![
2
1
)
(
'
0
]
2
,
1
[ )
(
ln
E
E
T
E
E
E dE
e
E
mes
Two approaches to generate the two energies: kVp and filter switch or only filter switch
Example of Low Energy spectra
Polychromatic spectra:
1mm Al
2mm Al
3mm Al
4mm Al
10
Dual-energy X-ray absorptiometry (DXA)](https://image.slidesharecdn.com/rinkel-isotdaq2015-230305154127-cc9191be/85/RINKEL-ISOTDAQ2015-ppt-10-320.jpg)
RINKEL-ISOTDAQ2015.ppt
- 1. Introduction to medical imaging Jean Rinkel International School on Trigger and Data Acquisition Systems Centro Brasileiro de Pesquisas Físicas, Rio de Janeiro 2015
- 2. X-rays Outline Computed Tomography (CT) Radiography Single Photon Emission Tomography (SPECT) Positron Emission Tomography (PET) Magnetic Resonance Imaging (RMI) Non-ionizing radiation techniques Echography Medipix detectors
- 3. Medical imaging modalities Source: Winnie Wong International School on Trigger and Data Acquisition Systems Wigner Datacenter Budapest, 2014
- 4. Beer-Lambert attenuation law: T E E N E N mat ) ( exp ) ( ) ( 0 Trade-off to set the energy: - High variability of τmat to separate soft issues at low energies - High dose to patient at low energies Principle of Radiography X-ray production Image source: http://www.arpansa.gov.au/radiationprotection/basics/xrays.cfm 4 Image source: The Essential Physics of Medical Imaging (textbook) T E E N E N mat mat ) ( exp ) ( ) ( 0 0 20 40 60 80 100 120 0 2 4 6 8 10 12 14 16 18 20 Nombre de photons absorbés Energie (keV) 0 20 40 60 80 100 120 0 100 200 300 400 500 600 Plein flux Nombre de photons émis Energie (keV) L mat Incident spectrum N0(E) Measured spectrum N(E) Interactions of X-rays with Matter T
- 5. Analysis limited to qualitative visual interpretation Detectors in Radiography Computed radiography - Absorbed x-ray energy is trapped in a photostimulable phosphor (PSP) screen (usually BaFBr) - The imaging plate is run through a special laser scanner, or CR reader, that reads and digitizes the image - Can be reused 5 http://medinfo.ufl.edu/othe r /histmed/klioze/index.html Conventional radiography (Films) - Based on a layer of light-sensitive emulsion consisting of silver halide (about 95% AgBr and 15% AgI) - When interacting with x-rays, Br- ions are liberated and captured by the Ag+ ions. - Process of developing with a chemical to form metallic silver (black) Image source: The Essential Physics of Medical Imaging (textbook) Main drawback: slow readout process
- 6. - Numerical detectors furnish an information which can be directly post processed - Enables quantification Digital radiography: flat panel detectors with real time display Detectors in Radiography 6 Direct conversion - Photoelectric absorption of the X-ray photons within the sensor (semi conductor, usually Amorphous selenium, a-Se – other materials tested: Si, CdTe, Ge) - Ion pairs are collected under applied voltage across the solid state converter Indirect conversion Flat panel detectors based on scintillators layer (usually caesium iodide (CsI) or gadolinium oxysulfide (Gd2O2S)) combined with a-Si photodiodes Trixell flat panel detector 3200×2304 pixels
