Optical Design of a Compact Image Acquisition Device for Mobile

Diabetic Retinopathy Screening

David Melo

1,2

, Jo

ao Costa

, Filipe Soares

and Pedro Vieira

Department of Physics, Faculdade de Ci

encias e Tecnologia, Universidade Nova de Lisboa,

Quinta da Torre, 2829-516, Caparica, Portugal

Fraunhofer Portugal AICOS, Rua Alfredo Allen, 455/461, 4200-135, Porto, Portugal

Keywords:

Diabetic Retinopathy, Fundus Camera, Optical System Design, Mechanical Prototyping.

Abstract:

Imaging the eye retina is critical to diagnose various pathologies, particularly Diabetic Retinopathy, which is

the leading cause of avoidable blindness in the world. Accessing the retina can be achieved through ophthal-

moscopes with small ﬁeld-of-view or Optical Coherence Tomography and fundus cameras, which are larger

and expensive. The image acquisition through tabletop fundus cameras is the preferred method for retinopa-

thy screening. However, these devices tend to be cumbersome and require expertise for operation, limiting its

broad application. In this paper, a compact optical system was designed for a handheld and smartphone-based

fundus camera prototype called EyeFundusScope, with the main goal of low-cost and high coverage screening.

The key features for the compact optical system are the mobile and non-mydriatic acquisition of fundus im-

ages by a smartphone camera, with high ﬁeld-of-view. The simplicity of the optical system was accomplished

by a three lens system setup, simulated using ray tracing software. The results reveal a system with only a few

aberrations in the periphery but with a good resolution at the center of 41

◦

ﬁeld-of-view. Besides the optical

system, a mechanical prototype was designed with the purpose of being 3D printed and easily portable.

1 INTRODUCTION

1.1 Diabetic Retinopathy

The eye retina is the only structure in the body where

vessels can be directly seen, without intrusive pro-

cedures. Imaging this structure is extremely impor-

tant in the diagnosis of various pathologies, particu-

larly Diabetic Retinopathy. This is a microvascular

disease caused by the diabetes mellitus condition, af-

fecting 76% of the diabetic patients for longer than

20 years (Cheung et al., 2010) and being the leading

cause of blindness in adults with working age (Bunce

and Wormald, 2006). It is characterized by the loss of

perycites and by a progressive capillary occlusion that

occurs mostly without symptoms. The capillary oc-

clusion can lead to retinal ischemia and to the break-

down of the blood-retinal-barrier (Tarr et al., 2013).

The disease can be divided in two different stages:

Non-proliferative and Proliferative (Kauppi, 2010).

The ﬁrst is characterized by abnormalities in the

blood vessels materialized in the leakage of sub-

stances from the lumen of the vessels to the reti-

nal epithelium. The leakages may be the blood it-

self leading to microaneurysms and intraretinal hem-

orrhages, and lipids leading to hard and soft Exudates

(Kauppi, 2010; Cunha-Vaz, 2007; Giancardo, 2012).

The Proliferative stage is characterized by the creation

of new blood vessels surrounding occluded regions

(neovascularization). The new blood vessels, being

more fragile than the previous ones, increase the risk

of bleeding and do not solve retinal ischemia (Gian-

cardo, 2012). In the Proliferative stage there is also

the formation of ﬁbrous tissue that while contract-

ing can provoke retinal detachment (do Prado et al.,

2002).

Several types of instruments can perform ophthal-

mological examination, but for the diagnosis of Dia-

betic Retinopathy the use of Fundus Camera is pre-

ferred (Salz and Witkin, 2015).

The asymptomatic proﬁle of the initial progres-

sion of diabetic retinopathy is problematic for diag-

nostic purposes. On the other hand, the success of

early treatment provides a large incentive to imple-

ment population-based screening programs for dia-

betic patients. In these programs, images of the pa-

tient retinas are acquired and assessed by qualiﬁed

technicians and ophthalmologists, which lead to high

Melo, D., Costa, J., Soares, F. and Vieira, P.

