AD Dual-Energy al Energ CT Based on A Double Do ble Layer La er Detector * All clinical images are courtesy of Hadassah Medical Center, The Hebrew University, Jerusalem Ami Altman, Ph.D., and Raz Carmi Ph.D., CT BU, PHILIPS Healthcare Content 1. The double-layer detector principle and operation 2 Advantages 2. Ad t and dd drawbacks b k off the th Double-Layer D bl L approach h 3. Material Decomposition method 4 The 4. Th Double-Layer D bl L Energy E spectra t compared d to t 80/140 kVp kV spectra t 5. The Effect of large noise in the Low_E image on material decomposition 6. “Spectral resolving power” and simulated results to compare various Dual-Energy CT method 7 D 7. Decomposing i specific ifi materials t i l from f mixtures, i t and d quantitative tit ti Iodine maps 8. Clinical applications and results Ami Altman & Raz Carmi Philips Healthcare 2 The Double-Layer Detector, Principle and Operation A 0.150-mm side-looking photodiode array y a 1.0-mm Tungsten g layer y shielded by X-Rays Coming from top 0.030 mm optical glue Top Scintillator, Inter-Layer Filter ~50% 50% Low Energy Raw data 1.0 mm 0.080 mm reflecting paint Bottom Scintillator: Y 2-mm GOS ~50% 50% E1 image + High Energy Raw data E2 image X = Weighted combined Raw data 1. 2. 3. 4. CT image For optimal performance the effective atomic number of the top scintillator is small without sacrificing light output (better than GOS) Top Scintillator thickness has been optimized for best energy separation and low-energy image noise The thin filter material and thickness has been optimized to attenuate < 3% of the intensity entering the detector, and yet, significantly increase the energy separation. Bottom scintillator is GOS, the thickness of which set to absorb 99.5% of the High-Energy spectrum (note that light collection is sideways) Ami Altman & Raz Carmi Philips Healthcare 3 Advantages and Drawbacks of The Double Layer Approach Main Advantages 1 1. Si lt Simultaneous and d equi-directional i di ti l sampling li off the th scanned d body b d in i the th 2 energy bands b d 2. Enables both projection-based and attenuation-space (image-space) material decomposition 3. The high g energy gy tail at the Low-Energy gy Spectrum, p , enables low-noise at the Low-Energy gy images g even for large patients. This has a significant advantage in the material spectral decomposition, compensating for the larger overlap between the two spectra (see next slides) 4. Enables a single-source dual energy CT with unlimited FOV for both axial and spiral scans, at all protocols. protocols 5. Can work in a conventional CT mode by multiplexing (analog MUX) the two layers at each detection pixel 6. Very simple side-looking photodiode arrays that enable any expansion of the detector array at all directions 7. Work at normal CT dose, with a potential for significant dose reduction (high light output of front scintillator) Drawbacks: 1. Energy overlap is larger than scanning with two kVp values due to the High-Energy tail at the Low-Energy Spectrum (80/140). However, this has also an advantage, as explained in (2) above 2. Requires more readout channels, and one more layer of scintillators, adding to DAS cost (partially compensated by simpler and inexpensive photodiode arrays) Ami Altman & Raz Carmi Philips Healthcare 4 Material Analysis Method With Dual Energy Spectral CT, Attenuation Space NOTE that the attenuation coefficient (at CT energy range) is linear with the density (concentration) for a specific effective atomic number and energy (away from K K-Edge) Edge) μ Z ,E Z Z3 ≅ A⋅ ⋅ ρ + B ⋅ 3 ⋅ ρ E E 1. On a µ-Space map, each material, characterized by its “effective atomic