Alessio, 1

Dose Reduction Strategies for SPECT/CT and PET/CT !

Adam Alessio, PhD, DABSNM aalessio@uw.edu

Department of Radiology University of Washington

http://faculty.washington.edu/aalessio/

DISCLOSURE: Dr. Alessio has received grant/research support from GE Healthcare

© Adam Alessio 2015, aalessio@uw.edu

Alessio, 2

Image Quality Tradeoffs in NM

0

10

20

0 10 20 30

Image Quality

Radiation Dose (a.u.)

Technique/Scanner 1

Technique/Scanner 2

New Technology ?

Diagnostic Utility Information Density

Technology Dose Savings

Scan Duration (minutes)

Alessio, 3

Image Quality Tradeoffs in NM

0

10

20

0 10 20 30

Image Quality

Radiation Dose (a.u.)

Technique/Scanner 1

Technique/Scanner 2

New Technology ?

Diagnostic Utility Information Density

Operational Dose Savings

Scan Duration (minutes)

Alessio, 4

Goal of Dose Optimization? A.  Make prettiest image possible

B.  Minimize radiation dose

C.  Maximize physician’s happiness

D.  Maximize technologist’s happiness (i.e., shortest acquisition time)

E.  Acquire with maximum image quality at minimum of dose

F.  Define a task and sufficient image quality to achieve task

Alessio, 5

Dose Optimization in Nuclear Medicine

•  Is all about –  Injected Activity?

OR

– Defining the desired task and the necessary image quality to achieve that task

– Dose Optimization = Rational Protocol Selection §  !! More than just a question of injected activity!! §  Appropriate protocol for the appropriate scanner, clinical resources,

study, and patient §  We need better approaches for rational protocol selection…

Alessio, 6

Diagnostic Reference Levels “Diagnostic reference levels (DRLs), which are a form of investigation levels, represent an important tool to optimize image quality and the radiation dose delivered to patients.” DRL’s help promote (not dictate) good practice for a more specific medical imaging task; and

•  Proposed 20 years ago. Used extensively in Europe for Quality Assurance

•  ICRP 73 (1996) •  NCRP, Report 172: Reference Levels and Achievable Doses in Medical and Dental Imaging:

Recommendations for the United States (National Council on Radiation Protection and Measurements, 2012).

Alessio, 7

Diagnostic Reference Levels •  DRLs are set at approximately the 75th

percentile of similar studies for similar patients •  Achievable doses, AD, represent the median

(50th percentile) of doses

•  ICRP 73 (1996) •  NCRP, Report 172: Reference Levels and Achievable Doses in Medical and Dental Imaging:

Recommendations for the United States (National Council on Radiation Protection and Measurements, 2012).

Alessio, 8

Diagnostic Reference Levels & Achievable Doses

Dose

•  75% of doses below Diagnostic Reference Level •  50% of doses below Achievable Dose (encourage

dose optimization for sites below 75% level)

DRL

75th % # Exams

AD

50th %

Alessio, 9

Diagnostic Reference Levels, NCRP 172 Nuclear Medicine Reference Levels

NCRP, Report 172: Reference Levels and Achievable Doses in Medical and Dental Imaging: Recommendations for the United States (National Council on Radiation Protection and Measurements, 2012).

“For nuclear medicine, the 75th percentile maximum RLs should be used as guidelines to limit unnecessary radiation dose as long as diagnostic-quality nuclear medicine studies are obtained, but not as absolute limits.”

NM Doses from a 2010 Survey of 9 Academic Centers Sample of Suggested Reference Levels

Examination DRL (mCi) AD (mCi)

Tc99m-Tetrofosmin (Stress) 39.0 25.0

Tc99m-Tetrofosmin (Rest) 29.0 18.0

99mTc-MAG3 10.0 7.5

Tc99m-MDP 32.0 23.0

F18-FDG 19.0 15.0

Alessio, 10

Diagnostic Reference Levels, NCRP 172 CT Reference Levels

NCRP, Report 172: Reference Levels and Achievable Doses in Medical and Dental Imaging: Recommendations for the United States (National Council on Radiation Protection and Measurements, 2012).

