Download Download SPECT/CT Basics, Technology Updates, Quality Assurance, and Applications S. Cheenu Kappadath, PhD...
SPECT/CT
Basics, Technology Updates, Quality Assurance, and Applications S. Cheenu Kappadath, PhD Department of Imaging Physics University of Texas M D Anderson Cancer Center, Houston, Texas
[email protected]
Educational Objectives 1.
2.
3.
Understand the underlying principles of SPECT/CT image acquisition, processing and reconstruction Understand current and future clinical applications of SPECT/CT imaging Familiarization with commercially-available SPECT/CT systems
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Outline
Review of SPECT principles Iterative SPECT reconstruction Hybrid SPECT/CT imaging SPECT/CT quality assurance Commercial SPECT/CT systems SPECT/CT clinical applications
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT Basics
Single Photon Emission Computed Tomography
Radio-pharmaceutical administration – injected, ingested, or inhaled Bio-distribution of pharmaceutical – uptake time Decay of radionuclide from within the patient – the source of information Gamma camera detects gamma rays and images (tomography) the radio-pharmaceutical distribution within the patient – SPECT Used for visualization of functional information based on the specific radio-pharmaceutical uptake mechanism
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT Hardware Anatomy of a Gamma camera 1. 2. 3. 4. 5.
Collimator Scintillation Detector Photomultiplier Tubes Position Circuitry Data Analysis Computer
© U of British Columbia
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT Back-Projection Model
g(s,) = f(x,y) along an in-plane line integral
© Bruyant, P. P., J Nucl Med 2002; 43:1343-1358
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Crystal Thickness
Thinner crystals spatial resolution
interactions occur at a better defined depth multiple interactions less likely less light spread interaction likelihood for higher energy ’s
Thicker crystals sensitivity
interaction likelihood (esp. for higher E ’s) likelihood of multiple interactions greater light spread spatial resolution
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Spatial Resolution
Intrinsic Spatial and Energy Resolution
# of scintillation photons, N Gamma-ray energy, E Spatial Resolution = 100 /N 1/N 1/E Energy Resolution = 100 FWHM/E 1/E
B Le H
Collimator Resolution System Resolution
S. Cheenu Kappadath, PhD
D ( Le H B ) Rg Le
Rs2 Ri2 Rg2
Le D
AAPM 2009 - Anaheim, CA
SPECT Acquisitions
SPECT acquires 2-D projections of a 3-D volume
© SPECT in the year 2000: Basic principles, JNMT 24:233, 2000 © Yale School of Medicine
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Radon transform angular symmetry violated in SPECT P() Anterior View
S. Cheenu Kappadath, PhD
≠
P(+) horizontally flipped Posterior
AAPM 2009 - Anaheim, CA
Radon transform angular symmetry violated in SPECT
Why ? Due to Differential Attenuation
L
b a i c I0 I(i+)
b
I(i)
I(i) = I0 e-a (L)dL c
I(i+) = I0 e-a (L)dL
Other mediating factors:
S. Cheenu Kappadath, PhD
distance-dependent resolution depth-dependent scatter AAPM 2009 - Anaheim, CA
SPECT Acquisitions
SPECT projections acquired over 360° Exception: Cardiac SPECT acquired over 180°
0° 0°
180°
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT images have isotropic voxel size 2-D filter of projections 3-D post-reconstruction filter No volume smoothing transverse
sagittal
coronal
Butterworth: 0.6 Nyquist, 10th order S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT Filtered BackProjection
FBP based on ideal Radon inversion formula
SPECT imaging systems are neither angularly symmetric nor shift-invariant
