Bioelectromagnetics 43:90--105 (2022)

Optical Scattering fromVitreous Floaters

Stuart W. Harmer,1* Andrew J. Lu¡,2 and Giampaolo Gini3
1
Department of Engineering, Computing and Design, University of Chichester,
Chichester, UK
2
Sapphire Eyecare, Southampton, UK
3
Western Sussex NHS University Trust, Chichester, UK
Vitreous “floaters” are a common entoptic phenomenon that can result in significant reduction in
quality of life in a proportion of sufferers. The authors use a computational mathematical model
based on Fourier optics and reflection and transmission coefficients calculated for a planar type II
collagen opacity suspended in aqueous to show that floaters are perceived by the patient through
interference effects that result in significant variations in intensity on the retina when viewing a
constant brightness surface. The model also predicts that backscattered intensity from floaters is ten
thousand to one million times lower than the variations in intensity produced on the retina, which
demonstrates that the visible effects of floaters for the patient can be highly significant, whereas
clinical observation of the vitreous may be entirely unremarkable. Importantly, the results also
demonstrate that floaters do not need to be opaque to cause symptoms, with only small differences
in refractive index between the floater material and the surrounding vitreous needed to produce
significant optical effects. The model predicts that pupil size is an important factor in determining
the severity of symptoms from floaters, with constricted pupils giving much greater effect than
dilated pupils. Finally, the authors’ model predicts that floaters degrade contrast sensitivity function,
with greatest degradation occurring in the 5–40 cycles per degree spatial frequency range and that
the effects of shadowing caused by floaters are very strongly correlated to the predicted degradation
of contrast sensitivity function. Bioelectromagnetics. 43:90–105, 2022. © 2021 The Authors.
Bioelectromagnetics published by Wiley Periodicals LLC on behalf of Bioelectromagnetics
Society.
Keywords: floaters; vision; scattering; symptomatic vitreous opacity; contrast sensitivity
function

INTRODUCTION traditional view that chronic vitreous opacity is trivial and

of no medical consequence; the scientifically unsubstan-
Vitreous floaters are a common entoptic phenom-
tiated belief that vitreous opacity will spontaneously
enon that can have a considerable detrimental effect on
migrate out of the sufferer's visual axis or that the patient
the quality of life in a small but significant proportion of
will become accustomed to or tolerant of the visual
those that are symptomatic [Sebag, 2011; Wagle et al.,
disturbances due to their opacity and the historical lack of
2011; Zou et al., 2013; Luff and O'Donnell, 2018]. The
safe and effective treatments for SVO, although this is
reason for the wide spectrum of tolerance to floater
symptoms is unknown, but psychological as well as
physical parameters are suspected to play a role Conflicts of interest: None.
[Cipolletta et al., 2012; Wu et al., 2020]. When vitreous *Correspondence to: Stuart W. Harmer, Department of
floaters produce symptoms, they are often referred to as Engineering, Computing and Design, the University of Chichester,
symptomatic vitreous opacity (SVO) [Ivanova et al., Chichester PO21 3HR, UK.
2016]. In dealing with the symptomatic patient, the Email: [email protected]
ophthalmic community is confronted with the difficulty in
Received for review 2 June 2021; Revised 7 December 2021;
reconciling often‐unremarkable clinical observations with Accepted 18 December 2021
the occasionally profound symptoms of the patient
presenting with SVO [Luff and O'Donnell, 2018]. The DOI:10.1002/bem.22386
difficulty is due to multiple factors, which include: a Published online 30 December 2021 in Wiley Online Library
(wileyonlinelibrary.com).
-----------------------------------------------------------------------------------------
This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium,
provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made.

© 2021 The Authors. Bioelectromagnetics published by Wiley Periodicals LLC on behalf of