Many pages of DRL’s for CT based primarily on ACR CT Accreditation Materials

Potential Diagnostic Ref Level

Alessio, 11

SAM Question: Diagnostic Reference Levels (DRLs) can be used in clinical practice to:

A.  Provide legal justification in event of malpractice law suit

B.  Set standards to identify normal, average doses C.  Set standards to identify unusually low doses D.  Compare local practice with peer institutions and

national levels E.  Provide required protocol settings for local practice

Alessio, 13

Technology Dose Savings Current technologies providing genuine improvements:

•  Improved collimators (SPECT) •  Improved solid-angle coverage (SPECT, PET) •  Improved detectors and electronics (SPECT, PET,

and CT) •  Improved data processing (SPECT, PET, and CT)

–  Iterative image reconstruction

•  Improved review software/workstations

Alessio, 14

Collimator Efficiency •  Collimators typically absorb well over 99.95% of

all incident photons.

•  Trade-off between spatial resolution and detection efficiency (sensitivity).

•  Collimator choices: LEGP, LEHR, MEGP, High Energy Ø  balance the trade-off Ø  used for different isotopes

Alessio, 15 From: Physics in Nuclear Medicine (Cherry, Sorenson and Phelps)

Collimator SensitivityPoint Source Geometric Efficiency in Air!

Alessio, 16

Collimator Design

NEMA planar LEHR system sensitivity

(cpm/µCi @ 10cm)

§  Unique septa design enables industry-leading NEMA sensitivity* (up to 26% higher)

NEMA planar LEHR system sensitivity

(cpm/µCi @ 10cm)

168160

195

185

175

165

155

Competitor 2 Competitor 1

Regular detectors -26%

-20%

§  Thicker septa lead to more attenuation

§  Low sensitivity

195

185

175

165

155

Symbia

202

Conventional Siemens AUTOFORM

*Vendor Statement: Slide provided from Siemens Healthcare

Alessio, 17

A Benefit of Application Specific Geometries: Solid-Angle Coverage

Focused geometries can provide significantly better

solid angle coverage à Many more counts

detected at a time

x % of sphere (counts) detected

x % of sphere (counts) detected

Parallel Collimator: Same detection efficiency,

Different resolution

Alessio, 18

Detectors: NaI vs Cadmium Zinc Tellurium (CZT)

Figure from GE Healthcare, Alcyone Technology White Paper

•  Inexpensive •  Energy Resolution ~9% •  Spatial Resolution ~4mm

•  Relatively expensive •  Energy Resolution ~5% •  Spatial Resolution ~2mm •  Compact

Alessio, 19

The Reconstruction Problem: An Inverse Problem

Observed data system matrix

Unknown image

Error in observations (noise, scatter, etc)

y = Px + n

x = P−1( y − n)DIFFICULT:

Requires Iterative Solution

1.  Each Vendor can have unique representation for y, P, x, n

2.  And, how they solve P-1

Main Point: Not all “OSEM” algorithms the same

Not all Vendors Recon algorithms are the same…

Alessio, 20

Iterative Image Reconstruction in SPECT

Faster, Better Images: 1.  Garcia et al, Cardiac Dedicated Ultrafast SPECT Cameras: New Designs and Clinical

Implications. J Nucl Med, 2011; 52. 2.  Borges-Neto et al. Clinical results of a novel wide beam reconstruction method for shortening

scan time of Tc-99m cardiac SPECT perfusion studies. J Nucl Cardiol. 2007.

Increasing Applications For Quantitative SPECT: 1.  Bailey, An Evidence-Based Review of Quantitative SPECT Imaging and Potential Clinical

Applications, JNM 2013 2.  Beauregard et al, Quantitative 177Lu SPECT (QSPECT) imaging using a commercially

available SPECT/CT system, Cancer Imaging 2011. 3.  Dewaraja et al, Accurate Dosimetry in 131I Radionuclide Therapy Using Patient-Specific, 3-

Dimensional Methods for SPECT Reconstruction and Absorbed Dose Calculation, JNM 2005.

Alessio, 21

Quantitative SPECT Reconstruction

Key Components: 1.  Attenuation correction 2.  Intra-Patient Scatter correction 3.  Accurate System Model (includes

collimator-resolution modeling) 4.  Intra-Collimator Scatter correction 5.  View-dependent decay correction

Ø  ALL incorporated into reconstruction algorithm

Scintillator

PMTs

collimator!

Alessio, 22 From: The Essential Physics of Medical Imaging (Bushberg, et al)!

Collimator Resolution Dependent on source-collimator distance

Alessio, 23

From: The Essential Physics of Medical Imaging (Bushberg, et al)!

Collimator Resolution Dependent on source-collimator distance

Ø All collimators suffer from depth dependent resolution response

Ø  Iterative reconstruction methods can model, and therefore somewhat compensate for, the resolution response of the collimator

Alessio, 24

Example of Iterative Recon Trial Stress/rest Tc-99m tetrofosmin single-isotope study

Full Duration, Filtered Backprojection

Half Duration, Wide Beam Reconstruction from UltraSPECT

Borges-Neto et al, Clinical results of a novel wide beam reconstruction method for shortening scan time of Tc-99m cardiac SPECT perfusion studies. J Nucl Card, 2007.

