Imaging for Physicists
Animated publication
ESTRO Course Book Imaging for Physicists
13 – 17 September, 2015 Leiden, The Netherlands
NOTE TO THE PARTICIPANTS
The present slides are provided to you as a basis for taking notes during the course. In as many instances as practically possible, we have tried to indicate from which author these slides have been borrowed to illustrate this course. It should be realised that the present texts can only be considered as notes for a teaching course and should not in any way be copied or circulated. They are only for personal use. Please be very strict in this, as it is the only condition under which such services can be provided to the participants of the course.
Faculty
Uulke van der Heide
Disclaimer
The faculty of the teachers for this event has disclosed any potential conflict of interest that the teachers may have.
Programme
Day 1
Sunday 13 September
09.00 – 09.50 Welcome and introduction
Uulke van der Heide, NL
09.50 – 10.40 10.40 – 11.10 11.10 – 11.50 11.50 – 12.40 12.40 – 13.10 13.10 - 14.10 14.10 – 15.00 15.00 – 15.40 15.40 – 16.10 16.10 – 18.00
MRI physics: basic principles
Eirik Malinen, NO
Coffee break
MRI physics: contrast formation MRI physics: space encoding
Tufve Nyholm, SE Gary Liney, AU
Questions and Answers (plenary) (MR physics)
Lunch
MRI physics : equipment
Gary Liney, AU
PET physics : basic principles
Daniela Thorwarth, DE
Coffee break
CASE / ASSIGNMENT
Day 2
Monday 14 September
08.30 – 09.20 09.20 – 10.10 10.10 – 10.30 10.30 – 11.20 11.20 – 12.10 12.10 – 12.40 12.40 – 13.40 13.40 – 14.30 14.30 – 15.10 15.10 – 15.40 15.40 – 17.00 08.30 – 09.20 09.20 – 10.10 10.10 – 10.30 10.30 – 11.20 11.20 – 12.10 Day 3
Uulke van der Heide, NL Uulke van der Heide, NL
MRI geometrical artifacts I MRI geometrical artifacts II
Coffee break
Pet physics : image reconstruction, contouring
Daniela Thorwarth, DE Cynthia Ménard, CA
Applications : MRI in brain
Questions and Answers (plenary) (MR physics)
Lunch
Functional imaging MRI
Gary Liney, AU Piet Dirix, BE
Applications: MRI in gynaecology RT
Coffee break
CASE / ASSIGNMENT
Tuesday 15 September
Applications: MRI in prostate
Cynthia Ménard, CA
Applications: MRI-guided radiotherapy for head-neck cancer
Piet Dirix, BE
Coffee break
CT physics 3D and 4D imaging
Koos Geleijns, NL Koos Geleijns, NL
CT physics and advanced applications
Free afternoon
Day 4
Wednesday 16 September
08.30 – 09.20 09.20 – 10.10 10.10 – 10.30 10.30 – 11.20 11.20 – 12.10 12.10 – 12.40 12.40 – 13.40 13.40 – 14.30 14.30 – 15.10 15.10 – 15.40 15.40 – 16.30 16.30 – 16.40 16.40 – 17.30
PET tracers and applications
Daniela Thorwarth, DE Daniela Thorwarth, DE
Guidelines for PET imaging in Radiotherapy
Coffee break
Dynamic PET and CT imaging
Eirik Malinen, NO Tufve Nyholm, SE
MR Safety
Questions and Answers (plenary) (PET)
Lunch
MRI physics: fast scanning, volume sequences
Gary Liney, AU
Parallel Discussion with the faculty
Coffee break
PRESENTATIONS : CASE / ASSIGNMENT
Break
PRESENTATIONS : CASE / ASSIGNMENT (10 presentations, break 10 minutes)
Day 5
Thursday 17 September
09.00 – 09.50 09.50 – 10.40
In-room imaging with cone-beam CT and MRI Applications : CT and PET for radiotherapy of lung cancer
Tufve Nyholm, SE
Uulke van der Heide, NL
10.40 - 11.00 Coffee break 11.00 – 12.40 PRESENTATIONS : CASE / ASSIGNMENTS
12.00 – 13.00
Short evaluation of the course and closing
Faculty
Uulke van der Heide
UMC Utrecht Utrecht, The Netherlands u.vd.heide@nki.nl Iridium Cancer Network Antwerp, Belgium piet.dirix@gza.be
Tufve Nyholm
Umeå University Umeå, Sweden tufve.nyholm@gmail.com
Piet Dirix
Koos Geleijns
Leiden University Medical Centre Leiden, The Netherlands K.Geleijns@lumc.nl Ingham Institute for Applied Medical Research Liverpool, Australia Gary.Liney@sswahs.nsw.gov.au DNR – Norwegian Radium Hospital Oslo, Norway eirik.malinen@fys.uio.no Princess Margaret Hospital Toronto, Canada cynthia.menard@rmp.uhn.on.ca Uniklinik für Radioonkologie Tübigen, Germany daniela.thorwarth@med.uni- tuebingen.de
Gary Liney
Eirik Malinen
Cynthia Ménard
Daniela Thorwarth
basic principles Eirik Malinen
• All clinical applications of MRI today are based on magnetic properties of the hydrogen nucleus • Body tissues contains lots of water and fat, and hence hydrogen
• Stern-Gerlach experiment: → Atomic nuclei has a quantized magnetic Otto Stern Walter Gerlach
• Consider charge q A
in circular motion: • Rotating charged sphere with uniform charge: Magnetic moment: r v q Current: r2 qv t q i π = ∆ ∆= mvr L , L m2 q iA = = =µ
q =µ
S
m2
Spin!
