Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 ·...

102
Photo-acoustic Imaging Vincent Jugnon September 7, 2012

Transcript of Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 ·...

Page 1: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Photo-acoustic Imaging

Vincent Jugnon

September 7, 2012

Page 2: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

1 Photo-acoustic ImagingImposed boundary conditionsView limitationAcoustic attenuationCluttered sound speedqPAT : an inverse problem for light propagation

2 Wave imaging and topological derivative

Page 3: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The photoacoustic effect

physical phenomenon observed when a medium is exposed to anelectromagnetic wave.the EM wave drops energy in the medium, which heats up, dilatesand emits an (ultra-sonic) acoustic wave.

Mary Evans Picture Library

Photoacoustic imaging

The sound of lightJun 4th 2009From The Economist print edition

Biomedical technology: A novel scanning technique that combines optics with ultrasound couldprovide detailed images at greater depths

IF LIGHT passed through objects, rather than bouncing off them, people might now talk to each other on“photophones”. Alexander Graham Bell demonstrated such a device in 1880, transmitting a conversation on abeam of light. Bell’s invention stemmed from his discovery that exposing certain materials to focused,flickering beams of light caused them to emit sound—a phenomenon now known as the photoacoustic effect.

It was the world’s first wireless audio transmission, and Bell regarded the photophone as his most importantinvention. Sadly its use was impractical before the development of optical fibres, so Bell concentrated insteadon his more successful idea, the telephone. But more than a century later the photoacoustic effect is making acomeback, this time transforming the field of biomedical imaging.

A new technique called photoacoustic (or optoacoustic) tomography, which marries optics with ultrasonicimaging, should in theory be able to provide detailed scans comparable to those produced by magnetic-resonance imaging (MRI) or X-ray computerised tomography (CT), but with the cost and convenience of ahand-held scanner. Since the technology can operate at depths of several centimetres, its champions hopethat within a few years it will be able to help guide biopsy needles deep within tissue, assist withgastrointestinal endoscopies and measure oxygen levels in vascular and lymph nodes, thereby helping todetermine whether tumours are malignant or not. There is even scope to use photoacoustic imaging tomonitor brain activity and gene expression within cells.

To create a photoacoustic image, pulses of laser light are shone onto the tissue being scanned. This heats thetissue by a tiny amount—just a few thousandths of a degree—that is perfectly safe, but is enough to cause thecells to expand and contract in response. As they do so, they emit sound waves in the ultrasonic range. Anarray of sensors placed on the skin picks up these waves, and a computer then uses a process of triangulationto turn the ultrasonic signals into a two- or three-dimensional image of what lies beneath.

The technique works at far greater depths (up to seven centimetres) than other optical-imaging techniquessuch as confocal microscopy or optical-coherence tomography, which penetrate to depths of only about amillimetre. And because the degree to which a particular wavelength of light is absorbed depends on the type

Economist.com http://www.economist.com/science/tq/PrinterFriendly.cfm?story_id=1...

1 of 3 15/06/2009 09:46

Page 4: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 5: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 6: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 7: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 8: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Interest for medical imaging

advantages of both ultrasonic imaging and pure optical imagingwithout their drawbacks

fast, cheap and harmless

Page 9: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The considered model

∂2p∂t2 (x , t)− c2

0∆p(x , t) = 0

p(x ,0) =β

CpA(x)

∂p∂t

(x ,0) = 0

where A(x) the ”instantaneous” energy deposition of the EM wave inthe medium, will be modeled later.

Page 10: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The inverse problem

Reconstruct :p(x ,0) = p0(x) =

β

CpA(x)

from boundary measurements∂p∂ν

(y , t) or p(y , t), y ∈ ∂Ω, t ∈ [0 T ].

