Transcript Detectors - CERN Indico

Introduction to medical imaging
Jean Rinkel
International School on Trigger and Data Acquisition Systems
Centro Brasileiro de Pesquisas Físicas, Rio de Janeiro
2015
Outline
X-rays
Radiography
Computed Tomography (CT)
Single Photon Emission Tomography (SPECT)
Positron Emission Tomography (PET)
Non-ionizing radiation techniques
Magnetic Resonance Imaging (RMI)
Echography
Medipix detectors
Medical imaging modalities
Source: Winnie Wong International School on Trigger and Data Acquisition Systems Wigner Datacenter Budapest, 2014
Principle of Radiography
X-ray production
Image source: The Essential Physics of Medical Imaging (textbook)
Image source: http://www.arpansa.gov.au/radiationprotection/basics/xrays.cfm
Interactions of X-rays with Matter
Incident spectrum N0(E) L
T
Plein flux
600
20
16
300
mat
200
Nombre de photons absorbés
Nombre de photons émis
N ( E )  N 0 ( E ) exp  mat ( E )  T 
18
500
400
14
12
10
N ( E )  N 0 ( E ) exp   mat ( E )   mat  T 
8
6
4
100
2
0
0
20
40
60
Energie (keV)
80
100
120
Beer-Lambert attenuation law:
Measured spectrum N(E)
0
0
20
40
60
Energie (keV)
80
100
120
Trade-off to set the energy:
-
High variability of τmat to separate soft issues at low energies
-
High dose to patient at low energies
4
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)
http://medinfo.ufl.edu/othe
r/histmed/klioze/index.html
Detectors in Radiography
Analysis limited to qualitative visual interpretation
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
Main drawback: slow readout process
Image source: The Essential Physics of Medical Imaging (textbook)
5
Detectors in Radiography
Digital radiography: flat panel detectors with real time display
Indirect conversion
Flat panel detectors based on scintillators layer (usually caesium
iodide (CsI) or gadolinium oxysulfide (Gd2O2S)) combined with a-Si
photodiodes
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
Trixell flat panel detector
3200×2304 pixels
- Numerical detectors furnish an information which can be directly post processed
- Enables quantification
6
Quantification in mammography
Objective
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
Detector
T
Acquisition
Scatter
Breast compression to ensure a constant thickness T
Incident beam
Breast
Direct
beam
Data processing
Compression paddle
1) Direct problem
Thickness equation:
X-ray measurement:
T  t gland  t fat
Att   ( fat ( E )t fat   gland ( E )t gland )dE   fat ( E )t fat   gland ( E )t gland
E
2) Per pixel inversion
With the monochromatic approximation:

Att -  gland ( E )T
t fat 
 fat ( E )   gland ( E )

t
 gland  T  t fat
7
Quantification in mammography
Rough mammogram
Thickness of glandular tissue
8
Dual-energy X-ray absorptiometry (DXA)
Principal application: measurement of bone mineral density (BMD), reflecting the strength of bones as represented by calcium content
0 ( E )
Incident Flux (energy E)
Mass attenuation coefficient

( ) cm 2 / g

Bone(b) / Soft tissue (st)
 ( E )  0 ( E )  e  
Transmitted flux
Attenuation
mesE  ln
  st ( E )  M st
0 ( E )
 M b  b ( E )  M st  st ( E )
 (E)
Measurements at 2 energies (L, H)
Reconstruction
b (E)  Mb
Projected mass (g/cm2)
mesL   bL  M b   stL  M st

H
H
 mesH   b  M b   st  M st
mesL  Rst  mesH
Mb 
 bL  Rst  bH
mesL  Rb  mesH
M st 
 stL  Rb  stH
 stL
with Rst  H
 st
 bL
with Rb  H
b
9
Dual-energy X-ray absorptiometry (DXA)
Two approaches to generate the two energies: kVp and filter switch or only filter switch
Polychromatic spectra:
1mm Al
2mm Al
3mm Al
4mm Al
E2
mes[ E1, E 2 ]   ln  0' ( E )  e  ( E ) T dE
E1
Example of Low Energy spectra
10
Dual-energy X-ray absorptiometry (DXA)
LE 2

 b ( E ) M b  st ( E )M st
'
dE
mesLE   ln  0 ( E )  e

LE 1
Non linear direct problem 
Polychromaticity
HE 2
mes   ln  ' ( E )  e  b ( E )M b  st ( E )M st dE
HE
0