Bioelectromagnetics Society.
 Optical E¡ects of Floaters 91

becoming less true with the surgical advances afforded by 10–25 nm in diameter, that are prevented from coales-
micro‐incisional vitrectomy surgery [Sommerville, 2015; cing by macro‐molecules maintaining the spacing of the
Milston et al., 2016; Zeydanli et al., 2020]. fibrils within the meshwork [Sebag, 1989]. Because the
Clinical inspection through the slit lamp of fibrils are much smaller (<1/10 th) than the wavelength
vitreous of patients presenting with SVO is often of visible light in the vitreous media, they do not scatter
unrevealing, although advances in OCT and B‐scan visible light as they form a composite medium with
ultrasonography are enabling the presence, position, mostly water, which can be effectively described by a
and severity of degenerative changes of the vitreous to single refractive index [Rytov, 1956], i.e. that of
be quantified more robustly [Mamou et al., 2015]. vitreous, ~1.335, in humans [Palanker, 2013]. In this
Because of the traditional downplaying of the effects case vitreous is transparent and optically homogenous,
of SVO, the lack of a clearly quantifiable disease, and allowing light to propagate without scattering from lens
the historically limited options for treatment, the to retina. However, degenerative processes due to
symptomatic patient has all too often been dismissed ageing, disease, inflammation, and trauma can result in
by the ophthalmic profession and informed that they the breakdown of vitreous structure and the aggregation
must adapt to life with their condition or advised that of the fine collagen fibrils, increasing their size until
their symptoms will improve through a combination electromagnetic homogeneity is no longer an accurate
of acceptance, adaptation, and/or movement of the description of the medium at optical wavelengths, with
opacity away from the visual axis. Over the last localised areas of liquefaction and expelled collagen
decade or two, the attitude of the ophthalmic clumping at significantly greater‐than‐wavelength di-
community to patients with SVO has altered some- mensions. When this occurs, the vitreous becomes
what. This shift in attitude is explained by the electromagnetically heterogenous and scattering of
significant advances in vitreo‐retinal surgical tools visible light occurs at the resulting discontinuities in
and techniques over the same time period, which now refractive index, so that visible light no longer
permit effective and safer removal of the vitreous and propagates undisturbed from lens to retina. Aggregations
any opacities therein [Ivanova et al., 2016; Luff and of type II collagen, which has a complex refractive index
O'Donnell, 2018] and by an increased appreciation of of ~1.49 + i 0.001 [Sekar et al., 2017; Scholkmann et al.,
the effect that vitreous opacity can have on the quality 2014; Sekar et al., 2017], are suspended in aqueous
of vision beyond that of visual acuity alone. Efforts humour, refractive index ~1.3335 [Palanker, 2013], and
have been made to understand the scattering effects result in scattering through refraction and diffraction of
that vitreous opacity has on the vision of the sufferer, incident optical energy. It should be noted that because
with contrast sensitivity function (CSF) [Pelli and of the very small extinction coefficient of collagen
Bex, 2013] being one measure employed to evaluate (10−3), very little optical energy will be absorbed by the
the degradation in vision caused by SVO [Garcia opacity and hence visual effects are almost entirely due
et al., 2016]. Importantly, reduction in contrast to the forward redistribution (scattering) of incident
sensitivity has been correlated with the presence of light. Although the word “opacity” is used medically,
SVO and has been used as a quantitative index for this is rather misleading as the material causing the
determining the appropriateness of potential surgical visual effects may be transparent, the visual effects
intervention [Mamou, 2015; Garcia, 2016; Milston, arising from interference caused by differences in
2016; Sebag, 2020, 2014]. However, contrast sensi- refractive indices between the “opacity” and the
tivity testing is not standardised or regularly used in surrounding vitreous. The authors' aims are to prove,
clinical evaluation. theoretically, that vitreous opacity can result in variations
To understand why it is that some patients with in the intensity distribution on the retina, and thus be
SVO present with severe symptoms, the scattering perceived by the patient with SVO; to show that
mechanisms of visible light, ~380–740 nm in free space backscattering from vitreous opacity is very much
(280–550 nm in the vitreous gel), by opacities suspended smaller than forward scattering and therefore explain
within the vitreous cavity need to be considered the difficulty in matching subjective symptoms with
quantitatively. While there is much published work on objective examination; and to explain the link between
electromagnetic scattering in other biomedical fields degradation in contrast sensitivity function with the
[Lin and Guy, 1974; Kim and Lin, 1998], there is, to presence of SVO. In providing this solid scientific
date, very scarce representation of the effects of explanation, based on physics, the authors hope to
scattering by opacity within the vitreous in the literature provide explanation and additional reassurance to
[Serpetopoulos and Korakitis, 1998]. The vitreous clinicians considering surgical intervention for SVO in
consists of a meshwork of ultrafine collagen fibrils, an eye with unremarkable clinical examination.

Bioelectromagnetics
92 Harmer et al.

MATHEMATICAL MODEL symptoms to those afflicted, but the vitreous is often

unremarkable upon clinical examination? What are
The authors use a Fourier optics model of a
the key parameters that determine the optical effect of
simplified human eye together with a mathematical
vitreous opacity (i.e. pupil diameter, opacity size,
model of human vision implemented in the commer-
position of opacity relative to retina)? Does SVO
cial software MATLAB (MathsWorks, Natick, MA).
degrade CSF? And is this degradation correlated to
The optical model is based on the solution of the
other symptoms of SVO?
Helmholtz equation [Goodman, 1996; Yariv, 1997;
The model uses the paraxial solutions of the
Mezouari and Harvey, 2003] in the paraxial limit and
Helmholtz equation, Equation (1),
we confine our interest to two dimensions for
computational speed. The paraxial limit is an excellent
approximation in the case of the human eye since the ∇ 2 E + k 02 n2E = 0 (1)
F‐number of the human eye is approximately between 2π
two and eight depending on the pupil diameter, giving where E is the electric field, k 0 = λ is the wave-
angular deviations from the visual axis of between 4° number in free space, λ is the wavelength in free
and 14°, which are very well approximated paraxially. space, and n is the refractive index of the medium in
Vitreous opacity, which acts as the scattering object, is which the wave is propagating; there are two different
represented by the complex reflection and transmis- media in the model: the vitreous has refractive index
sion coefficients of a single planar dielectric layer na and the opacity refractive index nb . We develop a
[Kapilevich et al., 2014]. In the model, the opacity is two‐dimensional model using z to represent distance
completely specified by the complex refractive index, along the visual axis and x to represent displacement
n′ + in″, of the opacity; the axial thickness and width on the retinal plane (see Fig. 1). The general solution
of this layer; the refractive index of the surrounding to Equation (1) in two dimensions is,
medium, na (the vitreous); the incident spectrum of ∞

light, here assumed white; the distance from scattering

E (x, z ) = ∫−∞ F (kx ) ei (k x+k z) dkx
x z (2)
object to the retina; and the focal length of the
where k 02 n2 = k 2 = kx2 + kz2 and kx and k y are the lateral
cornea/lens optical system. See Figure 1 for a diagram
and axial wavenumber components, respectively. Using
of the model eye used. It should be noted that any
the paraxial approximation, ∣kx∣ < < ∣k∣, we have
model depending on exact electromagnetic solutions k2
is unrealistic for this problem because of the widely kz ≈ k − 2kx ; applying this result to Equation (2), the
variable form that vitreous opacity may take. Here, the paraxial electric field distribution is well approximated by,
aim is to estimate the magnitude of visual effects and ∞ kx2 z
explain the main clinical observations for SVO, which E (x , z ) = eikz ∫−∞ F (kx ) e−i 2k eikx x dkx (3)
are: Why it is that vitreous opacity can result in severe
Equation (3) can be written as an inverse Fourier
transform,

