PET/CT and Fusion Issues Acknowledgements Jon A. Anderson Department of Radiology The University of Texas Southwestern Medical Center at Dallas • • • • • • • • • • • • American Associate of Physicists in Medicine 2003 Annual Meeting Why Are We Excited About PET/CT? Tammy Pritchett, RT, CNMT (UTSW) Michael Viguet, CNMT, (UTSW) Dana Mathews, MD (UTSW) Thomas Lane, PhD (UTSW) Jonathan Frey (Siemens) Jim McCann (Siemens) James Bland (CPS) Ken Halliday (CTI) Alex Ganin, PhD (GE) Jeff Simer (GE) Johann Fernando, PhD (Philips) and many others PET Fundamentals: Ideal Case Annihilation Event Positron-emitting Nucleus • Use of sequential CT and PET in the same scanner to obtain detailed anatomical and functional images yields γ2 Detector Ring e+ – “automatic” and reliable co-registration of the two images for improved diagnostic performance, – reduced acquisition time for PET study to enable better patient compliance and higher patient throughput, – lower noise in the attenuation correction step of the PET data processing – “knock your socks off” images that referring physicians can readily appreciate and use. γ1 Two Events ( ) in Detector at the Same Time* Define Line of Response (LOR) Used in Reconstruction *Same Time = within 6-12 ns (typ) PET Fundamentals: Real Case RANDOMS with false LOR γ2 suppress with small coincidence time and collimation γ1 to single-row vs multi-detector CT) 2D 3D γ2 γ1 collimators 2D γ2 γ3 scatter γ1 SCATTERS with misplaced LOR suppress with energy resolution and collimation Sensitivity TRUES with correct LOR 2D and 3D PET (Distinctly similar 3D Plane Collimators, restricted span between rings, lower sensitivity, reduced scatter No collimators, larger span between rings, high sensitivity, higher scatter 1 Corrections for Quantitative Studies (All PET is Quantitative) Attenuation Correction Factors (ACFs) − ∫ µ ds Raw Sinogram Data (T +S+ R) P1 = e γ1 Remove Randoms γ2 Normalize Detector Responses − P2 = e Correct for Deadtime s2 Correct for Attenuation T Sinogram Ready for Reconstruction How Big is the Attenuation Correction Factor (ACF)? d = 23 cm → ACF ≈ 9 Attenuation Correction Factor in Soft Tissue Attenuation Correction Factor 140 120 100 80 60 40 d = 35 cm → ACF ≈ 28 20 0 0 10 20 30 40 50 ∫ µ ds s2 Pcoincidence = P1 • P2 s1 Correct for Scatter s1 60 Path Length [cm] Basic Arrangement of a PET/CT Siemens Biograph (#6) HR+ PET (BGO) Somatom Emotion CT (single slice) − Probability of both photons escaping depends on the integral of the absorption coefficient along the entire LOR -- exactly the data from transmission CT! =e ∫ µ ds s1+ s2 ACF is inverse of this probability. In A Sense, We Have Had PET/CT For A Long Time, but …. Corrections to PET scans for attenuation and possibly scatter require an attenuation map (µ-map, AKA CT image) at 511 keV obtained by measuring transmission of 511 or 662 keV photons from isotopic sources. - obtained sequentially to emission scan - acquired on same time scale (requires about 1/3 of total imaging time, say 3 minutes out of 8 minute scan) PET µ-map images are CT scans that - are made at 511 keV - have low spatial resolution - visualize bone, tissue, and contrast differently than x-ray CT Current Implementations (August 2003) CPS (marketed by CTI and Siemens) Reveal RT/Reveal XVI Biograph LSO/Biograph Sensation 16 GE Discovery LS Discovery ST PET CT apologies to any vendor that was inadvertently omitted! Philips GEMINI 2 Available Technology Manufacturer CPS (CTI and Siemens) GE Philips Product CT Reveal RT Biograph LSO Siemens Reveal XVI Biograph Sensation 16 Siemens Somatom Emotion (2 Slice) Crystal Technology PET Detectors Reveal ECAT ACCEL BGO LSO GSO bismuth germanate lutetium oxyorthosilicate:Ce