1 (minutes) (mCi)

Low Energy All-Purpose 5,700 30 20

High Resolution 3,600 30 31

Ultra-High Resolution 1,500 60 33

* Activity in SPECT Phantom required for obtaining approximately 20 million counts within an 8 cm-thick
slice containing the cold rods. A single camera SPECT system has been assumed. Note that a “high”
resolution collimator (or “LEAP”, “LEUR” etc.) from a given manufacturer does not necessarily correspond
to the “high” resolution collimator of another manufacturer, either in sensitivity, high energy penetration or
resolution.

13
SPECT Phantom User Manual

APPENDIX II: Example protocols and phantom

specifications

GUIDELINES FOR SPECT PHANTOM IMAGES (Revised 8-96)

I. PHANTOM: Data Spectrum Corporation’s Deluxe Model ECT/DLX/P:

A. Diameters of Cold Rods: 4.8, 6.4, 7.9, 9.5, 11.1, 12.7 mm
B. Diameters of Cold Spheres: 9.5, 12.7, 15.9, 19.1, 25.4, 31.8 mm
C. Approximate scan time: 30-60 (minutes).

II. COLLIMATION: Use of your High or Ultra High Resolution Collimator (LEHR or
LEUR) is recommended.

III. COUNT DENSITY: The total counts-per-slice should be as follows:

A. For Cold Rods: 20.0 + 5.0 Million counts-per-slice
B. For Cold Spheres: 2.0 + 0.5 Million counts-per-slice

IV. PIXEL SLICE THICKNESS (Recommended values):

A. For Cold Rods: Approximately 8 cm (not mm) after slice addition.
B. For Cold Spheres: Approximately 6 to 12 mm.

V. OTHER ACQUISITION AND RECONSTRUCTION PARAMETERS:

You are free to choose other parameter values; however, the following values in
bold are recommended for high quality SPECT scanning.
A. No. of angular samples: 64, 90, 120 or 180
B. Angular range: 360 degrees
C. No. of linear samples per acquisition: 64, 128 (corresponding to 3 to 4 mm/pixel)
or other.
D. Reconstructed image matrix size: 64×64, 128×l28 or other.
E. Radius-of-Rotation: 13 cm.
F. Filter used: Ramp, Hanning, Hamming, Shepp-Logan, Butterworth or other.
G. Cut-off frequency: greater than or equal to 1.4 cycles/cm.
H. Collimator Used: High Res
I. Energy window: 15 – 20%

VI. RECONSTRUCTED IMAGE, REQUIRED FORMAT AND FILM EXPOSURE:

A. Image size: 1.5 + 0.25 inches (i.e., size of image of phantom on film).
B. Maximum count density must be recorded on transparency film as black (opaque).
C. Zero count density must be recorded as white (i.e., clear).
D. “Thresholding” or “windowing” must not be used!
E. Gray scale that was used must be recorded on film for calibrating actual count
densities to film densities.

14
 SPECT Phantom User Manual

SENSITIVITY MEASUREMENTS
Sensitivity can be measured using any of several methods, although each method may
provide slightly different results. Sensitivity measurements can be made with an isolated
point source in air, distributed (“flood”) sources or cylindrical phantoms. One widely used
method, described below, uses a water-filled, cylindrical phantom filled with a low
concentration of radioactivity. As one may expect, a number of variables should first be
considered before acquiring the data for the sensitivity measurements.
One debate revolves around the question of whether sensitivity should be stated in
terms of single slice sensitivity or total volume sensitivity. For scanners that acquire only
one slice at a time, the numbers are the same. Total volume sensitivity can be considered as
the sensitivity for a single slice multiplied by the total number of slices for a multi-slice,
camera based single photon or multiple ring positron ECT system.
Sensitivity may be expressed in terms of (cnts•sec-1)•(Ci•ml-1)-1. In order to
determine these measurements, first fill a 21 cm diameter cylinder with a measured volume
of water. No inserts (rods, spheres, etc.) should be in the phantom. Add—and mix well in
this water—a known amount of radioactivity of the radionuclide chosen for the
measurements. For example, 2.49 mCi (at scan time) was mixed into 7155 mL of water,
resulting in 0.348 Ci•ml-1. The SPECT data acquisition was for a total of 349 seconds,
using a low energy, high resolution collimator. The radionuclide was Tc-99m and the energy
discriminator was centered at 140 keV with a 20% (FWHM) wide window. The slice
thickness was 2.5 cm with a total of 11 simultaneous slices in one acquisition. This
particular study was performed with a dual-head SPECT system, a fact that must be taken
into account. Note that the amount of radioactivity was kept low mainly to eliminate the
effects of system dead-time on the sensitivity measurements.
When the projection data were reconstructed, the number of counts-per-slice for one
detector in the original data were as listed below:

