IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 11, NO.

4, JULY/AUGUST 2005 759

Recent Progress in Exact and Reduced-Order

Modeling of Light-Scattering Properties
of Complex Structures
Xu Li, Member, IEEE, Allen Taflove, Fellow, IEEE, and Vadim Backman

Abstract—An emerging research area in biophotonics with po- not focus on providing a spatial map of the macroscopic op-
tentially near-term clinical applications in early stage cancer de- tical properties of the tissue under test. Instead, they seek to
tection involves the investigation of possible correlations of the detect minute cellular physiological differences, such as those
elastic light scattering properties of tissues with alterations in their
cellular composition and nanostructure. Until recently, exploring between normal and cancerous/preinvasive cells, by examining
these correlations has been impeded by a lack of robust and ac- the changes in spectral, angular, and polarization characteristics
curate mathematical models of the light scattering properties of of light scattered from tissue. By providing information about
complex structures. In this paper, we review recent progress in this tissue and cellular structures on the micro/nano scales, light
area. Topics include: 1) development of accurate reduced-order scattering based techniques may offer improved diagnostic sen-
expressions for the total scattering cross section spectra of a wide
range of nonspherical and inhomogeneous particles; 2) rigorous sitivity in living tissue. This realization has motivated increased
finite-difference time-domain modeling results showing how the interest in utilizing elastic light scattering as a diagnostic mech-
backscattering of light can be sensitive to nanometer scale features anism, and incorporating the scattering features into traditional
embedded within micrometer-scale particles; and 3) development imaging based approaches to enhance their accuracy and sen-
of accurate reduced-order expressions for the backscattering depo- sitivity. Representative recent developments include optical co-
larization properties of a wide range of inhomogeneous particles.
These advances provide an improved science base for cellular level herence tomography enhanced by spectroscopic analysis [9] and
biophotonics, and have promise to accelerate the development of static light scattering microscopy [10].
novel corresponding clinical technologies. For any technique relying on elastic light scattering as a di-
Index Terms—Cancer diagnosis, finite difference time domain agnostic mechanism, data analysis and correlation of the light
(FDTD), light-scattering spectroscopy, scattering. scattering properties of tissues with alterations in their cellular
composition and nanostructure plays an essential role in the di-
agnostic process. Conventionally, the interpretation of measured
I. INTRODUCTION data involves extracting the size, refractive index, and density
URING the past decade, it has become apparent that the of tissue structures such as cell nuclei and other organelles by
D analysis of the elastic light scattering characteristics of liv-
ing tissue can provide valuable diagnostic information. Prelim-
comparing the observed scattering features with Mie theory,
which is an analytical solution of light scattering by homoge-
inary data has suggested that alteration in these light scattering neous spherical particles. This approach has been demonstrated
signatures may be used as a sensitive marker to detect neoplasia to be a powerful tool to provide quantitative information on cel-
at an earlier stage than possible by histological analysis, where lular structures [2], [11]. However, this approach clearly relies
some crucial information may be lost due to tissue fixing, stain- upon the approximation of the cellular scattering particles as ho-
ing, and the limited resolution of microscopes [1]. A number of mogeneous spheres. This approximation cannot account for the
techniques, such as elastic light-scattering spectroscopy [1]–[5], complexity of the structures of biological tissue. Furthermore,
angle-resolved low coherence interferometry [6], [7], and the re- it cannot account for the subtle and complex cellular changes
cently proposed coherent backscattering spectroscopy [8], have during the initial onset of the malignancy, including changes in
been developed to exploit the elastic light scattering properties the morphology and internal texture characteristics of the nuclei
of tissue for disease diagnosis. In general, these techniques do and organelles.
In order to fully utilize the potential of elastic light scattering
for disease diagnosis, it is important to understand the mecha-
Manuscript received December 31, 2004; revised August 4, 2005. This work nism by which changes in shape and internal structure of nuclei
was supported by the National Science Foundation under Grants BES-0238903 and cell organelles affect light scattering. This understanding
and ACI-0219925. The computational resource was provided by the NSF Tera-
grid under Project TG-MCB 040062N. will provide researchers with fundamental insights regarding
X. Li is with the Department of Biomedical Engineering and the Depart- the means to improve techniques and data analysis methods to
ment of Electrical Engineering and Computer Science, Northwestern University, maximize diagnostic sensitivity and accuracy. For example, it
Evanston, IL 60208 USA (e-mail: [email protected]).
A. Taflove is with the Department of Electrical Engineering and Com- is important to know which tissue properties can or cannot be
puter Science, Northwestern University, Evanston, IL 60208 USA (e-mail: determined by use of light scattering. From a clinical standpoint,
taflove@ece. northwestern.edu). if certain measures of the light scattering signal from tissues can
V. Backman is with the Department of Biomedical Engineering, Northwestern
University, Evanston, IL 60208 USA (e-mail: [email protected]). be unambiguously and reliably associated with disease-specific
Digital Object Identifier 10.1109/JSTQE.2005.857691 cellular changes, this knowledge will facilitate the development