k2z
E (x , z ) = eikz I−1⎧F (kx ) e−i 2k ⎫
x
(4)
⎨
⎩ ⎬
⎭
∞
where I−1{A (kx )} = ∫−∞ A (kx ) eikx x dkx . Equation (4)
provides an expression for F (kx ) in terms of the “input
field distribution,” E (x, 0), which is the field distribu-
tion at the plane z = 0; hence, we have the Fourier
transform of the input field distribution as Equa-
tion (5),
F (kx ) = I {E (x, 0)} (5)
1 ∞
where I {a (x )} = 2π ∫−∞ a (x ) e−ikx x dx . Now, we may
Fig. 1. Simplified human eye model, giving the definition of write Equation (4) as,
the angle used in plots throughout this article. The axial
length of the human eye, the distance from the front of lens k2z
to the retina zt = z1 + z2, is assumed to be 22.3 mm in this E (x , z ) = eikz I−1⎧I {E (x , 0)} e−i 2k ⎫
x
(6)
article [Bhardwaj and Gandhi Rajeshbhai, 2013]. ⎨
⎩ ⎬
⎭
Bioelectromagnetics
 Optical E¡ects of Floaters 93
Utilising the convolution theorem, we may write r (1 − ei2k0 nb h)
Equation (6) as, ⎧ ; ∣x∣ < a
R = 1 − r 2ei2k0 nb h (11)
kx2 z ⎨ 0 ; ∣x∣ ≥ a
eikz ⎩
E (x , z ) = E (x, 0) ⊗ I−1⎧e−i 2k ⎫ (7)
2π ⎨
⎩ ⎬
⎭
(1 − r 2) eik0 nb h
where ⊗ represents the one‐dimensional spatial ⎧ 1 − r 2ei2k0 nb h ; ∣x∣ < a
T= (12)
convolution operation. ⎨1 ; ∣x∣ ≥ a
{ }
kx2 z kx 2
Since I−1 e−i 2k = 2πk ei 2z , then Equation (7)
can be written as,
iz
⎩
n −n
where a is the pupil radius, r = na + nb , and h is the
a b
k thickness of the opacity. Equations (11) and (12) are
E (x, z ) = eikz E (x , 0) ⊗ D (x, z ) (8) obtained by considering the reflection and transmis-
i 2πz
sion of electromagnetic waves by a planar layer
kx 2
where D (x, z ) = ei 2z . Equation (8) provides a pre- through applying appropriate electromagnetic
scription for propagating a field distribution from the boundary conditions [Kapilevich et al., 2014].
plane z = 0 to the plane z. At the eye's pupil the input One should note that the refractive index of the
field is given by the plane wave, A (x ) = eik 0 xna sin θ , opacity may be a complex number, that is there is a
where, for simplicity, we have assumed an amplitude refractive index, nb′, and extinction coefficient, nb″, which
of unity and that the eye is receiving light from a point together constitute the complex refractive index,
source located at infinity. Within the eye, the spherical nb = nb′ + inb″. It is assumed that healthy vitreous is
wave propagates at angle of θ to the visual axis lossless and therefore has zero extinction coefficient.
(z‐axis). This field distribution is multiplied by the From Equations (10) and (12) the field distribution after
pupil function, P(x), defined to be unity within the the opacity is given by,
pupil (radius a) and zero outside and then multiplied
kx 2
again by the thin lens function, L (x ) = e−i 2f , where k
k = k 0 na . This function focuses the incident wave to a E (x , z1) = eikz1 ((P (x ) A (x ) L (x ))
i 2πz1
minimum beam width at the focal distance f from the
lens. Hence the field distribution immediately after the ⊗ D (x , z1)) T (x ) (13)
pupil and lens is given by,
And from Equations (10) and (11) the field
E (x, 0) = P (x ) A (x ) L (x ) distribution reflected by the opacity is given by,

⎧ ik (x sin θ− 2f ) ; ∣x∣ ≤ a
x2

= e (9)
E (x , z1) = eikz1
k
((P (x ) A (x ) L (x ))
⎨ 0 ; ∣x∣ > a i 2πz1
⎩
⊗ D (x , z1)) R (x ) (14)
This input field is then propagated from the pupil
(z = 0) to the opacity plane (z = z1) using Equation (8),
Finally, the field distribution on the z = h1 + h2
plane is obtained by taking Equation (13) as the input
E (x , z ) = eikz1
k
(P (x ) A (x ) L (x )) ⊗ D (x , z1) field distribution and appropriately applying Equation (8),
i 2πz1 to give the forward scattered field, EF, on the retinal plane.
(10)
k
EF (x ) = E (x, z2) = eik (z1+ z2)
At the opacity plane, the incident field is partially i 2π z1 z2
reflected from the opacity and partially transmitted
through the opacity; the forward and back scattered fields (((P (x ) A (x ) L (x )) ⊗ D (x , z1)) T (x )) ⊗ D (x, z2)
are obtained by multiplying the field distribution given by (15)
Equation (10) by further scattering function in the opacity
plane. The scattering function is given by R (Equation 11) Similarly, for the field back scattered at the lens
for back scattering and T (Equation 12) for forward plane (z = 0), EB,
scattering,

Bioelectromagnetics
94 Harmer et al.

EB (x ) = E (x , 0) [Lesmes et al., 2010] and a rapid tail‐off at higher

spatial frequencies that is a consequence of cone cell
k size and density and neuro modulation effects and is
= eik 2z1 (((P (x ) A (x ) L (x ))
i 2πz1 not explained by the optical physics of the eye.
(16) Mathematical models of human vision are numerous,
⊗ D (x, z1)) R (x )) ⊗ D (x, z1)
with a comprehensive description of various models
provided in Watson and Ahumada [2005]. The authors
The point spread function (PSF) at the retinal have chosen, somewhat arbitrarily, to use an asym-
and lens planes are obtained by taking the modulus metric parabolic function in log‐log space as there is
square of Equations (15) and (16) respectively, i.e. best fit numerical data of “normal” human vision
PSF (x ) = ∣E (x ) ∣2. It should be noted that the PSF in provided for this model [Chung and Legge, 2016].
this model is not isoplanatic, since Equations (15) and The MTF on the retina is calculated using Equation
(16) depend on whether the light passed through the (17) and the mathematical model of human vision is
opacity is determined by the angle of propagation θ given by the contrast sensitivity function, CSF . The
and if so, where it has passed through the opacity. contrast sensitivity is linked to the optical MTF of
From the PSF, Equation (15), the modulation Equation (17) by assuming the presence of an
transfer function (MTF) is obtained by, additional neural MTF term, MTFNeural , which char-
acterises spatial filtering from the retina to the brain,
MTF (γ ) = ∣I {PSF (x )} ∣ (17) Equation (18)

