GE Medical Systems Growth through Innovation

1
PET/CT Imaging and Fusion
Technology ACMP June 14th 2004
Phil Vernon PET/CT Research Manager GE Healthcare
2
Why PET/CT ?
Why PET ? Conventional Imaging - Shows only
major structural changes - Relatively poor at
finding neoplasia - Very poor at characterizing
tissue - Slow at showing metabolic change in
response to therapy FDG PET Imaging -
Lights Up cancer - Demonstrates metabolic
change in Rx
3
Why PET/CT ?
Why PET/CT ? But PET Imaging - Shows limited
anatomy - Has poor spatial resolution - Takes a
long time (relative to CT/X-ray) - Is not widely
understood by physicians surgeons PET/CT
Imaging - Adds anatomy to the PET scan -
Reduces Imaging time - Improves acceptability by
referring physicians
4
SENSITIVITY AND SPECIFICITY OF PET AND CT IN
CANCERS

• Sensitivity Specificity Change in
  Patient (Detect cancer?) ( Detect
  only cancer?) Management
• PET CT PET CT Due to PET
• Lung
• Diagnosis 98 67 73 -
• Staging 82 61 91 76 37
• Recurrence 98 72 92 95
• Colorectal
• Recurrence 93 77 85 68 31
• Lymphoma
• Staging 89 80 93 73 21
• Recurrence 83 95 10
• Melanoma
• Staging 84 88 91 75 26

5
Classification
Classification Classification
Correct Correct but
Incorrect
Equivocal TUMOR STAGE CT alone
23 (58) 8 (20) 9 (22) PET
alone 16 (40) 16 (40) 8 (20) Visual
Correlation 26 (65) 5 (12) 9 (22)
of PET and CT Integrated PET/CT 35 (88)
4 (10) 1 ( 2) NODE STAGE CT alone
22 (59) 2 ( 5) 13 (35) PET
alone 18 (49) 14 (38) 5 (14) Visual
Correlation 22 (59) 4 (11) 11 (30)
of PET and CT Integrated PET/CT 30 (81)
1 ( 3) 6 (16)
Lardinois,D et al. Staging of Non-small-cell
Lung Cancer with Intergrate PET/CT and CT. N
Engl J Med 32825 2003
6
. . .integrated PET-CT is superior to PET
alone, CT alone, or visual correlation of PET and
CT in determining the stage of disease in
non-small-cell lung cancer.
Since PET images have a fairly high resolution,
lesions that are less than 1 cm can be detected.
This is a critical advantage over conventional CT
and magnetic resonance imaging However tracing
focal abnormalities to specific lymph nodes is
difficult, or even impossible with the use of PET
alone. Several studies has demonstrated that
fusion of images of the trunk obtained by CT and
PET from different scanners is technically
possible.26-28 However, this approach did not
increase the accuracy of mediastinal staging over
that obtained by PET alone.
Lardinois,D et al. Staging of Non-small-cell
Lung Cancer with Intergrate PET/CT and CT. N
Engl J Med 32825 2003 26. Wahl,R et al.
Staging of mediastinal non-small-cell lung cancer
with FDG PET, CT and Fustion Images. Radiology
1994 191371-377 27. Vansteenkiste JF, et al.
FDG-PET in potentially operable non-small cell
lung cancer do Fusion images improve the
localisation of regional lymph node metastases?
Eur J Nucl Med 1998 251495-1501. 28. Magnani
P, Carretta A, Rizzo G et al. FDG PET and spiral
CT image fusion for mediastinal lymph node
assessment of non-small cell lung cancer patients
J Cardiovasc Surg 1999 40741-8
7
If PET/CT is Good What is the best technology
? What is it good for ?
8
If PET/CT is Good What is the best technology ?
How important are Resolution, Sensitivity, Scatte
r Fraction, Countrate Performance?
(Physical Performance defined by NEMA
standards) What about Crystal Technology ? Do
we need 64-slice CT ? What really matters ?
9
If PET/CT is Good What is PET/CT Good for ?
Staging (we have lots of good references) What
about - Screening for Cancer ? - Cardiology
? - Neurology ? - Radiation Treatment Planning
? - Evaluating Cancer Treatment ? Where does
PET/CT give the greatest clinical advantage ?
10
What is the best technology ?
How important are Resolution, Sensitivity, Scatte
r Fraction, Countrate Performance?
(Physical Performance defined by NEMA standards
for PET) Resolution - How big does a 1 mm
source appear to be. Limits ability to see
edges. Impacts ability to detect, and define
the shape of, small objects. Sensitivity -
How much information is collected per unit time
from a given quantity of tracer. Primary
determinant of signal-to- noise level. Principal
limitation on ability to locate small foci.
Scatter Fraction - What fraction of the data
collected is scatter (bad data). Degrades
signal-to-noise. Count Rate Performance As
the tracer dose is increased, how much more
information is collected.
11
What is the best technology ?
Resolution - Average resolution 5.5mm to 7.5mm
FWHM Newer systems have poorer resolution
than older designs Sensitivity - Varies
from about 3.5 to 9.2 cps/kBq without septa
(3D) Very large variation from one design to
another Scatter Fraction From about 37 to 50
without septa (3D), 16 with septa (2D) Count
Rate Performance Most meaningful measurement
is a graph of NECR vs tracer concentration.
NECR (Noise Equivalent) Countrate) is total
countrate corrected for scatter and random
events (both are noise), but also penalized for
the noise added by the correction.
NEMA NU2, 2001 (based on whole-body phantom)
(NU2 1994 -based on brain phantom - will appear
on some following slides because 2001 numbers are
not available from all vendors)
12
Spatial Resolution
Reveal Reveal Discovery Discovery
Gemini5 Hi-Rez1 RT2 ST3
LS4 Spatial Resolution _at_ 1
cm 4.6 6.3 6.1
4.8 4.9 _at_
10 cm 5.8 7.4 6.9
5.1 5.5
_at_ 20 cm 6.8 8.6
7.6 6.2
NEMA Resolution is NOT How small an object can
I see? but is
How big does a 1 mm object appear?
1 CTI Brochures distributed at RSNA 2 Reilly
Commnunications (www.reillycomm.com)2004 3
Mawlawi, O, et al. JNM June 2003, and in press 4
DeGrado, JNM 35, 8 1994 5 Reilly
Commnunications (www.reillycomm.com)2004
13
Sensitivity Amount of Information per milliCurie
of Tracer Injected
3D Sensitivity 2D Sensitivity Discovery
ST BGO 1280 9.1 300 2.1 Discovery LS
BGO 1060 6.5 220 1.3 Mfg A LSO
780 ? Mfg A LSO 800 4.5 Mfg B
GSO 660 ?
2001 cps/kBq
1994 kcps/uCi/ml
2001 cps/kBq
1994 kcps/uCi/ml
14
Scatter Fraction Percentage of collected data
that is Invalid (noise due to scatter)
3D SF 2D SF Discovery LS 30 10
Discovery ST 29 16 Mfg A
LSO 35 Mfg A LSO 30 Mfg B
GSO 30
Nema 1994
Nema 1994
Scatter from a 20 cm x 20 cm phantom
15
Discovery ST
NECR kcps
100000
DST 2D
Practical Clinical Range for Onocolgy
80000
60000
40000
DST 3D
Mawlawi et al (J Nuc Med, June 2003 and in press)
20000
70Kg patient (154 lb) 15mCi injection 45 min
uptake
kBq/cc
0
80
50
60
70
40
10
20
30
16
NECR Curves (Noise Equivalent Count Rate)
NECR kcps
100000
DST 2D
Mawlawi et al (J Nuc Med, June 2003 and in press)
80000
T2 TS2R
NECR
60000
T2 TSR
NECR
40000
NECR 92kcps _at_ 49 kBq/cc
20000
kBq/cc
0
80
50
60
70
40
10
20
30
17
COUNTS Determine Image Quality