- 7. Estimate an image of the mass composition of the two main components of breast soft tissues (fat and glandular tissues) to optimize breast cancer detection Objective Data processing 1) Direct problem 7 Detector Incident beam Direct beam Scatter Compression paddle T Acquisition Breast compression to ensure a constant thickness T Breast ) ( ) ( ) ( - Att fat gland gland fat gland fat t T t E E T E t 2) Per pixel inversion With the monochromatic approximation: fat gland t t T Thickness equation: gland gland fat fat E gland gland fat fat t E t E dE t E t E ) ( ) ( ) ) ( ) ( ( Att X-ray measurement: Quantification in mammography
- 8. Rough mammogram Thickness of glandular tissue 8 Quantification in mammography
- 9. ) ( 0 E st st b b M E M E e E E ) ( ) ( 0 ) ( ) ( Incident Flux (energy E) Transmitted flux Bone(b) / Soft tissue (st) st H st b H b H st L st b L b L M M mes M M mes Measurements at 2 energies (L, H) Mass attenuation coefficient g cm / ) ( 2 Projected mass (g/cm2) ) ( ) ( ) ( ) ( ln 0 E M E M E E mes st st b b E Attenuation Principal application: measurement of bone mineral density (BMD), reflecting the strength of bones as represented by calcium content H st L st st H b st L b H st L b R with R mes R mes M H b L b b H st b L st H b L st R with R mes R mes M Reconstruction 9 Dual-energy X-ray absorptiometry (DXA)
- 10. 2 1 ) ( ' 0 ] 2 , 1 [ ) ( ln E E T E E E dE e E mes Two approaches to generate the two energies: kVp and filter switch or only filter switch Example of Low Energy spectra Polychromatic spectra: 1mm Al 2mm Al 3mm Al 4mm Al 10 Dual-energy X-ray absorptiometry (DXA)
- 11. 11 2 5 2 4 3 2 1 0 2 5 2 4 3 2 1 0 HE LE HE LE HE LE b HE LE HE LE HE LE st mes b mes b mes mes b mes b mes b b M mes a mes a mes mes a mes a mes a a M Inversion by deformation of the linear approach: Calibration 2,5 3,0 3,5 4,0 4,5 3 3,5 4 4,5 5 5,5 6 6,5 mesure BE mesure HE 0g/cm2 Hy + Corr 0,4g/cm2 Hy + Corr 0,8g/cm2 Hy + Corr 1,2g/cm2 Hy + Corr 1,6g/cm2 Hy + Corr Calibration to estimate the coefficients of the polynomial Non linear direct problem 2 1 ) ( ) ( ' 0 2 1 ) ( ) ( ' 0 ) ( ln ) ( ln HE HE M E M E HE LE LE M E M E LE dE e E mes dE e E mes st st b b st st b b Polychromaticity 2 2 2 2 287 . 7 909 . 6 94 . 15 57 . 80 54 . 72 484 . 6 333 . 5 629 . 6 33 . 13 1 . 141 68 . 77 33 . 20 h l h l h l QRM h l h l h l Plexi mes mes mes mes mes mes L mes mes mes mes mes mes L Dual-energy X-ray absorptiometry (DXA)
- 12. Low energy acquisition High energy acquisition Combination Bone mineral density Soft tissues 12 Dual-energy X-ray absorptiometry (DXA)
- 13. Low energy acquisition BMD image Soft tissues image Backbone Hip Arm 13 Dual-energy X-ray absorptiometry (DXA)
- 14. High energy acquisition BMD Soft tissues Low energy acquisition Thoracic radiography 14 Dual-energy X-ray absorptiometry (DXA)
- 15. Low energy acquisition BMD Soft tissues zoom Encephalometry Dual-energy X-ray absorptiometry (DXA) 15
- 16. Introduction to computed tomography (CT) History - First CT scanner in 1972 (EMI) - Hounsfield awarded by the Nobel prize in Physiology or Medicine 1979 - 4 minutes /scan Outline Introduction Radiography CT SPECT PET Medipix Principle “Tomography” refers to a picture (graph) of a slice (tomo). Radiography : projection on a plane of superposition of anatomical structures CT: combination of various radiography at different angles to eliminate this superposition and display three-dimensional (3D) slices Image source: The Essential Physics of Medical Imaging (textbook) Units Tomography: linear attenuation coefficient μ (3D map) Hounsfield units: Radiography: attenuation 16
