Content uploaded by Angelos T. Vouldis
Author content
All content in this area was uploaded by Angelos T. Vouldis on May 08, 2014
Content may be subject to copyright.
Content uploaded by Konstantina Nikita
Author content
All content in this area was uploaded by Konstantina Nikita on Mar 11, 2014
Content may be subject to copyright.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 55, NO. 6, DECEMBER 2006 1975
Three-Dimensional Diffraction Tomography
Using Filtered Backpropagation and
Multiple Illumination Planes
Angelos T. Vouldis, Costas N. Kechribaris, Theofanis A. Maniatis, Member, IEEE,
Konstantina S. Nikita, Senior Member, IEEE, and Nikolaos K. Uzunoglu, Fellow, IEEE
Abstract—In this paper, a three-dimensional (3-D) extension
of the well-known filtered-backpropagation (FBP) algorithm is
presented with the aim of taking into account scattered-field-data
measurements obtained using incident directions not restricted
in a single plane. The FBP algorithm has been extensively used
to solve the two-dimensional inverse-scattering problem under
the first-order Born and Rytov approximations for weak scatter-
ers. The extension of this algorithm in three dimensions is not
straightforward, because the task of collecting the data needed to
obtain a low-pass filtered version of the scattering object, taking
into account all spatial frequencies within a radius of √2k0,
and of incorporating these data to the FBP algorithm, needs to
be addressed. A simple extension using incident field directions
restricted to a single plane (illumination plane) leaves a region of
spatial frequencies of the sphere of radius √2k0undetermined.
The locus of these spatial frequencies may be crucial for the
accurate reconstruction of objects which do not vary slowly along
the axis perpendicular to the illumination plane. The proposed
3-D FBP algorithm presented here is able to incorporate the data
collected from more than one illumination plane and to ensure the
reliability of the reconstruction results.
Index Terms—Acoustic scattering, biomedical imaging, inverse
problems, tomography.
I. INTRODUCTION
IN THE inverse-scattering problem, the aim is to reconstruct
the object function, which depends on the wave properties
of matter in an area of interest characterized by inhomogeneous
distribution of these properties. Such a reconstruction is impor-
tant for a variety of applications ranging from medical imaging
[1] to geoscience [2] and nondestructive testing [3].
During the past few years, there has been a general trend
to implement inverse-scattering algorithms that could be used
for three-dimensional (3-D) reconstructions in order to de-
scribe in a more realistic manner the corresponding practical
applications. It has been shown that two-dimensional (2-D)
approximations are not accurate for objects that depart from
the cylindrical model and vary along the third dimension
[4], [5]. The extension of known 2-D inversion methods in the
full 3-D is not a trivial issue mainly because of the heavy com-
putational burden that is imposed by the existence of an extra
Manuscript received August 15, 2005; revised August 8, 2006.
The authors are with the School of Electrical and Computer Engineer-
ing, National Technical University of Athens, 15773 Athens, Greece (e-mail:
knikita@cc.ece.ntua.gr).
Digital Object Identifier 10.1109/TIM.2006.884352
dimension. Especially, inversion algorithms based on an itera-
tive optimization methodology (see, e.g., [6]–[13]) can be very
time consuming for general 3-D cases.
In the frequency domain, when it is known beforehand that
the contrast of the scattering objects is within certain limits,
the linear approximations Born and Rytov can be utilized to
simplify the inverse problem considerably. In this case, the
object function can be reconstructed by employing diffraction-
tomography (DT) algorithms [14] which are based on the linear
relation between the Fourier transform of the scattered-field
data and the Fourier transform of the object function [15]. This
relation can be used to produce the final reconstruction in two
different ways, namely the direct Fourier interpolation (DFI) al-
gorithm [15], [16] and the filtered backpropagation (FBP) [17]
algorithm. The difference in the methodology of these al-
gorithms lies in the way they use to interpolate the Fourier
transform of the scattered-field data. In the DFI algorithm,
data are interpolated in the spatial frequency domain prior to
an inverse Fourier transform, while in the FBP algorithm, the
interpolation is performed directly in the spatial domain. The
FBP algorithm has been proven to be a more reliable inversion
method, avoiding artifacts associated with the interpolation in
the spatial frequency domain [17].
Other DT algorithms have also been developed to simulate
in an accurate way the practical requirements of clinical tomo-
graphic acquisition setup. The inclusion of evanescent waves
has been investigated in order to increase the resolution of
the reconstructed image [18], [19]. DT algorithms have been
proposed for the case of nonplane wave illumination and for
arbitrary measurement surface, implying the possibility of more
practical acquisition geometries [20]–[22]. The theoretical [23]
and experimental [24], [25] application of Rytov-based DT
algorithms proved its superiority to the first-order Born approx-
imation, provided that the scattered-field data are backpropa-
gated to the area of the scattering object. Pan and Anastasio [26]
showed that for 2-D problems, the complete rotation of the
incident wave in the (0,2π)interval is not necessary in order
to cover a circle of radius √2k0in the spatial frequency
domain and a rotation in the (0,3π/2) angle range is sufficient.