assumes a linear, shift-invariant system and angular symmetry of projections
SPECT projection data affected by attenuation, scatter, and spatial resolution that are all depth-or distance-dependent
Thus, FBP reconstruction cannot adequately model the physics of SPECT
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Conventional SPECT Corrections Attenuation: Chang post-processing algorithm L(x,y,i)
I(x,y) = SPECT image w/o AC I(x,y,i) = IAC(x,y).e-L(x,y,i) IAC (x,y) = I(x,y) / {(1/M).i
e-L(x,y,i)};
I(x,y,i) i
i = 1, M
IAC(x,y)
Scatter: Energy window subtraction
Lower Scatter Window
PhotoPeak Window
Upper Scatter Window STD in acrylic
20000
STD in air
Counts
P(x,y) = projections w/ scatter PLE(x,y) = projection at lower energy PHE(x,y) = projection at higher energy PSC (x,y) = P(x,y) – kL.PLE(x,y) – kH.PHE(x,y)
Energy Spectrum of Sm-153 30000
STD in acrylic with TEW Scatter Correction
10000
0 25
50
75
100
125
150
Photon Energy [keV]
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Outline
Review of SPECT principles Iterative SPECT reconstruction Hybrid SPECT/CT imaging SPECT/CT quality assurance Commercial SPECT/CT systems SPECT/CT clinical applications
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT Iterative Reconstruction Maximum Likelihood-Expectation Maximization (ML-EM) Accounts for the statistical nature of SPECT imaging Incorporates the system response p(b,d) – the probability that a photon emitted from an object voxel b is detected by projection pixel
d
voxel b
detector d
p(b,d) captures…
1. Depth-dependent resolution 2. Position-dependent scatter 3. Depth-dependent attenuation
Use a measured attenuation map along with models of scatter and camera resolution to perform a far more accurate reconstruction S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT Iterative Recon: Attenuation Modeling
b along a line integral … g(s,) = f(x,y) * pattn(x,y,s,) pattn(x,y,s,) = probability due to attenuation pattn(x,y,s,) = exp(-ab(x’,y’)x’,y’))
S. Cheenu Kappadath, PhD
a AAPM 2009 - Anaheim, CA
SPECT Iterative Recon: System Resolution Modeling Distance-dependent collimator beam ________ Rs = Ri2 + Rc2 r Pencil Beam (FBP) Intrinsic Detector Resolution Ri
- iterative) Fan Beam (2D
Cone Beam (3D iterative) S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT Iterative Recon: Resolution Modeling
2D: g(s,) = f(x,y) * pres(x,y,s,) 3D: g(s,) = f(x,y,z) * pres(x,y,z,s,) pres = probability due to resolution “fan of acceptance” (2D fan beam model) “cone of acceptance” (3D cone beam model)
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT Imaging: Scatter
Scatter compensation occurs before attenuation
the photopeak window contains scatter attenuation accounts for the removal of photopeak photons
Scatter contribution estimated as a weighted sum of one or more adjacent energy window images, Ci(x,y,) S(x,y,) = i ki × Ci(x,y,) Subtract scatter prior to reconstruction Pcorr(x,y,) P(x,y,) - S(x,y,) Incorporate scatter into forward projection P(x,y,) Pcorr(x,y,) + S(x,y,)
S. Cheenu Kappadath, PhD
SC techniques: DEW TEW ESSE
AAPM 2009 - Anaheim, CA
SPECT Iterative Reconstruction
True projection intensity = sum of true voxel intensities weighted by detection probabilities
Forward Projection
True voxel intensity = sum of true detector intensities weighted by detection probabilities
Back Projection
S. Cheenu Kappadath, PhD
B
y (d ) (b) p (b, d ) b 1
D
(b) y (d ) p (b, d ) d 1
AAPM 2009 - Anaheim, CA
Iterative Reconstruction Flow Diagram D
(b ) [k ]
[ k 1] (b )
d 1
y ( d ) p (b , d ) [k ] b '1 (b ') p (b ', d ) B
D
p (b , d ) d 1
In clinical practice, the stopping criteria is number of iterations (a time constraint) instead of a convergence criteria.