Optical Design of a Compact Image Acquisition Device for Mobile Diabetic Retinopathy Screening.

DOI: 10.5220/0006592200630070

In Proceedings of the 11th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2018) - Volume 1: BIODEVICES, pages 63-70

ISBN: 978-989-758-277-6

costs due to the required use of expensive and bulky

equipment, and the laborious task of manual analy-

sis by scarcely available medical personnel. The pro-

totype EyeFundusScope, currently under investiga-

tion by Fraunhofer Portugal AICOS, aims to address

these issues by researching on a self-contained solu-

tion comprising automated retinopathy detection with

a low cost optical attachment to a smartphone for reti-

nal image acquisition. The major goal is to improve

patient access to early treatment and decrease the bur-

den of screening actions on healthcare systems.

The present work is a contribution to a compact

optical system, proposed to achieve this goal. The

resolution of the optical system is intended to be sufﬁ-

cient to distinguish microaneurysms, the small struc-

tures that commonly appear in the earliest stages of

Diabetic Retinopathy and whose size varies from 25

to 125 micrometers (Li et al., 2006).

1.2 Fundus Camera

A Fundus Camera is a device that allows the observa-

tion of the structures and the blood vessels in the oc-

ular fundus, being employed in the diagnosis of sev-

eral pathologies (like Diabetic Retinopathy, described

in section 1.1) (P

erez et al., 2012).

When compared with other eye examination de-

vices, a Fundus Camera enables patient documenta-

tion and easy follow, as well as allowing analysis of a

great extent of the patient retina, due to the wide ﬁeld-

of-view empowered by the usage of indirect oph-

thalmoscopy principles, hardly achievable with direct

ophthalmoscopy methods (Benbassat et al., 2012),

(Phillips, 1984).

The importance of the fundus examination can be

seen in many medicine ﬁelds and not only in ophthal-

mology. Since the retina is the human body structure

where the vessels can more easily be seen with no

use of ionizing radiation (P

erez et al., 2012), ﬁelds

like Neurology and Cardiology can also use the capa-

bilities of a Fundus Camera (Patton et al., 2005). A

handheld portable Fundus Camera can also be a cru-

cial tool in the development of telemedicine (P

erez

et al., 2012).

In the present work, we propose a simple Fundus

Camera optical system, using only 3 lenses, allowing

a 40

ﬁeld-of-view with minimized aberrations and

no need of pupil dilation. Generally, to have good

capabilities, indirect ophthalmoscopes use a consid-

erably high number of lenses (Tran et al., 2012). In

this work, to diminish the production costs, only the

fundamental components of a Fundus Camera were

used. A ﬁeld-of-view of at least 40

◦

is desired, since

this is generally considered an adequate tradeoff be-

tween sufﬁcient retinal area imaged and enough reso-

lution for analysis of ﬁner retinal features, thus allow-

ing obtaining clinical meaningful conclusions about

eventual abnormalities. A diagram with these compo-

nents is presented in Figure 1. The software used for

the Optical Simulation was BEAM IV, an Optical Ray

Tracer developed by Stellar Software.

Ocular Lens

Light Source

Condenser Lens

Beamsplitter

Objective Lens

Eye Model

Figure 1: Demonstration of the fundamental components in

the Fundus Camera prototype.

1.3 Optical Principles

To reach an optimal optical system several lenses

types were tested. To simulate them according to

the characteristics supplied by the manufacturers, the

thin lens approximation was used. This approxima-

tion neglects the thickness of the lens and considers

that the unit planes pass through the axial point of the

inﬁnitely thin lens (Jenkins and White, 1957). Con-

sidering that the media on both sides of the lens is the

same, the following equation can be used to describe

it (Born and Wolf, 1999).

Lens-Maker’s Formula

lens

− n

(

−

) (1)

Where n

is the refractive index of the surrounding

medium, the air in this case, equal to 1, n

lens

is the re-

fractive index of the lens, R

is the radius of curvature

of the ﬁrst surface and R

is the radius of curvature of

the second surface. The P

lens

is the refractive power

in diopters.