1077-260X/$20.00 © 2005 IEEE

760 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 11, NO. 4, JULY/AUGUST 2005

and acceptance of optical methods for screening and diagnosis

of patients and for preclinical/basic biological studies.
Recent advancements in numerical methods of solving
Maxwell’s equations have made it possible to calculate light
scattering by inhomogeneous objects of arbitrary shape and
internal organization. In particular, the application of the finite-
difference time-domain (FDTD) method [12] has yielded fruit-
ful results in modeling light scattering by individual cells
[13]–[15]. However, significant computational resources re-
quired by the FDTD method, or other exact numerical solutions
of Maxwell equations for arbitrary geometries, do not allow for
the calculation of light scattering by macroscopic tissue struc-
tures. The large computational requirement also poses a signifi-
cant challenge in utilizing exact modeling to address the inverse
Fig. 1. Examples of Gaussian random spheres with fixed r0 = 1.75 µm.
problem, wherein multiple solutions of the forward problem (a)–(c) Gaussian spheres with increasing δr (Γ is fixed at 70◦ ). (d)–(f)
are often required. A more feasible approach to calculate light Gaussian spheres with decreasing Γ (δr is fixed at 0.1). (a) δr = 0.1, Γ =
scattering by complex tissue structure is to combine numerical 70◦ , (b) δr = 0.5, Γ = 70◦ , (c) δr = 0.9, Γ = 70◦ , (d) δr = 0.1, Γ = 90◦ ,
(e) δr = 0.1, Γ = 30◦ , (f) δr = 0.1, Γ = 10◦ .
modeling with reduced-order analytical models.
In this paper, we review recent progress in both exact numer-
ical modeling and reduced-order analytical approximations for r(ϑ,

ϕ) and normalized standard deviation of the radius δr =
light scattering by particles with complex shapes and internal (r(ϑ, ϕ) − R)2 /r0 . Using a modified Gaussian correlation
structure. Our objective here is to provide an overview of rele- function, the correlation between two radii over solid angle Ω is
vant principles, techniques, and current research by coherently  
sin2 (Ω/2)
integrating the body of published work with new results. In Sec- Cr (Ω) = exp − (1)
2 sin2 (Γ/2)
tion II, we introduce two stochastic models, the Gaussian ran-
dom sphere (GRS) model and the Gaussian random field (GRF) where Γ is the correlation angle of the Gaussian sphere, defined
model, for describing and synthesizing irregular shapes and as √
the angular displacement over which the correlation drops to
internal structures. These geometric models provide test beds 1/ e. The shape characteristics of a Gaussian random sphere
for the subsequent theoretical development. Section III presents are uniquely determined by the statistical parameters r0 , δr ,
the development of an accurate reduced-order expression, the and Γ.
equiphase-sphere (EPS) approximation, for the total scattering We generate the three-dimensional (3-D) geometry of Gaus-
cross section (TSCS) spectra of a variety of nonspherical and sian spheres using a computer program based on the code
inhomogeneous particles. We demonstrate that a wide range developed by Muinonen and Nousiainen [17]. As shown in
of particles with subwavelength perturbations in their shapes Fig. 1(a)–(c), increasing δr results in an increased deforma-
or internal inhomogeneities have TSCS spectra similar to their tion of the particle shape from a sphere. On the other hand,
equiphase-sphere counterparts. In Section IV, we present FDTD as illustrated in Fig. 1(d)–(f), reducing Γ leads to increased
modeling results demonstrating that the backscattering of light short-distance fluctuations (increased numbers of “valleys” and
can be sensitive to nanometer-scale features within the particles. “hills”) on the particle surface.
Finally, Section V presents the development of a reduced-order Similarly, the GRF model [18] mathematically describes a
expression for the backscattering depolarization properties of random process having a Gaussian probability density function.
inhomogeneous particles. Here, we consider the refractive index n (r) as a function of
spatial location r = (x, y, z). Each value of n (r) is a Gaussian
random variable with
mean n0 = n (r) and normalized stan-
II. STOCHASTIC METHODS FOR MODELING THE GEOMETRY OF dard deviation δn = (n (r) − n0 )2 /(n0 − 1). For a GRF
NONSPHERICAL AND INHOMOGENEOUS PARTICLES distribution with Gaussian function as the correlation model,
In order to investigate light scattering by particles with a wide the two-point correlation function Cn (r) is given by
variety of shapes and interior structures, statistical approaches
Cn (r) = e−r
2
/ (L c /2)2
(2)
are very useful for modeling the particle geometry [16]. This is
especially the case for investigating light scattering by biolog- where Lc is the characteristic correlation length representing
ical tissues, wherein shape irregularities and internal inhomo- the length scale over which the correlation drops to a negligible
geneities exist at a wide range of length scales. In this section, we level. For such choice of correlation function, the characteristics
introduce two stochastic models, the Gaussian random sphere of the spatial distribution of n (r) are determined by n0 , δn , and
(GRS) model to synthesize nonspherical shapes, and Gaussian Lc .
random field (GRF) model to synthesize inhomogeneous inter- Using the turning-band method [19], we create geometrical
nal textures. models of spherical particles with refractive index having GRF
The GRS model [16] is defined in spherical coordinates as distributions. Fig. 2 shows graphs of six representative parti-
having angle-dependent radius r(ϑ, ϕ) with mean radius r0 = cles with refractive index distributions synthesized by the GRF
LI et al.: RECENT PROGRESS IN EXACT AND REDUCED-ORDER MODELING OF LIGHT-SCATTERING PROPERTIES OF COMPLEX STRUCTURES 761