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Ordered Subset EM (OSEM)
Each OSEM iteration is a ML-EM iteration using an ordered subset of n (out of N) projections (eg: 4/36 views - 9 subsets, start with 0°,90°,180°,270° views) The next OSEM iteration starts with the result of the previous OSEM iterations but uses a different ordered subset of n projections (next set uses 10°,100°,190°,280° views) rate of convergence by using an ordered subset of all N projections for each iteration m OSEM iterations with n subsets each mn ML-EM iterations using all N each time
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
OSEM Iterative SPECT Reconstruction: Attenuation and Scatter Correction Un-Corrected
Corrected
Note the “hot-rim” artifact S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
OSEM Iterative SPECT Reconstruction: Collimator Resolution Modeling 99mTc
Bone Scan (osteosarcoma), LEHR Collimator
Standard Filtered Backprojection
2-D OSEM w/ fan beam modeling (m=12,n=10)
2-D pre-filter: Butterworth, fc = 0.6 Nyquist, order = 10
3-D OSEM w/ cone beam modeling (m=25,n=10) S. Cheenu Kappadath, PhD
3-D Gaussian Post-Filter (7.8 mm FWHM)
AAPM 2009 - Anaheim, CA
Outline
Review of SPECT principles Iterative SPECT reconstruction Hybrid SPECT/CT imaging SPECT/CT quality assurance Commercial SPECT/CT systems SPECT/CT clinical applications
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT/CT Hybrid Imaging: Why?
Non-uniform attenuation maps required
Previous methods used constant maps that work for brain but are problematic for thorax and pelvis radioactive source-based transmission CT – time penalty
Functional-anatomical overlay (image fusion)
Improve localization of uptake regions Increase confidence in interpretation
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
CT-based AC for SPECT/CT CT
CTAC
μ‐map
CT noise reduced
Smooth, re‐bin CT to match SPECT Register CT w/ SPECT
Apply bi‐linear transform on pixel‐by‐pixel basis
Reconstructed SPECT Transition Matrix
aijk
S. Cheenu Kappadath, PhD
Other factors: ‐SPECT projections ‐Scatter estimates ‐Collimator response AAPM 2009 - Anaheim, CA
CT-based values Material attenuation versus Energy Air
Muscle
Bone
Photoelectric effect Compton scatter dominant dominant
0.3
(cm2/g) 0.2
0.1 CT
0 0
100
200 Energy
S. Cheenu Kappadath, PhD
300
400
m = k ¥ CT-HU (simple but not accurate) Compton Scatter probability proportional to e- density Photoelectric effect probability proportional to (Z/E)3 Attenuation mismatch between PE and CS with energy for high Z
500
(keV) AAPM 2009 - Anaheim, CA
CT-based values - HU-to-cm-1 conversion - not linearly related - piece-wise linear - bi- or tri-modal - Effective energy differences - CT (~ 70 – 80 keV) - SPECT (nuclide dependent) eg: 140 keV for Tc-99m
CT Number-to-Tc-99m v alue Function 0.3
value (cm-1)
0.25 0.2 0.15 0.1 0.05
1000
200
0
-1000
0
CT Number (HU)
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT/CT Hybrid Imaging: Iterative Reconstruction
FBP w/ Butterworth 0.4/5
99mTc
EC-DG (NSCLC)
3-D OSEM w/ resolution modeling
3-D OSEM w/ resolution and attenuation modeling S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Outline
Review of SPECT principles Iterative SPECT reconstruction Hybrid SPECT/CT imaging SPECT/CT quality assurance Commercial SPECT/CT systems SPECT/CT clinical applications
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT/CT QA/QC
Planar (AAPM Reports 6 and 9; NEMA NU 1-1994)
SPECT (AAPM Report 22 and 52)
Inherently includes all planar gamma camera QA Energy/Spatial resolution, uniformity, deadtime, sensitivity, rotational uniformity, opposed-head registration, etc.
Uniformity and Contrast Resolution
SPECT/CT (AAPM TG 177: Jim Halama)
NM-CT registration CT-HU to linear attenuation () transformation
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
NM-CT Registration
Use Co-57 button sources w/ SPECT phantom
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
CT-HU to -map transformation
Use an electron density phantom
CIRS Inc.