The MTF is a measure of how spatial frequencies CSF = ∣MTFOptical MTFNeural∣ (18)
are filtered when passed through an optical system and are
conventionally plotted as a function of angle, θ, (degrees) With,
versus spatial frequency, μ, in units of cycles per degree
(cpd). Conversion from x‐coordinate in the retinal plane to
πz
θ‐coordinate is realised byx = 180t θ and conversion from log10 (CSF ′)
spatial frequency in radians per metre, γ , to spatial α − β 2 (log10 (μ) − log10 (μP ))2 μ < μP
frequency
z
in cycles per degree, μ, is realised by =⎧
μ = 360tn γ , where zt = z1 + z2. ⎨ α − δ 2 (log10 (μ) − log10 (μP ))2 μ ≥ μP
a
⎩
The model computes values over ten equally (19)
spaced wavelength steps, from 380 to 740 nm in free
space, with each wavelength step having the same where α, β, and δ are constants and μ and μP are the
intensity (white light). The light is assumed tempo- spatial frequency in cycles per degree and spatial
rally incoherent and accordingly intensities are frequency at which the CSF is maximum respectively.
summed over each wavelength step; thus the PSF is The values used in Equation (19) are those that gave
formed by adding the intensities of the 10‐point spread best fit with normal human vision [Chung and Legge,
functions calculated for each wavelength step. How- 2016]: α = 2.22, β = 0.68, δ = 1.28, and μP = 2.5.
ever, less significant wavelength‐dependent effects, Note that CSF is used in vision science to measure the
such as dispersion in the vitreous and opacity, are not sensitivity of the visual system as a function of spatial
included as these effects are likely to be very small frequency [Pappas et al., 2005; Pelli and Bex, 2013].
over the visible band. The spatial resolution of the We assume, as no specific information was given, that
model is set at λ/10 of the shortest wavelength of light the best fit data obtained were measured for pupils of
used in the simulations, giving a value of 38 nm, thus 4 mm diameter in eyes without vitreous opacity and
ensuring that interference effects are accurately that this gives a contrast sensitivity function CSF ′.
computed. Note, all equations given are valid for This assumption does not significantly affect the
continuous distributions and since MATLAB operates following results if it is altered to 2 mm or 8 mm,
discretely, integrals in equations are replaced with the which mark the extreme values for the human eye.
appropriate summations when implemented in code. Using Equation (18) we have,
Equation (17) provides the MTF at the surface of
the retina and does not include the modulation effects CSF ′ = ∣MTF ′Optical MTF∣Neural (20)
of the retina, optic nerve, and brain. These effects are
highly important as human vision has a peak spatial From Equations (18) and (20) and assuming that
frequency sensitivity around a few cycles per degree the neural MTF is independent of pupil diameter, we
Bioelectromagnetics
 Optical E¡ects of Floaters 95

can compute the CSF of an eye with opacity and a an emmetropic eye and the angle at which the light
different pupil diameter according to Equation (21) propagates through the eye, which is chosen to be along
the visual axis (θ = 0). Note, since in the simplified
MTFOptical mathematical model of the eye used, all refraction occurs
CSF = CSF ′ (21) at the front surface of the eye and so the axial length and
MTF ′Optical not the depth of the vitreous cavity is the important
parameter. With these fixed parameters, the variables are
In the calculations presented, the value of CSF ′ is limited to the distance of the opacity from the retinal
normalised to have a maximum value of unity, which plane, the thickness and width of the vitreous opacity, and
occurs at spatial frequency μP = 2.5cpd. MTF ′Optical is the pupil diameter. Further, by equating the opacity
the MTF calculated by Equation (17) for an eye with thickness and width to form a single variable of opacity
4 mm diameter pupil and in the absence of opacity. In “size,” the model is reduced to a manageable three
the model, the MTF and CSF step size is 0.0474 cpd. degrees of freedom, suitable for brief exploration and
Variations in CSF from eye to eye will have some effect discussion in a single article.
on the results predicted by application of Equation (21); Figures 2 and 3 display the computed PSF, MTF,
however, the authors do not have information on the and CSF for an eye with no opacity and a 25 µm size
variance of parameters α, β, δ, and μP that would enable opacity, located 2 mm from the retina, respectively. The
such variations in CSF to be explored. Therefore, effect of the centrally located vitreous opacity is clear,
predicted results are for average human vision. The with the greatest effect presenting for the smallest pupil
MATLAB code can be summarised as taking an incident diameter. The opacity results in the central maximum
planar wave and focusing this through a refractive intensity of the PSF decreasing and the intensity of the
optical system of focal length f. The amplitude side lobes (the oscillations that occur outside of the
distribution is then propagated to the plane in which central maxima of the PSF) increasing. The physical
the opacity is located, Equation (10), where the Fresnel explanation for these effects on the PSF is simply that
reflection and transmission coefficients, Equations (11) the opacity redistributes (scatters) the light in a forward
and (12) respectively, are applied to calculate the direction with the optical energy being deflected though
forward and backscattered amplitudes. These amplitudes larger angles, increasing the intensity in the side lobes
are then propagated onto the retina and lens planes, and a consequent reduction of intensity in the central
Equations (15) and (16) respectively, to give the maxima. When examined in spatial frequency space, the
amplitude distributions which are used to calculate the MTF is reduced in the spatial frequency range up to
effects of vitreous opacity. about 50 cpd, with significant reduction up to about
30 cpd, while being largely unaffected at much higher
spatial frequencies. The very high spatial frequency
SIMULATION RESULTS components present in the MTF are not present in
human vision due, predominantly, to the finite sampling
Quantifying the E¡ect of Vitreous Opacity onVision size of cone cells in the retina, with cone cell density
Modelling using the MATLAB code developed being maximum in the foveal region with a spacing of
by the authors allows the computation of PSF, the ~2.7 µm [Wells‐Gray et al., 2016]. This cut‐off effect is
spatial frequency domain analogue, MTF at the retina, clearly seen in the CSF plots in Figures 2 and 3, with
and the CSF for the whole human visual system to be CSF falling to about 1% of peak value at 30 cpd.
estimated. The computed PSF, Equation (15), is The reduction in the maximum intensity of the
normalised so that it has maximum value of unity in central PSF maxima caused by the presence of
the case of no vitreous opacity; this is done so that vitreous opacity can be usefully quantified by defining
comparison may be made of the effect on forward the reduction in relative intensity (RRI), which is
scattered intensity by the presence of vitreous opacity. expressed as a percentage decrease of the intensity
The MTF, Equation (17), is conventionally scaled, so without opacity, Equation (22).
that it has a value of unity at zero spatial frequency.
Even with the simplified human eye model max (PSF ) − max (PSFopacity ) ⎞
presented, there are issues with fully exploring the RRI = 100⎛ ⎜ (22)⎟