HE1

Inversion by deformation of the linear approach:
2
2
M st  a0  a1 mesLE  a2 mesHE  a3 mesLE mesHE  a4 mesLE
 a5 mesHE
2
2
M b  b0  b1 mesLE  b2 mesHE  b3 mesLE mesHE  b4 mesLE
 b5 mesHE
Calibration to estimate the coefficients of the polynomial
Calibration
4,5
mesure HE
4,0
0g/cm2 Hy + Corr
0,4g/cm2 Hy + Corr
3,5
0,8g/cm2 Hy + Corr
1,2g/cm2 Hy + Corr
1,6g/cm2 Hy + Corr
3,0
2,5
3
3,5
4
4,5
5
mesure BE
5,5
6
6,5
LPlexi  20.33  77.68  mesl  141.1  mesh  13.33  mesl  mesh  6.629  mesl2  5.333  mesh2
2
2
11LQRM  6.484  72.54  mesl  80.57  mesh  15.94  mesl  mesh  6.909  mesl  7.287  mesh
Dual-energy X-ray absorptiometry (DXA)
Low energy acquisition
High energy acquisition
Combination
Bone mineral density
Soft tissues
12
Dual-energy X-ray absorptiometry (DXA)
Low energy acquisition
BMD image
Soft tissues
image
Backbone
Hip
Arm
13
Dual-energy X-ray absorptiometry (DXA)
Thoracic radiography
Low energy acquisition
BMD
High energy acquisition
Soft tissues
14
Dual-energy X-ray absorptiometry (DXA)
Encephalometry
BMD
zoom
Low energy acquisition
Soft tissues
15
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
Principle
“Tomography” refers to a picture (graph) of a slice (tomo).
CT
SPECT
PET
Medipix
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
Units
Radiography: attenuation
Image source: The Essential Physics of Medical Imaging (textbook)
Tomography: linear attenuation coefficient μ (3D map)
Hounsfield units:
16
Dose in CT
http://www.radiologyinfo.org/en/safety/?pg=sfty_xray
17
Detectors in CT
Indirect conversion:
Image source: The Essential Physics of Medical Imaging (textbook)
18
Generations of CT
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
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
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
http://perso.telecom-paristech.fr/~bloch/ATIM/tomo.pdf
Evolution of acquisition time: 5 minutes / slice in 1972 vs 0.35 s / rotation today
19
Multi slice CT – dose and acquisition time
Number of pixel rows tend to increase (in the z-axis)
Image source: The Essential Physics of Medical Imaging (textbook)
-> Reduced acquisition time
-> Reduced dose due to the z-axis geometric efficiency
http://www.impactscan.org/download/msctdose.pdf
Today: most fabricants propose 64 slices scanners
20
Multi slice CT and scatter
Effect of detector size
Evolution of detector size
X-ray source
1 strip- 1972 (EMI)
Volume of the
patient producing
scatter
X-ray source
Volume of the
patient producing
scatter
64 strips- 2004 (Toshiba)
21
Multi slice CT and scatter
Incident X-rays flux
0 ( E )
+
Patient
Image of
transmitted
beam
=
Image of
scatter
Acquired image
4.5
4
Scatter image
3.5
3
Transmitted image
2.5
2
1.5
1
0.5
22
Multi slice CT and scatter correction
Source: Phys. Med. Biol. 52 (2007) 4633–4652
23
Classical reconstruction in CT
Radon transform and sinogram
50
50
100
100
y
t (pixels)
150
150
200
250
200
t
pqt
300
350
250
50
100
150
200
250
20
40
60
80
y
100
120
140
160
180
q (degrees)
x
t
u
50
50
100
100
t (pixels)
x
y
q
150
150
200
250
200
300
f
350
250
50
100
150
200
250
20
40
60
80
100
120
140
q (degrees)
x
Reconstruction: inversion of the Radon transform - finding f(x,y) fromPqt
Filter back projection algorithm (FBP):
 Exact
 Fast
 Not well adapted for noisy data
Ramp Filter
0
24
160
180
Iterative reconstruction in CT
Standard FBP vs iterative reconstruction
Example of the MLEM (Maximum Likelihood Expectation Maximization)
algorithm
The image at the iteration k+1, fik+1, is estimated by multiplying the
previous image, fik, 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.
Dushyant Sahani, Iterative Reconstruction
Techniques (ASIR, MBIR, IRIS): New Opportunities
to Lower Dose and Improve Image Quality SCBTMR 2015
25
Multi slice CT – iterative reconstruction
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
Dushyant Sahani, Iterative Reconstruction Techniques (ASIR, MBIR, IRIS):
New Opportunities to Lower Dose and Improve Image Quality SCBT-MR 2015
American Journal of Roentgenology. 2010;195: 713-719
26
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
Single photon emission computed tomography (SPECT)
-
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
Outline
Introduction