We assume that we have ideal point-like broadband acoustic receivers,with enough spatial density, so that we can neglect the error in :∫ T

0

∫∂Ω

f (x , t)dσ(x)dt ≈ ∆x∆t∑

i

∑j

f (xi , tj)

Page 11: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The inverse problem

Reconstruct :p(x ,0) = p0(x) =

β

CpA(x)

from boundary measurements∂p∂ν

(y , t) or p(y , t), y ∈ ∂Ω, t ∈ [0 T ].

We assume that we have ideal point-like broadband acoustic receivers,with enough spatial density, so that we can neglect the error in :∫ T

0

∫∂Ω

f (x , t)dσ(x)dt ≈ ∆x∆t∑

i

∑j

f (xi , tj)

Page 12: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The problem has been well studied in the case of :free spacefull view measurementsnon-attenuating mediumhomogeneous sound speed.

Page 13: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The problem has been well studied in the case of :free spacefull view measurementsnon-attenuating mediumhomogeneous sound speed.

Page 14: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The problem has been well studied in the case of :free spacefull view measurementsnon-attenuating mediumhomogeneous sound speed.

Page 15: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The problem has been well studied in the case of :free spacefull view measurementsnon-attenuating mediumhomogeneous sound speed.

Page 16: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 17: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

We can express the pressure p on the boundary of the medium Ωin terms of spherical means of p0.We can then use a wide range of inversion formulas for thespherical Radon transform to reconstruct p0.

Reconstruction using inverse spherical Radon approachTrue initial condition

Page 18: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

1 Photo-acoustic ImagingImposed boundary conditionsView limitationAcoustic attenuationCluttered sound speedqPAT : an inverse problem for light propagation

2 Wave imaging and topological derivative

Page 19: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 20: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 21: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 22: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The free-space hypothesis implies that ultrasonic waves ”do notsee” the boundary of the observed medium.In some imaging devices, we have to account for acousticboundary conditions :

1c2

0

∂2p∂t2 (x , t)−∆p(x , t) =

1c2

0δ′(t)p0(x)

p(x ,0) =∂p∂t

(x ,0) = 0

p(y , t) = 0 y ∈ ∂Ω

with boundary measurements∂p∂ν

(y , t) on some time interval(0,T ).

Page 23: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 24: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 25: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

we cannot express the measurements in terms of sphericalmeans anymore.

True initial condition Reconstruction using inverse spherical Radon approach

in fact, a duality approach allows the same quality ofreconstruction as in the free space case.

Page 26: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Consider the solutions of the wave equation in the free-space :

1c2

0

∂2v∂t2 (x , t)−∆v(x , t) = 0

with final conditions :

v(x ,T ) =∂v∂t

(x ,T ) = 0 x ∈ Ω

Page 27: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Testing the measures against those ”probe functions”, we get :∫ T

0

∫∂Ω

∂p∂ν

(y , t)v(y , t)dσ(y) =

∫Ω

p0(x)∂v∂t

(x ,0)dx

Page 28: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Consider plane wave test functions of the form :

v(x , t ; θ, s) = δ

(x .θc0

+ s − t)

Compute : ∫ T

0

∫∂Ω

∂p∂ν

(y , t)v(y , t)dσ(y) =∂

∂sR[p0](s, θ)

Page 29: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Now, choosing plane test functions of the form :

v(x , t ; θ, s) = 1− H(

x .θc0

+ s − t)

where H is the Heaviside function. we will have :∫ T

0

∫∂Ω

∂p∂ν

(y , t)v(y , t)dσ(y) = R[p0](s, θ)

In both cases, it is straightforward to obtain p0 using classicalback-projection algorithms.

p0 = R?BR[p0]

Page 30: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 31: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

1 Photo-acoustic ImagingImposed boundary conditionsView limitationAcoustic attenuationCluttered sound speedqPAT : an inverse problem for light propagation

2 Wave imaging and topological derivative

Page 32: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 33: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 34: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

We still consider the wave equation :∂2p∂t2 (x , t)− c2

0∆p(x , t) = 0

p(x ,0) = p0(x)∂p∂t

(x ,0) = p1(x)

p(y , t) = 0 y ∈ ∂Ω

and we want to reconstruct p0,p1 from the measurements :∂p∂ν

(x , t) x ∈ Γc .