⎝ max (PSF ) ⎠
multivariate space of the model simulations. To reduce
complexity, some parameters of the model are fixed with
appropriate values. The fixed parameters are the axial where PSF is the PSF with no opacity, PSFopacity is
length of the eye (22.3 mm) and the focal length of the the PSF with opacity, and max requires taking the
eye (also 22.3 mm), which is equal to the axial length for global maximum value of that function.
Bioelectromagnetics
96 Harmer et al.

Fig. 2. The point spread function, modulation transfer function, and contrast sensitivity function
predicted by the model for an emmetropic human eye with no vitreous opacity and with different
pupil diameters. The point spread function and modulation transfer function are determined by
the optical properties of the eye anatomy and are representative of effects at the retinal surface,
whereas the contrast sensitivity function is calculated from measurements made on the complete
human vision system and is representative of what is perceived. The contrast sensitivity function
is plotted in log‐log format as is conventional in vision science.

Fig. 3. The point spread function, modulation transfer function, and contrast sensitivity
function predicted by the model for an emmetropic human eye with a 25 µm collagen
vitreous opacity located 2 mm from the retinal plane. The opacity is located symmetrically
on the visual axis. The effects of the vitreous opacity are clearly seen in the point spread
function, modulation transfer function, and contrast sensitivity function plots, with greatest
effect occurring for the smallest pupil diameter. The contrast sensitivity is plotted in log‐log
format as is conventional in vision science.

Bioelectromagnetics
 Optical E¡ects of Floaters 97

Fig. 4. The difference between two contrast sensitivity functions, with and without vitreous
opacity and the definitions of low, mid, and high spatial frequency used to quantify the effect
of vitreous opacity on contrast sensitivity function. An eye with a 2 mm pupil dimeter and
with 50 µm vitreous opacity located 2 mm from the retina is modeled. The plots are given in
linear space (left) and log‐log space (right).

The RRI is used here as the principal measure of where CSF is the CSF with no opacity, CSFopacity is the
the severity of vitreous opacity because decrease in CSF with opacity, and the mean value of ΔCSF% is
intensity caused by vitreous opacity results in localised taken over the chosen spatial frequency ranges.
shadowing on the retina when a constant brightness Finally, we also wish to compare the effect of
background, such as a blue sky or white light computer opacity on the intensity distribution on the retina as a
monitor, is viewed. The RRI, Equation (22), gives a ratio to the reflected or backscattered intensity distribu-
direct measure of the maximum localised reduction in tion at the pupil plane, which is available for clinical
intensity produced by the presence of vitreous opacity. viewing. A ratio of the change in central intensity on the
However, opacity also scatters incident light to wider retinal plane produced by opacity to the maximum
angles and so a measure of the effect on overall vision backscattered intensity on the pupil plane gives a useful
(CSF) by vitreous opacity is also required. To quantify measure of the predicted anisotropy in forward and back
the degradation in CSF caused by vitreous opacity, the scattering. The forward to back scattering ratio FBR is
difference between CSF with and without vitreous given in Equation (24)
opacity is computed and divided by the CSF without max (PSF ) − max (PSFopacity )
vitreous opacity, Equation (23), to give the decrease in FBR = (24)
CSF as a percentage. These values are averaged over max (PSFB )
three spatial frequency ranges: low, mid, and high,
corresponding to spatial frequency ranges of 0.1–1, where PSFB is the backscattered intensity distribu-
1–5, and 5–40 cpd, respectively, so that the effect of tion obtained from taking the square of the
degradation of CSF due to vitreous opacity can be modulus of Equation (16). The FBR is large and
quantified broadly by spatial frequency range. See is better expressed as a decibel quantity,
Figure 4 for illustration. dBFBR = 10 log10 (FBR).
The described model was run over a range of
pupil diameters of 2, 4, and 8 mm, which cover the
CSF − CSFopacity ⎞ usual range of the human eye [Watson and Yellott,
ΔCSF % = 100⎛⎜ ⎟
(23)
2012]; a range of opacity sizes, from 10 µm through to
⎝ CSF ⎠ 96 in 5 µm steps; and a range of opacity separations
Bioelectromagnetics
98 Harmer et al.

Fig. 5. Contour plot of reduction in relative intensity (RRI) caused by opacity of different
sizes and at different distances to the retinal plane for an eye with three different pupil
diameters. 2 mm (top plot), 4 mm (middle plot), and 8 mm pupil (bottom plot). The values
given on the contours are the RRI as a percentage.