Reflection mode DT has been investigated [27] for use in cases
where the reconstruction of the discontinuities present in the
object function is sought. Multifrequency inversion schemes
within the context of reflection mode DT algorithms have been
considered in order to improve the quality of the reconstructed
0018-9456/$20.00 © 2006 IEEE
1976 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 55, NO. 6, DECEMBER 2006
image [28]. Hybrid algorithms aiming to solve 3-D problems
with 2-D algorithms and cross section by cross section recon-
struction have been presented [17], [29], [30].
In [31], a 3-D extended formulation of the DFI and FBP
algorithms was presented. It was observed that the restriction of
the incident direction in a plane does not result in full coverage
of a sphere of radius √2k0in the spatial frequency domain.
Numerical tests for objects satisfying the weak-scattering cri-
teria showed that the use of a single illumination plane, cur-
rently used in many experimental DT apparatus, can potentially
degrade the reconstruction outcome, especially if the recon-
structed object varies abruptly along the axis perpendicular
to the illumination plane. When a second illumination plane
perpendicular to the first is used, a complete sphere of radius
√2k0is covered in the spatial frequency domain. The incorpo-
ration of the scattered-field data from the second illumination
plane can be easily done in the DFI algorithm, whereas it is
not mathematically straightforward in the FBP algorithm. The
ability to incorporate multiple illumination planes in the FBP
would be desirable given that the FBP is generally a more
reliable inversion algorithm.
In this paper, we present a modified 3-D FBP algorithm
consisting of some extra steps compared to the one presented
in [31], which is able to incorporate data from more than one
illumination plane. Compared to the DFI algorithm using the
same set of data, the proposed method yields better reconstruc-
tion quality, thus confirming the inherent superiority of the FBP
algorithm in a 3-D context. The following analysis applies only
to 3-D scalar fields such as acoustic wavefields.
II. INVERSE PROBLEM FORMULATION
In the frequency domain, the Lippmann–Schwinger integral
equation [32] can be used to describe the wave propagation of
a scalar wavefield
ψ(r)=ψinc (r)+
V
g(r −r )o(r )ψ(r )dr(1)
where ψ(r)is the total field, ψinc (r)is the incident wave,
g(r |r )is the Green’s function in three dimensions, and o(r)
is the unknown object function which we aim to reconstruct
o(r)=k2
0n2(r)−1(2)
where n(r)is the refractive index and k0is the wavenumber in
the surrounding medium. Henceforth, it will be assumed that
the incident field is a plane wave; so that, ψinc(r) = exp(i
k·r)
and the time dependence of the wavefields is exp(−iωt).
When Born or Rytov approximation is invoked, a linear rela-
tion can be found correlating the Fourier transformed scattered-
field data with the Fourier transform of the object function. In
the case of the first-order Born approximation which assumes
that ψ(r)=ψinc (r)inside the scatterer, this relation is [33]
O(ωξ,ω
z,φ)=−2ik2
0−ω2
ξ−ω2
z
·e−ik2
0−ω2
ξ−ω2
zl0ΨB,l0,φ(ωξ,ω
z)(3)
Fig. 1. Geometry of the problem with the coordinate systems (ωx,ω
y,ω
z)
and (ωη,ω
ξ,ω
z).
Fig. 2. Ewald hemisphere coverage resulting from measurements of the
scattered field with the incident wave at direction ˆη.
where Ois the Fourier transform of the object function, and
ΨBis the Fourier transform of the scattered field when the
Born approximation holds. The propagation direction ˆηof the
incident field defines a new coordinate system (η, ξ, z)which
can be obtained by rotating the (x, y, z)coordinate system by
an angle φaround the z-axis (Fig. 1). The scattered field is
measured on the plane η=l0. Equation (3) implies that the
Fourier transform of the scattered data in the measurement
plane enables the determination of the Fourier transform of
the object function on the surface of a hemisphere with radius
k0centered at (−k0,0,0) in the (η, ξ, z)coordinate system
(see Fig. 2). If Rytov approximation is used, the corresponding
equation is [33]
O(ωξ,ω
z,φ)
=−2ik2
0−ω2
ξ−ω2
ze−ik2
0−ω2
ξ−ω2
zl0Pl0,φ(ωξ,ω
z)
e−ik0l0(4)
where Pis the Fourier transform of the complex phase of the
scattered field. Equations (3) and (4) form the basis of 3-D
VOULDIS et al.: 3-D DIFFRACTION TOMOGRAPHY USING FBP AND MULTIPLE ILLUMINATION PLANES 1977
Fig. 3. Spatial-frequency coverage obtained by using incident plane wave
directions only on the xy plane. The frequencies belonging to the white
(missing) region cannot be determined, although they belong to the sphere of
radius √2k0.
