Design of a smart contact lens for controlled retinal lighting in red light therapy

Abstract

Current red-light therapy devices lack control over light distribution on the retina, limiting their evaluation and use while presenting potential risks. We designed a scleral lens embedding 625 nm LEDs strategically placed to provide wide-angle retinal illumination while leaving the pupil free. Two configurations were tested: one with a metalized inner part to reflect light from the LEDs towards the retina and another with guiding micro-optics for better efficiency. Experimental irradiance maps obtained with both configurations demonstrate retinal illumination over 130°. We also discuss how this design could meet ophthalmologic standards and improve red-light therapy for myopia control.

1. Introduction

Red-light therapy is a noninvasive treatment that utilizes low-wavelength red light to stimulate cellular function and promote healing. It has garnered attention for its potential in a variety of medical applications, e.g. retinal neurodegeneration treatments, photo-stimulation of retinal mitochondria [1,2], or photo-biomodulation [3].

Recent studies [4–7] also suggest that Repeated Low-level Red Light (RLRL) therapy may play a role in controlling the progression of myopia. This is particularly significant given the global increase in myopia rates. The treatment is believed to influence eye growth and reduce elongation of the eyeball, a key factor in myopia progression. RLRL could thus offer a novel, non-invasive approach to manage and slow down the development of myopia, especially in children and adolescents.

Further researches and clinical trials are ongoing to better understand its efficiency and optimal use parameters and some recommend that clinicians reconsider its use until its efficiency is confirmed, in particular until safety standards are guaranteed. For instance, L. Ostrin et al [8] measured that red light devices approached or surpassed, after 3 min of continuous viewing, the Maximum Permitted Exposure (MPE), putting the retina at risk of photochemical and thermal damages.

This is mainly due to the devices currently used to apply the treatment, i.e, either desktop ones, such as Eyerising from Optoelectronics Technology [9], or hand-held ones, e.g., the device supplied by CH electronics which is based on a commercial DC torch with nine 670 nm LED mounted behind a light diffuser with an internally reflecting tube [10]. Apart from the fact that these devices require the user to do nothing else, it is clear that retinal illumination will vary due to possible misalignment of the light source with the eyes, or an involuntary or voluntary saccade or blink. To prevent possible photochemical damages, allow for full assessment of the efficiency of this therapy, and determine the best way to apply it (minimum exposure duration, frequency, etc.) the ideal device should allow to control and adjust finely, at any time of the treatment, the retinal illumination.

This inspired the design of the solution described here. Our idea is to embed red LEDs into a contact lens (Fig. 1(a)) around the pupillary axis, with or without additional projection optics. Such a configuration allows for the increase of the retinal area illuminated, while the orientation and number of LEDs and the projection optics provide control over light exposure parameters (e.g., uniformity, intensity, foveal illumination or not, etc.). The second key element is the remote control of the lighting parameters, by an eyewear communicating with the contact lens. Such a solution also allows to solve alignment issues, e.g., blink or voluntary or involuntary eye movements, that could impact the effectiveness of the treatment and raise doubts about how treatments are effectively applied. In addition, it does not block the user’s vision during the treatment, enabling him to maintain his mobility.

Fig. 1.

Principle of the SCL for RLRL-therapy. a) Schematic view of the SCL with 3 LEDs illuminating the retina. b) The eyewear with the NFC reader and the NFC primary coil, located in the temples and the rim. It is connected to a PC through a USB link for controlling the field intensity. The SCL contains the passive elements to bias the 3 LEDs from the received magnetic field. c) Block diagram of the system.

Because optoelectronics components encapsulation in a contact lens is a challenge, we chose to use edge-emitting LEDs coupled with a reflective element to redirect the LEDs’ light towards the retina. In this paper, two optical configurations are investigated: one where part of the inside of the contact lens is metalized to reflect the light from the LEDs towards the retina; a second where a guiding micro-optics is added into the lens, in front of the LEDs for better efficiency.

The paper is organized as follows. The Method section presents successively, the device (contact lens and associated driver), the lighting configurations tested, their optical modeling, the manufacturing process for both of them, and the test bench used to measure performances. The Results section presents the irradiance maps obtained with each optical configuration (including a video). Finally, we discuss how this approach could be made compliant with standard ophthalmologic requirements such as independence of pupil size variation, and wearable by children.