Slice 1 : 2,745 Slice 2 : 11,866 Slice 3 : 124,044

Slice 4 : 160,700 Slice 5 : 165,690 Slice 6 : 166,044

Slice 7 : 164,829 Slice 8 : 157,869 Slice 9 : 158,821

Slice 10 : 89,755 Slice 11 : 5,002

Note that the counts in the lower and higher slices are significantly lower in number
than those in the middle of the phantom (slices 4 through 7). The reason is because this
particular phantom did not span the entire length of the field-of-view along the axis-of-
rotation. This must also be taken into account. If we average the counts for the center slices
-1 -1
(5 through 7), we shall arrive at an average of 165,521 (counts•slice •detector ). Dividing
165,521 counts by 349 seconds (the scan time), and dividing this quotient by 0.348 Ci•ml-1
(the concentration you prepared) gives us the sensitivity for one slice (pixel thickness equal
to 2.5 cm) for one detector, or approximately 1363(cnts•sec-1)•(Ci•ml-1)-1. Recall that the
total volume sensitivity is the product of the single slice sensitivity (1,363) times the number

15
SPECT Phantom User Manual

of slices (11) times the number of detectors (2). Thus the total volume sensitivity of this
ECT system, using the collimators and energy window settings previously described, is equal
-1 -1 -1
to 29,986 (cnts•sec )•(Ci•ml ) . The single slice sensitivity for both detectors, for a pixel
-1 -1 -1
slice thickness of 2.5 cm, is equal to 2,726 (cnts•sec )•(Ci•ml ) . Sensitivity measurements
are required at low counting rates so that deadtime effects are avoided or negligible.
Sensitivity measurements are a critical part of routine quality control. It is a good
idea to perform this type of measurement immediately after initial installation, after software
upgrades and after hardware modifications. Remember that collimators, window-widths,
dead-time, gamma ray energy, slice thickness, field-of-view and other factors can directly
influence the resulting values.

LINE SPREAD MEASUREMENTS

It is often useful to determine the SPECT system's response when imaging a line
source placed either in an “air-filled” phantom or in a water-filled phantom. This data can
provide information on the accuracy of the attenuation compensation technique being used,
the effect of changing collimation and/or reconstruction filters on the spatial resolution, and
on the uniformity of resolution on and off, the axis-of-rotation.
These measurements can be made using the SPECT phantom, in conjunction with
user-supplied glass capillary tubes having a thin outside diameter (approximately 1.2 to 1.5-
mm). These tubes (for example, Kimble #46485) are usually available in bulk quantities
typically as long as four feet, from laboratory supply houses. They can then be cut to proper
length (12-cm if the cold rod insert is also installed) using a small triangular file. They may
be filled with a solution of radionuclide using capillary action by tilting one end of the tube
and dipping it into the radiotracer solution. By holding the other end closed, the tube may
then be removed from the solution and sealed with a sealer such as “SealEase” or with a
putty compound such as “Mortite”. The cold rod insert (or the hot spot insert) should be
properly oriented within the cylindrical body of the phantom. Also, the two nylon supporting
bolts used to secure the insert to the bottom plate of the phantom must be properly
positioned. One should be on the long central axis of the cylinder and the other should be
located at 3-3/16 inch from this axis. The cover plate should be properly installed so that the
two filler holes are aligned directly over the two support rods. One glass capillary tube may
then be carefully placed through the central filler hole and fitted into the small hole drilled
into the top of the corresponding support bolt. Similarly, a second capillary tube may be
placed through the other (off-axis) filler hole and positioned into the small alignment hole
that is drilled in the top of the second (off-axis) supporting bolt. Note also, that the two filler
caps have a small (1/16-inch diameter) hole axially located on the end facing the interior of
the cylinder. The end of the capillary tube, which should be protruding about 1/4-inch above
the top surface of the cover plate (if the tube was cut to the correct length), is to be carefully
positioned within the centrally located hole in the filler plug. The filler cap may then be
screwed into the filler hole with the capillary tube positioned within it (providing, of course,
that the capillary tube was not cut too long or too short). This will securely fasten the
capillary tube within the cylinder. If the outside diameter of the capillary tube is slightly too
large, the small holes may be slightly enlarged using a small diameter drill and bit of proper

