Journal of Theoretical Biology 419 (2017) 305–309

Contents lists available at ScienceDirect

Journal of Theoretical Biology

journal homepage: www.elsevier.com/locate/yjtbi

Non-extensive distribution of human eye photoreceptors MARK

Ali Mehri
Department of Physics, Faculty of Science, Babol Noshirvani University of Technology, Babol, Iran

A R T I C L E I N F O A BS T RAC T

Keywords: Spatial distribution of cone and rod photoreceptors in human eye retina, are studied by using non-extensive
Human eye statistics. We show that photoreceptors’ packing density obeys q-exponential and q-Gaussian distributions.
Photoreceptor Cones retinal distribution has higher non-extensivity in young subjects in comparison with old ones. Absolute
Tsallis statistics value of q-parameter for spatial distribution of both types of photocells gets higher values along horizontal
Non-extensivity
meridian. We also calculate Shannon's additive entropy and Tsallis’ non-additive entropy for cone photocells
Entropy
distribution in retina, and compare their results. Information entropy, especially Tsallis entropy, is capable to
discriminate between young and old subjects, properly. Our results confirm the ability of non-extensive
statistics to unveil retinal eye problems caused by abnormal photoreceptors distribution.

1. Introduction
Human eye, as a complex organ, provides and adapts a great
amount of visual information for us. Eye consists of three layers: outer,
middle and inner (Willoughby et al., 2010; Binder et al., 2009). Cornea
and sclera are the outer layers. The middle layer of the eye is composed
of iris, ciliary body and choroid. Retina, as inner layer of the eye, is a
complex and layered structure of neurons that transmits light to the
brain (Abrahams, 2009; Page, 2001).
The retina is a light-sensitive layered structure with several layers of
neurons interconnected by synapses at the back of the eye that covers
about two third of its interior surface. Optics of the eye creates an
image of the visual world on the retina. The retina and its correspond-
ing visual field are divided into quadrants. In this scheme, the surface
of the retina is subdivided by vertical and horizontal lines that intersect
at the center of the fovea. The vertical line divides the retina into nasal
and temporal divisions and the horizontal line divides the retina into
superior and inferior divisions (Purves, 2001; Bridgeman, 2013).
Retinal photoreceptor cells convert light into electrical signals that
can stimulate neural processes, leading to vision. By photon absorption
in a photoreceptor cell, chemical reactions change the potential of cell
membrane. In human eyes rod, cone, and ganglion photoreceptors are
involved in the phototransduction procedure (Hecht, 1987). Human
retina contains approximately six million cone cells, which are con-
centrated in fovea and its nearby. The cones detect the light with high
resolution and color sensitivity help us see different colors. Three types
of cones, with various sizes respond to different ranges of light
wavelength (short, medium and long), conferring colors recognition.
Cones require significantly brighter light in order to produce a signal.
On the other hand, rods are very thin and extremely sensitive, and can
be triggered by just a single photon. They detect the light with a higher
sensitivity to brightness for larger field sizes even in the peripheral
region, but without color sensitivity and lower resolution. So, at very
low light levels, people just can experience achromatic vision, because
solely rod cells are active. Almost 120 million rod cells are distributed
across the human retina. At most, two percent of three million
ganglions in human visual system are photosensitive. This type of
ganglion cells has important role in entrainment and reflexive re-
sponses to the brightness of light. They do not directly contribute to
sight. But they support persistence of circadian and pupillary light
reflex (Foster et al., 1991).
Several studies on human retinal photoreceptors have indicated
their spatial distribution. Curcio et al. (1990) measured cone and rod
photocells density in whole-mounted human retinas, obtained from
seven individuals between 27 and 44 years of age. They found that
cones packing density exponentially decreases with increasing distance
from the fovea. In contrary, rods density increases with increasing
retinal eccentricity to reach its maximal in rod ring, and then declines
slowly from the rod ring to the far periphery. It also found that, outside
the foveal center, the photoreceptors density decreased significantly
with increasing age (Curcio et al., 1993; Panda-Jonas et al., 1995). At
all ages, the retina showed meridional difference in cone densities, with
cone photoreceptor packing density decreasing faster with increasing
eccentricity in the vertical dimensions than in the horizontal dimen-
sions (Song et al., 2011). Rod and cone photoreceptors are frequently
affected in retinal degenerative diseases. Cone density follows a
symmetrical distribution between fellow eyes. A systematic distribution
of parafoveal cones between fellow eyes may provide an anatomical