2. Device description:

2.1. Contact lens and associated driver

The basic system has been already described elsewhere [11]. It consists of two elements: an eyewear containing a primary coil, with the driving electronics, and a scleral contact lens (SCL) containing a secondary coil and its electronics. The contact lens is in Poly Methyl MethAcrylate (PMMA) and is made up of three parts: a front surface, a rear surface (on the eye), and the optoelectronics sandwiched between these two surfaces.

The eyewear has two functions: power the electronic circuit by inductive coupling and control it through Near-field communication (NFC). The eyewear is equipped with an NFC reader located in one of the eyewear’s temples, while the primary coil is located in the rim, Fig. 1(b). The NFC reader is built around a PN7642 single-chip from NXP Semiconductors compatible with many NFC tag formats (ISO/CEI 14443A-B). The reader is powered via a USB-C port. The magnetic field intensity emitted by the primary coil can be varied using software provided by NXP (NFC Cockpit). The eyewear weighs 35 g (this could be easily reduced to 25 g by optimizing the empty space to reduce the amount of plastic material).

The second part of the system is thus the smart scleral contact lens (SCL). The light source is made of two or three LRVH9F 625 nm LEDs connected in parallel. They are powered by the NFC field received on the secondary coil. The LEDs act as a single wave rectifier on the sinusoidal NFC voltage generated on the secondary coil. Figure 1(c) shows a block diagram of the instrumented SCL. Figure 1(c) shows a block diagram of the smart SCL.

In this paper, we tested the NXP operating functionalities stand-alone. We did not encapsulate it in the various contact lens prototypes presented hereafter, to simplify the encapsulation challenges with the optics. The electronic driver used here was sufficient to power the contact lens and to modulate the LED intensity to assess the in-eye irradiance performance. The NXP will be justified in the future to bring more driving functionalities to the SCL but after testing what are the ones necessary for the considered red light therapy application.

2.2. Lighting principle and configurations investigated

The idea is to control the light power distribution incoming on the retina while enlarging the illuminated surface. This is achieved by using off-axis sources (the number and locations of which are discussed here). It has the first advantage of facilitating retinal illumination over a wide angle (130°), the second is to keep the central vision free, subject to variations of the pupil size (as discussed in §3.4), and the last to provide a better control of the retinal illumination, in the central retina, by adjusting the power of each light source.

The optoelectronic circuit (i.e., the PCB) limits the pupil size in case of mydriasis. In this study, we chose a PCB with an internal diameter of 5 mm. Although a larger size would be more appropriate for treatment targeting children, we consider this size to be large enough for a proof of concept and discuss this point further in the discussion section.

The encapsulation of a PCB with LEDs into a contact lens is not easy due to the lens curvature and the limited volume available (around 50 mm³). To facilitate it, we position the emitter on the front side and fold the optical path to illuminate the retina. Two configurations were thus designed and tested. In the first one, the inside part of the contact lens in front of the LEDs is etched and metalized. In the second one, a guiding optics is added to the lens, in front of the source for better efficiency. For the sake of simplicity, we use the following indexing: each configuration is abbreviated as M-n or TIR-n. M denotes the metalized configuration and TIR, the one using an additional guiding optics. We use the term TIR by analogy with augmented reality systems using Total Internal Reflection to guide light into a waveguide [12] even if reflections here are due to metallized surfaces . The following number denotes the number of LEDs. For instance, M-3 denotes a lens with 3 LEDs where the optical path folding is due to the lens interior surface metallization. An example of M-lens is depicted in Fig. 2(a), and an example of TIR optics, manufactured in Zeonex K22R and placed upon the SCL’s circuit, is in Fig. 2(b). In Fig2b, the projection optics is annular but only sections could be used as illustrated in Fig. 2(c). This projection optics slightly reduces the pupil (totally Fig. 2(b) or partially Fig. 2(c)) by an additional millimetre. The design and manufacturing of these elements are described in the next section.

Fig. 2.

Example of two contact lens configurations. a) M-3 configuration: the inside of the contact lens is etched and metalized to reflect the three LEDs’ light toward the retina. b) TIR-2 configuration: view from the top of the optical guide above the secondary coil, c) TIR-2 with TIR sections only.