Page 35: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Applying the previous algorithm to limited-view data fails.

−1 −0.5 0 0.5 1

−1

−0.5

0

0.5

1−1 −0.5 0 0.5 1

−1

−0.5

0

0.5

1

−1 −0.5 0 0.5 1

−1

−0.5

0

0.5

1−1 −0.5 0 0.5 1

−1

−0.5

0

0.5

1

Page 36: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Once again we are interested in obtaining integral information :

M(v0, v1) =

∫Ω

p0(x)v1(x)− p1(x)v0(x)dx

for example we have :

M(

0, δ(

x .θc0

+ s))

= R[p0](s, θ)

M(

0,eik .x)

= F [p0](k)

and evenM (0, δ(x)) = p0(x)

Page 37: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Once again we are interested in obtaining integral information :

M(v0, v1) =

∫Ω

p0(x)v1(x)− p1(x)v0(x)dx

for example we have :

M(

0, δ(

x .θc0

+ s))

= R[p0](s, θ)

M(

0,eik .x)

= F [p0](k)

and evenM (0, δ(x)) = p0(x)

Page 38: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

If we can find gv0,v1 such that v solution of :

1c2

0

∂2v∂t2 (x , t)−∆v(x , t) = 0

v(x ,0) = v0(x),∂v∂t

(x ,0) = v1(x)

v(y , t) = 0 y ∈ ∂Ω\Γcv(y , t) = gv0,v1 (y , t) y ∈ Γc

(1)

vanishes at time T :

v(x ,T ) = 0,∂v∂t

(x ,T ) = 0

Then we get what we want :∫ T

0

∫Γc

gv0,v1 (y , t)∂p∂ν

(y , t)dσ(y)dt = M(v0, v1)

Page 39: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

If we can find gv0,v1 such that v solution of :

1c2

0

∂2v∂t2 (x , t)−∆v(x , t) = 0

v(x ,0) = v0(x),∂v∂t

(x ,0) = v1(x)

v(y , t) = 0 y ∈ ∂Ω\Γcv(y , t) = gv0,v1 (y , t) y ∈ Γc

(1)

vanishes at time T :

v(x ,T ) = 0,∂v∂t

(x ,T ) = 0

Then we get what we want :∫ T

0

∫Γc

gv0,v1 (y , t)∂p∂ν

(y , t)dσ(y)dt = M(v0, v1)

Page 40: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Such a control exists :

Theorem J.L. LionsIf T and Γc satisfy controllability conditions, then for all (v0, v1) inL2(Ω)× H−1(Ω), there exists a control gv0,v1 solution of this controlproblem.

+ constructive proof.

Page 41: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Sharp controllability conditions

Theorem Bardos, Lebeau, RauchThe system is controllable if every ray of geometrical optics, starting at anypoint x ∈ Ω, at time t = 0, hits Γc before time T at a nondiffractive point.

⇒ No “glancing” nor “trapped” rays.

Uncontrollable and controllable geometries

Page 42: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The approach is computationaly expensive : O(n2d+1x ), but highly

parallelizable.

0 0.2 0.4 0.6 0.8 1

0

0.2

0.4

0.6

0.8

1

20 40 60 80 100 120 140 160

20

40

60

80

100

120

140

160

Theoretically, it is possible to get the same reconstruction as in the fullview case.

Page 43: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The approach is computationaly expensive : O(n2d+1x ), but highly

parallelizable.

0 0.2 0.4 0.6 0.8 1

0

0.2

0.4

0.6

0.8

1

20 40 60 80 100 120 140 160

20

40

60

80

100

120

140

160

Theoretically, it is possible to get the same reconstruction as in the fullview case.