E-mail address: alimehri@nit.ac.ir.

http://dx.doi.org/10.1016/j.jtbi.2017.02.030
Received 3 November 2016; Received in revised form 17 February 2017; Accepted 24 February 2017
Available online 27 February 2017
0022-5193/ © 2017 Elsevier Ltd. All rights reserved.
A. Mehri
basis for the involvement of the photoreceptor layer in the first step of
binocular spatial sampling (Lombardo et al., 2013). Furthermore, it has
confirmed that, age was significantly correlated with the density both in
metric and angular units at 2° superior. Axial length and age were
significantly correlated with parafoveal cone photoreceptor distribution
(Obata and Yanagi, 2014).
In spite of all successes of traditional extensive statistics to discover
the macroscopic behavior of systems from microscopic details, it is not
able to interpret the complexity in natural systems. Non-extensive
statistics appears as a powerful way to describe complex systems. In
this study, by using the data from the mentioned experiments (Curcio
et al., 1990; Song et al., 2011), we will show that photoreceptors
distribution in retina can be well described by non-extensive statistics.
Non-extensivity parameter will introduce as a convenient index of
aging, and photocells distribution along horizontal/vertical meridians
of retina. We will also calculate entropy for photocells spatial distribu-
tion. We think it can be used to indicate vision problems which are
related to anormal distribution of retinal photoreceptors.
The organization of the remainder of the article is as follows. In
Section 2, we briefly review non-extensive statistics and information
theory. Then, we report and describe our findings using related graphs
and tables, in Section 3. There in, entropy for photoreceptors distribu-
tion in retina and their non-extensivity parameters are calculated.
Finally, in Section 4, we present a brief discussion and summary of the
work.
2. Nonextensive statistics
Entropy (or uncertainty) and its complement, information, are
perhaps the most fundamental quantitative measures in description of
multi-state systems. Many applications of entropy can be found in the
literature, not only in statistical mechanics but also in other fields like
economics, biology, engineering, etc.
In traditional statistical mechanics two forms are considered to
describe the concept of entropy. It was shown by Boltzmann that this
quantity can be expressed as the logarithm of total number of
microstates in phase space: S = log(N ). N is the number of microstates
associated with a macrostate of isolated system. Gibbs put forward an
important step in statistical mechanics by introducing entropy for non-
isolated systems, in terms of the probability to occupy system's
microstates:
N
S = ln( pi−1) = ∑ pi ln( pi−1),
i =1
where pi denotes the probability of i-th microstate occurrence. This
entropic form, further discussed by von Neumann and Shannon, and
constituting the basis of traditional statistical mechanics, is additive.
Consider a system composed of two probabilistically independent
subsystems A and B (i.e., pijA + B = pi A pj B , ∀ (i , j )), then we will have
S (A + B ) = S (A) + S (B ). In the information theory, the entropy mea-
sures the uncertainty in an ensemble. Jaynes showed that all statistical
features of physical systems are immediate consequence of constrained
entropy maximization. This optimization yields exponential form for
occupation probability of i-th state of the system.
If the correlations between the elements are strong enough, then
the extensivity of entropy is lost, being therefore incompatible with its
thermodynamic notion. Many efforts have been devoted to overcome
this difficulty by proposing non-additive entropies, which enable the
generalization of traditional statistical mechanics. For a wide class of
physical systems which entail long-range interactions, it is necessary to
use non-additive entropy. These entropy definitions share the common
feature that they yield distributions of power law form under the
entropy maximization (Darooneh et al., 2010). The most studied non-
additive entropy, is introduced by Tsallis (1988):
Journal of Theoretical Biology 419 (2017) 305–309
N N
1 − ∑i =1 piq
Sq = ln q ( pi−1) = ∑ pi lnq ( pi−1) = ,
i =1
q−1 (2)
where ln q (x ) = (x1− q
− 1)/(1 − q ) referred to as q-deformation of tradi-
tional natural logarithm function. For the particular case of equal
probabilities, we obtain Sq = ln q (N ) for Tsallis entropy as its upper
bound. Real quantity q was introduced as non-extensivity parameter
for physical systems that presents the long range interactions Sq is
non-additive for any q ≠ 1. Hence, for two probabilistically indepen-
dent subsystems, it satisfies Sq (A + B ) = Sq (A) + Sq (B ) + (1 − q ) Sq (A) Sq (B ).
We further see that, additivity is asymptotically recovered in the limit
q⟶1.
The optimization of Tsallis entropy, in the presence of probability
normalization and expectation value constraints, leads to q-exponential
form for probability distribution of the microstates associated with a
macrostate of system:
1
expq (x ) = [1 + (1 − q ) x ]1− q . (3)
q-exponential function, as a power-law function in q ≠ 1, yields
traditional exponential form in q⟶1. Moreover, the generalization of
Gaussian distribution, which is proportional to expq (−x 2 ), properly
describes single peak non-Gaussian behaviors. This power law func-
tion, namely q-Gaussian, has been applied to many problems in the
literature. It also marginally reduces to traditional Gaussian distribu-
tion in q⟶1. This capability motivates people to use it in many natural
and artificial complex systems (Gell-Mann and Tsallis, 2004, Mehri and
Darooneh, 2011). Non-extensivity parameter (q), is known as a
convenient statistical tool to classify the correlation range in complex
systems (Kowalski et al., 2013). Due to complex structure of human
retina, in next section, we will try to describe the spatial distribution of
its photoreceptors using q-exponential function extracted from Tsallis
entropy.
3. The photoreceptors distribution in retina
In this section, we check the spatial distribution of human eye
photocells, by using data from the experiments carried out by Curcio
et al. (1990) and Song et al. (2011). Finite dimensions of retina limit
the size of related data, and forces us to deal with incomplete
information.
First of all, we will try to find a better distribution function
matching with descending behavior of cone photoreceptors packing
density, by moving away from fovea in different areas of retina. We fit
(1) the corresponding data from Curcio et al. (1990) with exponential,
power-law and q-exponential functions. As stated in Fig. 1 q-exponen-
Fig. 1. The number of cones as a function of eccentricity for retinal quadrants. q-
exponential function shows the best fitting result to packing data from Curcio et al.
(1990).
306
A. Mehri Journal of Theoretical Biology 419 (2017) 305–309