few wavelengths). Light scattering by particles within this size

range exhibits complicated dependencies on wavelength and
scattering angles, and cannot be characterized by Rayleigh or
Rayleigh–Gans approximations. Such particles are also of great
relevance to tissue optics, since many structures such as cell
nuclei are in this size range.
We have recently introduced the equiphase-sphere (EPS) ap-
proximation for calculating the TSCS spectra of nonspheri-
cal [20], [21] and inhomogeneous particles [22]. In the EPS
approximation, the wavelength dependent TSCS spectrum of a
(s)
particle is given by the sum of the “edge effect” term σs (λ)
(v )
and “volume diffraction effect” term σs (λ) [23], [24]

σs( λ) = σs(s) (λ) + σs(v ) (λ). (3)

Fig. 2. Examples of inhomogeneous spherical particles having GRF refractive-
index distributions with fixed n 0 = 1.1 and r0 = 2 µm. The x̂ − ẑ cross sec- (s)
tional cuts of the particles’ interior refractive-index distribution are mapped Here, σs (λ) can be approximated as [23]
in grayscale. (a)–(c) Inhomogeneous particles with increasing δn (L c is fixed
at 0.4 µm). (d)–(f) Inhomogeneous particles with decreasing L c (δn is fixed σs(s) (λ) ≈ 2S[2π(3V /4π 1/3 /λ ]−2/3 (4)
at 0.2). (a) δn = 0.12, Lc − 0.4 µm, (b) δn = 0.24, Lc − 0.4 µm, (c) δn =
0.32, Lc − 0.4 µm, (d) δn = 0.2, Lc − 1.2 µm, (e) δn = 0.2, Lc − 0.6 µm, where S is the particle’s maximum cross section area transverse
(f) δn = 0.2, Lc − 0.1 µm.
to the direction of the incident light, and V is the volume of the
particle.
model. In each example, we map the x̂ − ẑ cross sectional cut of Based on the Wentzel–Kramers–Brillouin (WKB) technique,
(v )
the particle’s interior refractive index distribution in grayscale. we approximate the volume term σs (λ) as [20], [22]
In Fig. 2(a)–(c), Lc is fixed at 0.4 µm while δn increases from
σs(v ) (γ) = 2S[1 − 2n0 sin ρ/ρ + 4n0 sin2 (ρ/2)/ρ2 ] (5)
0.12 [Fig. 2(a)] to 0.32 [Fig. 2(c)]. In Fig. 2(d)–(e), δn is fixed
at 0.2 while Lc decreases from 1.2 µm [Fig. 2(d)] to 0.1 µm where n0 is the volume averaged refractive index and ρ =
[Fig. 2(f)]. These examples demonstrate the capability of the 2πd(n0 − 1)/λ is the maximum phase-shift produced by the
GRF model to mimic refractive index fluctuations occurring “equiphase sphere” of the particle. For a nonspherical particle,
over a variety of geometrical scales. It is evident that δn de- the equiphase sphere is defined as the “best-fitting” ellipsoid of
scribes the magnitude of refractive index variability, while Lc the particle. We note that finding the best-fitting ellipsoid for an
characterizes the size of the internal features of the particle. arbitrary 3-D shape is generally a multiparameter optimization
Figs. 1 and 2 demonstrate that stochastic models such as problem with eight free parameters: three semiaxes (a, b, and
GRS and GRF are capable of representing complex shapes and c); three coordinates of the center (x0 , y0 , and z0 ); and two rota-
internal structures spanning a wide range of length scales which tional angles (θ0 and φ0 ). Here, we fix some of these parameters
are characteristic to cellular structures. These geometries are to simplify the optimization procedure. First, we specify semi-
also mathematically well parameterized, and thus are suitable axis c to be aligned with the incident wave vector ẑ. Thus we
to serve as test beds for generalized theoretical development. have θ0 = 0. The second constraint is to match the cross section
In our subsequent analyses and numerical experiments, we use area of the ellipsoid with the projected area of the particle in the
these geometries as generic representation of a wide range of x̂ − ŷ plane (Sp = π a b). Furthermore, the location of the geo-
nonspherical and inhomogeneous geometries. metric center (x0 , y0 , z0 ) is assigned to the center of mass of the
irregular particle. Therefore, we need to determine only three
III. DEVELOPMENT OF ACCURATE REDUCED-ORDER free parameters: the longitudinal semiaxis c, the aspect ratio of
EXPRESSIONS FOR THE TOTAL SCATTERING CROSS SECTION the cross section ηT = a/b, and the transverse rotational angle
SPECTRA OF NONSPHERICAL AND INHOMOGENEOUS φ0 (the angle between cross section major semiaxis a and x̂).
PARTICLES—THE EPS APPROXIMATION The objective of the optimization procedure is to minimize the
mean squared difference of the ẑ-directed light ray pathlength
The total scattering cross section (TSCS) is one of most im-
between the irregular particle and the corresponding ellipsoid.
portant parameters describing the light scattering properties of
Parameters ηT , c, and φ0 are chosen such that
particles. For a homogeneous sphere, the TSCS can be read-
ily calculated with Mie theory for any wavelength. For parti- arg(c, ηT , φ0 )|min{δ L r 2  } . (6)
cles with irregular shapes and internal structures, however, no
exact analytical solution is available to characterize the wave- After finding its best fitting ellipsoid and volume averaged
length dependence of their TSCS. For such particles, approxi- refractive index, the TSCS spectrum of the irregularly shaped
mation methods are desirable for providing practical solutions particle is then approximated by (3)–(5), where d = 2c.
to light-scattering problems. Our investigation has been focused In order to apply the EPS method in practice, it is important
on particles with sizes in the resonance range (on the order of a to determine the range of validity of this approximation. Also,
762 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 11, NO. 4, JULY/AUGUST 2005