- 18. Detectors in CT 18 Indirect conversion: Image source: The Essential Physics of Medical Imaging (textbook)
- 19. Generations of CT http://perso.telecom-paristech.fr/~bloch/ATIM/tomo.pdf 19 First generation: - One source and one detector - Pencil-like X-ray beam - For each slice: translation + rotation of the system « source + detector » - Duration of scan (average): 25-30 mins Second generation: - Narrow beam irradiating a few detectors (up to 30) - For each slice: translation + rotation of the system « source + detectors » - Duration of scan (average): less than 90 s Third generation: - Suppresion of the translation - Broad beam covers the whole volume of the patient - Until 700 detectors - For each slice: rotation of the system « source + detector » - Duration of scan (average): approximately 5s Fourth gneration: - Detectors: multiple (more than 2000) arranged in an outer ring which is fixed - The source only is rotating - Duration of scan (average): few seconds Evolution of acquisition time: 5 minutes / slice in 1972 vs 0.35 s / rotation today
- 20. Multi slice CT – dose and acquisition time http://www.impactscan.org/download/msctdose.pdf -> Reduced dose due to the z-axis geometric efficiency Today: most fabricants propose 64 slices scanners 20 Number of pixel rows tend to increase (in the z-axis) Image source: The Essential Physics of Medical Imaging (textbook) -> Reduced acquisition time
- 21. Effect of detector size Evolution of detector size 1 strip- 1972 (EMI) X-ray source Volume of the patient producing scatter Multi slice CT and scatter 64 strips- 2004 (Toshiba) X-ray source Volume of the patient producing scatter 21
- 22. Scatter image Transmitted image 4.5 4 3.5 3 2.5 2 1.5 1 0.5 ) ( 0 E = + Image of scatter Image of transmitted beam Acquired image Patient Incident X-rays flux Multi slice CT and scatter 22
- 23. Multi slice CT and scatter correction 23 Source: Phys. Med. Biol. 52 (2007) 4633–4652
- 24. Reconstruction: inversion of the Radon transform - finding f(x,y) fromPqt Radon transform and sinogram y x t q pqt f t u y x 50 100 150 200 250 50 100 150 200 250 y x 50 100 150 200 250 50 100 150 200 250 q (degrees) t (pixels) 20 40 60 80 100 120 140 160 180 50 100 150 200 250 300 350 q (degrees) t (pixels) 20 40 60 80 100 120 140 160 180 50 100 150 200 250 300 350 0 Ramp Filter Classical reconstruction in CT Filter back projection algorithm (FBP): Exact Fast Not well adapted for noisy data 24
- 25. Iterative reconstruction in CT Standard FBP vs iterative reconstruction Example of the MLEM (Maximum Likelihood Expectation Maximization) algorithm Dushyant Sahani, Iterative Reconstruction Techniques (ASIR, MBIR, IRIS): New Opportunities to Lower Dose and Improve Image Quality SCBT- MR 2015 25 The image at the iteration k+1, fi k+1, is estimated by multiplying the previous image, fi k, by a correction factor: - pj: measured projection - Rji radon transform operator The correction factor is the normalized retroprojection of the measured projection divided by the calculated projection Bayesian algorithm: optimizes the likelihood of the estimated image knowing the measurement under the hypothesis of a Poisson noise.