Page 44: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

1 Photo-acoustic ImagingImposed boundary conditionsView limitationAcoustic attenuationCluttered sound speedqPAT : an inverse problem for light propagation

2 Wave imaging and topological derivative

Page 45: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 46: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 47: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

We have to consider the attenuated wave equation :

1c2

0

∂2pa

∂t2 (x , t)−∆pa(x , t)−D(t) ∗ pa(x , t) =1c2

0δ′(t)p0(x)

where D takes the form :

D(t) =1√2π

∫R

(K 2(ω)− ω2

c20

)eiωtdω

Page 48: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

If not accounted for, attenuation generates a blurring in thereconstruction.

True initial condition

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Reconstruction using inverse spherical Radon on attenuated data

0

0.05

0.1

0.15

0.2

0.25

0.3

Page 49: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

We propose an algorithm to deal with a power-law attenuation modelof exponent 2 :

K (ω) =ω

c0+ ia|ω|2

which is close to the thermo-visquous model :

K 2(ω) =ω2

c20

11− aiω

at low frequencies.

Page 50: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

We link pa(x , t) to p(x , t) (solution with no attenuation) via anintegral operator. Inverting this operator, we get p and using usualalgorithms, we reconstruct p0.The inversion of the operator is achieved using an asymptoticexpansion (w.r.t. to a) given by a stationary phase theorem. Wewill write :

Lφ = Lkφ+ o(ak+1)

and :L−1ψ = L−1

k ψ + o(ak+1)

Page 51: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Going in the Fourier domain, we can write :(∆ + K 2(ω)

)pa(x , ω) =

iωc2

0p0(x)

(∆ +

ω2

c20

)p(x , ω) =

iωc2

0p0(x)

Green’s functions are known for both equations, so we get :

pa(x , t) =1

∫R

ω

c0K (ω)e−iωt

∫ ∞0

p(x , s)eic0K (ω)sdsdω

Page 52: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

We want to invert the operator :

Lφ(t) =1

∫R

ω

c0K (ω)e−iωt

∫ ∞0

φ(s)eic0K (ω)sdsdω

Page 53: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

We have K (ω) = ωc0

+ iaω2

2 and :

ω

c0K (ω)= 1− i

ac0

2ω +

∞∑k=2

ik (−1)k(ac0ω

2

)k

Lφ(t) =1

∫ ∞0

φ(x , s)

∫R

(1− i

ac0

2ω +

∞∑k=2

(−iω)k(a

2

)k)

eiω(s−t)e−12 c0aω2sdωds

=

(1 +

ac0

2∂

∂t+∞∑

k=2

(−

ac0

2

)k ∂k

∂tk

)(1√

∫ ∞0

φ(x , s)1

√c0as

e− 1

2(s−t)2

c0as ds

)

Page 54: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

We have to investigate the operator :

Mφ =1√2π

∫ ∞0

φ(x , s)1√

c0ase−

12

(s−t)2

c0as ds

Page 55: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Stationary phase theoremFor f and ψ sufficiently smooth, if :

Im(f (t0)) = 0 f ′(t0) = 0 f ′′(t0) 6= 0 f ′ 6= 0 in K\t0

then for ε > 0, we have the following approximation :∣∣∣∣∣∣∫

Kψ(t)ei f (t)

ε dt − eif (t0)/ε

(ε−1 f ′′(t0)

2iπ

)− 12 ∑

j<k

εjLjψ

∣∣∣∣∣∣ ≤ Cεk∑α≤2k

sup |ψ(α)(x)|

where the Lj can be explicitly expressed as differential operators.

Page 56: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

In our case, we obtain for k = 1:∣∣∣Mφ(t)−(φ(t) +

ac0

2(tφ(t))′′

)∣∣∣ ≤ Ca32∑α≤4

sup |φ(α)(t)|

At the end we have a third order ODE to solve :

pa(x , t) = Lp(x , t) =

(1 +

ac0

2∂

∂t

)(p(x , t) +

ac0

2

(t∂2p∂t2 (x , t) +

∂p∂t

(x , t)))

+ o(a)

for higher order asymptotics, we use :

Mφ(t) =k∑

i=0

(c0a)i

2i i!(t iφ(t))(2i)(t) + o(ak )

Page 57: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 58: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 59: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

In a bounded domain, we adapt the dual approach presentedpreviously.Testing the measurements against appropriate waves, we accessan integral operator of R[p0] that we invert using the stationaryphase theorem.