To simulate a Fundus Camera optical system, the

optical path taken by the rays is separated in two dif-

ferent ones by the usage of a beamsplitter (Tran et al.,

2012). The path that describes how the rays illu-

minate the retina is called illumination path and the

path describing how the rays go from the retina to the

smartphone camera is called imaging path. To sim-

ulate them two different approaches were tested, the

4-extreme model for the illumination, and the parallel

rays model for the imaging path (see Figure 2). The

4-extreme model assumes that the light source emits

BIODEVICES 2018 - 11th International Conference on Biomedical Electronics and Devices

from a single point with a certain aperture previously

declared by the manufacturer. The angles for which

the relative luminous intensity is bellow half the max-

imum intensity can be neglected.

Figure 2: Demonstration of the 4-extremes model along the

x-axis.

The parallel rays model assumes that when two

rays focused at some point reach a lens they are colli-

mated and leave the lens with the same direction and

parallel with each other. In Figure 3 one of the appli-

cations of this model is demonstrated, showing rays

focused at some point, leaving the pupil parallel.

Figure 3: Demonstration of the parallel rays model. In this

ﬁgure rays are focused on the retina, leaving the pupil col-

limated and parallel.

1.4 Related Work

Recently, the features of handheld Fundus Camera

prototypes have increased signiﬁcantly when com-

pared with the traditional tabletop fundus cameras

(Tran et al., 2012), proving as an helpful instrument

in the diagnosis of many pathologies related with

the retina and facilitating telemedicine applications

(Quellec et al., 2016; Jin et al., 2017). This improve-

ment of the capabilities of handheld devices led to a

variety of different approaches. Some examples that

reﬂect the recent scientiﬁc development are:

• Nonmydriatic Fundus Camera Based on the Rasp-

berry Pi



Computer: Uses the Raspberry Pi



camera module coupled with a Condenser Lens

to perform fundus imaging with a very low pro-

duction cost (Shen and Mukai, 2017).

• Eye-Selﬁe: By using internal ﬁxation points as

targets, it allows a self-performed acquisition of

the fundus photography, completely by the patient

(Swedish et al., 2015).

There are already, some fundus camera prototypes

available in the market but most, present one of the

following issues:

• Low Field Of View, as for the D-EYE ophthalmo-

scope (D-EYE S.r.l, ).

• Require pupil dilation, as for the Volk inView

(Volk Optical Inc., a).

• High price, as for the Volk Pictor Plus (Volk Op-

tical Inc., b).

The system we propose differs from the previous

approaches by using a smartphone for non-mydriatic,

high ﬁeld-of-view retinal image acquisition. The use

of a smartphone instead of custom electronic devices

for image capture and processing allows a substan-

tial decrease in costs while allowing for a very high

image quality and resolution, thus guaranteeing the

cost-effectiveness of the overall solution.

2 TOOLS AND METHODS

2.1 Eye Model

To guarantee a satisfactory ﬁeld-of-view, an accurate

model of the eye is needed. The eye has two refrac-

tive lenses, the cornea and the crystalline lens. Based

on the literature (Atchison and Smith, 2000) and fol-

lowing a similar approach to (Tocci, 2007), a model

of the eye was created in BEAM IV considering the

radius of curvature, diameter and asphericity coefﬁ-

cients of all the structures relevant for ray tracing. The

pupil has been designed with a 4 mm diameter to sim-

ulate a non-mydriatic acquisition with no visible light

and is coincident with the lens anterior surface. The

chromatic aberrations from the eye were neglected as

the change in diopters at different wavelengths were

not considered signiﬁcant in the scope of this work

(Atchison and Smith, 2000).