DT algorithms. Following the DFI method, the values of the
function Ofrom equispaced samples in the (ωξ,ω
z,φ)coordi-
nate system are interpolated to a cubic grid of the (ωx,ω
y,ω
z)
coordinate system and then an inverse fast Fourier transform
(FFT) is used to obtain the reconstructed object. It has been
noted that the interpolation process produces artifacts in the
final results, and thus degrades the reconstruction quality [17].
On the other hand, the FBP algorithm is based upon an integral
representation of the object function, thus avoiding the spectral
interpolation. Thus, the only necessary interpolation is made
on the spatial domain, and the accuracy of the reconstructed
images is improved considerably.
In the 2-D case, the rotation of the incident direction on
the xy plane and the measurement of the scattered field on a
line perpendicular to the incident direction in the transmitted
field region result in the coverage of a circle with radius
√2k0in the spatial frequency domain. In an analogous manner
in the 3-D case, if the incident direction is rotated along a
single plane and the scattered field is measured on a plane
perpendicular to the incident direction, a sphere of radius √2k0
is covered in the spatial frequency domain excluding a set
of frequencies near the axis perpendicular to the illumination
plane. Specifically, if the incident-direction vector is restricted
to the z=0 plane, then the Fourier transform of the object
function on the locus of spatial frequencies defined by the equa-
tion 4k2
0R2−R4−4k2
0ω2
z<0, where R=ω2
x+ω2
y+ω2
z,
remains undetermined (Fig. 3). A full coverage of a sphere of
radius √2k0could be obtained by combining measurements
from different illumination planes. Although the DFI method
can easily incorporate the data collected from more than one
illumination plane, in the FBP algorithm, this is not a trivial
issue. An extended version of the FBP algorithm that is able
to overcome this limitation will be presented in the following
section.
III. 3-D EXTENSION OF FBP ALGORITHM
The aim is to reconstruct a low-pass filtered version of the
object function
oLP(r)= 1
(2π)3
+∞
−∞
+∞
−∞
+∞
−∞
OLP(
k)·ei
krd
k(5)
where OLP(
k)=0if |
k|≥√2k0,else,OLP(
k)=O(
k).
A coordinate transformation is needed in order to exchange
the Cartesian coordinate representation for semispherical arcs
corresponding to the ensuing coverage of the spatial frequency
domain. The vector
k=(kx,k
y,k
z)can be expressed as [33]
k=k0(ˆs1−ˆs0)(6)
where
ˆs0=(sinθ0cos φ0,sin θ0sin φ0,cos θ0)(7)
is the propagation direction of the incident plane wave, and
ˆs1=(sinθ1cos φ1,sin θ1sin φ1,cos θ1)
=1
k0ωξˆ
ξ+ωzˆz+k2
0−ω2
ξ−ω2
zˆη(8)
is the vector pointing to the surface of the Ewald hemisphere
that corresponds to the transmission-region data. At this point,
it is assumed that the propagation direction of the incident wave
is restricted to the xy plane (θ0=π/2).
This transformation can be used to obtain the following
integral representation of the object function with respect to the
rotated coordinate system (ωη,ω
ξ,ω
z)[31]:
oLP(r)=1
2
k0
(2π)3
2π
0
k0
−k0
k0
−k0
|ωξ|
k2
0−ω2
ξ−ω2
z
OLP (k0(s1−s0))
·exp iωξξ+k2
0−ω2
ξ−ω2
z−k0
·η+ωzzdθ1dωξdωz.
(9)
However, an equation similar to (9), combining data from
more than one illumination plane cannot be derived.