Page 60: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

No correction, a=0.0005

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

Order 1 correction, a=0.0005

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Order 8 correction, a=0.0005

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Page 61: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

No correction, a=0.0025

0

0.05

0.1

0.15

0.2

0.25

0.3

Order 1 correction, a=0.0025

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Order 8 correction, a=0.0025

0

0.1

0.2

0.3

0.4

0.5

0.6

Page 62: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

1 Photo-acoustic ImagingImposed boundary conditionsView limitationAcoustic attenuationCluttered sound speedqPAT : an inverse problem for light propagation

2 Wave imaging and topological derivative

Page 63: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 64: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 65: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

In the case of an unknown cluttered sound speed, we look at :1

c(x)2∂2p∂t2 (x , t)−∆p(x , t) = 0

p(x ,0) = p0(x)∂p∂t

(x ,0) = 0

where c(x) randomly varies around a constant :

1c2(x)

=1c2

0

(1 + σµ

(xxc

))

Page 66: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

If we apply algorithms for homogeneous sound speed to this case, wemay observe artifacts.

True initial condition Reconstruction using inverse Radon

Page 67: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

We combine 2 ideas :Kirchhoff migration can be robustified using space and frequencycorrelations between the dataour reconstruction algorithms for extended targets are based onRadon transforms inversions. The last step of such inversions is aback-projection which is similar to the Kirchhoff migration.

Page 68: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Our methods of inversion read :

R[p0] =W[p]

p0 = R?BR[p0]

where R is the spherical or line Radon transform, R?, its adjoint, is abackprojection :

R?[f ](x) =

∫R

∫∂Ω

f (y , ω)e−iω|x−y |dσ(y)dω

resp.

R?[f ](x) =

∫R

∫Sn−1

f (y , ω)e−iωθ.xdθdω

Page 69: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Kirchhoff migration :

IKM(x) = R?[p](x) =

∫R

∫∂Ω

p(y , ω)e−iω|x−y |dσ(y)dω

The original CINT :

ICINT (x) =

∫R

∫|ω2−ω1|≤Ωd

∫∂Ω

∫|y2−y1|≤Xd

p(y1, ω1)e−iω1|x−y1|

p(y2, ω2)eiω2|x−y2|dσ(y1)dσ(y2)dω1dω2

Page 70: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Our extended target reconstruction :

Irad(x) = R?BW[p](x) =

∫∂Ω

BW[p](y , ω)e−iω|x−y |dσ(y)dω

The proposed CINT algorithm :

ICIR(x) =

∫R

∫R

∫∂Ω

∫∂Ω

e− (ω1−ω2)2

2Ω2d e

− |y1−y2|2

2X2d

BW[p](y1, ω1)e−iω1|x−y1|BW[p](y2, ω2)eiω2|x−y2|dσ(y1)dσ(y2)dω1dω2

Page 71: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Stability Analysis

In a slightly different setting :

finite bandwidth[ω0 −

B2, ω0 +

B2

]with B ω0

xc X0

we can carry out a stability analysis. We have :

SNRradon ∼ e−ω2

0τ2c

2√

1 + Bτc

√1 +

X0

Xc

and :SNRCIR ∼

1√Ω2

dτ2c +

X 2d

X 2c

Page 72: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Reconstruction using inverse Radon

Page 73: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 74: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

True initial condition Reconstruction using inverse Radon Reconstruction using CINT

Page 75: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

1 Photo-acoustic ImagingImposed boundary conditionsView limitationAcoustic attenuationCluttered sound speedqPAT : an inverse problem for light propagation

2 Wave imaging and topological derivative

Page 76: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

A second inverse problem

good reconstruction of the absorbed energy density p0(x) insidethe medium is not the true (relevant) piece of information.it is not intrinsic and depends on both the illumination and onoptical coefficients.need of light propagation model, in terms of intrinsic opticalcoefficients (µa, µs, ...).