Fig. 2. The cone density as a function of eccentricity. for nasal and temporal hemiretinae (Left), and for superior and inferior hemiretinae (Right). The fitting results show that q-
Gaussian distribution fits better that Gaussian distribution to empirical data.

Fig. 3. The cone density in retinal quadrants as a function of eccentricity for two age groups. The fitting results show that, young subjects have higher q-parameter value in comparison
with old ones. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)

tial model gives highest goodness of fit (r 2 = 0.999 ). It is clear that, q-
exponential function is fitted to packing density data better than two
others. In other words, this model function properly describes spatial
distribution of retinal photoreceptors, with incompleteness in the tail
of distribution for very small packing densities. Non-extensive statistics
is widely applied to investigate such incomplete systems (Tsallis, 2009).
Cones density in both horizontal and vertical directions along the
retinal meridian follows a pick function, with a maximum in the center
of fovea. Their spatial distribution in retina, from nasal to temporal
(left panel) and from superior to inferior (right panel), is depicted in
Fig. 2. We fit these data (from Curcio et al. (1990)) with Gaussian and
q-Gaussian model functions. It is clear that, q-Gaussian distribution
has better fitting result to mentioned data.
Figs. 1 and 2 confirm the validity of our proposal to apply non-
extensive statistics for describing photoreceptors’ distribution in hu-
man retina. Now, we want to extract non-extensivity (q-parameter) of
spatial distribution of cones using experimental data from Song et al.
(2011) for young and old subjects. By fitting the cones density data to
q-exponential function, one can obtain q-parameter. Fig. 3 illustrates
packing density of cone photocells, in nasal, temporal, superior and