16
 SPECT Phantom User Manual

size; although, remember that if the inside diameter of the tube is increased too much, the
resolution measurements could be affected due to “tilting” of the tube. As the glass capillary
tubes are very fragile, it is advisable to practice the procedures of cutting the tubes and
mounting them a few times, prior to actually using source-filled tubes. Of course, the
phantom is to be filled either with “air” for line source measurements made in air, or with
water for measurements made in water.
After the line sources have been mounted, the phantom is positioned so that the line
sources are parallel to the axis-of-rotation, and then the line sources are scanned with the
SPECT system. Of course, conventional measurements may be made by simply taking
planar views. Measurements are made by generating numerical (or graphical) profiles
(displayed either linearly or on a log scale) through the reconstructed images of the line
sources. Since the line source response function of most SPECT systems is not precisely
circularly symmetric (largely due to geometric collimation variations as a function of the
distance of the source from the collimator surface), it is recommended that more than one
profile be obtained. Typically a radial profile (i.e., consisting of a line drawn through both
the axis-of-rotation and the center of the line source image), and a tangential profile (i.e.,
consisting of a line which passes through the center of the line source image and is tangent to
an imaginary circle whose center is at the axis-of-rotation, and whose circumference passes
through the center of the line source image) are used. These profiles may be a few pixels
wide in order to encompass the central region of the image of the line source. With the
numerical data values for the pixels comprising these profiles, either modulation transfer
functions (MTFs) may be computed, or, more simply, the full-width-at-half-maximum
(FWHM) and full-width-at-tenth-maximum (FWTM) of the line source image may be
estimated. The latter calculations are illustrated in the sample profile presented below.
These data correspond to a SPECT system using

Profile Pixel Number Profile Pixel Value

xi (channel number) yi(counts•pixel-1)
59 173
60 508
61 3,704
62 13,424
63 25,218
64 27,614
65 17,748
66 6,054
67 1,133

-1
3.2 mm•pixel linear samples, 180 angular views (2 degree angular sampling), continuous
rotation of the gantry, a 15 cm radius-of-rotation (measured from the front surface of the
-1
collimator) and a Von Hann filter with a cut-off frequency equal to 1.6 cycles•cm .

17
SPECT Phantom User Manual

Note the following values:

0.5ymax  13807 (that is: 0.5 × 27614)

FWHM (in pixels) = 2pixels

25218  13807
 times1pixel 
25218  13424

17748  13807
 times1pixel 
17748  6054

or FWHM (pixels)  2  0.96  0.34  3.30 pixels

If each pixel is equal to 3.2mm  k  3.2mm  pixel 1  , then

FWHM mm  k  FWHM pixels   10.6mm

Similarly, the FWTM is estimated as follows:

. ymax  2761.4
01

FWTM(pixels) = 5 pixels

3704  2761
+ times1pixel 
3704  508

6054  2761
+ times1pixel 
6054  1133

or FWTM (pixels)  5  0.29  0.67  5.96 pixels

FWTM (mm)  k  FWTM  pixels   191

. mm

Note that the data values were simply linearly interpolated to obtain the fractional
pixel widths. A more precise interpolation method (e.g., five-point) may also be used.
Another approach would be to perform a "least-squares-fit" of the data values to an assumed
Gaussian function, and then use that function to analytically calculate the FWHM and
FWTM. Note that since the linear sampling is only 3.2 mm•pixel-1, the measured response
function will be a convolution of this sampling with the true response. Thus aliasing can