• Nuclear spin is a form of angular momentum • Nuclear spin, I , is quantized in units of ℏ • Nuclear quantum number depends on nuclear configuration; I=1/2, 1… • Hydrogen has spin I=1/2, with spin projection numbers m I =+1/2 , -1/2; spin ‘up’ or ‘down’ • Magnetic moment is μ =γ I
γ
Nucleus Unpaired Protons Unpaired Neutrons Spin γ (MHz/T) 1 H 1 0 1/2 42.58 2 H 1 1 1 6.54 31 P 1 0 1/2 17.25 23 Na 1 2 3/2 11.27 14 N 1 1 1 3.08 13 C 0 1 1/2 10.71 19 F 1 0 1/2 40.08
• In an external magnetic field, the potential energy is: → Two energy states are possible • Zeema ffect B ⋅ −= µ pot E B 2 1 Bm I γ = γ−= pot E
2 1
B γ+
B
2 1
B γ−
• Spin system under an external magnetic field exposed to electromagnetic radiation • Transitions from spin down to spin up or vice versa may occur if ℏ ω = ΔE pot = γ ℏ B ℏ ω ΔE pot = γ ℏ Β Isidor Isaac Rabi
= γ ℏ
B → ω=γ
B; resonance condition
• ℏ ω
• Resonance frequency, 1 H, B=1T: ω≈ → Without external field With external field With external field + electromagnetic radiation 43 MHz
• Spin transition probability is equal for up → down and down → up • How can a net energy absorption be observed? • Distribution of spins follows Boltzmann: • Difference increases with B kTB kT E e e N N pot / / γ − ∆− ↑ ↓ = =
• Population difference generates a net magnetization • The more spins, the stronger the magnetization • Torque exerted on a magnet by a magnetic field: M 0 BMM τ × = = γ dt d
d
BMM τ × = = γ
dt
Felix Bloch
0 Md , BM td Md z x y y x = γ−= γ= ⇒ td Md , BM td
ω = ⇒
0 x
0 y
cos( M)t(M
t) sin( M)t(M , t) ω =
x
L
y
L
• ω L = γΒ ; Larmor frequency • Set of equations describing a precession B 0 z M)t(M =
Joseph Larmor
z
z
x
x
y
y All spins in phase with same Larmor frequency Spins out of phase
• How can the magnetization be altered? • Introduce oscillating (RF) magnetic field in the xy-plane z x B y =B 1 cos( ω t) B 0 z x ω = ω L
• The degree of which the magnetization is tipped relative to B 0 due to an excitation pulse • From Bloch’s considerations: • t: duration of pulse • B1: ~RF power 1 B 2 πγτ =θ z x θ M xy M z
• Fluctuating magnetic fields from the molecular environment may have Larmor frequency→ stimulated transitions may occur • After an RF-pulse, the z-component of M relaxes back to equilibrium via such stimulated transitions • Longitudinal relaxation, Spin lattice relaxation, T1 relaxation • Rate of relaxation: R1=1/T1
• The transverse component of the magnetization also decays • Local, microscopic field inhomogeneities causes each spin to precess with a frequency slightly different from ω L • An excitation pulse initially causes all spins to precess in phase, but a dephasing then occurs • transverse- or spin-spin relaxation; T2
• However, transverse relaxation is also caused by B 0 inhomogeneities and tissue magnetic susceptibility • Actual T2 time is denoted T2*: • T2* z z 90 ° pulse x x y y y time Transversal x z Longitudinal T1 T2* x • Bloch’s equations expanded with relaxation components; M xy /T2* and (M z -M 0 )/T1 • May be shown that: ) e1(M)t(M 1T/t 0 z − − = T1=300ms M63.0 M 1 T t ∞ = ⇒ = z, z T2*=100ms T2* T2 *2T/t eM)t(M − = xy 0,xy xy,0 xy M37.0 M *2T t = ⇒ = Brain CSF Fat • Changes in magnetization give rise to a current in a wire loop (Faraday’s law of induction) • Receiver coil perpendicular to B 0 : x y z x y 90 ° pulse relaxation Coil signal coil coil • Envelope of FID describes the T2*-decay: Fourier transform Thank you for your attention! MRI physics: Contrast formation Tufve Nyholm Precession Magnetic field 42.576 MHz/T Flip RF puls Magnetic field 42.576 MHz/T Relaxation ( Rotating coordiante system T1 relaxation Parallel plane T2 relaxation Transversial plane B0 T1 relaxation • Spin-lattice or longitudinal relaxation • Restoring longitudinal magnetization after RF excitation • T1 – Time until 63% of the initial magnetization M0 is restored Adipose tissue – 240ms Spinal fluid – 4300ms Gray matter – 980ms White matter – 780ms Muscles – 880ms 980ms T2 relaxation • Spin-spin or transversial relaxation • Loss of transversial magnetization after RF excitation • T2 – time until 63% of the transversal magnetization is lost Adipose tissue – 70ms Spinal fluid – 2200ms Gray matter – 100ms White matter – 90ms Muscles – 50ms 100ms T2* relaxation Higher field Lower field Spin-Echo sequence T1 relaxation T2 relaxation Signal equation Constant depending on • Coils • Temperature • etc Proton density T2 contrast Minimize influence i.e. Long TR Focus TE T2 contrast Examples T2 Contrast Adipose tissue – 70ms Spinal fluid – 2200ms Gray matter – 100ms White matter – 90ms Muscles – 50ms TE=90ms T1 contrast TR 180 90 TE M Tissue with shorter T1 Tissue with longer T1 T1 Contrast Focus Minimize influence i.e. Short TE TR Intermediate T1 Long T1 Short T1 T1 contrast Adipose tissue – 240ms Spinal fluid – 4300ms Gray matter – 980ms White matter – 780ms Muscles – 880ms Examples T1 contrast