based on the WKB technique, we have derived the validity con-

ditions of the EPS approximation as functions of the statistical
parameters of the particle geometry and internal structures. For
an inhomogeneous spherical particle with diameter 2r0 , nor-
malized refractive index standard deviation δn , and correlation
length Lc , the validity condition of the EPS approximation is
given by
√

βn ≡ 4 2r0 (n0 − 1) Lc δn /λ < 1. (7)
Similarly, we have determined the range of validity of the
EPS approximation for particles with irregular shapes. For a
nonspherical particle radial standard deviation ∆ from its best
fitting ellipsoid and radius-angular correlation angle Γ, the va-
lidity condition of the EPS approximation is given by

√
βr ≡ 2 2/π(n0 − 1) Γ∆/λ < 1. (8)
We note from (7) and (8) that two effects contribute to the
validity and accuracy of the EPS approximation for inhomoge-
neous and nonspherical particles: 1) the magnitude of the pertur-
bation in the shape or the inhomogeneity of the particle, repre-
sented by ∆ and δn (n0 − 1) respectively; and 2), the geometri-
cal scale of the surface or internal features, denoted by Γ and Lc .
The fact that β is proportional to Γ and Lc indicates that when
inhomogeneity or surface irregularity occurs on length scales
that are small compared to the incident wavelength, they have
less impact on the validity of the EPS approximation. In other
words, particles with smaller-scale perturbations in their shapes
or internal inhomogeneities have similar TSCS spectra as their
equiphase-sphere counterparts. This effect can be observed from
Figs. 3 and 4, where we compare the TSCS spectra calculated by
the EPS approximation with accurate FDTD benchmark data for
a variety of nonspherical and inhomogeneous particles. These
examples also demonstrate that the EPS approximation (3)–(5)
are capable of modeling the TSCS spectra for a wide range
of nonspherical and inhomogeneous particles when the validity
conditions given by (3) and (4) are satisfied. Furthermore, the Fig. 3. Demonstration of the accuracy of EPS approximation applied to non-
mathematical simplicity of these reduced-order expressions also spherical particles. TSCS spectra calculated by EPS approximation are com-
makes them especially appealing for being applied in inverse- pared to FDTD benchmark data. The incident light propagates in the ẑ di-
rection. (a)–(c) Demonstration that when βr < 1, the EPS approximation can
scattering problems [21]. give reasonable accuracy for calculating the TSCS spectra. (d) When βr > 1,
the validity of the EPS approximation is not guaranteed. (a) ∆ = 0.3 µm,
IV. SENSITIVITY OF BACKSCATTERING SIGNATURES TO Γ = 50◦ , β̄ r = 0.45, (b) ∆ = 0.7 µm, Γ = 10◦ , β̄ r = 0.48, (c) ∆ = 0.6 µm,
Γ = 20◦ , β̄ r = 0.57, (d) ∆ = 0.8 µm, Γ = 70◦ , β̄ r = 1.5.
NANOARCHITECTURE OF SCATTERING STRUCTURES
Our analysis and numerical examples, shown in Section III,
indicates that the TSCS spectra of nonspherical and inhomoge- Here, we demonstrate the sensitivity of backscattering sig-
neous dielectric particles are not sensitive to shape and texture natures to nanoscale parameters using the FDTD method. Prior
perturbations at length scales much smaller than the wavelength. to this investigation, we developed and validated a simple yet
An important question then arises for researchers investigat- effective modification to significantly improve the accuracy
ing optical tissue diagnostic techniques: are there any light- of the FDTD near-to-far field (NTFF) transformation for cal-
scattering parameters that provide sufficient sensitivity to detect culating the backscattering of strongly forward-scattering ob-
nanometer-scale cellular changes? This question is of partic- jects [25]. We use the modified FDTD-NTFF approach to cal-
ular interest, since recent clinical evidence has indicated that culate the spectral and angular distribution of backscattered
light scattering signals are extremely sensitive to minute differ- light from inhomogeneous dielectric particles with identical
ences in tissue and cellular structures [1]. As we will discuss, sizes and volume-averaged refractive indices, and compare the
contrary to the TSCS, light-scattering signals in the backward calculated scattering patterns with their homogeneous coun-
direction contain signatures that are sensitive to such nanoscale terpart. This comparison is illustrated in the middle panels of
perturbations. Fig. 5, where we plot the scattering intensity distribution as
LI et al.: RECENT PROGRESS IN EXACT AND REDUCED-ORDER MODELING OF LIGHT-SCATTERING PROPERTIES OF COMPLEX STRUCTURES 763