Conclusion from this study: cardiac SPECT perfusion studies may be performed with the WBR algorithm using half of the scan time without compromising qualitative or quantitative imaging results.

Alessio, 25

SAM Question Technology dose savings can be achieved in

SPECT imaging through all the following except:

A.  Improved collimator designs B.  Higher resolution detectors made from materials

such as CzT C.  Improved data processing and reconstruction

algorithms D.  Faster rotation of detector heads

E.  Greater solid angle coverage

Alessio, 27

Trends in PET Technology •  Larger Bore Sizes (70cm towards 78cm and

more…)

•  More Reproducible Quantitation –  Better Calibration (ex: Siemen’s Quanti-QC) –  Respiratory Compensation –  Better image reconstruction (ex: GE’s Q.CLEAR)

•  Better Signal to Noise through: –  Better Time-of-Flight (ex. Philip’s Digital PET) –  Larger axial sampling –  Better image reconstruction

Alessio, 28

Digital Photon Counting PET

Converts scintillation light directly to a digital signal, with zero analog noise. Allows for Faster Timing Resolution

Vendor Statements from Philips Healthcare

Alessio, 29

Variations in resolution loss vs. size and smoothing M

ean

Max

FBP OSEM

Increasing smoothing

Alessio, 30

How to reduce partial volume effect in PET? PSF-Based Iterative Reconstruction

Detectors Scanner bore

Locations of point sources

�

sv = 2mm

�

sv = 348mm

Measured profile in black, parameterized profile in red �

s

�

s

Eve

nts

(a.u

.)

PPSF (sv ,s)

Each radial location has blur in radial direction

Improve System Model

Alessio, 31

Measured Spatially Variant System Modeling (PSF) in Iterative Reconstruction Contrast Recovery vs. Size

Images with “clinical” reconstruction parameters: 2.7mm/pixel, 7mm post-filter, 28 subsets

4 it 8 it

Contrast Recovery vs. “True” Noise across 50 scans

Proposed Method

Observations: Addition of PSF… •  Leads to roughly 7% bias improvement at

matched true noise levels across all sphere sizes

Prior Method OSEM+LOR

Proposed Method OSEM+LOR+PSF

FDG PET Exam, 109kg patient

Alessio et al, “Application and Evaluation of a Measured Spatially Variant System Model for PET Image Reconstruction,” IEEE Trans Med Imaging, 2010.

Prior “best” method

Alessio, 32

Time-Of-Flight PET •  Measures time difference of detection of photons

–  If time difference =0, annihilation at center of field of view

•  Timing resolution 500 ps = 7.5 cm

Conventional backprojection

TOF backprojection

Alessio, 33

Time-of-Flight PET

Contrast recovery coefficient versus noise for 27 cm diameter cylinder

Contrast recovery coefficient versus noise for 35cm diameter cylinder

Karp et al, Benefit of Time-of-Flight in PET: Experimental and Clinical Results, JNM 49:3, 2008.

TOF

non-TOF

Alessio, 34

TOF gain as a function of patient mass

TOF gain for matched noise levels, averaged over 6–9 lesions (1- to 2-cm diameter) for each patient, is plotted as function of patient mass. Error bars reflect the range of TOF gains seen for this patient.

Karp et al, Benefit of Time-of-Flight in PET: Experimental and Clinical Results, JNM 49:3, 2008.

Alessio, 35

Question:

Time-of-flight PET is especially beneficial for:

A.  High resolution brain imaging B.  Smaller pediatric patients C. Depth of interaction detectors D. Obese patients Answer: D. Larger patients will have more signal to noise gains than smaller objects.