The deﬁned structures of the eye, as represented

in Figure 4, are:

• Corneal Anterior surface:

Diameter = 11.50 mm

Radius of Curvature = 7.75 mm

Asphericity coefﬁcient = -0.2

• Corneal Posterior surface:

Diameter = 11.50 mm

Radius of Curvature = 6.8 mm

Asphericity coefﬁcient = 0

Optical Design of a Compact Image Acquisition Device for Mobile Diabetic Retinopathy Screening

• Pupil/ Lens Anterior surface:

Diameter = 4 mm

Radius of Curvature = 10 mm

Asphericity coefﬁcient = -0.94

• Lens Posterior Surface:

Diameter = 9 mm

Radius of Curvature = -6 mm

Asphericity coefﬁcient = 0.96

• Retina:

Diameter = 24 mm

Radius of Curvature = 12 mm

Asphericity coefﬁcient = 0

Anterior Cornea

Posterior Cornea

4 mm Pupil

Retina

Lens

Figure 4: BEAM IV simulation of the eye model used with

collimated rays being focused on the retina.

2.2 Illumination Path

For the illumination path the main goal is a 40

◦

ﬁeld-

of-view with a uniform illumination of the retina. The

diagram of the illumination path is represented in Fig-

ure 5. The image is obtained using a white visible

LED but, to allow a non-mydriatic acquisition, a Near

Infrared (NIR) LED is used, helping the examiner to

perform alignment of the device with the eye and to

ﬁnd the area of the retina to be imaged. As the NIR

LED is simply used for guidance, only the Visible

LED imaging and illumination capabilities were eval-

uated. As in most fundus cameras, there is a lens

above the light source to collimate the rays and an-

other lens to focus the rays (Tran et al., 2012). This

lens that focuses the rays before reaching the eye is

called objective lens and it is where the simulations

performed began. To obtain a ﬁeld-of-view of 40

◦

there is a constraint that the relationship

W D

2 f

should

be equal to or larger than sin(20) (see Figure 6), where

W D stands for Working Diameter and f means the ef-

fective focal length of the lens.

The type of objective lens to choose should min-

imize spherical aberrations. This condition, coupled

with the required numerical aperture, makes Aspheric

lenses the only suitable option for the focusing of

the rays when reaching the retina. After searching

PCX Lens

Beamsplitter Aspheric Lens

Light Source

Eye Model

Figure 5: Illumination path and the description of the com-

ponents used.

Working

Diameter

Effective Focal Length

20º

Figure 6: For a 20

◦

half-angle the relationship

W D

2 f

should

be equal to or larger than sin(20).

for a lens that ﬁts these requirements we opted to

use a Thorlabs aspheric lens with 50.00 mm diame-

ter, 40.00 mm focal length and SLAH-64 glass type,

placed 25 mm ahead to the right of the center of the

beamsplitter. For the collimation of the rays coming

from the light source, the used Condenser lens was an

Edmund Optics Plano-Convex Lens with 25.4 mm di-

ameter, 38.1 mm focal length and N-BK7 glass type,

placed 47 mm below the center of the beamsplitter.

To check the distance between the objective lens

and the human eye, the plane where the rays were in

focus was calculated. This plane is called the focal

plane and is where the circle of confusion is mini-

mum.

In a theoretically aberration free-system, this is

where the pupil should be placed. However, after test-

ing it, was observed that due to spherical aberrations,

the intermediate rays reached the optical axis further

than the extreme rays, so only 40% reached the retina.

This problem was solved by putting the pupil 5 mm

after the focal point of the extreme rays. Using this

different conﬁguration, 90% of the emitted rays reach

the retina and the illumination proﬁle is uniform, as

can be observed in Figure 7. The half-angle on the

retina calculated was 20.65

◦

leading to a total ﬁeld-

of-view of 41.3

◦

In the previous diagrams, the simulated light

source emits on a single wavelength (486 nm). As

the white light emitted by the LED has a continuous

emission spectrum, measurements at the other end of

BIODEVICES 2018 - 11th International Conference on Biomedical Electronics and Devices

Figure 7: The retinal area illuminated, also called retinal

illumination proﬁle, is described, for the conﬁguration rep-

resented in Figure 5. The uniformity is noticeable.

the visible spectrum are needed. It was concluded that

for red light (656 nm) the system is also optimal with

92% of the rays reaching the retina with an half-angle

of 20.57

◦

, leading to a total ﬁeld-of-view of 41.14

◦

2.3 Imaging Path

For the imaging path the key features desired are the

almost complete fulﬁllment of the Smartphone Cam-

era sensor and the minimization of aberrations.