number”, is represented along a straight line, the angle of which depends on its “effective Z” for a given energy set 2. Angular difference between the representing lines of two specific materials with given atomic number depends on the mean-energy difference between the two spectra 3. The statistical “line width”, namely, the distribution of points along it, depends on the separate spectra image noise, noise as well as on the overlap between the two spectra. spectra Basically, an “effective atomic number” spectrometer Iodine Conventional CT Image zoom µ_E_L Low (HU) μE-Low (HU) Zeff_1 1 > Zeff Zeff_2 2 > Zeff_3 Zeff 3 Zeff_1 Zeff_2 Zeff_3 Calcium water Water (E_low & E_high = 0 HU) μE-High (HU) µ_E_High (HU) A phantom with different concentrations of Calcium and Iodine contrast agent Ami Altman & Raz Carmi Philips Healthcare 5 Double-Layer Detector - Energy Spectra With / Without 35-cm Water Absorber Energy Windows Obtained In A Double‐Layer CT Detector From a 140 kVp X Ray Tube Air Only From a 140 kVp X‐Ray Tube, Air Only 900000000 Δ<E>=31 keV 800000000 700000000 The drawback becomes an advantage: dN/d dE 600000000 500000000 400000000 300000000 <E_Low >= 63 keV 200000000 <E_High >= 94 keV 100000000 0 0 20 40 60 80 100 120 140 X‐Ray Energy (keV) Spectrum_Low_E Spectrum_High_E 2. Compare with 600 mAs 80 kVp scans on adults, adults where images are very noisy, reducing significantly tissue & material separation Low and High Energy Spectra 35 cm Water Absorber Low and High Energy Spectra, 35 cm Water Absorber 3500000 3000000 Δ<E>=26 keV 2500000 dN/dE 1. The high-energy tail in the Low_E spectrum, enables good IQ (low-noise), (low noise) even for large patients. 2000000 1500000 <High_E>=101 keV 1000000 <Low_E>=75 keV 500000 0 0 20 40 60 80 100 120 140 X‐Ray Energy (keV) Spectrum_E_low Spectrum_E_high Ami Altman & Raz Carmi Philips Healthcare 6 Compare With Dual kVp, 80 VS. 140 kVp – Spectral Difference (No extra filter on 140 kVp beam) 140 VS. 80 kVp Spectra in Air The CT image noise, for the same mAs, obtained in the 80 kVp image, with 35-cm water cylinder, is 9 times larger than that of the 140 kVp image!! 40000000 dN/dE (# #/keV) 35000000 30000000 Δ<E>=18 keV 25000000 20000000 <E_low>=53 keV 15000000 10000000 <E_high>=71 keV 5000000 0 0 20 40 60 80 100 120 140 X-Ray Energy (keV) 140kVp_Specrum 80kVp_Specrum 80 VS 140 kVp Spectra - 36 36-cm cm Water ALSO: A reasonably-seemed protocol of 200mAs at 140 kVp + 650mAs at 80 kVp would still result in 3 times more noise in the 80 kVp image g (for ( 35-cm water cylinder). y ) 30000 dN/dE (#/k keV) 25000 Δ<E>=27 keV 20000 15000 <E_high>=88 keV This would reduce severely the material separation capability 10000 <E_low>=61 keV 5000 0 0 50 100 150 X-Ray Energy (keV) 140kVp_Specrum 80kVp_Specrum Ami Altman & Raz Carmi Philips Healthcare 7 The Effect of Higher Noise in The Low-Energy Image 10 mM/L Iodine, SD=10 HU both E_Low and E_High; Gaussian fit 12 to the noise: Separation is possible 10 8 10 8 6 4 2 Equivalent concentration of Ca to get the same HU 70 mM/L I di in Iodine i Water 6 1140 1120 1100 1080 1060 Low HU 1040 E 4 2 1140 0 1120 1200 1100 1150 1080 1100 1060 1050 1040 1000 1000 High HU 1050 1100 1150 1200 E 10 mM I, I SD=15 HU E E_Low, Low =10 HU E E_High; High; Gaussian fit to the noise: 10 mM/L Iodine in Water Separation is almost impossible 12 10 70 mM/L I 8 Iodine 6 10 8 6 4 2 Ca 4 1160 1140 1120 1100 1080 1060 1040 1020 Low HU E 2 1150 1100 0 1200 1150 1200 1150 1100 1050 HighE HU 10 mM I 1100 1050 1050 1000 1000 This is why in any Tube-Based Dual Energy CT, CT one might be forced to use 100/140 kVp instead of 80/140 kVp (a use of filter on the high kVp, improves