Bioelectromagnetics
 Optical E¡ects of Floaters 99

Fig. 6. Contour plot of mean decrease in contrast sensitivity function in the low spatial
frequency range caused by opacity of different sizes and at different distances to the retinal
plane for an eye with a 2 mm pupil diameter. The values given on the contours are in
percent.

Fig. 7. Contour plot of mean decrease in contrast sensitivity function in the mid spatial
frequency range caused by opacity of different sizes and at different distances to the retinal
plane for an eye with a 2 mm pupil diameter. The values given on the contours are in
percent.

Bioelectromagnetics
100 Harmer et al.

Fig. 8. Contour plot of mean decrease in contrast sensitivity function in the high spatial
frequency range caused by opacity of different sizes and at different distances to the retinal
plane for an eye with three different pupil diameters. 2 mm (top plot), 4 mm (middle plot),
and 8 mm pupil (bottom plot). The values given on the contours are in percent.

Bioelectromagnetics
 Optical E¡ects of Floaters 101

Fig. 9. Contour plot of forward to backscatter ratio from opacity of different sizes and at
different distances to the retinal plane for an eye with three different pupil diameters. 2 mm
(top plot), 4 mm (middle plot), and 8 mm pupil (bottom plot). The values given on the
contours are the FBR in dB.

Bioelectromagnetics
102 Harmer et al.

TABLE 1. Correlation Coefficients Between RRI and Mean CSF Degradation at the Three Chosen Spatial Frequency Ranges

Low Mid High

Spatial frequency range
pupil diameter mm 0.1–1 cpd 1–5 cpd 5–40 cpd
2 0.646 to 0.748 0.643 to 0.747 0.832–0.884
3 0.357 to 0.519 0.338 to 0.503 0.971–0.980
4 0.014 to 0.212 0.130 to 0.321 0.973–0.982
5 −0.093 to 0.109 0.047 to 0.244 0.950–0.967
6 −0.173 to 0.027 0 to 0.198 0.921–0.946
7 −0.208 to −0.01 −0.028 to 0.172 0.897–0.930
8 −0.207 to 0 −0.049 to 0.152 0.877–0.916
Note: The correlation coefficients listed are the ranges for 95% confidence. A value of 0 represents no correlation, while a value of 1
indicates perfect positive correlation and a value of −1 perfect negative correlation. Values above 0.8 are highlighted in bold.

from the retinal plane, from 1 mm through to 19 mm Figure 8, are very similar in shape to the corresponding
in 1 mm steps. The RRI, Equation (18), contrast RRI plots, Figure 5, whereas the plots for low and mid
degradation, Equation (23), and FBR, Equation (24) spatial frequency CSF degradation, Figures 6 and 7,
were computed for each of these 3 × 19 × 20 = 1140 look quite different. This similarity can be quantified by
scenarios, and then these data were plotted as contour calculating the 95% confidence correlation values
plots to show the dependence of these three measures between RRI and CSF degradation; the results are
on the variables of pupil diameter, opacity size, and tabulated in Table 1. The correlation between the
distance of opacity from the retinal plane; see principal symptom of vitreous opacity, quantified by
Figures 5‐9. RRI, and degradation of CSF is very strong for the
Although the collagen opacity absorbs almost high spatial frequency region, but is weak or
none of the optical energy incident upon it (i.e. it is uncorrelated for the low and mid‐spatial frequency
nearly transparent), there is still a significant optical regions. Smaller diameter pupils (2 and 3 mm) display
effect at the retinal plane. Figure 5 shows that some degree of correlation in the mid and low spatial
shadowing, resulting from reduced intensity from the frequency regions, but this vanishes for pupil diameters
opacity as compared to no opacity, can be consider- of 4 mm and larger. The close similarity of degradation
able and that this effect is strongly dependent on of CSF with RRI is to be expected from the principle
opacity size and the distance of the opacity from the of conservation of energy, with virtually no absorption
retinal plane. For a constant opacity size, the RRI falls of optical energy within the opacity and extremely
off rapidly with increasing distance from the retinal low backscattering; the reduction in intensity of the
plane, making the visual effects of SVO decrease with central PSF maxima produced by the opacity must be
increasing distance from the retina. As expected, for balanced by increased intensity in the PSF side‐lobes,
constant distance from the retinal plane, increasing leading to degradation of CSF. What is less obvious is
opacity size results in increasing RRI, and so larger that the frequency dependence of this degradation
opacity results in greater visual effects. The pupil size should be greater in the high spatial frequency range,
strongly influences the effects of vitreous opacity, and symptoms of shadowing caused by vitreous
with smaller pupil diameters giving greater RRI for opacity should correlate strongly only to degradation
the same values of retinal separation and opacity size. in high frequency CSF. To our knowledge, these
Figures 6‐8 display the mean degradation in predictions, founded on the physics of electromagnetic
CSF. Importantly, the degradation of CSF is predicted scattering, are unique in the literature, and further
to be spatial frequency‐dependent, with quite different work could be undertaken to investigate clinically
trends being apparent for low, mid, and high spatial whether the predicted magnitudes of CSF diminution
frequency regions; compare Figure 6 and Figure 7 and the correlation between floater severity and
with Figure 8. For the 2 mm diameter pupil, the degradation of high frequency (5–40 cpd) CSF is
greatest degradation in CSF is to be found in the high observed in practice.
spatial frequency region, with values ~50% being Finally, Figure 9 depicts the forward to back
reached for opacity close to the retina, whereas in scatter ratio; there is a huge difference between the
the mid‐spatial and low spatial frequency regions maximum change in the forward scattered intensity
these values only reach ~6% and ~1%, respectively. with and without opacity to the maximum back
The CSF in the high spatial frequency ranges, scattered intensity with opacity. Values above 40 dB