2.3. Optical design and modeling

Simulations were performed with OpticStudio in Non-Sequential mode in a Liou-Brennan eye model [13]. We chose this model because it is a well-known one that considers many realistic factors that other models do not (e.g. off-centre pupil, curved retinal surface and an optical index gradient for the crystalline lens). Since the light from the LEDs hits the reflective element before the front surface of the contact lens, this front surface was not considered in the modeling. The pool of water created by the scleral lens (since it covers the entire cornea and rest on the sclera) was taken into account in our simulations.

Our objective here is only to obtain a homogeneous level of illumination, so this eye model is sufficient even if it perhaps overestimates the level of retinal illumination [14]. By homogeneous illumination we mean here a uniform light distribution, without variations in intensity and will use both terms interchangeably.

The coil inside the SCL has an outer diameter of 10.50 mm and an inner one of 5.0 mm. Previous tests have shown that when encapsulated, the circuit undergoes a conical deformation of around 10°, which is considered in our model.

The LED sources considered are LRVH9F Firefly 0402 from OSRAM, emitting at 625 nm, with 300 µm height and a FWHM of 140°. They are positioned on a radius 3 mm from the center of the coil, arranged in an equilateral triangle if there are three, or face to face when there are two. We only considered two and three LEDs as a reasonable compromise between volume constraints and the need for homogenous retinal illumination.

Regarding the M-configuration, the curvature of the surface to be metalized was modeled by 4 segments with 3 connection points (Fig. 3(a)). Considering three LEDs, the position of each of these three points and the reflector curvature were optimized to maximize retinal illumination and uniformity in a ± 35 degrees cone around the optical axis, for a total illuminated surface of 130°.

Fig. 3.

Optical design of the projection optics. A) Cross-section of the inner lens surface to be metalized in the M-configuration. B) Annular TIR guide (TIR - 1 axis). C) Projection optics with 2 radii of curvatures (TIR – 2 axes).

Regarding the TIR configuration, the advantage is that because it can be manufactured independently from the contact lens, more sophisticated diamond turning tools can be used and thus more complex surfaces. Two sub-cases were investigated. In the first one (TIR-1axis), the optics is annular with a radial curvature as depicted in Fig. 3(B). Basically, the curved surface plays the same role as the segmented one in the M-configuration. In the second case (TIR-2axes), a biconical surface is positioned over each LED as depicted in Fig. 3(C). For both cases, the radii of curvature were also optimized to maximize retinal illumination and uniformity in a ± 35 degrees cone around the optical axis, considering three LEDs. In the future, the benefit of more complex optical designs, such as freeform optics [15] could be investigated.

2.4. Optical simulations

The illumination uniformity norm [16,17] refers to the distribution of the irradiance on a specified surface and how close these values are to each other. We used three parameters to describe the uniformity: the standard deviation of irradiance, $U_1=\frac{E_{avg}}{E_{max}}$ , $U_2=\frac{E_{min}}{E_{max}}$ where U stands for uniformity and E for irradiance.

Irradiance in the absence of an eye, produced by a M-2 and M-3 lens, at a distance equivalent to that of the retina in the eye, is depicted in Fig. 4(a)-(b). This simulation is presented here for later comparison with experimental results (section 3). The retinal irradiance in the Liou-Brennan eye produced by a M-2 and M-3 lens is depicted in Fig. 4(c)-(e). When comparing Fig. 4(a)-(b) to Fig. 4(c)-(d), a slight focusing effect, due to the presence of the eye can be observed. The associated retinal illumination values are reported in Table 1. As expected, the irradiance is too disparate when only two sources are used. Three sources offer a good compromise between a limited number of LEDs and uniform illumination. Simulations with four lenses were also carried out but are not reported here as they did not allow to significantly improve the irradiance homogeneity when compared to the three LEDs case.

Fig. 4.

Simulated Irradiance (130°) with an M-2 (a) or M-3(b) configuration without the eye at retinal distance. c) Simulated retinal illumination (130°), with a M-2 (c) or M-3 (d) configuration. The encircled area (±35°) is the one chosen for optimization and depicted in more detail in (e). Irradiance scales are graduated in mW.cm^-2 and coordinates in mm.