Assuming that the scattered-field data from two perpendicu-
lar illumination planes z=0and y=0are available, it can be
easily observed that the sphere of radius √2k0is fully covered
in the spatial frequency domain. This approach will result in
a single coverage for the regions defined by the inequalities
4k2
0R2−R4−4k2
0ω2
z<0,4k2
0R2−R4−4k2
0ω2
y<0and in
double coverage for the rest of the sphere. Using the mapping
of the (ωξ,ω
z,φ)coordinate system to the (ωx,ω
y,ω
z)coor-
dinate system, we can infer whether each Fourier-transformed
1978 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 55, NO. 6, DECEMBER 2006
Fig. 4. Block diagram illustrating the proposed 3-D FBP algorithm.
scattered-field data corresponds to a single or double covered
region and accordingly multiply it with a suitable weighting
function
σ(ωx,ω
y,ω
z)=
1,for 4k2
0R2−R4−4k2
0ω2
z<0
and R≤√2k0
1,for 4k2
0R2−R4−4k2
0ω2
y<0
and R≤√2k0
0.5,for 4k2
0R2−R4−4k2
0ω2
z≥0
and 4k2
0R2−R4−4k2
0ω2
y≥0
and R≤√2k0
0,elsewhere. (10)
The value 0.5 in (10) compensates for the areas of double
coverage.
Applying independently the FBP algorithm to each set of
weighted data and subsequently averaging could theoretically
reconstruct the object function. The practical implementation of
this algorithm however did not yield satisfactory results due to
the mismatch between the points defined on the (ωξ,ω
z,φ)grid
and the corresponding (ωx,ω
y,ω
z)grid, the consequences of
which are more pronounced near the boundaries between single
and double coverage regions.
The proposed approach consists of initially applying the FBP
algorithm defined by (9) for the reconstruction of oxy(r)using
z=0illumination plane and the reconstruction of oxz(r)using
y=0illumination plane. Then, using the FFT on oxy (r)and
TAB L E I
MEAN-SQUARE ERROR FOR THE TWO-LAYERED
SPHERE RECONSTRUCTION
Fig. 5. Relative error profile for the two-layered sphere along z-axis.
oxz(r), the functions Oxy and Oxz are derived, respectively.
The value of the Fourier transform of the object function will
be found from a weighted average of Oxy and Oxz
O(ωx,ω
y,ω
z)=w(Oxy,O
xz).(11)
The weighting function w(Oxy,O
xz)can be defined equal to σ
or, alternatively, it can be set as a maximum operator selecting
the highest value from Oxy,Oxz at every spatial frequency
point (ωx,ω
y,ω
z). Inverse FFT yields the final reconstruction.
This algorithm is depicted graphically in Fig. 4.
IV. RESULTS
In this section, the proposed FBP algorithm will be evaluated
using a set of scattering objects. The aim is to investigate the
enhancement provided by the incorporation of two perpendicu-
lar illumination planes compared to the DFI algorithm using the
same data set and the single plane illumination FBP method.
To this end, three different scattering objects are considered:
1) two-layered spherical scatterer centered at (0, 0, 0) with inner
radius r1=1.5λ, outer radius r2=3λ, and refractive indexes
n1=1.03,n2=1.05, respectively, 2) a two-layered cubic
scatterer with inner side length l1=2λand outer side length
l2=4λ, centered at (0, 0, 0) with refractive indexes n1=1.05
and n2=1.03, respectively, and 3) two nonconcentric cubes
VOULDIS et al.: 3-D DIFFRACTION TOMOGRAPHY USING FBP AND MULTIPLE ILLUMINATION PLANES 1979
Fig. 6. Reconstructed images at the plane z=0for the case of the two-layered sphere.
TAB L E I I
MEAN-SQUARE ERROR FOR THE TWO-LAYERED CUBE RECONSTRUCTION
centered at (−1.5λ, −1.5λ, −1.5λ),(1.5λ, 1.5λ, 1.5λ)both
with side lengths l=3λand their respective refractive indexes
being n1=1.05 and n2=1.03.
Synthetic scattered-field data were generated using the con-
jugate gradient FFT method [34]. A total of 24 different in-
cident field propagation directions are used with a step of
∆ϕ=15
◦. The measurement planes are defined to be 25λ
away from the origin. The scattered field is measured for every
incident direction on a set of 64 ×64 points placed on a
rectangular grid with 0.5λgrid spacing. Two scattered-field
data sets are used for each scattering object: one with the
propagation direction restricted to the xy plane (xy data set)
and another with the propagation direction restricted to the xz
plane (xz data set).
Rytov approximation is used, and the field is backpropagated
to a plane passing through the origin before applying the
reconstruction algorithm. A considerable improvement of the
reconstruction results was observed after applying a phase un-
Fig. 7. Relative error profile for the two-layered cube along z-axis.
wrapping procedure on the 2-D phase function ps(ξ,z). The use
of the maximum selective operator as the weighting function
wyielded the best quality reconstruction, and therefore, it was
adopted for the presented simulation results.
As a figure of merit for the reconstruction quality, the total
reconstruction error is defined by
err =
D
[nreconstructed(r)−ncorrect (r)]2dr
D
n2
correct(r)dr
(12)
where Dis the volume occupied by the scattering objects. The
relative error along planes z=constant was also found to be
a sufficiently informative quantity, and it will be used for the
evaluation of the reconstruction results.