18
 SPECT Phantom User Manual

occur and make the measurements somewhat inaccurate. Unfortunately, this effect will be
increased if one uses 6.4mm sampling (i.e., 64×64 reconstructions). Fortunately, most
SPECT camera manufacturers are supplying 128×128 capability. Ideally, even finer
sampling (e.g., 256×256 or 512×5l2) would be desirable for these measurements. Sometimes
a “zoom-mode” spatial resolution is a complicated function of intrinsic and extrinsic camera
performance, sensitivity variations (non-uniformities), spatial non-linearity’s and distortions,
collimation characteristics, centering, gain, asymmetric image size, radius-of-rotation, linear
and angular sampling intervals, reconstruction filters, attenuation compensation algorithm,
amount of scattering material present, etc. Thus, if one would like to use the above described
procedure to compare the SPECT system’s performance over a period of time, one must be
careful to ensure that none of these parameters have changed.

IMAGE STATISTICAL (COUNT DENSITY) NOISE

Using the reconstructed image of the uniform cylindrical portion of the SPECT
phantom, it is possible to measure the noise level (expressed here as the “percent-root-mean-
square,” or %rms, noise power). Assuming that the total number of counts (i.e., detected
gamma photons), or gamma events used to reconstruct the slice of interest can be determined
from the SPECT system (also various other parameters are required, such as the detailed
shape of the reconstruction filter, pixel size, and the diameter, either in pixels or in
centimeters), it is possible to directly compare the %rms noise level with values predicted
from equations published in the scientific literature relating to SPECT imaging. This makes
it possible to estimate whether the noise propagation characteristics of the complete SPECT
system are within reasonable agreement with theoretically expected performance.
In the example presented below, the following definitions and equations are used:

I. Sample Mean Count Density Yavg (counts•pixel-1):

N
1
Yavg 
N
y
i 1
i

where N is the number of pixels (picture elements) within a region of interest (ROI)
near the center of the reconstructed image of the uniform cylindrical portion of the
SPECT phantom, and yi is the number of counts per pixel in the ith pixel. It is assumed
for the computation that the SPECT image was reconstructed using an appropriate (e.g.,
Chang) attenuation compensation algorithm.

II. Sample Standard Deviation Sdev:

 y 
N
1 2
S dev   Yavg , or
N 1
i
i 1

19
SPECT Phantom User Manual

N
1
S dev 
N 1
( y
i 1
2
i )  N  Y avg
2

III. Sample Percent Root-Mean-Square Noise %rms:


% rms  S dev / Yavg 100 
IV. Example Data Values:
The following pixel values - expressed here as real fractions - were measured within a
5×5 pixel (each pixel equals 3.175 mm) ROI located near the center of the reconstructed
image of the uniform cylindrical portion of the SPECT phantom (using a 128×128 image
matrix and a generalized Von Hann filter having a cutoff frequency equal to 1.6 cycles per
cm):

32.233 34.913 33.054 30.085 33.766

31.912 34.966 34.815 39.553 37.188

33.327 31.854 27.700 30.004 30.600

27.815 33.685 35.303 29.958 30.965

32.334 32.224 31.866 29.129 32.209

Using the equations and definitions presented above, the measured parameters are:

Yavg  32.5(counts  pixel 1 )

Sdev  2.8

2.8
Fractional root-mean-square (rms) noise   0.086 , or
32.5
%rms  0.086100  8.6%

It should be noted that for a pixel “signal” value (i.e., the difference between that
pixel value and the average background Yavg) to be statistically significant, this signal value
should be at least 3 to 4 times greater than the measured standard deviation (Sdev). Also note
that the derivation presented above corresponds to noise related to regions of the size of a
“pixel” however, it is possible to extend this methodology to reflect the noise level found in
larger regions (i.e., such as 1 to 2 cm, etc.), although care, of course must be exercised.
Finally, remember that the noise spectral density of reconstructed images is not “white,” or
Poisson, but is actually structured due to finite angular sampling and other characteristics of
the reconstruction process.