- 26. Multi slice CT – iterative reconstruction American Journal of Roentgenology. 2010;195: 713-719 Dushyant Sahani, Iterative Reconstruction Techniques (ASIR, MBIR, IRIS): New Opportunities to Lower Dose and Improve Image Quality SCBT-MR 2015 Recent Integration of iterative reconstruction methods: GE: ASIR (Adaptive Statistical iterative Reconstruction) - 2009 Siemens: IRIS (Iterative Reconstruction in Image Space) - 2010 Philips: iDose - 2011 Toshiba: AIDR (Adaptive Iterative Dose Reduction) - 2011 26
- 27. Dual energy CT - Quick kV switch to achieve dual energy acquisition in a single acquisition - Dual source CT Dual-energy tomography vs Subtraction angiography (invasive technique) Source: Acad Radiol 2012; 19:1149–1157 27
- 28. Single photon emission computed tomography (SPECT) Outline Introduction Radiography CT SPECT PET Medipix Trade-off between spatial resolution and the detection efficiency Spatial resolution depends strongly on the collimator and the source position. - Injection of gamma emitting radionuclide into the patient with specific binding to certain types of tissues - Gamma-ray detection by a nuclear camera of emissions from the patient from a series of different angles around the patient (Typically: 1282 pixels with 120 or 128 projection images) - Tomographic reconstruction of emission images - Functional information: distribution of the radioactive agent Typical 180-degree cardiac orbit RAO (right anterior oblique) to LPO (left posterior oblique) Parallel collimation Typical values for low-energy high-resolution parallel hole collimators : - Tangential resolution for the peripheral sources :7 to 8 mm FWHM - Central resolution: 9.5 to 12 mm - Radial resolution for the peripheral sources: 9.4 to 12 mm 28
- 29. Radionuclide: Example of active molecule: fluorodeoxyglucose (FDG): analog to glucose indicate the tissue metabolic activity explore the possibility of cancer metastasis Detectors Positron Emission Tomography (PET) Comparison to SPECT: - Better spatial resolution: typically 5-mm FWHM in the center of the detector ring - PET usually more expensive than SPECT scans: able to use longer-lived more easily obtained radioisotopes than PET. Principle: detection of pairs of gamma rays emitted indirectly by a positron-emitting radionuclide. - The emitted Positron scatter through matter - It annihilates with an electron resulting in two 511-keV photons emitted in opposite directions - Simultaneous detection means that an annihilation occurred on the line connecting the two detectors Typical time windows - BGO detectors: 12 ns - LSO detectors: 4.5 ns 29
- 30. PET - CT 30
- 31. Dual modality (SPECT / CT and PET CT) Example of PET CT images: PET information superimposed in color on grayscale CT images Source: http://www.chups.jussieu.fr/ext/OLDceb/protected/protectedCNEBMN/atelierDreuilleTEP/ArticleEMC_PET.pdf Simultaneous acquisition of PET or SPECT and low dose CT: - CT provides - good depiction of anatomy - attenuation correction - help for patient alignement - SPECT or PET provides functional information 31
- 32. Cathode Semi condutor Photon - - Charge Time Anodes Pitch threshold Counting 30 keV Spectroscopy s( t) time ΔT 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Counting s(t) time 80keV 30keV 10keV 90keV 40keV 60keV 20keV 30keV 30keV 20keV 60keV 50keV 50keV 40keV 60keV 70keV ΔT Spectroscopy Direct conversion detectors Outline Introduction Radiography CT SPECT PET Medipix 32
- 33. Medipix overview Outline Introduction Radiography CT SPECT PET Medipix - Medipix3 Timepix3 - Pixel size: 55x55um - Format 256x256= 65536px - Framing rate: 1000Hz http://medipix.web.cern.ch/medipix/pages/medipix3.php - Medipix3 (Frame based) - Single pixel mode: High dynamic range (24 bits) - Charge summing mode (12 bits) - Spectrometric mode: 110 μm pixels / 8 energy thresholds (12 bits) - Timepix3 (Data-driven) - Counting - Time over Threshold (ToT) - Time of Arrival (ToA) Source: T. Poikela et al, 15th INTERNATIONAL WORKSHOP ON RADIATION IMAGING DETECTORS, 2013