1980 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 55, NO. 6, DECEMBER 2006
Fig. 8. Reconstructed images at the plane z=0for the case of the two-layered cube.
Fig. 9. Line profile of the object function for the two-layered cube.
A. Two-Layered Sphere
In this case, a notable improvement in the reconstruction
results occurs when both illumination planes are incorporated.
This is demonstrated from the values of the total reconstruction
error presented in Table I. Figs. 5 and 6 depict the relative
error along planes z=constant and the reconstruction results
at the plane z=0, respectively, using various inversion algo-
rithms and data sets. The superiority of FBP over DFI and the
enhancement of the final result with the inclusion of multiple
illumination data are clearly shown here. All methods recon-
struct equally well the object boundaries. But, the FBP using
two illumination planes is more accurate with respect to the
refractive index of the internal structures.
Fig. 10. Relative error profile for the two-cube case along z-axis.
TABLE III
MEAN-SQUARE ERROR FOR THE TWO-CUBE RECONSTRUCTION
VOULDIS et al.: 3-D DIFFRACTION TOMOGRAPHY USING FBP AND MULTIPLE ILLUMINATION PLANES 1981
Fig. 11. Reconstructed images at the plane z=−1.5λfor the case of the two cubes.
Fig. 12. Reconstructed images at the plane z=1.5λfor the case of the two cubes.
B. Two-Layered Cube
In order to evaluate the proposed algorithms with scatter-
ing objects of higher spatial frequency content, a two-layered
cube is considered. Once more, the FBP algorithm using two
illumination planes produced the lowest total reconstruction
error as shown in Table II. Moreover, the relative error profile
along planes z=constant shown in Fig. 7 demonstrates the
benefits of using more than one illumination plane with FBP.
1982 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 55, NO. 6, DECEMBER 2006
Fig. 13. Mean-square error along the lines parallel to the z-axis for the two-cube case.
Although the improvement in the shape of the reconstructed
object is not obvious from a mere visual inspection of Fig. 8, the
internal structure of the object is more accurate using the FBP
with two illumination planes. Finally, based on the line profile
shown in Fig. 9, it can be concluded that the FBP with two
illumination planes yields a better reconstruction result, which
is more apparent in the inner layer of the scattering object.
C. Two Nonconcentric Cubes
In this case, we examine a nonconcentric pair of cubes. In the
relative error profile (Fig. 10), the FBP with two illumination
planes outperforms the other reconstruction methods around
the center (z=0) of the reconstruction area. In the regions
around the centers of the cubes where the relative error profile is
comparatively lower, the reconstruction methods which utilize
the xz set of data are obviously more accurate compared to
the FBP algorithm with data from the xy illumination plane.
Overall, the total reconstruction error (Table III) is lower using
the FBP with two illumination planes. This is also evident
in Figs. 11 and 12 where the reconstructed profiles at planes
passing through the centers of the cubes are depicted. In Fig. 13,
the mean-square error along the lines parallel to the z-axis is
plotted in order to emphasize the improvement in the boundary
reconstruction observed when the xz set of data is used.
V. D ISCUSSION
The modified version of the 3-D FBP inversion algorithm
presented here seems to keep the advantages that the FBP
presents compared to the DFI algorithm while making pos-
sible the incorporation of scattered-field-data measurements
acquired when the incident direction was placed in different
planes.
In all the simulations carried out, the comparison between
the results obtained with one illumination plane, and the results
produced when the incident direction was rotated in two dif-
ferent planes proved the importance of incorporating as much
data as possible in the reconstruction process. In some cases,
the inclusion of a second set of data is necessary in order to
capture the essential features of the imaged object, while in
other cases, the improvement was marginal. Nevertheless, the
tradeoff between accuracy, computational time, and the difficul-
ties entailed in the practical implementation of data acquisition
from multiple illumination planes has to be always taken into
account. Furthermore, similar studies can be performed for
reflected mode DT algorithms where the importance of the
inclusion of a second set of data could be even more important.
Future work will include a noise analysis and an investigation
with respect to acquisition geometry parameters.
VI. CONCLUSION
A modified 3-D extension of the FBP inverse-scattering
algorithm was presented, and it was found that its capability
of taking into account the scattered-field data collected from
more than one incident direction improves the quality of the
reconstructed image. Thus, the main shortcoming of the FBP
algorithm compared to the DFI noted earlier [31] is overcome,
and the comparison of these two algorithms when a complete
sphere of radius √2k0is covered in the spatial frequency
domain was made possible, confirming the superiority of FBP.