20
 SPECT Phantom User Manual

IMAGE CONTRAST - example using sphere

Reconstructed images through the “cold” spheres portion of the SPECT phantom can
provide valuable information relating to the capability of the SPECT system to detect
realistic “lesions.” For example, by simply obtaining a series of SPECT scans with a
decreasing number of total detected events per slice (typical slice thickness equal to about 1
cm), it is possible to estimate (by visual inspection of the images) the minimum number of
gamma-photons required to detect a “lesion” (sphere) of a given size or diameter. This might
be quite useful when attempting to determine which collimator should be used, for example.
It is also useful in case it is expected that the performance of the SPECT system has
deteriorated. This can be easily accomplished by comparing current reconstructed images of
the spheres with earlier “benchmark” results. Remember, however, that all reconstruction
parameters (such as radius-of-rotation, reconstruction matrix size, collimator(s), total counts
in slice, slice thickness, reconstruction filter, angular and linear sampling intervals, etc.) must
not have been modified when comparing previous results.
In addition to qualitative inspection of reconstructed SPECT images, it is possible to
perform quantitative measurements by computing the sphere’s image contrast Cimage:

Counts  Pixel   Counts  Pixel 

1 1


sphere bkg

Counts  Pixel 
C im age 1
= Signal cnts/background cnts.
bkg

For example, the following pixel values (expressed here simply as normalized
integers) were measured within the reconstructed image of a large sphere using a small ROI
(3×4 pixels):

17,348 15,539 16,074

15,883 14,236 14,779

15,333 14,192 14,841

16,497 1,965 16,736

Note that the mean counts per pixel within this ROI or the counts per pixel is simply 15618.
A background ROI was selected adjacent to the sphere of interest, and the following pixel
values were measured within an ROI having the dimensions of 4×4 pixels:

22,154 23,010 24,637 25,114

23,993 24,387 24,938 24,773

27,127 26,653 26,220 25,788

27,764 27,454 27,160 26,941

21
SPECT Phantom User Manual

Note that the estimated mean counts per pixel within the selected background region
of interest is equal to 25507 counts per pixel. The sphere’s image contrast is thus equal to:

C im age 
15618  25507  = Signal cnts/Normal background cnts.
25507

 0.39 or 39% 

It has been pointed out in scientific literature that two conditions must be satisfied in
order that a lesion located within a relatively uniform background be reliably detected. First,
the signal (i.e., the numerator in the contrast equation) must be greater than 3 or 4 times the
standard deviation, that is, the value of Sdev as defined in the last section. This is at least
valid for small “resolution-sized” lesions. For larger lesion diameters this criterion may be
relaxed somewhat. Furthermore, even if this first condition is satisfied the lesion's image
contrast must still be greater than the observer’s visual contrast threshold defined as the
minimum contrast of an object of the shape of a small sphere that an observer can reliably
detect on a “screen or film”.
This latter value has been estimated to be equal to an image contrast of about 12%, or
equal to an optical density of about 0.09 when using a typical nuclear medicine film. SPECT
images of small lesions that satisfy both of the above conditions may or may not be detected.
If the SPECT system were perfect, then it would be possible to detect the small lesions.
However, if there are systematic errors such as non-uniformities, etc., then the small lesions
may not be detected.
It can finally be noted that this derivation can be extended to determine an estimate of
signal-to-noise (S/N) ratios for arbitrarily sized defects having arbitrary radionuclide uptake
ratios compared to surrounding background activity. However, extreme care must be taken
in appropriately determining the regions used for the noise power - as it will now depend, not
only on the various reconstruction parameters, but also on the size of the simulated lesions.
These calculations are beyond the scope of this manual. Needless to say, it is also imperative
that the slice selected to perform these quantitative measurements on should be precisely
centered through the diameter of the sphere, since partial volume effects will effect the
measured results.

22
 SPECT Phantom User Manual

APPENDIX III: Sample Images

23
SPECT Phantom User Manual

SPECT Phantom

RODS SPHERES

UNIFORM LINE
SOURCES

The SPECT image of the Cold Rods had an effective slice thickness
of approximately 8 cm, while the effective slice thickness for the Spheres was approximately
12 mm. The radius-of-rotation was equal to approximately 13 to 14 cm. Over 30 million
counts were used to reconstruct the image of the Cold Rods. Approximately two to three
million counts were acquired for the slice containing the Spheres. The SPECT system was
reasonably well-calibrated, although as indicated by the image of the Uniform Section, the
value used for the attenuation coefficient was slightly too great, resulting in an apparent
increase in activity towards the central region of the Phantom.