In order to perform ray tracing analysis of the

imaging path, two pairs of parallel rays were consid-

ered, one pair parallel with the optical axis and the

other with 20

◦

inclination. The distance between the

rays, on each pair, was equal to the size of the pupil,

4 mm.

The system was optimized for a LG Nexus 5X

camera whose relevant speciﬁcations are:

• Horizontal Angle of View : 68.2

◦

• Vertical Angle of View : 53.1

◦

• Sensor Size : 1/2.3” (6.17 x 4.55 mm)

The ﬁnal optical system must guarantee that the

Vertical Angle of View is mostly ﬁlled with the retinal

image, in order to allow the highest possible retinal

resolution, essential for the clinical analysis of ﬁne

features. The ﬁrst solution tested was with the As-

pheric Lens, described in section 2.2, as an Objective

Lens and with a Plano-Convex Lens with 40.0 mm of

focal length and 25.4 mm diameter serving as an Oc-

ular. The maximum diameter for the Plano-Convex

lens to ﬁt the scope of a compact system was deﬁned

to be of 25.4 mm. The system presented too much

aberrations and the rays did not reach the Smartphone

Camera parallel. To correct the aberrations, as the

Objective Lens already fulﬁlled the requirements for

the illumination path, the Ocular Lens was changed.

A Best-Form Lens with 40.0 mm focal length, 25.0

Diopters and 25.4 diameter was tested. The aberra-

tions were almost eliminated and the rays reached the

smartphone camera parallel to each other, with an in-

clination of approximately 22

◦

, leading to an angular

ﬁeld-of-view of 44

◦

. The diagram can be seen in Fig-

ure 8. The angle in this conﬁguration is inferior to the

achieved with the Plano-Convex lens as an ocular, but

still wide enough to avoid a signiﬁcant crop.

Figure 8: The imaging path with a Best-Form Lens as the

Ocular Lens and an Aspheric Lens as the Objective, using

the parallel rays model.

In this setup the distance between the Smartphone

CMOS sensor and the Best-Form Lens is 22 mm. The

distance between the surface of the Best-Form Lens

with less curvature and the planar surface of the As-

pheric Lens is 50 mm and between the Aspheric lens

and the eye is 33 mm.

2.4 Imaging Path for Eyes with

Refractive Errors

Eyes with refractive errors present different optical

characteristics and so the distance between lenses in

the optical system must be adjustable to compensate

this.

As the smartphone camera is able to change its

focus target distance, the refractive errors were only

simulated in the range -5D to +5D. Since one of the

possible cause of refractive errors is the size of the

eyeball, for the modulation of Myopia the retina was

moved 3 mm away from the refractive center of the

eye. Concerning the modulation of Hyperopia the

eyeball was shortened 3 mm.

In Figure 9, the system conﬁguration for an eye

without any refractive error is shown. In Figure 10

and Figure 11, diagrams showing the adjustments

done to compensate these refractive errors are shown.

Figure 9: The distances between each components to image

an eye without refractive errors.

Optical Design of a Compact Image Acquisition Device for Mobile Diabetic Retinopathy Screening

For the Myopic eye the error is corrected by mov-

ing the Objective Lens 5 mm away from the eye.

Figure 10: The distances between each components to im-

age an eye with refractive errors (Myopia).

Concerning the Hyperopic eye, the Objective lens

is approximated 5 mm to the eye.

Figure 11: The distances between each components to im-

age an eye with refractive errors (Hyperopia).

2.5 Mechanical Prototyping

The optical system previously described is to be im-

plemented in a 3D printed prototype. The design of

the mechanical prototype was developed using Solid

Works. The important goals for this prototype are to

allow the arrangement of the desired lens, a support

for the smartphone that ensures that the camera is cen-

tered with the optical path, and the adjustment of the

objective lens.