TR=450ms Inversion-recovery (IR) TR 180 180 180 90 TI TE M Intermediate T1 Long T1 Short T1 IR IR Summary T1 contrast T2 contrast TE - Short TR – Optimized TE - Optimized TR – Long Inversion recovery TI - Optimized • Use for anatomical imaging • For pathology together with contrast agent • Use for pathology • Use for anatomical imaging Proton contrast Minimize infludence i.e. Long TR Minimize influence i.e. Short TE Focus Turbo spin echo Fast spin echo 180 180 180 90 Gradient echo (T2*) TR α α TE α α Gradient TR Gradient echo Small angle - reduces T1 weighting and yielding proton density weighting Large flip - yields T1 weighting Short TR - increases T2* weighting (residual transverse magnetization is dominant) Long TR - enhances T1 weighting Short TE - reduces T2* weighting and increases T1 or PD weighting Long TE - enhances T2* weighting What kind of contrast? A. T2w B. T1w C. T1w + contrast Different contrasts T2w T1w T1w+GD Why is it often difficult to see gold markers in T2w spin echo sequences of the prostate? A. T2w images do often have low signal poor signal to noise B. The gold gives the same signal as the prostate at long echo-times C. T2w images do often have a too large voxel size D. The artifact is often small in spin-echo sequences 25% 25% 25% 25% The gold gives the same ... T2w images do often hav... The artifact is often small .. T2w images do often have.. T2w T1w Gradient echo Summary again • T1 Weighting • Maximizing T1 short TR • Minimizing T2 short TE • Maximizing T2 long TE • Minimizing T1 long TR • Minimizing T2 short TE • Minimizing T1 long TR • T2 Weighting • Proton weighting Thank you MRI Physics: Space Encoding A/Prof Gary Liney 13 th September 2015 ESTRO Imaging for Physicists Introduction • MRI extremely flexible spatial localisation Orientation easily altered • Gradients used to modulate phase and frequency In-plane directions always ‘phase’ and ‘frequency’ • Signal is reconstructed with 2D or 3D Fourier Transformation Spin Echo Sequence 180 ° 90 ° RF G z G y Slice Selection Phase Encoding G x Signal Frequency Encoding TE An axial image.. Fourier Transform (FT) • Time signal can be decomposed into sum of sinusoids of different frequencies, phases and amplitudes • Fourier series may be represented by frequency spectrum • Time and frequency domain data can be thought of as FT pairs s(t) = a 0 + a 1 sin( ω 1 t + ϕ 1 ) + a 2 sin( ω 2 t + ϕ 2 ) + … Fourier Transform (FT) S1 has amplitude a and frequency f S2 has a /2 and 3 f S3 = S1 + S2 S3 is two sine waves of different frequency and amplitude The FT is shown 0.5 1.5 2.5 3.5 S1 S1 S2 S1 2 S3 3 -3.5 -2.5 -1.5 -0.5 A FT Pairs Delta ‘Top Hat’ FT Sinusoid Sinc FT Time Frequency FT Pairs Gaussian Lorentzian FT Gaussian Exponential FT Time Frequency Gradients B γ 0 ω = • Recall that the resonant frequency is proportional to field strength 0 dy dB G 0 dz dB G 0 dx dB G 0 = x • Magnetic gradient changes B 0 strength over distance • In MRI a linear gradient changes the resonant frequency in a given direction field = y ( ) x xG B + = = ω γ z 0 Slice Selection isocentre B 0 y z x B 0 ω 0 Slice Selection Gradient in z-direction G z isocentre B 0 y z x B 0 + ∆ B B 0 B 0 - ∆ B + ∆ ω ω 0 - ∆ ω ω 0 ω 0 Slice Selection • Gradient used to change resonant frequency in slice direction • Excite spins using ( sinc-shaped ) 90 ° RF pulse containing a bandwidth of frequencies • Only a particular section of spins are excited into transverse plane • Signal has been discriminated in one dimension • Can change orientation, slice thickness and position Slice Selection frequency G z G z RF pulse length/Bandwidth Centre frequency z Slice thickness Slice positions Slice Selection Frequency Encoding • Need to discriminate signal in-plane • Another gradient is used to produce changes in resonant frequency Phase Encoding • Need to still encode signal in remaining direction (y) Use changes of phase • When a gradient is applied the spins will be at different phases once the gradient has been turned off • This is the role of the phase encoding gradient (next) Phase Encoding ω = γ B 0 G phase G y Y-direction → time → Phase Encoding ω = γ (B 0 +yG y ) z G phase G y Y-direction → time → Phase Encoding Y-direction → ω = γ B 0 G phase G y time → Initially, all spins have same frequency Apply phase encoding gradient slower unchanged faster After PE Gradient turned off All spins have same frequency again, but different phase Faster unchanged slower Apply Frequency Encoding Gradient Phase Encoding • Each pixel is assigned a unique phase and frequency • FT decodes unique frequency but only measures summation of phase • Individual phase contributions cannot be detected • Need multiple increments of PE gradient to provide enough information about phase changes • Number of PE increments depends on image matrix Spin Echo Sequence 180 ° 90 ° Resonance condition ω = γ (B 0 + zG z ) RF G z G y G z G x z Spin Echo Sequence z Increment gradient after RF pulse and before read-out 180 ° 90 ° RF G y G z G y G x Spin Echo Sequence 180 ° 90 ° z Apply gradient during read-out RF G x G z G y G x Signal Scan Time • Frequency encoding done at time of echo • Phase encoding done over many TRs • Scan time is given by N PE × TR × NEX Multi-Slice Imaging • Period between the echo and the next RF pulse is called dead time • Used to excite a separate slice • Multiple slices are acquired in each TR • Slice profiles are not rectangular leading to cross-excitation • Slices are acquired with gaps or interleaved ‘3D’ MRI • True 3D volume rather than multiple 2D slices • A slab or multiple-slabs are selected • Phase encoding also in the ‘slice’ dimension Through-plane resolution can be comparable to in- plane Phase wrap in ‘slice’ direction • SNR is improved, scan time longer N PE × TR × NEX × N s What is k-space? • ‘k’ is wave-number: number of cycles per unit distance Spatial analogue to ‘cycles per second’ (frequency) • k-space is the raw data An array of numbers whose FT is the MR image • Each row in k-space corresponds to the echo data obtained from a single application of the PE gradient Rows near centre correspond to low-order PE steps (small gradients) Rows at edges correspond to high-order steps What is k-space? FT k-space and image-space of the brain What is k-space? k PE k FE Phase encoding increments 1 2 3 2 3 1 Frequency encoding gradient k-space • All of k-space needs to be filled to create an image Centre: bulk signal/contrast information Edge: image detail • Individual cells do not correspond one-to-one with individual pixels in image • Each cell has information about every image pixel: explains why motion artefacts propagate through whole image k-space k y FOV k x ∆ k FOV = 1/ ∆ k ∆ x = 1/FOV k k-space k y FOV k x ∆ k FOV = 1/ ∆ k ∆ x = 1/FOV k k-space Full k-space Centre k-space Edge k-space k-space trajectories k y k x GRE or SE : one line of k- space per TR (usually 256, 512 lines) Image time = N phase × TR EPI : all lines of k-space per TR (typically 64 or 128) Image time = TR k-space trajectories Radial k-space : Centre oversampled Motion compensation Partial k-space These techniques acquire part of k-space and ‘fill-in’ the rest due to conjugate symmetry k y Just over half data collected k x Partial Fourier : Also called fractional NEX Collects half of phase-encode steps and speeds up imaging Partial Echo : Collects half of echo reducing the shortest possible TE MRI Physics: Equipment A/Prof Gary Liney 13 th September 2015 ESTRO Imaging for Physicists Installation of New Scanner RF Cage • MRI inherently low (RF) signal technique • Faraday cage All 6 sides enclosed in copper Electromagnetic shielding Examples microwave oven, coax cable • Integrity must be maintained Penetration panel Mesh window, waveguide Closed scan room door, no fluorescent lights RF Cage Construction Waveguides Penetration Panel Mesh Window Door surround The MRI Controlled Area Quench pipe O 2 alarm 5 Gauss Line Control Room Pressure release hatch The Inner Controlled Area Scan Room 30 Gauss Cabinet (Equipment) Room RF Heat Exchanger Gradients MNS MRI Equipment: Overview Plus: Peripheral equipment RT Specific equipment Test Objects Patient Bore short & wide bore music internal lighting and ventilation panic button & intercom windings in cryostat shielding gradients vacuum detachable table RF body coil +shim coils Example Specifications Shielding Passive and active 0.2 (40 cm DSV) Homogeneity (ppm) Field stability (ppm/hr) <0.1 Cooling Liquid helium only Magnet specifications for Siemens Avanto 1.5 T Boil-off (l/hr) Helium refill Shim plates Active shim 0 10 years 16 x 15 3 linear terms (20 coils) 5 2 nd order (32 coils) Mass (tonnes) 5.5 2.5 4.0 Radial (x,y) 5 Gauss Axial (z) 5 Gauss Minimum area (m 2 ) <30 Example Specifications RF channels 8,18,32 Bandwidth (MHz) 1 RF & Gradient specifications for Siemens Avanto 1.5 T Gradient amplitude (mT/m) 33, 40, 45 Slew rate (mT/m/ms) 125-200 Host computer Memory (GB) 2 x Pentium IV 2 Hard drive (GB) 73 GB (images) Computer specifications for Siemens Avanto 1.5 T Image processor speed 2.2 GHz Reconstruction (ips 256 2 matrix) 1002 Magnet • Application Whole body (head only) & peripheral systems • Type Permanent, resistive, superconducting 1987: Elscint’s Gyrex System • Orientation Horizontal, vertical field • Design Tunnel-short & wide bore Open Philips’ vertical HFO System ‘Open-ness’ Dedicated systems Whole-body systems ..shorter & wider Field Strength ‘NMR’ systems Static Field (B 0 ) • Low sensitivity requires high field • 1 Tesla = 10,000 Gauss 0.3-0.7 G Earth’s field • Projectile effect • Mostly superconductors field decay: 5-10 G y -1 field stability: <0.1 ppm h -1 y z B 0 x Superconductors • Niobium-Titanium • Cryostat Double dewar with nitrogen/helium Cryoshielding helium only • Cryogens Cryostat Quench Pipe Replenish due to Boil off zero boil off/cryogen free • Quench Expensive & safety risk! Vent pipe, oxygen monitor Homogeneity Uniform imaging volume at isocentre Off-centre imaging? RT Planning? 