Alessio, 36

CT: Instrumentation/Processing

•  CT Detectors •  Improved data/image processing

– Discussion of CT image enhancement –  Iterative image reconstruction

Alessio, 37

CT Detectors Modern Systems use Solid State Scintillation Detectors

–  Scintillation Materials: CdW04, Gd2O2S, HiLight™, GEMSStone™, CsI

–  Coupled to photodiodes

•  Flat-panel detectors usually use Cesium-Iodide (CsI) coupled to amorphous silicon photodiodes

–  Originally developed for angiography –  Used in

§  C-Arm Conebeam CT Systems §  Philips BrightviewXCT SPECT/CT

–  Cons: Low contrast resolution and Slow acquisition

–  Pros: High spatial resolution and Large area

X-ray

Visible light

scintillator

photodiode

Electrical signal

Conventional 3rd Generation CT Detector: 2-4 cm axial, ~55° arc

Flat panel detector: 35 cm x 45 cm

Alessio, 38

CT Iterative Recon Summary

•  Each vendor is offering “iterative” methods to reduce image noise, effectively allowing for reduced dose acquisitions at matched image quality

•  “Iterative” data enhancement methods can be applied at any step in imaging chain

•  Image-Based Iterative Methods: –  Philips: iDose –  Siemens: IRIS (iterative reconstruction in image space) –  Toshiba: AIDR (adaptive iterative dose reduction) –  Third Party Solutions: Clarity™ CT from Sapheneia (Sweden);

ContextVision Inc. (Sweden)

•  “More Fully” iterative reconstruction methods –  Toshiba: AIDR3D (adaptive iterative dose reduction) –  GE Healthcare: ASIR (Adaptive Statistical Iterative Recon), ASIR-V, Veo –  Philips: iDose4

–  Siemens: SAFIRE (Sinogram Affirmed Iterative Recon)

all essentially Image-Based Iterative Methods

Tomographic Image Reconstruction

Raw CT Data (projection

data) Images

Alessio, 39

Image-Based Enhancement

Marketing Brochure for Clarity CT Solutions, www.claritysolutions.org

Alessio, 40 40

FBP GE:ASIR GE:VEO

120 kVp, variable mAs (NI=36), 1.375 pitch. 0.625/0.8 mm slice: Width = 400, Level = 40 HU 65 YO female, 83.7 kg, 160 cm, BMI = 32.7

So#  Tissue  Conspicuity  Increased noise Increased conspicuity

Alessio, 41

FBP Siemens: SAFIRE Reconstruction of half dose data

Moscariello et al, Coronary CT angiography: image quality, diagnostic accuracy, and potential for radiation dose reduction using a novel iterative image reconstruction technique—comparison with traditional filtered back projection. Eur Radiol, 2011.

Coronary CT angiography

Alessio, 42

Dose Reduction Techniques

CT: Operational

•  Factors that affect radiation dose with CT •  Appropriate protocols

– Diagnostic CT vs Localization CT vs Attenuation Correction CT

Alessio, 43

Factors Affecting CT Dose

♦ X-ray beam energy (kVp)

♦ X-ray tube current (mA)

♦ Rotation or exposure time

♦ Slice thickness

♦ Object thickness

♦ Pitch or spacing

♦ Dose-reduction techniques

♦ X-ray source to isocenter distance

Direct Influence on Dose

Indirect Influence on Dose

♦ Reconstructed slice thickness image statistics require higher kVp and/or mAs in thinner slices to achieve equivalent level of noise as in thicker slices.

♦ Reconstructed image resolution algorithms enhancing spatial resolution also increase image noise- higher kVp and/or mAs may be used to compensate.

Alessio, 44 Adaped from McNitt-Gray, “Radiation Dose in CT”, Radiographics, 2002, 22:1541-1553.

CTDIw measured in 16cm head & 32 cm body phantoms Factors Affecting CT Dose

0

10

20

30

40

50

60

80 100 120 140

CTD

I_w

(mG

y)

tube voltage (kVp)

Head

Body

300mAs, 10mm

0

10

20

30

40

50

60

100 200 300 400

CTD

I_w

(mG

y)

tube current (mAs)

Head

Body

120kVp, 10mm

0

10

20

30

40

50

60

70

80

90

0 0.5 1 1.5 2 2.5

CTD

I_vo

l(mG

y)

helical pitch

Head

Body

120kVp, 300mAs

0

10

20

30

40

50

60

70

4 8 12 16 20

CTD

I_w

(mG

y)

multislice collimation (mm)

Head

Body

120kVp, 300mAs

Dose varies linearly with tube current

Dose reduces more than linearly with tube voltage

As you decrease dose, you increase noise (usually decrease image quality) – No Free Lunch

Alessio, 45

Automatic Exposure Control •  Modulate Tube Current based

on patient specific information in the –  Longitudinal (z-axis)

for example: AutoMA (GE), Z-DOM (Philips), CareDose (Siemens)

–  Angular Direction

•  On average, can achieve ~20% (3-45%) dose reductions at matched quality*

*McCollough CH, et al. CT dose reduction and dose management tools: overview of available options. Radiographics. 2006;26:503-512.