the poor spectral separation of the 100/140 kVp combination.) Water 8 Ami Altman & Raz Carmi Philips Healthcare Spectral Resolving Power – An Objective Measure of the Material Decomposition Quality in Dual-E CT Following conventions in 2D mass spectroscopy, and in a combined bi d Mass-TOF M TOF spectroscopy, t we define d fi a “Spectral “S t l Resolving Power” 10 8 6 4 2 1140 1120 1100 1140 1120 1080 1100 1060 1080 1060 1040 LowE HU 1040 1020 1000 1020 1000 HighE HU Thus fit to the data from a standard phantom (with low concentrations of Iodine and Calcium ((see previous p slide), ), Two 2D Gaussian functions: 12 10 8 6 4 Than the Spectral Resolving power is defined: 2 0 1150 1100 1150 1100 1050 1050 1000 1000 950 950 RESOLVING_ POWER= A+ B ∫ ( Gaussian# 1 U Gaussian# 2 ) 2D A B ile of Pr Where A and B are the non overlapped volumes of the two G Gaussian i functions f ti , and d the th denominator d i t is i the th total t t l volume l off the union of the the two Gaussian functions ne Li This takes in account all relevant factors: Image Noise, Mean Energy gy Difference,, patient p size,, spectra p overlap, p, Mean Energy gy of each spectrum etc. Note that the resolving power ≤ 1 Ami Altman & Raz Carmi Philips Healthcare 9 Simulations and Comparison Conditions (GEANT4 [GATE] full CT Simulation) 1. Dual-Energy methods: i. Double-Decker Brilliance geometry and detector sizes with X-DFS, 2320 views, Single-Slice CT, axial 360-deg, scans 250 mAs, 140 kVp ii. Dual-Source CT has been simulated using 2 scans with Brilliance geometry and detector sizes with X-DFS, 2320 views, Single-Slice CT, axial 360-deg, with 130 mAs at 140 kVp and 670 mAs at 80 kVp (Note that the dose per mAs at 80 kVp is ~5.2 times less than in 140 kVp). Dual source CT has been simulated with and without a Tin (Sn) filter (0.35 –mm thick). iii. kVp Switching has been simulated with the same Brilliance geometry and parameters as above, Using 1/8 scheme (1 view of 80 kVp every 8 views of 140 kVp), which is one of the best modes to overcome the sampling sparsity, with 130 mAs at 140 kVp and 670 mAs at 80 kVp (No Tin filter has been used) iv. Photon Counting (for reference) 150 mAs (this the equivalent dose to ~250 mAs in Current Integration), 2 Energy Windows with no overlap has been used. Same geometry and conditions as above 2. Phantoms i. 20-cm Water Cylinder with 4 test tubes as shown in slide 8 ii. 36-cm Water Cylinder with the same 4 test tubes Ami Altman & Raz Carmi Philips Healthcare 10 Few Results Obtained From GEANT4 (GATE) Simulations * Method Spectral Resolving Power: Spectral Resolving Power: 20-cm Phantom 36-cm Phantom (10 mM/L I) (10 mM/L I) Comments Dual-Source 0.61 ± 0.02 0.22 ± 0.02 80;140 kVp with Tin filter Same dose for all methods Same dose for all methods Dual-Source 0.42 ± 0.02 0.30 ± 0.02 Energy separation is low ~19 keV with th Ti the Tin filt filter 0.54 ± 0.02 0.51 ± 0.02 All modes are possible for FOV up to 500mm Fast kVp Switching 80;140 kVp (no filter) 0.41 ± 0.02 0.21 ± 0.02 Cardiac questionable; Sparse sampling affects both IQ and material decomp.; Tin filter cannot be used, poor energy separation p Fast kVp switching 100;140 kVp 0.22 ± 0.02 0.18 ± 0.02 Almost useless without a filter Photon Counting CdTe, CdZnTe 0.75 0.66 2-Energy windows only; 100;140 kvp with Tin filter Double-Decker Detector 1-mm Top ScintillatorÆ 0.025-mm Tin Æ 2-mm GOS 80 kVp image noise is a serious li it ti iin M limitation Medium-large di l patients; ti t Hard to apply to gated\tagged CCTA; Limited FOV Assuming 10% energy resolution, and no rate limit *Attenuation & Beam Hardening corrections have been applied for all methods (See