Table 1. Comparison between three contact lens configurations in Liou-Brennan eye.

Optical system	Ratio on retina	Ratio on iris	Mean ±1 STD ( $\mu W.c{m}^{-2}$ )	$U_1=\frac{E_{avg}}{E_{max}}$	$U_2=\frac{E_{min}}{E_{max}}$
M-2	0.24	0.47	42.5 ± 25.2	0.34	0
M-3	0.23	0.44	40.8 ± 12.4	0.45	0.1
TIR-3 / 1 axis	0.20	0.44	38.3 ± 11.5	0.59	0.2
TIR-3 / 2 axes	0.29	0.41	53.4 ± 14.3	0.51	0.05
TIR-2 / 1 axis	0.20	0.44	39.7 ± 13.9	0.52	0.02

According to Table 1, the achieved uniformity criteria is good but as could be expected, it is difficult to direct all the energy through the iris, and one-quarter of the incoming flux hits the retina while one-half hits the iris.

The remaining flux is either located outside of the lens or lost within the lens due to numerous internal reflections. From Table 1, we notice that adding a TIR guide improves uniformity, and adding a second axis to converge light slightly improves further efficiency and reduces the light amount hitting the iris (to prevent potential tissue damage). This is explained by the fact that the TIR will guide the light coming out of the LED as soon as it exits the LED. In the absence of a guide, part of this light will not be directed effectively towards the retina.

These simulations were carried out with a 3 mm pupil. If the iris size increases to 4 or 5 mm, then more light enters the eye and the efficiency (flux impacting the retina on sources’ flux) increases from 15% to 25% to 36%. Controlling the illumination concerning pupil size is an important factor further discussed in Section 3.4.

In the absence of an eye, i.e. on a curved plane at distance equivalent to that of the retina (Fig. 4(a)-(b)), the mean irradiance is similar but standard deviation is divided by a factor two. Similarly

U1 and U2 increase respectively by 20% and 80%.

2.5. Prototypes manufacturing

We chose to use scleral contact lenses because their large dimension offers a larger volume and because their better stability on the eye is important to guarantee constant retinal lighting.

Lenses were manufactured by the company Laboratoire LCS using diamond turning. The main steps of the encapsulation process are depicted in Fig. 5. The front and back surfaces of the lens are lathed in two different PMMA pucks, with a possible inner surface metallization. Then the electronic circuit and the optics are positioned between the two pucks which are glued together with a biocompatible UV glue (Loctite AA 3301). The assembly is then lathed to obtain the final lens. In the case where the rear of the front surface is metalized, a metallization step is added. A matt polymer stencil is then fixed inside to restrict the metalized surface to ensure that there is no annular continuity in the metal which could cause an induced parasitic current in the metal, impacting the energy used to power the light sources. The surfaces were metalized by evaporating an aluminum thin layer of 200 nm thickness.

Fig. 5.

Manufacturing of the contact lens. a) disk lathing, b) metallization, c) UV curing, d) lens machining, e) view of diamond turning, f) circuit and optics positioning in the puck, g) gluing assembling the pucks. Third and fourth rows: contact lens full assembling process: h) the upper puck after etching, i) after metal deposition, j) optoelectronic circuit to be encapsulated, k) circuit positioned in the metalized puck, l) top view of the final contact lens, m) back view of the final contact lens.

Regarding the TIR configuration, a TIR-1axis (Fig. 3(b)), was manufactured by Optics Concept in Zeonex K22R (Fig. 6). Aluminum was deposited on the curved surface only. The metalized annular reflector was cut to prevent parasitic inductive coupling and heating between the contact lens and eyewear coils.

Fig. 6.

Designed and manufactured TIR guide. A) and b) are different views of the annular TIR guide (1axis). Outer diameter is 6 mm, inner diameter is 3,24 mm and maximum height is 0.5 mm.