Fig. 5. Demonstration of the sensitivity of backscattering signatures to nanoar-

chitecture of scattering structures. The particle geometries are shown in the
left panel. FDTD-calculated backscattering intensity distribution over wave-
length and scattering angle are displayed in the middle panel. The forward
scattering signatures are displayed in the right panel. All three particles have
n 0 = 1.1 and r0 = 2 µm. (a) Homogeneous particle. (b) Inhomogeneous par-
ticle with Lc = 50 nm}, δn (n 0 − 1) = 0.02. (c) Inhomogeneous particle with
Lc = 100 nm δn (n 0 − 1) = 0.02.

particles, the forward and small angle scattering is the major

contribution to the TSCS spectra, which are not sensitive to
small scale structures, as shown in Section III.

V. QUANTITATIVE ANALYSIS OF DEPOLARIZATION OF LIGHT

BACKSCATTERING BY INHOMOGENEOUS PARTICLES
Fig. 4. TSCS spectra calculated using rigorous FDTD numerical modeling In addition to spectral and angular properties, the polariza-
and EPS analyses for inhomogeneous particles. The Spatial distribution of the
particle refractive index in the x̂ − ẑ cross sectional cut is displayed in the left tion properties of light backscattered by biological tissue can
panel. (a)–(c) Demonstration that when βn < 1, the EPS approximation can give also be used as diagnostically valuable markers. Recently, we
reasonable accuracy for calculating the TSCS spectra. (d) When βn > 1, the have investigated the depolarization effect of dielectric particles
validity of the EPS approximation is not guaranteed. (a) δn = 0.2, Lc = 50 nm,
β̄ n = 0.36, (b) δn = 0.16, Lc = 200 nm, β̄ n = 0.57, (c) δn = 0.16, Lc = having complex internal structures by examining the distribution
600 nm, β̄ n = 0.98, (d) δn = 0.32, Lc = 1000 nm, β̄ n = 2.5. of optical paths and the associated phase changes of light rays
propagating in the particle [26]. Our theoretical analyses and
numerical experiments demonstrate that the backscattered lin-
functions of wavelength and scattering angle (centered at the ear depolarization ratio is directly associated with the statistical
backscattering direction). Distinct scattering features are evident parameters of the particle’s internal geometry. Specifically, for
in the scattering fingerprints from the two inhomogeneous par- an inhomogeneous particle having an internal refractive index
ticles, although their inhomogeneities have characteristic sizes distribution with standard deviation σn = δn (n0 − 1) and cor-
much smaller than the illumination wavelength (Lc = 50 and relation length Lc , the linear depolarization in the backscattered
100 nm, respectively, while λ̄ = 750 nm). These numerical re- light δλ ≡ I⊥ /I is given by
sults strongly support the hypothesis that there exit signatures
in backscattered light that are sensitive enough to detect alter- δλ ≈ C (2π/λ2 )(LC σn )2 (9)
ations in nanoscale architectures. Importantly, this sensitivity is
not limited by the diffraction limit. Potentially, backscattering where C is a constant independent of the distribution of the
signatures can serve as biomarkers to detect and characterize internal refractive index.
slight alterations in tissue structure. For comparison, we also Fig. 6 shows three representative results of our numerical ex-
plot the scattering intensity distribution in the forward direc- periments where backscattered intensities in both polarizations
tion. Here, no distinct scattering features are observed among (I and I⊥ ) are calculated for inhomogeneous particles. Ignoring
the different scatterers. The similarity in the forward scattering the oscillatory structures due to resonances in the backscattered
pattern is not surprising. For these strongly forward-scattering spectra, it is clear that the overall level of I⊥ , and therefore the
764 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 11, NO. 4, JULY/AUGUST 2005