Graph of tube current superimposed on CT projection radiograph showing longitudinal modulation*

varies between anatomic regions (Fig 3). Detailsregarding implementation by several manufactur-ers are given in Table 3.

Angular-LongitudinalTube Current ModulationThe simultaneous combination of angular andlongitudinal (x-, y-, and z-axis) tube currentmodulation involves variation of the tube currentboth during gantry rotation and along the z-axisof the patient (ie, from the anteroposterior direc-tion to the lateral direction, and from the shoul-ders to the abdomen). The operator must stillindicate the desired level of image quality by oneof the methods described earlier. This is the mostcomprehensive approach to CT dose reductionbecause the x-ray dose is adjusted according tothe patient-specific attenuation in all three planes.Details regarding the implementation of this dosemodulation technique by several manufacturersare given in Table 4. A graphic illustration of thisapproach is shown in Figure 4.

Automatic Exposure ControlAEC is analogous to acquisition timing in generalradiography. The user determines the image qual-ity requirements (as regards noise or the contrast-to-noise ratio), and the CT system determines theright tube current–time product. In practice, it isrelatively straightforward for the system to deliverthe desired image quality, once that has been de-fined. However, it can be quite difficult to achieveagreement on the image quality requirement forthe various CT examination types and patient agegroups.

In defining the required image quality, the userneeds to remember that pretty pictures are notneeded for all diagnostic tasks, but, rather, achoice can be made between low noise and a lowdose, depending on the diagnostic task. The CTsystem will then adjust the tube current duringthe gantry rotation, during movement along the

z-axis, or during movement in all three dimen-sions, according to the patient’s body habitus andthe user’s image quality requirements. Thus, wedifferentiate between the modulation of the tubecurrent to achieve a defined image quality, andthe prescription of the desired image quality bythe user. Together these tasks are referred to asAEC.

Image QualitySelection Paradigms

Each manufacturer of CT systems uses a differentmethod of defining the image quality in the userinterface. However, the reference value, index, orimage can be stored with a specific protocol in allmanufacturer-implemented AEC systems.

GE Healthcare uses a concept known as thenoise index. The noise index is referenced to thestandard deviation of CT numbers within a re-gion of interest in a water phantom of a specific

Table 3Longitudinal Tube Current Modulation Systems

Characteristic GE Healthcare Philips Siemens Toshiba

Product name Auto mA . . . . . . Real E.C.Requires CT projection

radiograph Yes . . . . . . YesModulation algorithm Attenuation based . . . . . . Attenuation basedOnline feedback No . . . . . . No

Figure 3. Graph of tube current (in milliamperes)superimposed on a CT projection radiograph illustratesthe concept of longitudinal dose modulation, with vari-ation of the tube current along the z-axis. The curve isdetermined by using attenuation data from the CT pro-jection radiograph and the manufacturer-specific algo-rithm.

506 March-April 2006 RG f Volume 26 ! Number 2

Ra

dio

Gra

ph

ics

Alessio, 46

Comparison of Typical Doses Hybrid CT Acquisitions

Study   Technique   Effective Dose (mSv)  

NM Bone Scan 20 mCi Tc99m MDP (740 MBq) 4.2

High-low Myocardial Perfusion Stress:Rest Tc99m trofosmin (40:10 mCi) 13

CT for diagnostic purposes [110-200] mAs1 CTDIvol = [8-14] mGy 11-20

CT for anatomic localization [30-60] mAs3 CTDIvol = [2-4] mGy 3-6

CT for attenuation correction only  

[5-10] mAs4  CTDIvol = [0.3-1.0] mGy  

 0.5-1.0  

For ease of comparison, all CT studies performed with 120kVp, pitch 1.375, 40mm collimation, 900 mm scan range, average tube current-time product is presented

NM Dosimetry: ICRP. Radiation dose to patients from radiopharmaceuticals: (Addendum 2 to ICRP Publication 53) ICRP Publication 80 Approved by the Commission in September 1997. 1998.

CT Dosimetry: CT dosimetry tool. London: ImPACT, St. George's Healthcare NHS Trust; 2007.

Alessio, 47

Summary Dose Reduction is possible through •  Operational Dose Savings:

–  Rational Protocol Selection –  Potentially use reference levels to align with peer institutions

•  Technology Dose Savings: –  Application-specific geometries (SPECT, PET) –  Improved collimators (SPECT) –  Improved detectors (SPECT, PET, and CT) –  Improved image reconstruction (SPECT, PET, and CT)

•  Review software is critical part of realizing potential of hybrid devices

Thank You Adam Alessio

aalessio@uw.edu