VOULDIS et al.: 3-D DIFFRACTION TOMOGRAPHY USING FBP AND MULTIPLE ILLUMINATION PLANES 1983
REFERENCES
[1] N. Duric, P. Littrup, A. Babkin, D. Chambers, S. Azevedo, A. Kalinin,
R. Pevzner, M. Tokarev, E. Holsapple, O. Rama, and R. Duncan, “Devel-
opment of ultrasound tomography for breast imaging: Technical assess-
ment,” Med. Phys., vol. 32, no. 5, pp. 1375–1386, May 2005.
[2] G. Fornaro, F. Serafino, and F. Soldovieri, “Three-dimensional focusing
with multipass SAR data,” IEEE Trans. Geosci. Remote Sens., vol. 41,
no. 3, pp. 507–517, Mar. 2003.
[3] J. C. Bolomey, “Recent European developments in active microwave
imaging for industrial, scientific and medical applications,” IEEE Trans.
Microw. Theory Tech., vol. 37, no. 12, pp. 2109–2117, Dec. 1989.
[4] P. M. Meany, K. D. Paulsen, S. D. Geimer, S. A. Haider, and
M. W. Fanning, “Quantification of 3-D field effects during 2-D mi-
crowave imaging,” IEEE Trans. Biomed. Eng., vol. 49, no. 7, pp. 708–720,
Jul. 2002.
[5] S. Y. Semenov, A. E. Bulyshev, A. E. Souvorov, A. G. Nazarov,
Y. E. Sizov, R. H. Svenson, V. G. Posukh, A. Pavlovsky, P. N. Repin,
and G. P. Tatsis, “Three-dimensional microwave tomography: Experimen-
tal imaging of phantoms and biological objects,” IEEE Trans. Microw.
Theory Tech., vol. 48, no. 6, pp. 1071–1074, Jun. 2000.
[6] M. Gustafsson and S. He, “A wave-splitting based optimisation approach
to multi-dimensional time-domain electromagnetic inverse problems,”
Math. Comput. Simul., vol. 50, no. 5/6, pp. 541–551, Dec. 1999.
[7] N. Joachimowicz, C. Pichot, and J. P. Hugonin, “Inverse scattering: An
iterative numerical method for electromagnetic imaging,” IEEE Trans.
Antennas Propag., vol. 39, no. 12, pp. 1742–1752, Dec. 1991.
[8] S. Caorsi, M. Donelli, and A. Massa, “Detection, location, and imaging of
multiple scatterers by means of the iterative multiscaling method,” IEEE
Trans. Microw. Theory Tech., vol. 52, no. 4, pp. 1217–1228, Apr. 2004.
[9] A. Massa, M. Pastorino, and A. Randazzo, “Reconstruction of two-
dimensional buried objects by a differential evolution method,” Inverse
Probl., vol. 20, no. 6, pp. 135–150, Dec. 2004.
[10] A. E. Souvorov, A. E. Bulyshev, S. Y. Semenov, R. H. Svensson,
A. G. Nazarov, Y. E. Sizov, and G. P. Tatsis, “Microwave tomography:
A two-dimensional Newton iterative scheme,” IEEE Trans. Microw. The-
ory Tech., vol. 46, no. 11, pp. 1654–1659, Nov. 1998.
[11] T. A. Maniatis, K. S. Nikita, and N. K. Uzunoglu, “Two-dimensional di-
electric profile reconstruction based on spectral-domain moment method
and nonlinear optimization,” IEEE Trans. Microw. Theory Tech., vol. 48,
no. 11, pp. 1831–1840, Nov. 2000.
[12] P. M Van den Berg and A. Abubakar, “Inverse scattering algorithms based
on contrast source integral representations,” Inverse Probl. Eng., vol. 10,
no. 6, pp. 559–576, 2002.
[13] H. Z. Takashi, T. Takenaka, and T. Tanaka, “Three-dimensional recon-
struction of a shallowly buried mine using time-domain data,” Microw.
Opt. Technol. Lett., vol. 39, no. 4, pp. 276–280, 2003.
[14] R. K. Mueller, M. Kaveh, and R. D. Inverson, “A new approach to
acoustic tomography using diffraction techniques,” Acoust. Imaging,
vol. 8, pp. 615–628, 1980.
[15] E. Wolf, “Three-dimensional structure determination of semi-transparent
objects from holographic data,” Opt. Commun., vol. 1, no. 4, pp. 153–156,
Sep. 1969.
[16] M. Kaveh, M. Sumekh, and J. Greenleaf, “Signal processing for dif-
fraction tomography,” IEEE Trans. Sonics Ultrason., vol. SU-31, no. 4,
pp. 230–239, Jul. 1984.