40 cm e.g. DSV 40cm = 0.2 ppm (at 1.5 T): 0.2 x 63.87 MHz = 12.8 Hz http://en.wikipedia.org/wiki/File:Finite_Length_So lenoid_field_radius_1_length_1.jpg#/media/File: Finite_Length_Solenoid_field_radius_1_length_ 1.jpg • Magnet is shimmed at installation- additional (dynamic) shimming may be required Shim Demo Demo Real Time Demonstration of Shimming Fringe (stray) Field • Scanner ‘footprint’ > 30 G Stainless steel, non- ferromagnetic objects < 30 G ECG monitors, unrestrained ferromagnetic objects < 10 G Credit cards, x-ray tubes Credit cards erased at 10 G Safety limit is ‘five gauss line’ • 7 Tesla scanner has 23 m 5 G line Passive & Active shielding • Radial & axial components Typically axial 1.6 times larger • May be measured with handheld gaussmeter < 5 G Pacemakers, general public < 3 G Moving cars etc < 1 G TVs, CT & PET scanners < 0.5 G Railways, gamma cameras 5 G line with Active shielding + Passive shielding Gradients (db/dt) G z = dB 0 /dz isocentre • 3 orthogonal or in combination • High amplitudes -Resolution -DTI • Fast switching rates -Faster scans y z x B 0 ω 0 B 0 - ∆ B - ∆ ω B 0 + ∆ B + ∆ ω ω 0 ω 0 Gradients • Gradient waveform trapezoidal • Amplitude, Rise time, Slew rate e.g. 10-50 mT/m, 200 µ s & 20-150 T/m/s • Linearity Distortion for RT planning? GradWarp or similar in 2D/3D ? Max amplitude plateau Slew Rate (T/m/s) = Amplitude (mT/m) Rise Time ( µ s) Rise time Gradients Maxwell Pair Separation = r √ 3 Golay Coils Linear between central arcs Optimised inductance ‘fingerprint’ coils Gradients Manufacturer Field Strength (T) SPL (dB(A)) • Eddy currents degrade imaging pre-emphasis (compensation) Active shielding • Lorentz force causes vibrations Philips 1.5 1.5 1.5 3.0 3.0 112 106 110 118 113 Siemens GE Varian Bruker First Field Noise & reduction methods thuMb Motion seCond Current Peripheral Nerve Stimulation (PNS) cardiac • Faster switching = faster imaging • Stimulation real issue • Reilly estimates (right) • Solutions: Parallel imaging (‘Coil Encoding’) Twin gradients nerve RF Coils (B 1 ) • Coil Usage: Transmit and/or receive at resonance • Properties 36.5 36 Cable loss, loading 35.5 Core Skin Q factor ( ω / ∆ω ) Efficiency 1-Q L 35 33.5 Temperature ( o C) 34 34.5 /Q 0 Filling factor 33 32.5 • RF heating effects (SAR) 32 0 5 10 15 20 25 Time (min) RF Chain DAC Turns digital signal to analogue for RF transmission Double balanced mixer Produces amplitude modulated RF waveform RF power amplifier RF Coil(s) transmit/receive signal Pre-Amp Phase sensitive demodulator Removes RF waveform from detected signal Low Pass Filter ADC Digitisers signal to be processed by computer RF Coils: Signal Characteristics 2 a x a I π 2 0 µ ( ) 2 2 4 2/3 B = Theoretical cylinder coils + π SNR a Body Coil: Poor SNR Excellent uniformity surface coils Surface Coil: Excellent SNR close to coil Poor uniformity Distance Finite Element Modelling used for complicated designs RF Coil Designs • Surface coils • Cylindrical coils Sinusoidal currents around surface gives homogenous B 1 Saddle, birdcage (‘distributed capacitance’) with more conductors approximate this • Solenoid useful in vertical fields (Philips HFO) B 1 saddle surface birdcage solenoid RF Coils • Typical Scanner Configuration: Integrated body coil Head coils (linear for QA) Torso Coil Surface coil Specialist coils e.g. wrist, breast Coil Arrays • Extend surface coil coverage Small coil excellent SNR • Overlap to prevent mutual inductance • Separate Rx channels Noise not correlated, further increase SNR • Can be used in parallel imaging* * Covered in ‘fast scanning, volume sequences’ Quadrature Coils imaginary • Linear polarisation- only half RF power effective • Circularly polarised Orthogonal coils at 90 ° phase • Efficient transmission Power halved (RF heating) • Receiver coils SNR increases √ 2 real RF Coils: Other • Dual Tuned Multi-nuclear spectroscopy proton MRI & other MRS Broadband amplifier e.g. GE’s OpTix system Digitised in scan room, optical transmission SNR increase by 27% • Optical RF Chain B 1 Uniformity • Surface coil uniformity problematic (ER coil prostate) • Commercial correction methods (e.g. SCIC) • In-house method: PD-W image to divide out inhomogeneites original corrected Dielectric Effect • At 3T λ ≈ 26 cm, comparable to patient • Conductive/resonance effect ‘B 1 Doming’ • Dielectric pad, test objects Body imaging restored (right) • Dual transmit body coil RT Specific Equipment Increase in sophistication from ‘making do’ to dedicated equipment Couch: Flat table-top (?RF coil in table) Magnet: wide bore RF coils: Use of diagnostic and/or dedicated equipment External lasers in MRI room Associated