Bioelectromagnetics
 Optical E¡ects of Floaters 103

and reaching 65 dB are predicted, corresponding to and more diffuse shadowing effect, perhaps perceived
FBR values of between approximately 104 and 106. as more of a ‘clouding' in vision. The improvement in
The very large asymmetry between reflected intensity symptoms for some patients complaining of SVO
and variations in forward scattered intensity is of following posterior vitreous detachment may be
significance as it is consistent with the sometimes‐ explained by this effect as the vitreous opacity is
noted difficulty in observing vitreous opacity on moved further from the retina by the collapsing gel
clinical examinations at the slit lamp. The predictions structure [Ivanova et al., 2016].
are also suggestive that absence of such clinical A significant prediction of the model is the very
observation can in no way be taken as a guarantee that strong dependence on shadowing of pupil diameter.
the patient cannot observe significant effects when Narrow pupils give much greater effects than wider
“looking through” vitreous opacity. pupils. Thus, one would expect that SVO symptoms
would become very noticeably worse in brightly lit
environments, where light‐induced miosis [Maqsood,
CONCLUSIONS 2017] would increase the depth of shadowing
Vitreous opacity is a common clinical presenta- perceived by vitreous opacity, and this is indeed often
tion to optometrists and ophthalmologists. In some reported [Luff and O'Donnell, 2018]. The effect also
cases, sudden onset of floaters can be a sign of retinal suggests that symptoms of SVO should be improved
tear following acute posterior vitreous detachment and by mydriatic drugs, such as atropine eye drops, and,
other potentially vision‐threatening retinal patholo- indeed, such an effect has been reported in the
gies. However, persistent vitreous opacity in the literature and used as a non‐surgical treatment for
absence of retinal pathology is common and in a SVO [Kaymak, 2017].
small but significant population can cause visual Degradation of contrast sensitivity function has
dysfunction and a reduction in quality of life. The been linked with SVO, and this is important since CSF
simplified human eye model described here and can be measured clinically and thus could be used as a
computationally implemented offers a physical ex- metric to indicate objectively that surgical interven-
planation as to the significant visual effects that are tion had made a measurable improvement. Currently,
reported by a small, but significant, subset of those most surgical intervention for SVO is based entirely
with SVO; explains why clinical observation with slit on the reported, and hence subjective, symptoms of
lamp alone is an unreliable diagnostic tool for SVO; the SVO patient and this lack of objective measure
theoretically links degradation of CSF with the may deter surgeons from offering surgery to the SVO
presence of SVO; and demonstrates that degradation patient. By utilising a mathematical model of human
of CSF in the spatial frequency range 5–40 cpd is very vision, the optical predictions of the model presented
strongly correlated to the degree of shadowing caused here can be used to predict the effect of SVO on
by vitreous opacity. contrast sensitivity function. CSF is predicted to be
The model implemented accounts for common degraded by SVO, with the greatest effects in the high
symptoms of SVO, i.e. the perception of well‐defined spatial frequency range of 5–40 cpd. High spatial
shadowing (RRI) in the visual field (see Fig. 5), with frequency CSF degradation shows the same trends as
patients often reporting dark shapes that move through RRI: with greatest effect observed for large opacity,
their visual field on eye movement. The severity of opacity that is close to the retina, and where the pupil
this shadowing is dependent on both the size of the diameter is smallest (see Fig. 8). The similarity of the
opacity and the distance that the opacity is located trends of high‐frequency CSF degradation and RRI is
from the retina. Not surprisingly, larger opacity results quantified by correlating these two measures; see
in greater shadowing (this effect will be both in depth Table 1. Since RRI is hard to measure directly,
of shadowing and also in the perceived angular size of requiring the shadows cast due to vitreous opacity to
the floater). Opacity suspended closer to the retina be located and measured on the retina, correlation of
produces greater shadowing effects than the same this with a clinically measurable quantity is important.
opacity located further from the retina. However, The very high degree of correlation between RRI and
opacity that is closer will produce shadowing with CSF suggests that degradation and subsequent post‐
smaller angular size than the same opacity located surgical improvement in high frequency CSF may be a
further away, so the expected results of opacity useful clinical metric for SVO.
located close to the retina is a well‐defined floater Collagen has a small extinction coefficient and a
with significant shadowing effects, while opacity refractive index which is not vastly different from
located further from the retina will produce a larger vitreous, yet even 10‐µm sized opacity composed of

Bioelectromagnetics
104 Harmer et al.