Radiography
CT
SPECT
PET
Typical 180-degree cardiac orbit
RAO (right anterior oblique) to LPO (left posterior oblique)
Parallel collimation
Medipix
Trade-off between spatial resolution and the detection efficiency
Spatial resolution depends strongly on the collimator and the source position.
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
Positron Emission Tomography (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
Radionuclide:
Example of active molecule: fluorodeoxyglucose (FDG):
analog to glucose indicate the tissue metabolic activity
explore the possibility of cancer metastasis
Detectors
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.
Typical time windows
- BGO detectors: 12 ns
- LSO detectors: 4.5 ns
29
PET - CT
30
Dual modality (SPECT / CT and PET CT)
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
Example of PET CT images: PET information superimposed in color on grayscale CT images
31
Source: http://www.chups.jussieu.fr/ext/OLDceb/protected/protectedCNEBMN/atelierDreuilleTEP/ArticleEMC_PET.pdf
Direct conversion detectors
Pitch
Counting
Time
Anodes
30 keV
-
Spectroscopy
threshold
Semi condutor
Outline
Cathode
Charge
Photon
Introduction
Radiography
CT
s(
t)
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
SPECT
PET
Medipix
1
Counting
ΔT
time
s(t)
80keV
30keV
70keV
60keV
30keV
10keV
90keV
60keV
50keV
50keV
40keV 30keV
Spectroscopy
60keV
40keV
20keV
20keV
ΔT
time
32
Medipix overview
Outline
-
Introduction
Radiography
CT
-
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)
SPECT
PET
Medipix
Medipix3 Timepix3
- Pixel size: 55x55um
- Format 256x256= 65536px
- Framing rate: 1000Hz
http://medipix.web.cern.ch/medipix/pages/medipix3.php
Source: T. Poikela et al, 15th INTERNATIONAL
WORKSHOP ON RADIATION IMAGING DETECTORS, 2013
Tomography at the brazilian synchrotron
Micro tomography based on indirect conversion
(scintillator + CCD)
Direct conversion (Medipix)
Best signal to noise ratio
compared to classical systems
Counting
Higher efficiency
Lower spatial resolution
- Pixel size limited by the CMOS technology 130 nm
- Charge sharing effect
Counting rate
•
•
•
Cellular structure
and features
Cell types and
orientation
Biology
1mm
1mm
Toward spectral tomography
Timepix used at 3 different threshold positions: between and after K edges of Cd and Cu
Source: Enrico Jr Schioppa et al 2012 JINST 7 C10007
35
Conclusion
Through examples of medical applications using ionizing radiation:
-
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/
Single chip assembly
Project of module based on 12 Medipix3RX
Pilatus 100k Dectris detector
Pixel
55 x 55 μm2
Area
24cm2
Number of Pixels
786K pixels
Pixel
172 x 172 μm2
Area
28cm2
Number of Pixels 95K pixels
Possibility to add various modules
(3 modules equivalent in size to Pilatus 300k)
http://photonscience.desy.de/research/technical_group
s/detectors/projects/lambda/index_eng.h
tml
Project of module based on 12 Medipix3RX
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
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
Project with NI PXI-7952R
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
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
Global clock inputs
On-board memory
1 Gb/s for LVDS
1 LVTTL, 1 LVDS
128 MB ; 800MB/s
Project with NI PXI-7952R
1 main board + 9 adapters
Communication and powering of 9 Timepix chips + synchronization with external equipments
Detector 1
.
.
.
Detector 9
Adaptor 1
.
.
.
Adaptor 9
Main Board
Project with NI PXI-7952R
Objective: full parallel readout of 9 chips Medipix3RX / TIMEPIX connected to one NI PXI-7952R
Limitation
Number of LVDS pairs for
data output / chip
Required Total Number of
LVDS pairs
1
59
2
68
4
86
8
122
=>Maximum number of LVDS pairs data outputs : 1 (instead of 8 in an optimized configuration)
with 66 LVDS pairs
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
Comparison and conclusion
Number of chips
Number of pixels
Total area
Operation mode
Readout time
Frame rate
Image size
LVDS pairs
Analog inputs
Analog output
Transfer rate
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
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
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