devices- markers, MR applicators… RT Planning Scans MRI often compromised by available equipment Dedicated System (MR Simulator) Dedicated RF coils Laser bridge, Wide bore Flat table top, Coils in bed Flex coils (2 x 4), table coil (32) plus long cabled body coil (18) Improved SNR & coverage H&N The Future • Higher field strength • More RF channels • Increase in MR-Simulators • Hybrid MR-Linac systems 7 Tesla system 64 channel H&N coil The Australian MR-Linac Positron Emission Tomography Physics - Basic Principles ESTRO Teaching Course on Advanced Imaging Technologies September 13 – 17, 2015 in Leiden, NL Daniela Thorwarth Section for Biomedical Physics, University Hospital for Radiation Oncology, Tübingen Emission Tomography (PET) PET imaging adds molecular information to morphology and function imaged with CT and/or MRI. [ 68 Ga]DOTATOC PET/MR T2w MRI and PET (BrainPET) show small satellite in dorsal area of frontal sinus (detected on PET). Boss et al, JNM 2010; 51. 2 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Basic principle of PET Positron emitters (β+) used as biomarkers Positron-electron annihilation ⇒ Two γ-quanta with 511 keV each are emitted under approx. 180 ° Coincidence detection in a detector ring Radioactive Isotope 511 keV γ-quant γ 1 β + Electron γ 2 511 keV γ-quant PET 3 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Today: Combined PET/CT PET/CT: combination of structure and function First clinical PET/CT prototype (mid 1990s) [3] 4 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles State-of-the-art PET/CT Designs Gemini series, Philips Healthcare Systems Biograph series, Siemens Healthcare Solutions Discovery series, GE Healthcare Aquiduo series, Toshiba Medical Systems Sceptre series, Hitachi Medical Systems Anyscan series, Mediso ! All PET/CT tomographs combine diagnostic PET and CT components and a dedicated patient support system. ESTRO Teaching Course Advanced Imaging: PET – Basic Principles 5 ! Courtesy T. Beyer, PET/CT 6 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Sinograms: Measurement of the activity distribution of a radioactive tracer. Raw data stored in sinograms φ y = φ line of response (LOR) x = offset 7 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Radial Sampling Mapping from sampling projections to sinograms [2]. Transaxial field of view of a PET tomograph is defined by the acceptance angle in the plane. 8 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles 2D-/3D-PET 2D-PET • Geometric collimation with septa 3D-PET • Projections at polar angles θ>0° measured • Increased sensitivity • Higher scatter fraction • Special reconstruction algorithms are necessary • Data sampling only with θ=0° • Lower overall sensitivity • Lower fraction of scattered photons 9 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles 2D/3D-PET acquisition 2D acquisition: the entire FOV is sampled. 3D acquisition: truncation of projections occurs for θ > 0°. This results in loss of data corresponding the ends of the tomograph [2]. 10 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Axial Sensitivity Restriction of max. acceptance angle 2D axial sensitivity profile for a line source in air in a 16-ring tomograph [2]. Axial sensitivity variation for 3D acquisition in a 16-ring tomograph [2]. 11 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Image Formation (1) Emission scan (2) Normalization scans (one per plane in 2D) to correct for differential detector efficiencies and geometric effects related to the detector ring (3) Set of sinograms of attenuation correction factors to correct of photon attenuation (self- absorption or scattering) by the object 12 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Radiation detection Desired for PET: 1. High stopping efficiency 2. Good energy resolution Scintillation detectors • Inorganic crystal that emits visible light photons after interaction of photons with detector. • # of scintillation photons is proportional to the energy deposited in the crystal. • Important properties for application in PET: • Stopping power for 511 keV photons • Signal decay time • Light output • Intrinsic energy resolution 13 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles applied in PET NaI(Tl): sodium iodide deoped with thallium BGO: bismuth germanate (Bi 4 Ge 3 O 12 ) LSO: lutetium oxyorthosilicate doped with cerium(Lu 2 SiO 5 :Ce) YSO: yttrium oxyorthosilicate doped with cerium(Y 2 SiO 5 :Ce) GSO: gadolinium oxyorthosilicate doped with cerium(Gd 2 SiO 5 :Ce) ESTRO Teaching Course Advanced Imaging: PET – Basic Principles New: LYSO 14 Photo-multiplier tubes (PMTs) PMTs: photo-detectors used in scintillation detectors for PET (1) Incoming photon deposits its energy at the photocathode, release of a photo-electron (2) Applied electric field accelerates the electron to the first dynode (3) Emission of multiple secondary