As can be seen in Figure 12, the movements of the

Objective are made with a Rack and Pinion system to

allow the examiner to precisely search for the working

distance, thus allowing the best possible focusing of

the retina.

Other solution to allow the movement of the lens

would be by the use of threaded surfaces in both sides,

so the rotation of the objective ensures a change in

the working distance. This approach was rejected be-

cause it is expected the future implementation of a

piece leaning against the patient forehead, to guaran-

tee the centering with the Optical Path. The rotation

of this piece, in contact with the patient, would not be

comfortable or, possibly, safe.

Figure 12: Mechanical prototype showing both the smart-

phone support and the rack and pinion mate.

Figure 13: Section view of the prototype. In blue is the

PCX Condenser Lens, in yellow the Best-Form Lens, in or-

ange the Aspheric Lens and in green the Beamsplitter. The

Beamsplitter is in the form of a cube only for simulation,

to simplify the ﬁxation in the mechanical case. The small

piece highlighted by the red arrow limits the adjustment be-

tween the aspheric lens and the beamsplitter. The conﬁgu-

ration presented is for an eye with myopia.

3 RESULTS AND DISCUSSION

3.1 Complete System Designed

The Complete System presented in Figure 14 has the

following elements:

• Light Source (Visible or Near Infra-red LED).

• N-BK7 Plano-Convex Lens, 38.1 mm Focal

Length, 25.4 mm φ, VIS-NIR Coated, 44,00e,

Edmund Optics.

• S-LAH64 CNC-Polished Aspheric Lens, 40.0

mm Focal Length, 50 mm φ, 392,00e, ThorLabs.

• Beamsplitter 50R/50T 35x35 mm, (≈ 35,00 e).

• N-BK7 Best-Form Lens, 40 mm Focal Length,

25.4 mm φ, 39,00e, Thorlabs.

The utilization of aperture stops is dependent on

the LEDs beam angle. The aperture can be used to

stop the rays emitted by the LED at a wider angle than

the necessary for a ﬁeld-of-view of 40

◦

, preventing

reﬂections and the imaging of undesired areas.

BIODEVICES 2018 - 11th International Conference on Biomedical Electronics and Devices

Smartphone Camera

Best-Form Lens

Light Source

PCX Lens

Beamsplitter

Aspheric Lens

Eye Model

Figure 14: Complete Optical System.

The main features of the developed system are:

• Around 40

◦

ﬁeld-of-view.

• Non-Mydriatic Acquisition, for a 4 mm pupil size,

achievable by using the NIR LED for guidance.

• No signiﬁcant aberrations (Spherical and Chro-

matic).

• Uniform Illumination of the Retina.

• Simple and affordable lens system.

3.2 Imaging Path Practical Tests

Some practical tests were performed on the Imag-

ing Path using an Optic Table. A 3.5 mm Iris was

used to simulate the pupil and a 38.1 mm focal length

Plano-Convex lens was used replacing the eye refrac-

tive center. For these tests, only the lights in the room

were used to illuminate the target, and this target was

a millimeter paper placed at 38 mm from the 38.1

mm Plano-Convex Lens. A black paperboard was

used, to prevent the loss of light to the environment

and undesired reﬂections. The smartphone used was

a Microsoft Lumia 650 with 63.4

◦

horizontal ﬁeld-

of-view, 49.7

◦

vertical ﬁeld-of-view and a 3.6 x 2.7

mm sensor size with an 8 Megapixels camera. It was

placed at 20 mm from the Best-Form Lens. The re-

sults, as shown in Figure 15, demonstrate a system

with only a few aberrations in the periphery but with a

good resolution at the center showing a ﬁeld-of-view

of about 43

◦

The spatial resolution was also assessed. Consid-

ering an 8MP resolution camera, the captured pic-

ture has 3272 × 2454 pixels. As the horizontal res-

olution is superior, the calculations are presented for

the vertical resolution. Due to the crop in the vertical

ﬁeld-of-view, the vertical half-angle of view will be

of about 17.5

◦

, which would correspond to 6.0 mm

in the retina. Thus, the spatial resolution is 4.9 µm,

meaning that each pixel contains 4.9 µm of the reti-

nal area, which is sufﬁcient to observe the expected

smallest microaneurysms (25 µm). Still based in the

Figure 15: Image obtained by the Microsoft Lumia 650

smartphone demonstrating that at least the centered 15 mm

of the target could be imaged.