this material is predicted to cause symptoms if the Lin JC, Guy AW. 1974. A note on the optical scattering
opacity is located close to the retina. While collagen characteristics of whole blood. IEEE Trans Biomedical
aggregations are one likely cause of SVO, even the Engineering 21:43–45.
Luff AJ, O'Donnell C. 2018. Contemporary management options
liquefaction of vitreous with corresponding small for symptomatic vitreous floaters. Available from https://
alterations in refractive index could give rise to www.opticianonline.net/cet-archive/5143 [Last accessed 28
symptoms. The term vitreous opacity is therefore September 2021].
somewhat of an inaccurate description, as symptoms Mamou J, Wa CA, Yee KM, Silverman RH, Ketterling JA,
may be produced by entirely transparent media that Sadun AA, Sebag J. 2015. Ultrasound‐based quantifica-
tion of vitreous floaters correlates with contrast
result in localised changes in refractive index. sensitivity and quality of life. Invest Ophthalmol Vis
Although unlikely to be adopted, SVRI (Symptomatic Sci 56:1611–1617.
Vitreous Refractive Inhomogeneities) is a more Maqsood F. 2017. Effects of varying light conditions and
accurate acronym. refractive error on pupil size. Cogent Med 4:1338824.
Backscattering by opacity is extremely low when Mezouari A, Harvey AR. 2003. Validity of Fresnel and
Fraunhofer approximations in scalar diffraction. J Opt A
compared to the variation in intensity caused on the Pure Appl Opt 5:86–91.
retina by the opacity, with values of 104–106 Milston R, Madigan MC, Sebag J. 2016. Vitreous floaters:
predicted; see Figure 9. Perhaps one of the most Etiology, diagnostics, and management. Surv Ophthalmol
significant problems of SVO is often the difficulty in 61:211–227.
matching reported symptoms with clinical examina- Palanker D. 2013. Optical properties of the eye. Available from
https://www.aao.org/current-insight/optical-properties-of-
tion. With the forward to backscatter ratios predicted eye [Last accessed 28 September 2021].
by the model presented, it is apparent that unremark- Pappas TN, Safranek RJ, Chen J, Bovik A. 2005. Handbook of
able clinical examination is to be expected even in the Image and Video Processing, second edition. New York,
presence of the most debilitating of symptoms. NY: Academic Press. 939–959.
Pelli DG, Bex P. 2013. Measuring contrast sensitivity. Vision Res
90:10–14.
Rytov SM. 1956. Electromagnetic properties of a finely stratified
medium. Sov Phys JETP 2:466–475.
REFERENCES Scholkmann F, Kleiser S, Metz AJ, Zimmermann R, Pavia JM,
Wolf U, Wolf M. 2014. A review on continuous wave
Bhardwaj V, Rajeshbhai GP. 2013. Axial length, anterior chamber functional near‐infrared spectroscopy and imaging instru-
depth—A study in different age groups and refractive mentation and methodology. Neuroimage 85:6–27.
errors. J Clin Diagn Res 7:2211–2212. Sebag J. 2020. Vitreous and vision degrading myodesopsia. Prog
Chung ST, Legge GE. 2016. Comparing the shape of contrast Ret Eye Res 79:100847.
sensitivity functions for normal and low vision. Invest Sebag J. 2011. Floaters and the quality of life. Am J Ophthalmol
Ophthalmol Vis Sci 57:198–207. 152:3–4.
Cipolletta S, Beccarello A, Galan A. 2012. A psychological Sebag J. 1989. The Vitreous: Structure, Function, and Patho-
perspective of eye floaters. Qual Health Res 22:1547–1558. biology. New York, NY: Springer‐Verlag. pp. 85–90.
Garcia GA, Khoshnevis M, Yee KMP, Nguyen‐Cuu J, Nguyen Sebag J, Yee KM, Wa CA, Huang LC, Sadun AA. 2014.
JH, Sebag J. 2016. Degradation of contrast sensitivity Vitrectomy for floaters: Prospective efficacy analyses and
function following posterior vitreous detachment. Am J retrospective safety profile. Retina 34:1062–1068.
Ophthalmol 172:7–12. Sekar SKV, Bargigia I, Mora AD, Taroni P, Ruggeri A, Tosi A,
Goodman J. 1996. Introduction to Fourier Optics, 2nd edition. Pifferi A, Farina A. 2017. Diffuse optical characterization
New York, NY: McGraw‐Hill, 38–39. of collagen absorption from 500 to 1700 nm. J Biomed Opt
Ivanova T, Jalil A, Antoniou Y, Bishop PN, Vallejo‐Garcia JL, Patton 22:015006.
N. 2016. Vitrectomy for primary symptomatic vitreous Serpetopoulos CN, Korakitis RA. 1998. An optical explana-
opacities: An evidence‐based review. Eye 30:645–655. tion of the entoptic phenomenon of ‘clouds' in posterior
Kapilevich BY, Harmer SW, Bowring NJ. 2014. Non‐Imaging vitreous detachment. Ophthalmic and Physiol Opt 18:
Microwave and Millimetre‐Wave Sensors for Concealed 446–451.
Object Detection. Boca Raton, FL: CRC Press. Sommerville DN. 2015. Vitrectomy for vitreous floaters:
pp. 156–160. Analysis of the benefits and risks. Curr Opin Ophthalmol
Kaymak H. 2017. Are there alternatives for treating floaters? Available 26:173–176.
from https://www.healio.com/news/ophthalmology/20171016/ Wagle AM, Lim WY, Yap TP, Neelam K, Au, Eong KG. 2011.
are-there-alternatives-for-treating-floaters [Last accessed 28 Utility values associated with vitreous floaters. Am J
September 2021]. Ophthalmol 152:60–65.
Kim JK, Lin JC. 1998. Successive order scattering transport Watson AB, Ahumada AJ, Jr. 2005. A standard model for foveal
approximation for laser propagation in whole blood detection of spatial contrast. J Vis 5:717–740.
medium. IEEE Trans Biomedical Engineering 45:505–510. Watson AB, Yellott JI. 2012. A unified formula for light‐adapted
Lesmes LA, Lu ZL, Baek J, Albright TD. 2010. Bayesian adaptive pupil size. J Vision 12:112.
estimation of the contrast sensitivity function: The quick Wells‐Gray EM, Choi SS, Bries A, Doble N. 2016. Variation in
CSF method. J Vis 10:1–21. rod and cone density from the fovea to the mid‐periphery in

Bioelectromagnetics
 Optical E¡ects of Floaters 105

healthy human retinas using adaptive optics scanning laser Zeydanli EO, Parolini B, Ozdek S, Bopp S, Adelman RA, Kuhn F,
ophthalmoscopy. Eye 30:1135–1143. Gini G, Sallam AB, Aksakal N. 2020. EVRS floaters study
Wu RH, Jiang JH, Gu YF, Moonasar N, Lin Z. 2020. Pars plana group, management of vitreous floaters: An international
vitrectomy relieves the depression in patients with sympto- survey the European vitreoretinal society floaters study
matic vitreous floaters. Int J Ophthalmol 13:412–416. report. Eye 34:825–834.
Yariv A. 1997. Photonics: Optical Electronics in Modern Zou H, Liu H, Xu X, Zhang X. 2013. The impact of persistent
Communications. New York, NY: Oxford University Press. visually disabling vitreous floaters on health status utility
pp. 30–33. values. Qual Life Res 22:1507–1514.

Bioelectromagnetics