Fig. 7. Confirmation of the validity of (9); i.e., that δλ is approximately pro-

portional to (2π/λ)(LC σ n )2 . Here, averaged linear depolarization ratio are
calculated from FDTD simulations conducted on 20 inhomogeneous dielectric
spheres as a function of the geometrical parameter β ≡ C (2π/λ̄ 2 )(L C σ n )2
with L c ranging from 0.05 to 1.2 µm, and σ n ranging from 0.005 to 0.035.
Fig. 6. Demonstration that the overall intensity of cross-polarized backscat-
In the calculation of β, we choose C = 12 and λ̄ = 750 nm. The grayscale
tering light increases as L c σ n becomes greater. Here, copolarized and cross-
level of each data point represents the magnitude of σ n with the brighter shades
polarized backscattering light intensity calculated by FDTD method for in-
corresponding to greater σ n s.
homogeneous dielectric spheres with n 0 = 1.1, r0 = 2 µm, and a variety of
L c and σ n . (a) L c = 0.05 µm, σ n = 0.02; (b) L c = 0.4 µm, σ n = 0.016;
(c) L c = 1.2 µm, σ n = 0.035.

istic of cellular structures. Using these model geometries as test

linear depolarization ratio δλ , increases as the product of Lc and beds, we have developed a reduced-order expression, the EPS
σn becomes larger. approximation, for calculating the TSCS spectra of complex
The dependence of δλ on the geometric characteristics of shapes. We have also developed a reduced-order expression to
the particle’s internal refractive-index distribution is most ev- quantitatively analyze the depolarization of light backscattering
ident in Fig. 7, where we summarize our numerical experi- by inhomogeneous particles. These analytical approximations
ments conducted on 20 inhomogeneous dielectric spheres with have simple mathematical forms, and thus can be use to address
Lc ranging from 0.05 µm to 1.2 µm and σn ranging from the inverse problem to extract certain geometrical parameters of
0.005 to 0.036. Here, we graph δλ averaged over the 500– the scatterers from measured scattering data.
1000 nm incident wavelength range against the geometric pa- The validity range of the EPS approximation indicates that
rameter β̄ ≡ C(2π/λ̄2 )(LC σn )2 in a log-log plot to cover a a variety of particles with subwavelength perturbations in their
wide range for both parameters. The constant C is chosen em- shapes or internal inhomogeneities have TSCS spectra similar to
pirically as C = 12 to give the best linearity. The 20 data points their equiphase-sphere counterparts. However, as demonstrated
are plotted with symbols representing different Lc and grayscale by our FDTD numerical modeling, there exist scattering signa-
levels corresponding to different σn for each particle geometry. tures in backscattered light that are sensitive to nanometer-scale
We also cross reference three data points with their correspond- features embedded within micron-scale particles. These results
ing backscattered spectra shown in Figs. 6(a)–(c). The relation- substantiate previous experimental observations that backscat-
ship given in (9) is plotted in dashed line to compare against tered light signals may be used to detect minute changes in
the data points. The reasonable data fitting confirms the validity cellular structures occurring in the early stage of carcinogenesis.
of (9); i.e., that δλ is proportional to (LC σn )2 /λ2 for the first The advances presented in this paper may provide an im-
order of approximation. proved science base for cellular-level biophotonics, and have
promised to accelerate the development of novel corresponding
clinical technologies by providing insights on the correlations of
VI. SUMMARY AND DISCUSSION
the elastic light scattering properties of tissues with alterations
In this paper, we have reviewed recent progress in analytical in their cellular composition and nanostructure.
and numerical modeling of light scattering by particles hav-
ing complex shapes and internal structures. We have introduced
stochastic geometric descriptions, namely the Gaussian random
ACKNOWLEDGMENT
sphere (GRS) and the Gaussian random field (GRF) models, for
characterizing and synthesizing irregular shapes and internal X. Li would like to thank Dr. Y. L. Kim and Y. Liu for helpful
structures in a wide range of length scales which are character- discussions on elastic light scattering for tissue diagnosis.
LI et al.: RECENT PROGRESS IN EXACT AND REDUCED-ORDER MODELING OF LIGHT-SCATTERING PROPERTIES OF COMPLEX STRUCTURES 765