R. Carmi, A. Altman, G. Naveh MIC IEEE 2005) Ami Altman & Raz Carmi Philips Healthcare 11 Materials Decomposition (e.g. Contrast Agents) in Mixtures µE1 ((HU)) 1. Any material concentration varies along the specific material spectral line (water at the origin in HU scale) Z1 2. Image locations with 2-material mixtures of Z1 and Z2 (easily generalizeable to more than 2 materials) can be quantified easily through simple vector r r r calculations X µE1 Z2 α 1α 1E1 W X =α + β β µE2 (HU) 1β 3. Add adaptive diffusion filter and proper statistical noise analysis to refine material separation (assuming Gaussian noise in the spectral map) µE2 1E2 Iodine + Carbon Iodine Calibration: 100% Iodine Image 1. Accurate quantification of Iodine contrast agent in Iodine+Carbon Mixtures 2. Carbon-based polymer mixed with Clinical Iodine Contrast (Ultravist) have been used 3. Measured in the Dual-Layer CT using a 25-cm Plexiglas phantom diameter, with inserts 51.8% 0% 1 2 2.3% 8 3 Carbon Image 7 Iodine Calcium 100% 4 6.3% Carbon 5 9.3% 6 0% 15.4% CT image Ami Altman & Raz Carmi Philips Healthcare12 Dose/Noise Effect on Material Decomposition I di iimages Energy Iodine E M Map 800 mAs The same p phantom,, different scan dose 50 mAs 15 mAs Ami Altman & Raz Carmi Philips Healthcare 13 Clinical Images, Obtained with A Dual-Layer Detector Spectral CT 2. 2 3. A Philips Brilliance-64 with a Double-Layer Detector operates routinely in Hadassah Medical Center, at the Hebrew University in Jerusalem. D lE Dual Energy scans are performed f d att 140 kV kVp with ith conventional ti l dose, d ≤ 250 mAs A ffor all ll protocols t l All images are courtesy of Hadassah MC, and Dr. Jacob Sosna, Head of the CT unit there. 140 kV 250 mAs Separation line 1. Iodine HU o of E1 1. 2. Calcium 3. H 2O HU of E2 Iodine-tagged blood well separated from blood-vessel calcifications and bones Soft-tissue (muscles) are well separated even from low lowconcentrated Iodine regions Different materials / tissues are overlayed with colors on the anatomic image Soft tissue separation from Iodine contrast and from bones: Soft Tissue Iodine-tagged Blood Calcium & Bones Fat Spectral Analysis Map HU of E1 Calcium Iodine Soft tissue Conventional CT Image Spectral CT Image, Dual-E HU of E2 Ami Altman & Raz Carmi Philips Healthcare 14 Virtual Non Contrast Image Generation (for algorithms & methods see L. Goshen, A. Altman & R. Crami MIC2008 IEEE) 250 mAs 250 mAs Ami Altman & Raz Carmi Philips Healthcare 15 Advanced Iodine Perfusion Maps, Tissues/Material Decomposition Main Procedure: Dual Energy Images Noise Removal, preserving Spectral Map information Noise Level Estimate Raw energy map 1800 Noise free energy map 1800 1700 1700 1600 1600 1500 1500 1400 1400 1300 1300 1200 1200 1100 1100 1000 Estimate of Material Response Vector Iodine Color Map Iodine Map Generation 1000 900 900 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 A tiny lung nodule detected on an Iodine-map Iodine map image (b), obtained with a PHILIPS Dual-Energy CT a b Note that on the conventional CT Image (obtained simultaneously during the same scan), the LungN d l looks Nodule l k as a normall Iodine-Tagged I di T d blood bl d vessel Conventional CT Spectral Iodine Maps Conventional CT Iodine Image A detected non-Perfused Lung Nodule (Tumor) Ami Altman & Raz Carmi Philips Healthcare 16 Towards Prepless CT Colonoscopy with Dual-Energy CT Nominal Virtual-Colonoscopy scan protocol and dose a b The colon is partially filled with stool and b th Iodine both I di and d Barium contrast agents Corrupted colon wall d c Non-cleansed residuals Conventional-CT electronic cleansing with high g and low HU thresholds