gadolinium oxyorthosilicate:Ce Decay Time ns 300 40 60 Attenuation cm 1.05 Length 1.16 1.43 Material LSO Somatom Sensation (16 Slice) Discovery LS GE Lightspeed (4, 8, 16 slice) Advance NXI BGO Discovery ST GE Lightspeed (4, 8 slice)* Discovery ST GEMINI Philips MX8000 Allegro (2 slice) GSO Relative Light Out % 15% 75% 25% Energy Resolution % 10.2 10 8.5 (single crystal) DS Bailey et al in Valk et al Positron Emission Tomography (Springer, 2003) *16 expected by end of 2003 PET Configurations Product Crystal Size and (#) Axial Field of View mm Transverse Field of View PET Performance # Image Planes Axial Sampling Ring and (Port) Diameter mm cm 3.4 82.7 (70) Product Sensitivity (Trues) Mode Axial Resolution Axis/10 cm 27** 3D kcps/kBq/mL Transverse Resolution Axis/10 cm Energy Resolution PET Xmis’n mm mm % 5.8/7.1* 6.3/7.4* 25 (recently announced 15 w new electronics) No cm cm 6.45 x 6.45 Reveal RT x 25 Biograph LSO (9216) Reveal XVI Biograph Sensation 16 16.2 58.5 Discovery LS 4 x 8 x 30 (12096) 15.2 55 35 4.25 92.7 (70/59)* Discovery LS 5.4** 31** 2D 3D 4.8/5.4** 6.0/6.3** 4.8/5.4** 4.8/5.4** 20 Yes Ge-68 Discovery ST 6.3 x 6.3 x30 (10080) 15.7 70 47 3.27 88.6 (70) Discovery ST 8.1** 35** 2D 3D 6.2/6.8** 6.2/6.8** 6.2/6.8** 6.2/6.8** 17 No 4 x 6 x 20 (17864) 18 90 2 90 (70/63)* 25** 3D 5.0/6.1* 4.9/5.5* 16 Yes Cs-137 GEMINI 47 Reveal XVI Biograph Sensation 16 * CT port/ PET port CT Configuration Product Generator Anode Heat Rating Minimum Slice Thickness Field of View kW MHU mm cm Reveal RT Biograph LSO 40 3.5 1 50 Reveal XVI Biograph Sensation 16 60 5.3 1 50 Discovery LS 53.2 6.3 .625 50 Discovery ST 53.2 6.3 .625 50 60 6.5 0.5 50 GEMINI Reveal RT Biograph LSO GEMINI (t+s), 410 LLD * NEMA 2001 ** NEMA 1994 Noise Equivalent Count Rate: NECR Noise equivalent count rate (NECR) is used to compare the effective count rate performance of systems and to determine the optimal operating regime of particular combinations of hardware and data collection strategy • It represents the count rate in a perfect system (no randoms or scatters) that would have the same SNR as the measured count rate in the system • Under certain T2 assumptions, can NEC = be calculated as: (T + S + nfR) n = 2 for direct subtraction of randoms and 1 if variation reduction scheme is applied to randoms; f = fraction of field of view filled by phantom 3 Bone Differs From Other Tissues When Changing µ70keV →µ511keV Specific Issues for PET/CT • Derivation of proper attenuation corrections from CT image • Recognition of PET artifacts due to – Attenuation correction model failures – Patient motion – CT artifacts Mass Attenuation Coefficients for Principal Tissue Types 10.00 Soft Tissue Total Soft Tissue PE Adipose Tissue Total Bone Total Bone PE 2 µ/ρ [cm /g] Adipose Tissue PE CT Energy 70 keV 1.00 • QA and test procedures • Design features of PET/CT facilities PET Energy 511 keV 0.10 Photoelectric Component of Attenuation Coefficient Compton Component Accounts for Almost All of Attenuation Coefficient 0.01 10 100 1000 Energy [keV] Conclusions Regarding the Relation of µ70keV to µ511keV Initial Scheme for Converting µ70keV →µ511keV 1) The attenuation coefficient at 511 keV can be obtained from that at 70 keV by simple multiplication by a constant, independent of tissue type, for all tissues having attenuation dominated by Compton scatter at CT energies (≈ 70 keV). 2) For materials having substantial contribution from photoelectric absorption, the scaling factor will be dependent on the nature of the material. 