CT image: -790 to 235 HU
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Outline
Review of SPECT principles Iterative SPECT reconstruction Hybrid SPECT/CT imaging SPECT/CT quality assurance Commercial SPECT/CT systems SPECT/CT clinical applications
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Commercial SPECT/CT systems Siemens SymbiaT (1-, 2-, 6, 16-slice CT) GE Hawkeye (1- or 4-slice CT)
S. Cheenu Kappadath, PhD
Philips BrightView (Flat-panel CT)
AAPM 2009 - Anaheim, CA
GE – Millennium VG Hawkeye
NM
3/8” and 1” NaI(Tl) crystals 16 simultaneous energy windows Slip-ring gantry Body-contouring based on infrared-based transmitters
CT
Co-planar, dental tube, 4-slice 20 mm beam no additional real estate needed Resolution: 3.5 or 1.75 mm (transaxial); 5 or 10 mm (axial) Time-averaged: 23 s per rotation (slow-scan) kVp: 120 – 140; mA: 1 – 2.5
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Phillips – BrightView XCT
NM
3/8” and ¾” NaI(Tl) crystals Energy-independent flood calibration (up to 300 keV) 15 simultaneous energy windows Body-contouring based on tissue impedance
CT
Co-planar, flat-panel detector, 14 cm axial FOV no additional real estate needed High-resolution: 0.33 mm isotropic voxels Time-averaged: 12 s or 24 s per rotation (slow-scan) kVp: 120; mA: 5 – 80
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Siemens - SymbiaT
NM
3/8” and 5/8” NaI(Tl) crystals Energy-independent flood calibration (up to 300 keV) 6 simultaneous energy windows Body-contouring based on infrared-based transmitters
CT
Diagnostic CT scanner kVp: 80/110/130; mA: 20 – 345 (T16) & 30 – 240 (T6) Scan time: 0.5, 0.6, 1, 1,5 s per rotation 1-, 2-, 6-, and 16-slice CT scanners
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Outline
Review of SPECT principles Iterative SPECT reconstruction Hybrid SPECT/CT imaging SPECT/CT quality assurance Commercial SPECT/CT systems SPECT/CT clinical applications
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Clinical SPECT/CT Imaging
Stress/Rest Myocardial Perfusion Imaging
Stress: 99mTc-sestaMIBI or 99mTc-Tetrafosmin Rest: 99mTc-labeled agents or 201Tl-chloride
99mTc-MDP:
bone diseases, bone metasteses 99mTc-sestaMIBI: parathyroid adenomas 99mTc-sulphur colloid: liver/spleen, lymphoscintigraphy 111In-Pentetreotide: neuroendocrine cancers 111In-ProstaScint: prostate cancer 123I/131I-MIBG: pheochromocytoma, neuroblastoma 131I-NaI: thyroid cancer
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Clinical SPECT/CT Imaging
99mTc-CEA:
colorectal cancer 99mTc-RBCs: hemangioma 99mTc-HMPAO, -ECD: brain perfusion 111In-WBC: infection 67Ga-citrate: inflammation, lymphoma 201Tl-chloride: tumor perfusion
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Clinical Benefits of SPECT/CT
Visualization, diagnosis and interpretation of primary and metastatic diseases
higher sensitivity and contrast than Planar imaging CT scan increases confidence in interpretation of SPECT examination
Surgical planning and IMRT treatment planning 90Y-microspheres radio-embolotherapy (selective internal RT or micro-brachytherapy) Internal radio-pharmaceutical therapy planning
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT/CT: Limitations
Patient motion
Contrast CT
between SPECT and CT scans respiratory and cardiac motion during SPECT acquisitions contrast introduces electron density-material mismatch map algorithms do not yet account for contrast CT
Absolute quantification (Bq/ml) not yet fully developed
radionuclide-dependent acquisition/reconstruction technique-dependent calibration techniques not yet standardized
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
SPECT/CT: Future Applications
Whole body SPECT/CT (analogous to PET/CT) Quantification of absolute activity (like PET) Compensation for CT contrast in map Compensation for respiratory, cardiac motion SPECT/CT-based 3-D dosimetry/treatment planning
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Future: Whole-body Bone SPECT/CT
Tc-99m MDP Bone Imaging
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
99m-Tc MDP SPECT/CT: Fused Coronal views
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA
Future: Multi-nuclide SPECT/CT
Maximum Intensity Projection (MIP) of a dual-isotope (Tc-99m and I-123) SPECT/CT mouse study. Published by the Molecular Imaging Center for Excellence newsletter, SNM publication Volume 2, 2008
S. Cheenu Kappadath, PhD
AAPM 2009 - Anaheim, CA