Page 77: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

The RTE

Consider the radiative transfer equation (RTE) :

s.∇xϕ(x , s) + µt (x , s)ϕ(x , s) =

∫Sn−1

k(x , s′, s)ϕ(x , s′)dσ(s′)

with µt = µa + µs, µs(x , s) =

∫Sn−1

k(x , s, s′)dσ(s′).

In this case we have : p0(x) ∝∫Sn−1

µa(x , s′)ϕ(x , s′)dσ(s′).

Page 78: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

A theoretical result for the RTE

We define an Albedo operator A :

L1(∂Ω−) → L1(Ω)

ϕ(x , s) 7→ Aϕ(x) =

∫Sn−1

µa(x , s′)ϕ(x , s′)dσ(s′)

Page 79: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Using the singularities of A, we prove the following stability estimates :

Theorem (G.Bal, A.Jollivet, VJ)

If (µt , µa, k) generates the Albedo A and (µt , µa, k) generates A then :

‖µt − µt‖L∞(Sn−1,W−1,1(X)) +‖µa − µa‖L∞(Sn−1,L1(X))

≤ C‖A− A‖L(L1(∂Ω−),L1(Ω))

If µt , µt ∈ L∞(Sn−1,W r ,p(X ))

‖µt − µt‖L∞(Sn−1,W s,p(X)) ≤ C2‖A− A‖r−s

p(1+r)

L(L1(∂Ω−),L1(Ω))

Assuming a model for k , we can also get stability estimates onanisotropy.

Page 80: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Conclusion

The standard problem of photoacoustics is well-posed, and has beenthoroughly treated.

Consideration of non-idealities may introduce some difficulties and/orill-posedness.

For each non-ideality, we proposed a model, analysed it, developedcorrections based on adapted mathematical tools, and validated themethods on numerical experiments.

Another important issue is to separate intrinsic coefficients fromillumination.

different models can be used to model light.we derived stability estimates for the RTE.still much to be done for this problem (need for theory andalgorithms).

The proposed methods need to be validated on real experimental data.

Page 81: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

1 Photo-acoustic ImagingImposed boundary conditionsView limitationAcoustic attenuationCluttered sound speedqPAT : an inverse problem for light propagation

2 Wave imaging and topological derivative

Page 82: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Ω

∂Ω

ρ=1K=1

KK, ρ

Page 83: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Helmholtz problem with inclusion D = za + δB: ∇.(

1χΩ\D + ρ−1χD(x)∇u)

+ ω2(

1χΩ\D + K−1χD(x))

u = 0∂u∂ν

= g on ∂Ω

background solution : ∆U + ω2U = 0∂U∂ν

= g on ∂Ω

Page 84: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Topological derivative

Choose Ds, ρs, Ks, and compute the quadratic misfit :

E(Ds; ρs,Ks) =

∫∂Ω|(us − umeas)(x)|2dσ(x)

we consider the effect of the addition of asymptotically smallinclusions.the volume derivative of E should be the most negative when Ds isclose to D.

Page 85: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Ω

∂Ω

ρ=1K=1

E(∅ )

K, ρ

Page 86: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Ω

∂Ω

ρ=1K=1

E(D1) ∼ E(∅ )

K’, ρ’

K, ρ

Page 87: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Ω

∂Ω

ρ=1K=1

E(D2) ∼ E(∅ )

K’, ρ’K, ρ

Page 88: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Ω

∂Ω

ρ=1K=1

E(D3) << E(∅ )

K’, ρ’

K, ρ

Page 89: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Classic asymptotic results yield :

Theorem Boundary PerturbationSuppose ωδ 1, then for any x ∈ ∂Ω :