electrons due to increased electron energy (4) … Good signal-to-noise ratio (SNR) Low quantum efficiency (QE) ~ 25% 15 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Detector Designs used in PET One-to-one coupling: • Single crystals glued to individual photo-detector Spatial resolution limited by discrete crystal size • Block detector design: • Rectangular scintillator block sectioned by partial saw cuts of different depth into discrete elements Usually 4 attached PMTs Standard block detector design scheme (from [2]) • • Anger positioning 16 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Detector Designs used in PET Anger detector: • Large scintillator crystal glued to array of PMTs Weighted centroid positioning algorithm used to estimate interaction position within the detector • Block detector Siemens-CTI ECAT 951, 8x8 block BGO with 4 PMTs (from [2]) Block detector system + Anger logic [3] 17 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Detection Coincidence time window: 2 τ 18 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Question: Timing Resolution 19 ESTRO Teaching Course Advanced Imaging: PET I Answer: Timing Resolution 20 ESTRO Teaching Course Advanced Imaging: PET I Time-of-flight (TOF) PET • In addition to the coincidence information, the difference in flight time is registered =∆ 2 x ∆ t c • Better SNR • Example: Δt=500ps Δx=? =∆ 5.7 5.0 t c ∆⋅ ⋅ cm = (a) Localization without TOF, (b) with TOF. Townsend, PMB 2008 [3] 21 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Question: Time-of-flight Assuming the detector system of a PET scanner has a timing resolution of 500 ps. How long is the corresponding sector on a LOR, where the annihilation event was located? 1. 75 cm 2. 15 cm 3. 7.5 cm 4. 1.5 cm 22 ESTRO Teaching Course Advanced Imaging: PET I Question: Time-of-flight Assuming the detector system of a PET scanner has a timing resolution of 500 ps. How long is the corresponding sector on a LOR, where the annihilation event was located? 1. 75 cm 2. 15 cm 3. 7.5 cm 4. 1.5 cm 23 ESTRO Teaching Course Advanced Imaging: PET I Detected Events in PET Detection event is valid (= prompt event) if ● Two photons are detected in coincidence window ● LOR is within valid acceptance angle ● Energy of both photons within selected energy window B A C τ ≈10 ns 2 True coincidence (A) , scattered coincidence (B) , random coincidence (C) 24 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Prompt Events Single event ● single photon is counted by detector (1-10%) True coincidence ● event derives from single positron-electron annihilation ● both photons reach tomograph without interaction Random coincidence ● two nuclei decay at approximately the same time ● random event count rate (R ab ) between two detectors a and b : 2 ∝ ⋅ N NN R b a 2 = τ ab 25 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Prompt Events (II) Scattered events ● One or both photons detected have undergone a Compton interaction ● Loss in energy and change in direction ● Due to poor energy resolution, many scattered photons cannot be discriminated ● Wrong LOR assigned Multiple (triple) events ● Three events from two annihilations detected ● Event is disregarded ● Proportional to count rate 26 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Performance of PET Systems Sensitivity counts At c S = ⋅ = sec kBq ⋅ ● Good systems reach S=7-9 counts/(sec.kBq) Spatial Resolution Energy Resolution Count Rate Performance Scatter Fraction 27 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Spatial Resolution Determined by full width half maximum (FWHM) of point spread function (PSF): )2 ln( 8 σ = FWHM 28 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Energy Resolution Statistical uncertainty of energy determination due to limited light yield of scintillator crystal Two methods for determination of energy resolution: ● Single event energy resolution ● Coincidence (i.e. both events) energy resolution FWHM = 16.4/19.6/21.6% Energy spectra for single photons for a BGO PET system [2]. 29 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Count Rate Performance Ratio Trues/Randoms unbalanced for high Activities Processing of detected photons takes finite time Noise Equivalent Count Rate: T 2 T : trues S : scattered R : randoms f : rel. area of object on proj. surface NECR = 2 ++ fR ST Count rate curves for true, random and multiple coincidences. Derived curves for expected and noise equivalent count rate [2]. Data recorded on a CTI ECAT 953B PET Maximum image quality Low image contrast using a 11 C-filled cylinder in water. Noisy image 30 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Scatter Fraction Fraction of the total coincidences recorded in the photopeak window that have been scattered Sources of scattering ● Scattering within the object containing the radionuclide ● Scattering off the gantry components (lead septa/side shields) ● Scattering within the detectors ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Log-lin count rate profiles of line source in air/water show additive scatter component outside of central peak [2]. 