1) Adjust resolution in CT to agree with resolution of PET 2) Convert CT numbers to µ70 keV, µ = ((CT/1000)+1)*µH2O,70 keV 3) Scale all µ values corresponding to CT values below ≈ 200 − 300 by a factor of µH2O,511 keV/µH2O,70 keV 4) Scale all µ values corresponding to CT values above ≈ 200 − 300 by a factor of µbone,511 keV/µbone,70 keV CT + 1024 Histogram of CT image, showing separation of tissue types [Kinahan et al.] Used in Siemens/CTI Scanners “hybrid segmentation” Kinahan et al.,Med Phys 25, 2046-2053 (1998) A Second Algorithm for Converting µ70keV →µ511keV Comparison of Attenuation Correction Methods 0.20 µPET= [((CT/1000)+1)*µH2O,80 keV]* µH2O,511 keV µH2O,80 keV Equivalent to formula in Kinahan for CT# ≤ 0 HU CT# > 0 HU: µbone,511 keV- µH2O,511 keV µPET= µH2O,511 keV + (CT/1000)*µH2O,80 keV* µ bone,80 keV- µH2O,80 keV Tissue is treated as a mixture of air and water below 0 HU and a mixture of water and bone above 0 HU. “GE Approach” Burger et al. EJNM 29 922-927(2002) Attenuation Coefficient at 511 keV 0.18 CT# ≤ 0 HU: Exact forms depend on choice of parameters; for this plot Hybrid Segmentation Approach Mixture Approach 0.16 0.14 Hybrid Segmentation 0.12 0.10 threshold=200 HU PET/CT<200 = 0.5 PET/CT>200 = 0.41 0.08 0.06 Mixture 0.04 0.02 0.00 -1000 -500 0 CT Number (HU) 500 1000 threshold = 0 HU µH2O,80 keV = 0.184 cm-1 µbone,80 keV = 0.428 cm-1 µH2O,511 = 0.096 cm-1 µbone,511 keV = 0.172 cm-1 4 Principal Artifacts Specific to PET/CT Failure of the Attenuation Correction Model • Artifacts due to fundamental failure of the attenuation correction model • Artifacts due to patient motion between the CT and PET scans • Truncation artifacts due to patient extending outside the CT field of view The Origin of Contrast and Metal Artifacts in PET/CT • There is no way to intrinsically deconvolve the measured attenuation coefficients µ into µ µ = ∑ ∗ ρ i i ρ i so we don’t know what we’ve got or how much, only what the total attenuation is at CT energies • Materials not in the model will scale differently and give the wrong attenuation correction • Incorrect attenuation values give incorrect activities, too high or too low Artifacts from Oral Contrast CT AC Mass Attenuation Coefficients for Tissue and Contrast Materials 10.00 Soft Tissue Bone Barium Sulfate Sodium Iodide CT Energy 70 keV NonAC 2 µ/ρ [cm /g] 1.00 PET Energy 511 keV Bolus of contrast in bowel generates apparent area of high uptake on attenuationcorrected image 0.10 The mass attenuation coefficients for contrast materials are significantly larger than those for tissues (including bone) due to the large photoelectric component; the scaling to 511 keV is different and the attenuation correction will be overestimated! 0.01 10 100 1000 Energy [keV] Artifacts from Hardware: Mediports Non-attenuationcorrected image reveals little contrast wrt surrounding tissues Artifacts from Metal: Orthopedic Hardware Intense activity shown on PET/CT (SUV = 6) is associated with metallic hardware having CT# > 3000HU Hot spot on PET Correlates to Mediport placement No anomalous uptake in nonattenuation corrected image 5 Artifacts from Metal: Orthopedic Hardware, continued Artifacts from Patient Motion Artifacts arise because the CT and PET scans are taken - at different times (i.e. sequentially) - on different time scales (seconds for CT, minutes for PET) Voluntary movements: Patient shifts position between CT scan and PET scan; principally movements of head and arms Attenuation Corrected PET Non-attenuation Corrected PET Breathing Artifact Involuntary movements: CT catches snapshot of patient position, but PET averages over minutes - CT is not normally done with breath hold (contrary to normal CT practice), but with shallow breathing - PET artifact not seen on conventional PET because both emission and transmission averaged over similar periods Quality Assurance Program: • QA for PET -- Breathing during