REFERENCES [21] X. Li, Z. Chen, J. Gong, A. Taflove, and V. Backman, “Analytical tech-
niques for addressing forward and inverse problems of light scattering by
[1] H. K. Roy, Y. Liu, R. K. Wali, Y. L. Kim, A. K. Kromin, M. J. Goldberg, irregularly shaped particles,” Opt. Lett., vol. 29, no. 11, pp. 239–1241,
and V. Backman, “Four-dimensional elastic light-scattering fingerprints 2004.
as preneoplastic markers in the rat model of colon carcinogenesis,” Gas- [22] X. Li, Z. Chen, A. Taflove, and V. Backman, “Equiphase sphere approx-
troenterology, vol. 126, pp. 1071–1081, 2004. imation for light scattering by stochastically inhomogeneous particles,”
[2] V. Backman, M. B. Wallace, L. T. Perelman, J. T. Arendt, R. Gurjar, Phy. Rev. E, vol. 70, pp. 056611–056618, 2004.
M. G. Muller, Q. Zhang, G. Zonios, E. Kline, T. McGillican, S. Shapshay, [23] Z. Chen, A. Taflove, and V. Backman, “Equivalent volume-averaged light
T. Valdez, K. Badizadegan, J. M. Crawford, M. Fitzmaurice, S. Kabani, scattering behavior of randomly inhomogeneous dielectric spheres in the
H. S. Levin, M. Seiler, R. R. Dasari, I. Itzkan, J. Van Dam, and M. S. Feld, resonant range,” Opt. Lett., vol. 28, no. 10, pp. 765–767, 2003.
“Detection of preinvasive cancer cells,” Nature, vol. 406, no. 6791, pp. 35– [24] Z. Chen, A. Taflove, and V. Backman, “Concept of the equiphase sphere
36, 2000. for light scattering by nonspherical dielectric particles,” J. Opt. Soc. Amer.
[3] J. R. Mourant, T. M. Johnson, S. Carpenter, A. Guerra, T. Aida, and A, vol. 21, no. 1, pp. 88–97, 2004.
J. P. Freyer, “Polarized angular dependent spectroscopy of epithelial cells [25] X. Li, A. Taflove, and V. Backman, “Modified FDTD near-to-far
and epithelial cell nuclei to determine the size scale of scattering struc- field transformation for improved backscattering calculation of strongly
tures,” J. Biomed. Opt., vol. 7, no. 3, pp. 378–387, 2002. forward-scattering objects,” IEEE Antennas Wireless Propag. Lett., vol. 4,
[4] Y. L. Kim, Y. Liu, R. K. Wali, H. K. Roy, M. J. Goldberg, A. K. Kromin, no. 1, pp. 35–38, 2005.
K. Chen, and V. Backman, “Simultaneous measurement of angular and [26] X. Li, A. Taflove, and V. Backman, “Quantitative analysis of the depolar-
spectral properties of light scattering for characterization of tissue mi- ization of backscattered light by stochastically inhomogeneous dielectric
croarchitecture and its alteration in early precancer,” IEEE J. Sel. Topics particles,” Opt. Lett., vol. 30, no. 8, pp. 902–904, 2005.
Quantum Electron., vol. 9, no. 2, pp. 243–256, Mar./Apr. 2003.
[5] L. B. Lovat, K. Johnson, M. R. Novelli, M. O’Donovan, S. Davies, C.
R. Selvasekar, S. Thorpe, I. J. Bigio, and S. G. Bown, “Optical biopsy Xu Li (S’99–M’03) received the M.S. degree in
using elastic scattering spectroscopy can detect high grade dysplasia and biomedical engineering in 2000 and the Ph.D. de-
cancer in Barrett’s esophagus,” Gastroenterology, vol. 126, no. 4, pp. A22– gree in electrical and computer engineering in 2003,
A22, 2004. both from the University of Wisconsin-Madison.
[6] A. Wax, C. H. Yang, M. G. Muller, R. Nines, C. W. Boone, V. E. Steele, From 2003 to 2005, she was a Postdoctoral Fel-
G. D. Stoner, R. R. Dasari, and M. S. Feld, “In situ detection of neoplastic low and subsequently a Research Associate in the
transformation and chemopreventive effects in rat esophagus epithelium Department of Biomedical Engineering, Northwest-
using angle-resolved low-coherence interferometry,” Cancer Res., vol. 63, ern University, Evanston, IL. She is now an Assistant
no. 13, pp. 3556–3559, 2003. Professor in the Department of Biomedical Engineer-
[7] J. W. Pyhtila, R. N. Graf, and A. Wax, “Determining nuclear morphology ing and the Department of Electrical and Computer
using an improved angle-resolved low coherence interferometry system,” Engineering, Northwestern University. Her research
Opt. Express, vol. 11, no. 25, pp. 3473–3484, 2003. interests include computational and experimental electromagnetics, microwave
[8] Y. L. Kim, Y. Liu, V. M. Turzhitsky, H. K. Roy, R. K. Wali, and imaging and sensing techniques for biomedical applications, biophotonics, and
V. Backman, “Coherent backscattering spectroscopy,” Opt. Lett., vol. 29, electrodynamics of nanophotonic devices.
no. 16, pp. 1906–1908, 2004.