307
A. Mehri Journal of Theoretical Biology 419 (2017) 305–309

Fig. 4. The cone density as a function of eccentricity along the horizontal (left) and vertical (right) retinal meridians for two age groups. Blue dotted curves correspond to fitting q-
Gaussian function to youngs group data, and red dashes lines belong to fitting to olds data. (For interpretation of the references to color in this figure legend, the reader is referred to the
web version of this article.)

young and old subjects. Its value is greater for cones distribution in
young subjects’ retina. It worth noting that, q-parameter has greater
values for cones retinal distribution along horizontal meridians.
We repeat fitting process for rod photoreceptors data form Curcio
et al. (1990). The fitting results with q-exponential function are
presented in Fig. 5. It is evident that, the distribution of rods in retina
is well described by q-exponential function. Non-extensivity parameter
has negative values for this type of photocells. Its absolute values are
greater in nasal and temporal regions, along horizontal meridian.
We also calculate information entropy for spatial distribution of
cones photoreceptors in retinal quadrants. In this procedure, spatial
probability distribution is defined as normalized population of photo-
cells in different distances from foveal center. Shannon (Gibbs) entropy
and Tsallis entropy can be obtained by using relations (1) and (2),
respectively. We need to know q-parameter value in Tsallis entropy
calculation. So, we apply non-extensivity parameters obtained from
Fig. 5. The distribution of rods as a function of eccentricity in for retinal quadrants fitting q-exponential function to cones data, which are discussed in
(Curcio et al., 1990). Fitting processes result in q-parameter for rods spatial distribution
Fig. 3. Table 1 contains our results in entropy calculation. We see that,
in retina.
both of the additive and non-additive entropies discriminate between
old and young subjects. It worth noting that, difference between Tsallis
Table 1
Shannon entropy (S) and Tsallis entropy (Sq) for spatial distribution of cones at four entropy of olds and youngs is greater than distance between Shannon
meridians of the retina. Old subjects have greater entropy values than youngs. entropy of them. This means that, Tsallis entropy has better results in
aging distinction by using cones distribution.
meridian age S Sq

Nasal old 2.38 1.72 4. Conclusion

young 2.32 1.65
Temporal old 2.37 1.49
young 2.30 0.89 In summary, in this article we evaluate spatial distribution of human
Superior old 2.33 1.75 retinal photoreceptors. It is found that q-exponential and q-Gaussian
young 2.29 1.60 functions appropriately describe packing density of cone and rod
Inferior old 2.32 1.71 photoreceptors (Figs. 1 and 2). We extracted non-extensivity parameter
young 2.29 1.67
(q) for their distribution in four meridians of the retina, using empirical
data from Curcio et al. (1990), Song et al. (2011). Nonextensivity of
inferior regions of retina, as a function of retinal eccentricity, for young cones retinal distribution has greater values for young subjects in
(blue diamonds) and old (red squares) subjects. All data are fitted with comparison with old ones. We also found that, absolute q-parameter
q-exponential model function (dashed and dotted curves), to find non- of both types spatial distribution along the horizontal meridian is higher
extensivity parameter. As can be seen in the figure, non-extensivity than that along vertical meridian (Figs. 3–5).
parameter has greater value in all parts of retina for young subjects. We calculated Shannon and Tsallis entropies for cones density
Hence, this parameter is an appropriate index of subject's age. distribution in retina (Table 1). It is shown that, Tsallis non-additive
As mentioned earlier, cones density varies along retinal meridian in entropy has better results in distinction between old and young
horizontal and vertical directions. We find non-extensivity parameter, subjects. This study confirms capability of non-extensive statistics in
again with fitting q-Gaussian function to the measured data for both investigating retinal photocells. Some eye diseases cause to the
young and old subjects (Song et al., 2011). The fitting results are abnormal distribution of photoreceptors in retina. Nonextensivity
presented in Fig. 4. Here again q-parameter discriminates between parameter as a statistical index can potentially be applied to detect
the problems in human retina.