31 Scatter Fraction Fraction of the total coincidences recorded in the photopeak window that have been scattered Sources of scattering ● Scattering within the object containing the radionuclide ● Scattering off the gantry components (lead septa/side shields) ● Scattering within the detectors Scatter fraction can be reduced by ● TOF ● Usage of a powerful iterative reconstruction method ● Shielding of scattered photons by septa and endshiels ● Small coincidence window ● Good energy resolution 32 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Summary • Positron emitter used as radioactive tracer in the patient • Use of collimators to generate ‚image slices‘ • Detection of γ-quanta in scintillation crystal • Better resolution due to Anger logic • Signal amplification using PMTs • Image reconstruction Literature [1] DL Bailey, JS Karp, S Surti. Physics and Instrumentation in PET. In: Positron Emission Tomography: Basic Science and Clinical Practice. Editors: PE Valk, DL Bailey, DW Townsend, MN Maisey. Springer London 2003, pp. 13-39. [2] DL Bailey. Data Acquisition and Performance Characterization in PET. In: Positron Emission Tomography: Basic Science and Clinical Practice. Editors: PE Valk, DL Bailey, DW Townsend, MN Maisey. Springer London 2003, pp. 41-61. [3] DW Townsend. Multimodality imaging of structure and function. Phys Med Biol 2008; 53: R1-R39. [4] B Sattler, JA Lee, M Lonsdale, E Coche. PET/CT (and CT) instrumentation, image reconstruction and data transfer for radiotherapy planning. Radiother Oncol 2010; 96: 288-297. Review. 34 ESTRO Teaching Course Advanced Imaging: PET – Basic Principles Imaging for Physicists Artifacts 1 Uulke van der Heide Artifacts in MRI Artifacts in MRI • An image artifact is any property or effect observed in an image that does not appear in the original object • Images can be distorted in many ways – Signal loss – Deformations – Poor resolution – Ghosting – Aliasing – ….. • Consequences for use – Interpretation is difficult – Geometrical accuracy may be compromised Outline Lecture 1 • Origin of geometrical artifacts – Fold-over artifacts – Ringing – Impact of field distortions Lecture 2 • Measurements for characterizing geometrical accuracy – Phantom design – Characterizing gradient errors • Examples • Practical consequences • Summary Origin of various artifacts • Sampling of k-space – Sample k-space in too large steps – Don’t sample high k-values • Magnetic field errors – Inhomogeneous B 0 – non-linear gradients – Susceptibility – Chemical shift field • Motion Imaging artifact • T 1 -weighted SE image of a brain • What is wrong? Sampling the MR signal • Nyquist criterium: signal must be sampled at at least twice the rate of the highest frequency component Sampling the MR signal • Nyquist criterium: signal must be sampled at at least twice the rate of the highest frequency component is not > 2f 2 F sampling • If the signal contains higher frequency components, aliasing occurs Resolve aliasing by increasing sampling frequency • Nyquist criterium: signal must be sampled at at least twice the rate of the highest frequency component • By increasing the sample frequency, the higher frequency components can be resolved and aliasing is avoided Field Of View covers entire object: no fold-over k y k x FOV Question: field of view • What happens with the distance between lines in k-space, if you reduce the field of view by a factor of 2? 1. The distance between k-lines is increased by a factor of 2 2. The distance between k-lines is reduced by a factor of 2 3. The distance between k-lines remains the same (but the extent of k-space is reduced by a factor of 2 Field Of View covers entire object: no fold-over k y k x 1/FOV FOV FOV too small: fold-over k y k x 1/FOV FOV How to suppress fold-over artifacts? k y k x 1/FOV FOV • If NSA>1: Measure all k-lines (full FOV), but reconstruct only half of the image How to suppress fold-over artifacts? k y k x 1/FOV FOV • If NSA>1: Measure all k-lines (full FOV), but reconstruct only half of the image How to suppress fold-over artifacts? k y REST slab k x 1/FOV FOV REST slab • If NSA=1: Saturate the signal from outside the field-of-view with REST slabs Saturate signal from outside FOV k y k x 1/FOV FOV Imaging artifact • T 1 -weighted SE image of a brain • What is the difference between the left and right image? Imaging artifact • T 1 -weighted SE image of a brain • What is the difference between the left and right image? Truncation errors (ringing) • Imaging sharp edges requires high frequency components in k-space • Cutting off the high-frequency k-lines causes oscillations in the image
Made with FlippingBook