[17] A. J. Devaney, “A filtered backpropagation algorithm for diffraction to-
mography,” Ultrason. Imag., vol. 4, no. 4, pp. 336–350, Oct. 1982.
[18] P. S. Carney and J. C. Scotland, “Inverse scattering for near-field mi-
croscopy,” Appl. Phys. Lett., vol. 77, no. 18, pp. 2798–2800, Oct. 2000.
[19] T. J. Cui, W. C. Chew, X. X. Yin, and W. Hong, “Study of resolution
and super resolution in electromagnetic imaging for half-space problems,”
IEEE Trans. Antennas Propag., vol. 52, no. 6, pp. 1398–1411, Jun. 2004.
[20] A. J. Devaney and G. Beylkin, “Diffraction tomography using arbitrary
transmitter and receiver surfaces,” Ultrason. Imaging, vol. 6, no. 2,
pp. 181–193, Apr. 1984.
[21] D. Nahamoo, S. X. Pan, and A. C. Kak, “Synthetic aperture dif-
fraction tomography and its interpolation-free computer implementa-
tion,” IEEE Trans. Sonics Ultrason., vol. SU-31, no. 4, pp. 218–229,
Jul. 1984.
[22] M. A. Anastasio and X. Pan, “An improved reconstruction algorithm for
3-D diffraction tomography using spherical wave sources,” IEEE Trans.
Biomed. Eng., vol. 50, no. 4, pp. 517–521, Apr. 2003.
[23] C. N. Kechribaris, K. S. Nikita, and N. K. Uzunoglu, “Reconstruction
of two-dimensional permittivity distribution using an improved Rytov
approximation and nonlinear optimisation,” J. Electromagn. Waves Appl.,
vol. 17, no. 2, pp. 183–207, 2003.
[24] N. Sponheim and I. Johansen, “Experimental results in ultrasonic tomog-
raphy using a filtered backpropagation algorithm,” Ultrason. Tomography,
vol. 13, pp. 56–70, 1991.
[25] N. Sponheim, L. J. Gelius, I. Johansen, and J. J. Stamnes, “Quantitive
results in ultrasonic tomography of large objects using line sources and
curved detector arrays,” IEEE Trans. Ultrason., Ferroelectr., Freq. Con-
trol, vol. 38, no. 4, pp. 370–379, Jul. 1991.
[26] X. Pan and M. A. Anastasio, “Minimal-scan filtered backpropagation
algorithms for diffraction tomography,” J. Opt. Soc. Amer. A, Opt. Image
Sci., vol. 16, no. 12, pp. 2896–2903, Dec. 1999.
[27] M. A. Anastasio and X. Pan, “A new reconstruction approach for reflec-
tion mode diffraction tomography,” IEEE Trans. Image Process.,vol.9,
no. 7, pp. 1262–1271, Jul. 2000.
[28] T. J. Cui and W. C. Chew, “Diffraction tomographic algorithm for the
detection of three-dimensional objects buried in lossy half-space,” IEEE
Trans. Antennas Propag., vol. 50, no. 1, pp. 42–49, Jan. 2002.
[29] O. R. Halse, J. J. Stamnes, and A. J. Devaney, “Three-dimensional dif-
fraction tomography by two-dimensional sectioning,” Opt. Commun.,
vol. 224, no. 4–6, pp. 185–195, Sep. 2003.
[30] M. A. Anastasio and X. Pan, “Computationally efficient and statistically
robust image reconstruction in three-dimensional diffraction tomogra-
phy,” J. Opt. Soc. Amer. A, Opt. Image Sci., vol. 17, no. 3, pp. 391–400,
Mar. 2000.
[31] A. T. Vouldis, C. N. Kechribaris, T. A. Maniatis, K. S. Nikita, and
N. K. Uzunoglu, “Investigating the enhancement of three-dimensional
diffraction tomography by using multiple illumination planes,” J. Opt.
Soc. Amer. A, Opt. Image Sci., vol. 22, no. 7, pp. 1251–1262, Jul. 2005.
[32] B. A. Lippmann and J. Schwinger, “Variational principles for scattering
processes,” Phys. Rev., vol. 79, no. 3, pp. 469–480, Aug. 1950.
[33] Á. C. Kak and M. Slaney, Principles of Computerized Tomographic Imag-
ing. New York: IEEE Press, 1988.
[34] D. T. Borup and O. P. Gandhi, “Fast-Fourier-transform method for calcu-
lation of SAR distributions in finely discretized inhomogeneous models
of biological bodies,” IEEE Trans. Microw. Theory Tech., vol. MTT-32,
no. 4, pp. 355–360, Apr. 1984.