[9] D. C. Adler, T. H. Ko, P. R. Herz, and J. G. Fujimoto, “Optical coher-
ence tomography contrast enhancement using spectroscopic analysis with
Allen Taflove (F’90) received the B.S., M.S. and
spectral autocorrelation,” Opt. Express, vol. 12, no. 22, pp. 5487–5501, Ph.D. degrees in electrical engineering from North-
2004.
western University, Evanston, IL, in 1971, 1972, and
[10] A. K. Popp, M. T. Valentine, P. D. Kaplan, and D. A. Weitz, “Microscopic
1975, respectively.
origin of light scattering in tissue,” Appl. Opt., vol. 42, no. 16, pp. 2871–
He has been a Professor in the Department of
2880, 2003. Electrical and Computer Engineering, Northwestern
[11] L. T. Perelman, V. Backman, M. Wallace, G. Zonios, R. Manoharan,
University since 1984. Since 1972, he has pioneered
A. Nusrat, S. Shields, M. Seiler, C. Lima, T. Hamano, I. Itzkan, J. Van Dam,
basic theoretical approaches and engineering applica-
J. M. Crawford, and M. S. Feld, “Observation of periodic fine structure in
tions of finite difference time domain computational
reflectance from biological tissue: A new technique for measuring nuclear
electrodynamics. He coined the FDTD acronym in a
size distribution,” Phys. Rev. Lett., vol. 80, p. 627, 1998.
1980 IEEE paper, and in 1990 was the first person
[12] A. Taflove and S. Hagness, Computational Electrodynamics: The Finite-
to be named a Fellow of the IEEE in the FDTD area. He has authored or co-
Difference Time-Domain Method. 3rd ed. Boston, MA: Artech, 2005. authored five books, 20 book chapters and articles, over 100 refereed journal
[13] A. Dunn and R. Richards-Kortum, “Three-dimensional computation of
papers and 300 conference papers, and 14 U.S. patents. He has been the adviser
light scattering from cells,” IEEE J. Sel. Topics Quantum Electron., vol. 2,
or co-advisor of 20 Ph.D. recipients and one postdoctoral fellow, five of whom
no. 4, pp. 898–905, Dec. 1996.
(including four women) are now tenured or tenure-track university professors.
[14] R. Drezek, A. Dunn, and R. Richards-Kortum, “ight scattering from He is currently an elected member of Northwestern’s General Faculty Commit-
cells: Finite-difference time-domain simulations and goniometric mea-
tee, and is the graduate program Chair of his department. He is also the faculty
surements,” Appl. Opt., vol. 38, no. 16, pp. 3651–3661, 1999.
advisor to the Undergraduate Design Competition, the Honors Program in Un-
[15] D. Arifler, M. Guillaud, A. Carraro, A. Malpica, M. Follen, and
dergraduate Research, and the student chapters of Eta Kappa Nu and Tau Beta
R. Richards-Kortum, “Light scattering from normal and dysplastic cer- Pi. His efforts on behalf of students at all levels were recognized by Northwest-
vical cells at different epithelial depths: Finite-difference time-domain
ern in 2000, when he was named a Charles Deering McCormick Professor of
modeling with a perfectly matched layer boundary condition,” J. Biomed.
Teaching Excellence. His research interests span much of the electromagnetic
Opt., vol. 8, no. 3, pp. 484–494, 2003.
spectrum.
[16] K. Muinonen, “Light scattering by stochastically shaped particles,” in
Light Scattering by Nonspherical Particles: Theory, Measurements, and
Applications, M. I. Mishchenko, J. W. Hovenier, and L. D. Travis, Eds.
San Diego, CA: Academic, 2000. Vadim Backman received the Ph.D. degree in med-
[17] K. Muinonen and T. Nousiainen, G-Sphere, available under the GNU ical engineering and medical physics from Harvard
General Public License, 2002, http://www.meteo.helsinki.fi/˜tpnousia/ University and the Massachusetts Institute of Tech-
gsphere/index.html [Online]. nology, Cambridge, MA, in 2001.
[18] R. J. Adler, The Geometry of Random Fields. New York: Wiley, 1981. He is an Assistant Professor of biomedical engi-
[19] A. T. A. Wood and G. Chan, “Simulation of stationary Gaussian processes neering at Northwestern University, Evanston, IL. His
in [0, 1],” J. Comput. Graph. Stat., vol. 3, no. 4, pp. 409–432, 1994. research interests include biomedical optics, spec-
[20] X. Li, Z. Chen, A. Taflove, and V. Backman, “Equiphase-sphere approxi- troscopy, development of theoretical approaches to
mation for analysis of light scattering by arbitrarily-shaped nonspherical describe light propagation in biological media, and
particles,” Appl. Opt., vol. 43, no. 23, pp. 4497–4505, 2004. optical diagnostic imaging.