only Electronic cleansing with dual-energy analysis 11 Ami Altman & Raz Carmi Philips Healthcare 17 Towards Prepless CT Colonoscopy with Dual-Energy CT (cont.) Compare Mode Conventional Electronic Dual Energy Electronic Cleansing Cleansing Bowel is still full of stool Ami Altman & Raz Carmi Philips Healthcare 18 Towards Prepless CT Colonoscopy with Dual-Energy CT (cont.) Conventional Electronic Cleansing Dual Energy gy Electronic Cleansing A false polyp caused by residual stool Ami Altman & Raz Carmi Philips Healthcare 19 Quantifying Composites of Tissues Mixtures, Soft-Plaque Characterization Vulnerable Plaque in Carotids Purple indicates high lipidic component in plaque Lumen Calcification Soft Plaque Ami Altman & Raz Carmi Philips Healthcare 20 Kidneys Stones Identification and Quantification + Brushite + Calcium Oxalate y = monohydrated Brushite + Cystine + Urique acid Uric Acid In-Vivo Kidneys Stone analysis, using a Calibration Phantom Ami Altman & Raz Carmi Philips Healthcare 21 Simultaneous Multi-Phase Imaging Using Contrast Agents Mix, injected in Separate times 1. Plaque-induced NWZ Rabbits, through cholesterol-rich diet 2. Early injection of targeted Iodine-Loaded nano-particles contrast agent, highly up-taken by Macrophages (N1177 NPC, developed and y NanoScan Imaging, g g, Lansdale,, PA)) manufactured by 3. Late injection (hours) of Gd Contrast-Agent (Magnevist 280) 4 Scanning to image simultaneously both plaque and lumen 4. Ami Altman & Raz Carmi Philips Healthcare 22 Demonstrating Material separation: Iodine vs. bone + gadolinium A) N1177 (iodine) + gadolinium. Scan 4 hours Scan: ho rs after N1177 injection and immediately after gadolinium injection B)) Material separation p with dual-energy gy spectral analysis shows the differentiation between iodine to gadolinium and bone B A bone and gadolinium iodine Gadolinium contrast material in the heart Iodine nanoparticles contrast material in the spleen Gadolinium contrast material in the heart Iodine nanoparticles contrast material in the spleen Note that the spleen is rich with Macrophages ! Ami Altman & Raz Carmi Philips Healthcare 23 A) N1177 (iodine nanoparticles) Scan: 2 hours after injection A A first example showing possible plaque in the aorta B) N1177 (iodine) + gadolinium. C) Material separation with dual-energy spectral analysis shows the differentiation Scan: 4 hours after N1177 between iodine in the soft plaque and injection and immediately gadolinium in the aorta lumen after gadolinium injection B C 139 HU Iodine nanoparticles captured in soft plaque. Max. intensity after 2 h, reducing g after 4 h soft plaque (iodine) 122 HU Gadolinium enhancement of the aorta lumen - can help in areas where th lumen the l walls ll are less l clear 134 HU gadolinium bone and gadolinium iodine Ami Altman & Raz Carmi Philips Healthcare 24 A Second Example, Soft-Plaque Imaging Simultaneously with The Lumen A) N1177 (iodine nanoparticles) Scan: 2 hours after injection A B) N1177 (iodine) + gadolinium. Scan: 4 hours after N1177 injection and immediately after gadolinium injection C) Material separation with dual-energy spectral analysis shows the differentiation between iodine to gadolinium and bone B bone C bone bone and gadolinium iodine Possibly some captured iodine inside plaque in the aorta walls iodine nanoparticles concentrated in the spleen iodine nanoparticles concentrated in the spleen Possibly some captured iodine inside plaque in th aorta the t walls ll and d gadolinium enhancement of the aorta lumen probably: plaque / lumen (iodine / gadolinium) dif f erentiation Ami Altman & Raz Carmi Philips Healthcare 25 25 Philips Healthcare, 2008 26