previous calculus, the usage of a smartphone camera,

instead of a higher resolution camera, is justiﬁed by

the eye diffraction limit. A much higher resolution

would be more expensive and the spatial resolution

would start to be close to the eye diffraction limit: for

a 5 mm pupil eye receiving light of about 500 nm,

this diffraction limit is 1.5 µm (Roorda and Duncan,

2015), close to the resolution calculated.

The quality of the image would be improved if the

used millimeter paper could be bent to describe more

accurately the retina and diminish spherical aberra-

tions.

3.3 Comparison with other Fundus

Cameras

The compact optical system described has a total

components cost of about e800, considering the ex-

ternal adapters for the good functioning of the optical

system. Fundus Cameras with similar ﬁeld-of-view

and non-mydriatic acquisition, like the Volk Pictor

Plus, cost about e10000.

The features of the prototype, coupled with its

simplicity and relatively low price compare favorably

with the products currently available, can facilitate the

provision of healthcare world-wide, and make it an in-

teresting solution for under-developed countries.

4 CONCLUSIONS

It can be concluded that the optical system designed

showed satisfactory capabilities in experimental tests,

being able to detect the smallest lesions associated

with Diabetic Retinopathy. Moreover, it is expected

that in the future this device can be a reliable tool

in early-stage diagnosis and, indirectly, contribut-

ing to an improvement in the treatment of Diabetic

Retinopathy cases all over the world.

Optical Design of a Compact Image Acquisition Device for Mobile Diabetic Retinopathy Screening

The 3D-printing of the mechanical prototype de-

scribed in section 2.5 is expected to be further devel-

oped in future work, so that the implementation of the

optical system can be performed. After the prototype

construction, several ”in vivo” tests are expected.

A light hazard measurement, regarding the ISO

15004-2 (ISO 15004-2:2007, 2007) and ISO 10940

(ISO 10940:2009, 2009) norms, for this kind of in-

struments is also mandatory, being the ﬁrst for light

hazard protection of Ophthalmologic Instruments and

the second speciﬁcally for Fundus Cameras.

ACKNOWLEDGEMENTS

We would like to acknowledge the ﬁnancial support

obtained from North Portugal Regional Operational

Programme (NORTE 2020), Portugal 2020 and the

European Regional Development Fund (ERDF) from

European Union through the project Symbiotic tech-

nology for societal efﬁciency gains: Deus ex Machina

(DEM), NORTE-01-0145-FEDER-000026.

REFERENCES

Atchison, D. and Smith, G. (2000). Optics of the Human

Eye. Butterworth-Heinemann, page 259.

Benbassat, J., Polak, B. C. P., and Javitt, J. C. (2012). Ob-

jectives of teaching direct ophthalmoscopy to medical

students. Acta Ophthalmologica, 90(6):503–507.

Born, M. and Wolf, E. (1999). Principles of Optics (7th

Ed). Cambridge University Press.

Bunce, C. and Wormald, R. (2006). Leading causes of cer-

tiﬁcation for blindness and partial sight in England &

Wales. BMC public health, 6:58.

Cheung, N., Mitchell, P., and Wong, T. Y. (2010). Diabetic

retinopathy. The Lancet, 376(9735):124–136.

Cunha-Vaz, J. (2007). Characterization and relevance of

different diabetic retinopathy phenotypes. Develop-

ments in Ophthalmology, 39:13–30.

D-EYE S.r.l. D-EYE Ophthalmoscope.

do Prado, R. S., Figueiredo, E. L., and Magalhaes, T. V. B.