• Noise in PET images is dominated by the counting
  statistics of the coincidence events detected.
• Noise in images can be reduced by filtering, but
  at the cost of resolution.

18
What is the Best way to Acquire Data ?
19
2D Acquisition
All events used to reconstruct a slice are
detected within that slice.
20
2D Acquisition
Events that cross several slices are rejected.
21
3D Acquisition
Events which cross several transaxial slices are
collected. These events are used in the
reconstruction of each of the slices crossed.
22
But the world is full Randoms and Scatter.
23
But the world is full Randoms and Scatter. Both
Scatter and Randoms can be significantly reduced
by Septa.
24
But Septa also remove those rays that Cross
several slices, and are in used in 3D
imaging. SEPTA can only be used in 2D imaging
25
3D Compared to 2D
SENSITIVITY 3D Better (more events) RESOLUTION
No Difference SCATTER FRACTION 3D Worse (no
septa) RANDOMS FRACTION 3D Worse (no
septa) COUNTING RATE 3D Worse (more
randoms scatter)
26
Patient 5 2D
"Courtesy of The University of Texas, M.D.
Anderson Cancer Center, PET Facility, Division of
Diagnostic Imaging"
Weight - 50 kg 110 lb BMI - 20.0 Activity at
Acq Start 11.0 mCi 3 min/ FOV 6 FOVs
27
Patient 5 3D
"Courtesy of The University of Texas, M.D.
Anderson Cancer Center, PET Facility, Division of
Diagnostic Imaging"
Weight - 50 kg 110 lb BMI - 20.0 Activity at
Acq Start 9.7 mCi 3 min/ FOV 6 FOVs
28
What about Crystals ?
The primary detector is a scintillation crystal.
It must stop the 511 keV gamma rays and convert
the energy to light. The light is detected by
photo- multiplier tubes.
PMT
29
(No Transcript)
30
Sensitivity
DST BGO
OEM 1 LSO
OEM 2 GSO

