Transcript Presentation on presenting - The College of Engineering at the

Bio-Electromagnetic Modeling:
Challenges and Observations
MILICA POPOVIĆ
D E P A RT M E N T O F E L E C T R I C A L A N D C O M P U T E R E N G I N E E R I N G
MCGILL UNIVERSITY
M ONTRÉAL, CA NADA
McGill University
James McGill (1744 – 1813)
2
• McGill’s first structure, the
Arts Building was completed
in 1843 and still serves as
the focal point of the
downtown campus.
 Redpath Museum, commissioned in 1880
and opened in 1882, is the oldest building
built specifically as a museum in North
America. Its natural history collections boast
material collected by the same individuals
who founded the collections of the Royal
Ontario Museum and the Smithsonian.
Research Support
3
 McGill University
 Natural Sciences and Engineering
Research Council of Canada – NSERC
 Le Fonds Québécois de la recherche
sur la nature et les technologies
Research Group
4
 Houssam Kanj, PhD (June 2008)
 Amir Hajiaboli, PhD (June 2009)
 Guangran Zhu (Kevin), PhD candidate
 Emily Porter, M. Eng program
 Zahra Al-Roubaie, M.Eng. (October 2008)
 Yi Zhang, M.Eng. (October 2007)
 Negar Tavassolian, M. Eng. (October 2006)
 Chun Yiu Chu, M. Eng. (October 2005)
 Lawrence Duong, M. Eng. (October 2005)
 Nicholas Yak, M. Eng. (October 2005)
 Qingsheng Han (Ted), M. Eng. (October 2004)
Outline
5
 Motivation
 Methods
 Challenges
 Results
Detecting breast cancer with microwaves
6
• In the microwave frequency range, tumors have a
higher water content than the surrounding fatty
tissue.
 tumors are “visible” to microwaves
• The losses of microwave propagation in the fatty
tissue are low (< 4dB/cm).
 low-power microwave signal can penetrate
through the fatty tissue without diminishing too
quickly
Tumor vs. breast tissue:
electrical properties in the microwave range
7
• Older data 
Parameters
at 6GHz
Fat
Tumor
Relative
permittivity εr
Electric
conductivity σ
(S/m)
9
0.4
50
7
• Recent data: [Lazebnik et al, 2007, Physics in
Medicine and Biology] :
Contrast still exists, but is much lower. (~10%)
Microwave-tissue interaction
8
 Tumor causes microwave scatter
 Tumor absorbs microwave energy
 Techniques:




Microwave tomography
Radar-like pulse imaging (backscatter)
Microwave-induced thermo-acoustic
Acoustic source, detecting Doppler
shift with microwaves
Challenges in antenna design
9
 Signal: pulse centered




around 6 GHz
Antenna transmits and
receives all key Fourier
components of the
microwave-centered pulse
 broadband operation
Simple to manufacture
Cost-effective
Small in size
Antenna design I
10
•
•
•
a = c = 5mm
d = 35mm
Rs varied between 50 & 800 /□
Antenna design II
11
• Compact:
34.25mm × 20mm × 1.3mm
• Thin film resistive loading is
used (standard values) 
• Repeatable
• Accurate
• Design was optimized for:
• Return loss
• Efficiency
• Fidelity
Antenna design III
12
• TWTLTLA =
•
•
Traveling Wave Tapered & Loaded
Transmission Line Antenna
Merge the guiding structure with
the radiating element to form a
single transitional structure
Use resistive loading to achieve
traveling wave characteristics
Antenna design - final
13
Uniplanar and very compact:
14mm × 17.5mm
Low profile
Microstrip
Antipodal
Tapered GND
Ultra Broadband
Easy transition to co-ax
Antenna design - final
14
Computational challenges: example
15
 Finite-difference time-domain (SEMCAD-X) simulations
 Realistic, MRI – based, detailed numerical models
 huge FDTD mesh
 computational cost
Computational challenges: example
16
•
•
•
To decrease the problem size:
we use the regression tree analysis method
(with the user-defined penalty factor)
we cluster the voxels of similar property into
bigger brick-like solids
result: manageable computational problem
with sufficient representation of anatomical
complexity
Computational challenges: example
17
Electrical permittivity distribution, 6 GHz
Direct, manual mapping
from MRI voxels
Simplified model through
regression tree algorithm
Computational challenges: example
18
Electrical conductivity distribution [S/m], 6 GHz
Direct, manual mapping
from MRI voxels
Simplified model through
regression tree algorithm
Computational challenges: example
19
The difference in the fields computed near the
antennas for the finer and the brick-approximated
model was very, very small!
Microwave-induced thermo-acoustic
20
2-D model of the central MRI-derived horizontal slice
Microwave-induced thermo-acoustics
21
 Faraday’s Law
 Simple force Equation
H
1
   E
t