(2002). Retinal detachment in preeclampsia. Arquivos

brasileiros de cardiologia, 79(2):183–186.

Giancardo, L. (2012). Automated fundus images analysis

techniques to screen retinal diseases in diabetic pa-

tients Docteur de l ’ universit

e Automated Fundus Im-

ages Analysis Techniques to Screen Retinal Diseases

in Diabetic Patients.

ISO 10940:2009 (2009). Ophthalmic instruments - Fun-

dus cameras. Standard, International Organization for

Standardization, Geneva, CH.

ISO 15004-2:2007 (2007). Ophthalmic instruments - Fun-

damental requirements and test methods.Part 2: Light

Hazard Protection. Standard, International Organiza-

tion for Standardization, Geneva, CH.

Jenkins, F. and White, H. (1957). Fundamentals of optics.

McGraw-Hill.

Jin, K., Lu, H., Su, Z., Cheng, C., Ye, J., and Qian, D.

(2017). Telemedicine screening of retinal diseases

with a handheld portable non-mydriatic fundus cam-

era. BMC Ophthalmology, 17(1):89.

Kauppi, T. (2010). Eye Fundus Image Analysis for Auto-

matic Detection of Diabetic Retinopathy.

Li, H., Esquivel, A., Davis, G., and Krupinski, E. (2006).

Evaluation of digital resolution for viewing diabetic

retinopathy microaneurysms.

Patton, N., Aslam, T., MacGillivray, T., Pattie, A., Deary,

I. J., and Dhillon, B. (2005). Retinal vascular image

analysis as a potential screening tool for cerebrovas-

cular disease: a rationale based on homology be-

tween cerebral and retinal microvasculatures. J Anat,

206(4):319–348. 15817102[pmid].

erez, M. A., Bruce, B. B., Newman, N. J., and Biousse,

V. (2012). The use of retinal photography in non-

ophthalmic settings and its potential for neurology.

Neurologist, 18(6):350–355. 23114666[pmid].

Phillips, C. I. (1984). Dilate the pupil and see the fun-

dus. Br Med J (Clin Res Ed), 288(6433):1779–1780.

6428541[pmid].

Quellec, G., Bazin, L., Cazuguel, G., Delafoy, I., Coch-

ener, B., and Lamard, M. (2016). Suitability of a low-

cost, handheld, nonmydriatic retinograph for diabetic

retinopathy diagnosis. Translational Vision Science &

Technology, 5(2):16.

Roorda, A. and Duncan, J. L. (2015). Adaptive op-

tics ophthalmoscopy. Annu Rev Vis Sci, 1:19–50.

26973867[pmid].

Salz, D. A. and Witkin, A. J. (2015). Imaging in diabetic

retinopathy. Middle East African journal of ophthal-

mology, 22(2):145.

Shen, B. Y. and Mukai, S. (2017). A portable, inexpensive,

nonmydriatic fundus camera based on the raspberry

 computer. Journal of Ophthalmology, 2017(3):5.

Swedish, T., Roesch, K., Lee, I., Rastogi, K., Bernstein, S.,

and Raskar, R. (2015). eyeselﬁe: Self directed eye

alignment using reciprocal eye box imaging. ACM

Trans. Graph., 34(4).

Tarr, J. M., Kaul, K., Chopra, M., Kohner, E. M., and Chib-

ber, R. (2013). Pathophysiology of Diabetic Retinopa-

thy. ISRN Ophthalmology, 2013:1–13.

Tocci, M. (2007). How to model the human eye in zemax.

Tran, K., Mendel, T. A., Holbrook, K. L., and Yates, P. A.

(2012). Construction of an inexpensive, hand-held

fundus camera through modiﬁcation of a consumer

”point-and-shoot” camera. Investigative ophthalmol-

ogy & visual science, 53 12:7600–7.

Volk Optical Inc. Volk InView.

Volk Optical Inc. Volk Pictor Plus.

BIODEVICES 2018 - 11th International Conference on Biomedical Electronics and Devices