Angelos T. Vouldis was born in Athens, Greece, in 1976. He received the
Diploma degree in electrical and computer engineering from National Tech-
nical University of Athens (NTUA), Greece, in 2000, where he is currently
working toward the Ph.D. degree in the School of Electrical and Computer
Engineering.
Since 2001, he has been a Researcher with the Biomedical Imaging and Sim-
ulations Laboratory, School of Electrical and Computer Engineering, NTUA,
Greece. His main research interests include the development of algorithms for
solving forward and inverse-scattering problems in acoustics and electromag-
netics, and parallel computing.
Costas N. Kechribaris was born in Athens, Greece, in 1972. He received the
B.Sc. degree in physics from National and Kapodistrian University of Athens in
1995, the M.Sc. degree in biomedical engineering from University of Aberdeen,
U.K., in 1996, and the Ph.D. degree in biomedical engineering from National
Technical University of Athens in 2002.
He has been a Senior Engineer with the Spectrum Monitoring Department
of National Telecommunications and Post Commission of Greece since 2003.
His research interests include inverse scattering, ultrasonics and computational
electromagnetics, and medical imaging.
Theofanis A. Maniatis (M’00) was born in Athens, Greece, in 1967. He
received the Diploma degree in electrical engineering from National Technical
University of Athens in 1990, the M.A.Sc. degree from University of Toronto,
ON, Canada, in 1993, and the Ph.D. degree from National Technical University
of Athens in 1998.
From 1999 to 2003, he was a Researcher with the Institute of Communica-
tion and Computer Systems of the National Technical University of Athens.
Currently, he is with the National Communications and Post Commission
of Greece. His research interests include inverse scattering, computational
electromagnetics, and medical imaging.
Dr. Maniatis is a member of the Technical Chamber of Greece and the
Hellenic Society of Biomedical Engineering.
1984 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 55, NO. 6, DECEMBER 2006
Konstantina S. Nikita (M’96–SM’00) received the
Diploma degree in electrical engineering and the
Ph.D. degree from National Technical University of
Athens (NTUA), Greece, in 1986 and 1990, respec-
tively, and the M.D. degree from the Medical School,
University of Athens, in 1993.
Since 1990, she has been working as a Researcher
with the Institute of Communication and Computer
Systems, NTUA. In 1996, she joined the Department
of Electrical and Computer Engineering, NTUA,
where she is currently a Professor. Her current
research interests include medical imaging, biomedical signal and image
processing and analysis, biomedical informatics, health telematics, simulation
of physiological systems, computational bioelectromagnetics, biological ef-
fects, and medical applications of electromagnetic waves. She has authored or
coauthored 90 papers in refereed international journals and chapters in books,
and more than 150 papers in international conference proceedings. She was
the coauthor of one book in Greek and the coeditor of one book in English
published by Springer. She holds two Greek patents. She has been the Technical
Manager with several European and National Research and Development
Projects in the field of biomedical engineering.
Dr. Nikita is a member of the Technical Chamber of Greece, the Athens
Medical Association, and the Hellenic Society of Biomedical Engineering.
She was the recipient of the 2003 Bodossakis Foundation Academic Prize
for exceptional achievements in “Theory and Applications of Information
Technology in Medicine.”
Nikolaos K. Uzunoglu (F’06) was born in Constantinople, Turkey, in 1951. He
received the B.Eng. degree in electronics from Istanbul Technical University,
Turkey, in 1973, the M.Sc. degree in quantum electronics and the Ph.D. degree
in electrical engineering science from University of Essex, U.K., both in 1976,
and the Academic degree of Doctor of Science in the fields of electrooptic
systems, in 1981.
From 1977 to 1984, he worked with the Research Center of the Hellenic
Navy. He was elected as an Associate Professor in 1984 and a Professor in 1987
with the School of Electrical Engineering at the National Technical University
of Athens, Greece. From 1988 to 1994, he served as the Dean with the School
of Electrical and Computer Engineering. From 1991 to 1999, he was the first
Director with the Institute of Communication and Computer Systems, which is
an Academic Research Institute of the National Technical University of Athens.
He has been a Coordinator with several European Commission and Domestic
Research Projects. He has supervised 60 Ph.D. candidates in his research
fields including electromagnetic theory applications, telecommunications, radar
systems, biomedical engineering, and telematics systems. He has authored and
coauthored 215 papers in international reviewed journals, eight books, and
several conference papers.
Dr. Uzunoglu is a member of the National Academy of Sciences of Re-
public of Armenia and Honorary Professor of State Engineering University of
Armenian. He was awarded the International Award G. Marconi in 1981 for his
work on scattering from inhomogeneities inside optical fiber waveguides.