 Ampere’s Law
u
1
  p
t
o
 Continuity of Mass
E 1
 (  H  J   E )
t 
p

 K (  u  SAR  ap)
t
c
2-D Quantities : duality
Ex
p
Hy
Hz
uz
u y


 1/ K

a
Jx
SAR
Microwave-induced thermo-acoustics
22
 Challenge: multi-physics modeling
 The time constants of the microwave, thermal
and acoustic processes are very different
 Acoustic properties of tissue (literature)
– not very recent
 If we want to build a phantom model for
experimental verification, challenge is to find
materials that can mimic the tissues both in the
electrical and in the acoustic sense, simultaneously
Small break, one of my favorite quotes:
23
I keep six honest serving-men
(They taught me all I knew);
Their names are What and Why and
When And How and Where and Who.
Rudyard Kipling
The Elephant's Child (1902)
Indian-born British author (1865 - 1936)
Human Eye and the Retina
24
http://thefutureofthings.com/articles/57/shedding-light-on-blindness.html
Retina
25
 Does the morphology of the
photoreceptor outer-segment
affect its optical filtering
properties?
 Does the geometry of a
photoreceptor help
discriminate between different
wavelengths?
From: Neurobiology: Bright blue times,
Russell G. Foster. Nature February 2005
Retinal rods and cones
26
Magnified image of the rods and cones of the human eye. © Omikron.
Reproduced by permission of Photo Researchers, Inc.
www.faqs.org/health/Sick-V1/Color-Blindness.html
Rods and cones: tri-chromatic color vision
27
Spectral sensitivities of the three cone types and the rod.
The tri-chromatic color vision theory was first introduced by Young in 1802.
Retinal rods
28
Photoreceptor
biochemistry:
Hyperpolarization by
rhodopsin
decomposition
Retinal rods and cones: outer-segment
29
Alan Fein and Ete Z. Szuts , Photoreceptor: Their Role in Vision, Cambridge Univ. Press 1982
Cone: outer-segment FDTD model (I)
30
 L=6.75µm
outer-segment
length
 The disks’
radius
decreases along
the central axis
of the
photoreceptor
τ1=τ2=15 nm
D1=4.1µm
D2= variable
Cone: outer-segment FDTD model (II)
31
εr-cytoplasm=1.85
εr-membrane=2.01
εr-intercellular=1.79
membrane = 37.38 S/m
Challenges:
• Reliable electrical
property values
• Our plans: to include
dispersion in future work
Cone FDTD model: computational challenge
32
Cell size
Δx=5 nm
Δy=Δz= 30nm
Time step
Δt=1.46×10-2 fs
Boundary
condition
10 layers Uniaxial
perfectly matched
layer
Total FDTD space
1500×200×200=
60 Mcells
 FDTD computation accelerated with a GPU which is
optimized for parallelizing and memory bandwidth.
Cone FDTD model: light excitation
33
E y  Ae

( t-t0 ) 2
2σ 2
sin ( 2πf(t  t 0 ))
Type
Gaussian modulated
Polarization
Y-polarized
Modulation
frequency
f=622THz
Time delay
t0=400Δt
Amplitude
A=0.2V/m
Mean variation
σ=100 Δt
|Ey| in the outer-segment
t=21.9fs, 43.8fs, 87.6fs, 131.1fs.
34
D2=4.1μm
|Ey| in the outer-segment
t=21.9fs, 43.8fs, 87.6fs, 131.1fs.
35
D2=1.2μm
Energy spectrum vs. the free-space wavelength
36
The results are normalized to the maximum value of energy spectrum at D2=4.1μm.
Bulk (averaged permittivity) cone model
37
Bulk (averaged permittivity) cone model
38
Cone illumination for varying angle of incidence
39
Φ=π/45
Poynting power distribution across SF
Cone illumination for varying angle of indicence
40
Φ=π/12
Poynting power distribution across SF
Cone illumination for varying angle of incidence
41
Φ=π/10
Poynting power distribution across SF
What is the total power available
to the photo - pigment molecules?
42
Laminar (folded membrane layers)
vs
bulk (averaged,) structure
Thank you!
43
Questions? Comments?
[email protected]
Energy Spectrum Calculation
V(x, t) 
D/ 2
 E (x, y, z  0, t)dy
y
 D/ 2
Induced voltage at each time step
after 43.8fs along outer-segment
V(x, f)  F{V ( x, t )}
Induced voltage
versus frequency
S(x) 
1081THz
S (v)  F {S ( x)}
211THz
Signal energy
over 211-1081THz
2
|
V
(
x
,
f
)
|
df

Energy spectrum
versus spatial frequency
University of Adelaide, Adelaide, Australia
44