(umeas − U) (x) = −δd(∇U(za)M(ρ−1,B)∇zNω(x , za)

+ω2(K−1 − 1)|B|U(za)Nω(x , za))

+ o(δd )

(us − U) (x) = −δds

(∇U(zs)M(ρ−1

s ,Bs)∇zNω(x , zs)

+ω2(K−1s − 1)|Bs|U(zs)Nω(x , zs)

)+ o(δd

s )

where Nω is a (geometry dependent) Neumann function : ∆Nω(x , z) + ω2Nω(x , z) = −δz∂N∂ν

= 0 on ∂Ω

Page 90: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Volume derivative of E at each point zs :

ITD[U,g](zs) = −∂E [U,g](zs)

∂(δs)d

∣∣∣∣δs=0

Using the asymptotic result :

ITD[U,g](zs) = δdπω4(K−1 − 1)(K−1s − 1)

Re

U(zs)

∫∂Ω

Nω(x , zs)Nω(x , za)dσ(x)U(za)

+δd4π2 (1− ρ)(1− ρs)

(1 + ρ)(1 + ρs)

Re∇U(zs)

∫∂Ω

∇zNω(x , zs)∇zNω(x , za)T dσ(x)∇U(za)

+o(δd )

Page 91: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Volume derivative of E at each point zs :

ITD[U,g](zs) = −∂E [U,g](zs)

∂(δs)d

∣∣∣∣δs=0

Using the asymptotic result :

ITD[U,g](zs) = δdπω4(K−1 − 1)(K−1s − 1)

Re

U(zs)

∫∂Ω

Nω(x , zs)Nω(x , za)dσ(x)U(za)

+δd4π2 (1− ρ)(1− ρs)

(1 + ρ)(1 + ρs)

Re∇U(zs)

∫∂Ω

∇zNω(x , zs)∇zNω(x , za)T dσ(x)∇U(za)

+o(δd )

Page 92: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Volume derivative of E at each point zs :

ITD[U,g](zs) = −∂E [U,g](zs)

∂(δs)d

∣∣∣∣δs=0

Using the asymptotic result :

ITD[U,g](zs) = δdπω4(K−1 − 1)(K−1s − 1)

Re

U(zs)

∫∂Ω

Nω(x , zs)Nω(x , za)dσ(x)U(za)

+δd4π2 (1− ρ)(1− ρs)

(1 + ρ)(1 + ρs)

Re∇U(zs)

∫∂Ω

∇zNω(x , zs)∇zNω(x , za)T dσ(x)∇U(za)

+o(δd )

Page 93: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

TheoremThe following identity holds. For any x ∈ ∂Ω, for any z ∈ Ω :(

− I2

+KωΩ)

[Nω(., z)](x) = Γω(x − z)

where Γω is the fundamental solution of the Helmholtz equation in thefree-space.

Define the new misfit function :

E2(Ds; ρs,Ks) =

∫∂Ω

∣∣∣∣(− I2

+KωΩ)

(us − umeas)(x)

∣∣∣∣2 dσ(x)

Page 94: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

ITD2[U, g](zs) = C(ω,K ,Ks, δ)Re

U(zs)

∫∂Ω

Γω(x − zs)Γω(x − za)dσ(x)U(za)

+C(ρ, ρs, δ)Re

∇U(zs)

∫∂Ω

∇zΓω(x − zs)∇zΓω(x − za)T dσ(x)∇U(za)

+o(δd )

We have that :∫∂Ω

Γω(x − zs)Γω(x − za)dσ(x) ∼ 1ω

Im Γω(zs − za)

∫∂Ω

∇zΓω(x − zs)∇zΓω(x − za)T dσ(x) ∼ ωIm Γω(zs − za)(

za − zs

|za − zs|

)(za − zs

|za − zs|

)T

Page 95: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

ITD2[U, g](zs) = C(ω,K ,Ks, δ)Re

U(zs)