7 of photons escape
11 of photons escape
24 of photons escape
20 mm
30 mm
20 mm
93 of photons stopped
76 of photons stopped
89 of photons stopped
30mm of BGO stops 93 of incident photons Rel
Sensitivity 1.00
25mm of LSO stops 88 of incident photons Rel
Sensitivity 0.79
20mm of GSO stops 76 of incident photons Rel
Sensitivity 0.67
30mm of BGO provides the highest sensitivity
31
Detector Scintillator Materials TheoryBench
Performance of Single Crystals in Ideal Conditions
GSO BGO LSO
Rel. Light Output 35 22 75
75-80 More Light means we can improve
energy resolution (reduce scatter)
connect more crystals/PMT (cut cost) Decay
Constant (ns) 60 300 40 40
Short (Fast) means we can reduce the
number of randoms accept a higher
count rate per crystal
m(cm-1) 0.67 0.95
.85 0.80 Photo fraction
28 40 35 35 High
stopping power and photo fraction
means higher sensitivity (more counts)
32
Theory and Practice of PET Scanners
In theory, there is no difference between
theory and practice but in practice, there
is.
Yogi Berra
33
Detector Scintillator Materials
PracticePerformance of Clinical Scanners
Engineering Design
GSO1 BGO2 LSO3 LYSO
Rel. Light Output 35 22 75
75-80 Energy ResolutionA 16 17
25 16 Scatter FractionB
30 29 35 30 Decay
Constant (ns) 65 300 40 Random
Rate (rel) 66 100 50
40 Peak NECRC (kcps) 48
62 58
m(cm-1) 0.67 0.95 .85
.80 Photo fraction 25 40 35
35 SensitivityD (kcps/uCi/ml) 700
1280 780 1100
Not All Scintillators Deliver on Their Promise
1 Philips Allegro 2 GE Discovery ST 3
Siemens Biograph LSO
A Manufacturers specifications
www.reillycomm.com B NEMA NU2 - 1994 from
www.reillycomm.com C NEMA NU2 - 2001 from Mfrs
data sheets D NEMA MU2 - 1994 from
www.reillycomm.com
34
With and Without Septa for 60 kg (130 lb)
PatientMayo Clinic - Abdomen/Liver
FDG
Geometry
Final Image
Crystal Depth
LSO/PICO 63 100 70 NECR 17
No Septa 3D
Trues 126K Random 314K Scatter 83K
15mCi (555 MBq), scanned 1hour 18 min post
injection 8.7mCi at Acq. time
523 K counts/sec
Trues 24 RS 76 NECR 30.4 kcps
Crystal Depth 30mm
Septa - 2D
Trues 39.3K Random 4.7K Scatter 6.0K
15mCi injected, scanned 51 min post
injection 10mCi at Acq. time
50 K counts/sec
Trues 79 RS 21 NECR 30.9 kcps
Crystal Depth 30mm
35
1895 The first Car uses new Piston Engine
36
1954 - Gas Turbine Engine Replaces Piston
Engine in Cars
"By 1965 the deluxe open-road car will probably
be 20 feet long, and powered by a gas turbine
engine -Leo Cherne, 1955.
37
1979 - Wankel Rotary Engine Replaces Piston
Engine in Cars
1979
1986
"The Wankel engine will. . .dwarf such major
post-war technological developments as
xerography, the Polaroid camera and color
television." -General Motors, 1969.
1993
38
2004 - Piston Engine Replaces new competitors
in 99.99 of All Cars
.01 sum of Electric, hybrid, and Rotary
39
Forget the Headlines ! Its not about the
Crystals
Do you buy a car by the engine name ? . . . . or
by performance and capability ?
Did you know the Hemi isnt a hemi ?
40
What about CT ?
How Much CT do I need in a PET/CT ? Largely
doing whole-body Imaging, so the CT must be
reasonably fast (but the PET takes 20
min!) Thin slices useful if doing RT planning,
or if using CT CAD to map lung
nodules Multi-Slice important for gating (heart
and respiration).
41
Thin Slices Drive DRR Quality
42
2D, 3D -gt 4D Imaging
4D provides Respiratory gating - Provides more
precise tumor contouring for Radiation Therapy
Planning
43
Our PET is growing up