308
A. Mehri
Acknowledgments
We express our deep sense of gratitude to M. Jamaati and R.
Khanbabaie for their useful discussions. We are also very much
thankful to the referees for their valuable comments.
References
Abraham, P., 2009. A Complete Guide to How the Body Works – The Atlas of The Human
Body. Thunder Bay Press, San Diego.
Binder, M.D., Hirokawa, N., Windhorst, U., 2009. Encyclopedia of Neuroscience.
Springer-Verlag, Berlin.
Bridgeman, B., 2013. Lecture Note on Behavioral Neuroscience. University of California,
Santa Cruz.
Curcio, C.A., Sloan, K.R., Kalina, R.E., Hendrickson, A.E., 1990. Human photoreceptor
topography. J. Comp. Neurol. 292, 497–523. http://dx.doi.org/10.1002/
cne.902920402.
Curcio, C.A., Millican, C.L., Allen, K.A., Kalina, R.E., 1993. Aging of the human
photoreceptor mosaic: evidence for selective vulnerability of rods in central retina.
Invest. Ophthalmol. Vis. Sci. 34, 3278–3296.
Darooneh, A.H., Naeimi, G., Mehri, A., Sadeghi, P., 2010. Tsallis entropy, escort
probability and the incomplete information theory. Entropy 12, 2497–2503. http://
dx.doi.org/10.3390/e12122497.
Foster, R.G., Provencio, I., Hudson, D., Fiske, S., Grip, W., Menaker, M., 1991. Circadian
photoreception in the retinally degenerate mouse (rd/rd). J. Comp. Physiol. A 169,
39–50. http://dx.doi.org/10.1007/BF00198171.
Gell-Mann, M., Tsallis, C. (Eds.), 2004. Nonextensive Enropy-interdisciplinary
Applications, Oxford University Press, New York.
309
Journal of Theoretical Biology 419 (2017) 305–309
Hecht, E., 1987. Optics 2nd ed.. Addison-Wesley, Reading, Massachusetts.
Kowalski, A.M., Rossignoli, R., Curado, E.M., 2013. Concepts and Recent Advances in
Generalized Information Measures and Statistics, Bentham e-Books. http://dx.doi.
org/10.2174/97816080576031130101.
Lombardo, M., Lombardo, G., Lomoriello, D.S., Ducoli, P., Stirpe, M., Serrao, S., 2013.
Interocular symmetry of parafoveal photoreceptor cone density distribution. Retina
33 (8), 1640–1649. http://dx.doi.org/10.1097/IAE.0b013e3182807642.
Mehri, A., Darooneh, A.H., 2011. Keyword extraction by non-extensivity measure. Phys.
Rev. E 83, 056106. http://dx.doi.org/10.1103/PhysRevE.83.056106.
Obata, R., Yanagi, Y., 2014. Quantitative analysis of cone photoreceptor distribution and
its relationship with axial length, age, and early age-related macular degeneration.
PLoS One 9 (3), e91873. http://dx.doi.org/10.1371/journal.pone.0091873.
Page, M., 2001. Human Body: An Illustrated Guide to Every Part of the Human Body and
How it Works. Dorling Kindersley, London.
Panda-Jonas, S., Jonas, J.B., Jakobczyk-Zmija, M., 1995. Retinal photoreceptor density
decreases with age. Ophthalmology 102, 1853–1859. http://dx.doi.org/10.1016/
S0161-6420(95)30784-1.
Purves, D., et al., 2001. Neuroscience 2nd edition. Sinauer Associates, Sunderland.
Song, H., Chui, T.Y.P., Zhong, Z., Elsner, A.E., Burns, S.A., 2011. Variation of cone
photoreceptor packing density with retinal eccentricity and age. Invest. Ophthalmol.
Vis. Sci. 52, 7376–7384. http://dx.doi.org/10.1167/iovs.11-7199.
Tsallis, C., 1988. Possible generalization of Boltzmann-Gibbs statistics. J. Stat. Phys. 52,
479–487. http://dx.doi.org/10.1007/BF01016429.
Tsallis, C., 2009. Introduction to Non-extensive Statistical Mechanics. Springer, New
York.
Willoughby, C.E., Ponzin, D., Ferrari, S., Lobo, A., Landau, K., Omidi, Y., 2010. Anatomy
and physiology of the human eye: effects of mucopolysaccharidoses disease on
structure and function – a review. Clin. Exp. Ophthalmol. 38, 2–11. http://
dx.doi.org/10.1111/j.1442-9071.2010.02363.x.