∫∂Ω

Γω(x − zs)Γω(x − za)dσ(x)U(za)

+C(ρ, ρs, δ)Re

∇U(zs)

∫∂Ω

∇zΓω(x − zs)∇zΓω(x − za)T dσ(x)∇U(za)

+o(δd )

We have that :∫∂Ω

Γω(x − zs)Γω(x − za)dσ(x) ∼ 1ω

Im Γω(zs − za)

∫∂Ω

∇zΓω(x − zs)∇zΓω(x − za)T dσ(x) ∼ ωIm Γω(zs − za)(

za − zs

|za − zs|

)(za − zs

|za − zs|

)T

Page 96: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation
Page 97: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Stability Study : Electronic Noise

Situation with 1 inclusion located at za. Computation of themeasurements.Addition of electronic noise

captor position

pote

ntia

l u

captor position

pote

ntia

l u

measure without noisemeasure with electronic noise

Application of the localization algorithms : topological derivative(with or without pre-processing), MUSIC, Kirchhoff migration,backpropagation⇒ estimation zest of za.large number of realizations (300)⇒ estimation of STD(|za− zest |)as a function of the noise level.

Page 98: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

0 5 10 15 20 25 300

0.1

0.2

0.3

0.4

0.5

0.6

0.7

electronic noise level sigma (%)

std(

|zes

t−z a|)

topological derivative with −I/2+KΩtopological derivativeMUSICKirchhoff MigrationBackpropagation

Page 99: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Stability Study : Clutter Noise

Situation with 1 inclusion located at za.Realization of one clutter noise, computation of themeasurements.

Example of a realisation of ρ (or K)

captor position

pote

ntia

l u

measure without noisemeasure with clutter noise

Application of the localization algorithms : topological derivative(with or without pre-processing), MUSIC, Kirchhoff migration,backpropagation⇒ estimation zest of za.large number of realizations (300)⇒ estimation of STD(|za− zest |)as a function of the noise level.

Page 100: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

0 1 2 3 4 5 6 7 8 9 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

clutter noise level sigma (%)

std(

|zes

t−z a|)

topological derivative with −I/2+KΩtopological derivativeMUSICKirchhoff MigrationBackpropagation

Page 101: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

Conclusion

Topological derivative algorithm is based on a good heuristic.Use of asymptotic tools to understand and improve it.Approach can be extended to elastic and electromagnetic waves,as well as to the EIT problem.Possibility to use higher-order asymptotics and multi-frequencymeasurements.Solutions needed to deal with limited-view acquisition.

Page 102: Photo-acoustic Imaging - MIT Mathematicsmath.mit.edu/~vjugnon/pdf/MIT_slides.pdf · 2012-09-07 · 1 Photo-acoustic Imaging Imposed boundary conditions View limitation Acoustic attenuation

H. Ammari, E. Bossy, VJ and H. KangMathematical models in photo-acoustic imagingSIAM Review (Vol.52, No.4).

H. Ammari, M. Asch, L. Guadarrama Bustos, VJ and H. KangTransient imaging with limited view data(under revision) to appear in SIAM Journal on Imaging Sciences.

H. Ammari, E. Bretin, VJ and A. WahabPhotoacoustic imaging for attenuating acoustic mediaMathematical Modeling in Biomedical Imaging II, Lecture Notes in Mathematics, to appear.

H. Ammari, E. Bretin, J. Garnier and VJCoherent interferometric algorithms for photoacoustic imagingsubmitted to SIAM Journal on Imaging Sciences.

G. Bal, A. Jollivet and VJInverse transport theory of photoacousticsInverse Problems, 26 025011.

H. Ammari, E.Bossy, VJ and H. KangQuantitative photoacoustic imaging of small absorbers(under revision) to appear in SIAM Journal on Applied Mathematics.

H. Ammari, J. Garnier, VJ and H. KangStability and resolution analysis for a topological derivative based functionsubmitted to SIAM Journal on Imaging Sciences.