helical CT acquisition can lead to “floating liver” artifact (variously called banana or mushroom artifact) which can then be reflected in PET scan. – Daily check per mfg with line or volume source to assure that system performance is not drifting – Period Physics check per standard PET practice • QA for CT -– Daily air cals and image quality check per mfg – Period Physics check per standard CT practice • Registration check -- check for proper registration of PET and CT images on periodic basis Examples: QA Procedures CPS (Siemens-CTI): Obtained with volume phantom (1.5 mCi 68Ge); summary report presented and logged; sinogram viewer for details. Gantry Offset Correction/Check: Phantom Geometry GE: Obtained with rotating line source in gantry. Report generated with fan-sum views and comparisons to previous trends. 6 Registration Test Stability of Registration: An Example The reproducibility of the acquisition process was better than σ = 0.1 mm (registration repeated without removing the dual line phantom) Registration Stability 1 0.5 CT FUSION PET In our case, Siemens/CT Gantry Offset Procedure could be used to numerically evaluate x (horizontal) y(vertical) and z(axial) registration without modifying machine calibration. Results can also be evaluated visually or with other programs Examples of PET Testing Tools Provided for PET/CT Machines • CPS (Siemens, CTI) – Uniformity and Sensitivity test software shipped with machine, based on volume 68Ge source used for daily QC • GE – Will provide physicist with testing software (NEMA NU-2 2001) Correction [mm] 0 Z-900 mm -1 -1.5 X -2 Y -2.5 -3 2/23 3/5 3/15 3/25 4/4 Workflow at the PET Center (FDG Whole Body Scans) (You may get different answers from different folks!) NEMA NU-2 2001 Spatial resolution Scatter fraction, count losses, randoms Sensitivity {now part of scatter fraction} NEMA NU-2 1994 Spatial resolution (xverse,axial) Scatter fraction Sensitivity Deadtime,count-rate losses Uniformity Scatter correction accuracy Count-rate correction accuracy Attenuation correction accuracy {now part of image quality test} Count rate correction accuracy {now part of image quality test} Image quality (scatter, AC accuracy) 94 Scatter fraction test 94 Count rate tests Pt instruction and prep 30-60 min Injection of Pt Workload Estimation PET Facility Throughput Example: 1 Hour Uptake, 30 Minute Scan Assay of dose Uptake of pharmaceutical Have Pt empty bladder Transport Pt to scanner 10 min Position Pt 5-30 min Scan Release Pt Many of these tests were developed to compare classes of machines rather than to provide practical acceptance test procedures. NU2-94 may be more appropriate for brain scanners and NU2-01 for WB oncology scans. Receive doses * * * * steps with highest technologist exposure QA Check of Scan 4/24 The stability of the system over the first 45 days was characterized by σ = 0.5 mm or better for all 3 axes. What Tests are Important? 15 Patient Number Arrival of patient 4/14 Date • Philips – Will ship NEMA NU-2 2001 software with machine The reproducibility of the registration procedure (multiple replacement of the phantom) was better than σ = 0.4 mm. -0.5 13 11 9 7 Phase 5 Uptake Scanning 3 1 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 Time of Day Read study Print for file and referring; distribute to PACS #pts/day = (Twork - Tuptake)/Tscan_rm # uptake rooms = Tuptake/Tscan_rm 7 Effective Dose Comparison: Conventional PET vs PET/CT Whole Body Scan, HR+ in 2D mode 12 mCi 18F FDG 1.3 rem transmission scan (average exposure of 56 mR on axis, free air, 6 bed scan, 8 min/bed,) Whole Body Scan, Biograph 10 mCi 18F FDG 1.1 rem WB CT scan (130 kVp, 120 mAs, 5 mm, p = 1.5) 1.3 rem Total 2.4 rem The End 8