24
 SPECT Phantom User Manual

DELUXE COLD RODS

Ultra-High Resolution High Resolution Low Energy, All-

Collimator Collimator Purpose Collimator
Rel. Sensitivity: 0.24 Rel. Sensitivity: 0.64 Rel. Sensitivity: 1.0

EFFECT OF COLLIMATION
Data Spectrum's Deluxe Phantom (Model ECT/DLX/P) was used. Each SPECT image
shown has an effective slice thickness equal to approximately 8 cm. Radius-of-rotation
(ROR) was equal to approximately 13 to 14 cm. Approximately a total of 20 to 30 million
events were acquired to reconstruct each section shown. The SPECT system was
“reasonably” well-calibrated. As a result of its relatively large loss of geometric resolution
as a function of distance from the collimator surface, the Low Energy, All Purpose collimator
may not be appropriate for many SPECT imaging situations, except perhaps for very low
count density studies. Although the images obtained with the Ultra-High Resolution
Collimator are good, it has a very low relative sensitivity, as compared with the other
collimators illustrated.

25
SPECT Phantom User Manual

STANDARD COLD RODS

HIGH RESOLUTION LOW ENERGY, ALL PURPOSE
COLLIMATOR COLLIMATOR.
Approximately 8 million cnts Approximately 10 million cnts.

ROR
14cm

ROR
22cm

EFFECT OF RADIUS-OF-ROTATION

Data Spectrum’s Standard Phantom (Model ECT/STD/P)) was used. The SPECT system was
moderately well-calibrated. The fall-off with distance for both collimators are illustrated.
SPECT system (not clinical) performance evaluation is best performed with the collimator
positioned as close as possible to the phantom (i.e., using the 13 to 14 cm radius-of-rotation,
or ROR). The image quality of the Low Energy All-Purpose Collimator degrades more
rapidly as the radius-of-rotation is increased.

26
 SPECT Phantom User Manual

BENCHMARK COLD RODS

11 Million 5.5 Million 2 Million 1. Million

EFFECT OF TOTAL COUNTS IN SLICE

Data Spectrum’s Benchmark Phantom (Model ECT/BEN/P) was used. Radius-of-rotation

was equal to 14 cm. The actual scan times ranged from 4 to 22 minutes using a dual head
SPECT system. The effective slice thickness varied from 12.8 mm (for the lowest count
density image) to 25.4 mm (for the highest count density image). The numbers given are the
total number of detected events acquired and used to reconstruct each transaxial sectional
image shown (i.e., the sum of all counts for all projections for that particular slice). The
SPECT system was “moderately” well-calibrated. It is suggested that at least 8 Million
counts per slice (for the Cold Rods) be used in evaluating the SPECT system for image
quality and artifact susceptibility, although this is not, of course, neither a definite lower
value nor a definite upper limit. Even higher count density studies might be useful,
dependent on the particular application. To increase the counts acquired per slice, the pixel
slice thickness may be increased by summing several adjacent transverse images through the
rods.

27
SPECT Phantom User Manual

DELUXE COLD RODS

EFFECT OF CENTER-OF-ROTATION CALIBRATION

Data Spectrum's Deluxe Phantom (Model ECT/DLX/P) was as used. Radius-of-rotation was
equal to ~14 cm. High count density projection data were acquired with a high resolution
collimator. With the center-of-rotation properly calibrated (left image), five of the six cold
rod sectors can be visualized. To demonstrate the effect of a center-of-rotation error on
image quality, the projection data were shifted by one pixel (middle) image and two pixels
(right image). As indicated below the images, these shifts correspond to 3.2 mm and 6.4 mm
center-of-rotation calibration errors, respectively. The loss in spatial resolution is readily
observed when there is a center-of-rotation error as small as 3.2 mm (middle image). With
the larger 6.4 mm center-of-rotation error there artificially appears to be “hot rods” observed
in the reconstructed transaxial slice (right image).

28