- 34. Tomography at the brazilian synchrotron Best signal to noise ratio compared to classical systems - Counting - Higher efficiency 1mm 1mm • Cellular structure and features • Cell types and orientation • Biology Micro tomography based on indirect conversion (scintillator + CCD) Direct conversion (Medipix) Lower spatial resolution - Pixel size limited by the CMOS technology 130 nm - Charge sharing effect Counting rate
- 35. Toward spectral tomography Timepix used at 3 different threshold positions: between and after K edges of Cd and Cu 35 Source: Enrico Jr Schioppa et al 2012 JINST 7 C10007
- 36. Through examples of medical applications using ionizing radiation: Conclusion - Quantification approaches enabled by emergence of numerical detectors (combined with applied mathematics and calibration procedures) - Recent innovations to optimize acquisition time and dose for a given image quality and associated problems and limitations: - Multi slice CT - Direct conversion (example of Medipix and Timepix) Suggested reading Books: - J.Bushberg et al., The Essential Physics of Medical Imaging, 3rd Ed., 2012 - J. Jan, Medical Image Processing, Reconstruction and Restoration, Concepts and Methods, 2006 Medipix: - http://medipix.web.cern.ch/medipix/
- 38. Pixel 55 x 55 μm2 Area 24cm2 Number of Pixels 786K pixels Project of module based on 12 Medipix3RX Pixel 172 x 172 μm2 Area 28cm2 Number of Pixels 95K pixels Pilatus 100k Dectris detector Possibility to add various modules (3 modules equivalent in size to Pilatus 300k) http://photon- science.desy.de/research/technical_group s/detectors/projects/lambda/index_eng.h tml
- 39. Project of module based on 12 Medipix3RX Main numbers: - Total Readout time (12 chips Medipix3RX / 24bits counters and 350MHz) = 562 us - Frame rate (for acquisition time = 10% of readout time) = 1615 f/s - Required Bandwidth: 7.6 Gbytes/s - 2.4 Mbytes images Single chip inputs / outputs 8 LVDS pairs receivers 10 LVDS drivers (8 for data) 2 analog inputs (generated from DACs) 1 analog output (to be connected to an ADC) Full parallel readout of 12 chips: 161 LVDS pairs (using 8 lines in parallel /chip for data) 24 analog inputs 12 analog outputs
- 40. Project with NI PXI-7952R Number of general-purpose channels 132, configurable as 132 single-ended, 66 differential, or a combination of both1 Maximum I/O data rates Single-ended 400 Mb/s for LVDCI25 Differential 1 Gb/s for LVDS Global clock inputs On-board memory 1 LVTTL, 1 LVDS 128 MB ; 800MB/s NI PXI-7951R – $ 3,415 NI PXI-7952R NI PXI-7953R NI PXI-7954R NI PXI-7961R NI PXI-7962R NI PXI-7965R - $ 10,020.00 NI PXI-7966R - $ 10,320.00
- 41. 1 main board + 9 adapters Communication and powering of 9 Timepix chips + synchronization with external equipments Adaptor 1 . . . Adaptor 9 Detector 1 . . . Detector 9 Main Board Project with NI PXI-7952R
- 42. Objective: full parallel readout of 9 chips Medipix3RX / TIMEPIX connected to one NI PXI-7952R Limitation =>Maximum number of LVDS pairs data outputs : 1 (instead of 8 in an optimized configuration) with 66 LVDS pairs Number of LVDS pairs for data output / chip Required Total Number of LVDS pairs 1 59 2 68 4 86 8 122 Main numbers: - Total Readout time (9 chips Medipix3RX / 24bits counters and 350MHz) = 4.5 ms - Frame rate (for acquisition time = 10% of readout time) = 202 f/s - Required bandwidth: 715 Mbytes/s - 1.8 Mbytes images Project with NI PXI-7952R
- 43. PROJECT WITH 12 CHIPS 12 786.432 24,0 cm² Parallel 8 lines / 24bits 563 us 1615 f/s 2,4 MB 161 24 12 7.6 GB/s PROJECT WITH NI 7952R 09 589.824 18,0 cm² Parallel 1 line / 24bits 4,5 ms 202 f/s 1,8MB 59 0 1 715 MB/s Number of chips Number of pixels Total area Operation mode Readout time Frame rate Image size LVDS pairs Analog inputs Analog output Transfer rate Comparison and conclusion DETECTOR OF SAME AREA THAN PILATUS 6M 840 55.050.240 1.680,0 cm² Parallel 8 lines / 24bits 563 us 1615 f/s 165,1 MB 11.270 1680 840 532 GB/s For National Instrument solution being competitive with 12 chips project: - Need for a Board with 161 LVDS pairs - Data have to be transmitted without lost through long flat cables (a few meters) at 350 MHz - On-board memory: 2,4 MB - 7.6 GB/s