2.6. Test bench

To compare the power received at the retina to the power emitted at the contact lens output, this last one was measured with the lens placed directly on the collecting surface of the photodetector (Thorlabs PM100D). The lens was then placed on the bench presented in Fig. 7. A plastic holder allows the lens to be held by its front surface to facilitate measurements. The lens is positioned 15 mm from the glass frame. The photodetector mounted on a rotation platform makes it possible to measure the irradiance produced by the lens at the level of the retina (at 24 mm from the lens) in the sagittal plane. The lens and eyewear can be rotated in the axial plane, allowing full retinal illumination mapping. Measurements in the sagittal plane were made over ±50 degrees with a step of 5° and in the axial plane over ± 180° with a step of 10° for a total of 360 measurements.

Fig. 7.

a) schematic view of the test bench. The scleral lens is held by its front surface. It is positioned 15 mm from the glass frame. Behind the lens. A photoreceptor mounted on a rotation platform allows us to measure the produced irradiance., b) test bench.

3. Results

The experimental results were obtained using the test bench described in the Methods section.

Three configurations were tested. Prototypes M-2 and M-3 were fully encapsulated (M-3 is depicted in Fig. 2(a)). For the TIR-2-1axis configuration, measurements were made with the guide and circuit assembly outside the lens. This limitation is discussed hereafter.

We present first the retinal irradiance map obtained with the configurations M-2/3 and TIR-2. Finally, since the LEDs are positioned back to the retina, we present the reflection efficiency by comparing the power emitted towards the retina to the one sent in front of the lens. These values can be compared with the simulations in section 2.

3.1. Homogeneity of irradiance map

The iso-intensity curves (Fig. 8) have been determined considering the smallest detectable intensity for cones, i.e., here $E_{avg}\ast 0.14$ (where $E_{avg}$ stands for the average energy over the retina in a 35° wide cone). The homogeneity measurements associated with Fig. 8(a)-(b) are presented in Table 2.

Fig. 8.

a,b,c) : heat map (left) and iso-map (right) of the irradiance measured for three configurations: M-2 (a), M-3 (b), and TIR-2 (c). d): examples of backward (left image) and forward (right image) illumination with the M-3 configuration. The lens is positioned at 15 mm from the eyewear. It is powered by inductive coupling and a video of the functioning device is available in Visualization 1 ^{(10.8MB, mp4)} .

Table 2. Irradiance homogeneity for three contact lens configurations: when the reflective element is due to the metallization (M) of part of the interior of the contact lens or the use of a TIR guide.

	Mean ± STD (µW.cm-2)	U1	U2
M-2	43.43 ± 9.51	0.76	0.42
M-3	40.64 ± 5.02	0.82	0.55
TIR-2 / 1 axis	20.33 ± 1.38	0.94	0.71

Experimental results confirm the benefit of using 3 LEDs rather than 2, in terms of optical power and distribution uniformity on the retina. Compared with simulations, results for M-3 agree in terms of (Mean ± STD) but U1/U2 values are significantly higher. This is first explained by the fact that simulations were calculated in a Liou-Brennan eye model and experimental measurements made in air (due to the lack of a proper phantom eye). Also, in practice, the lathed surface is rougher than expected and so more diffusive.

Illumination inhomogeneity in all irradiance maps is partly because biasing with the same signal in the LEDs in parallel does not guarantee that they generate the same light. Deviations may also be due to the dispersion of component electrical and manufacturing parameters. I-V characteristics may slightly differ from one LED to another. Any difference in the forward voltages gives rise to large current differences. Hence, unlike in the case of two antiparallel LEDs, the overall current does not split up into 3 identical parts. In addition, there are also some tolerances on the packaging, e.g., the transparent part of the package is 0.5 µm ± 0.1 thick. Moreover, soldering small surface-mounted components is a difficult process when done manually and mispositioning may have occurred. These points would be easily addressed when getting to more industrial stage.

With regards to results with TIR-2, we have used in this case the optoelectronics circuits with the annular projection optics depicted in Fig. 2(b). The micro-optics is annular with one curvature radius infinite and another designed to maximize reflections towards the retina. The index adaptation between the LED and the annular optics has been obtained using index-matching gel G608N3 from Thorlabs. The homogeneity measurements associated with Fig. 8(c) are presented in Table 2.

Table 2 shows a significantly better uniformity with the annular TIR guide (TIR-2/1axis), whether in terms of low standard deviation or high uniformity values. These results are consistent with the fact that the reflective surface of the annular TIR guide is of better quality and homogeneity than the one directly etched in the contact lens puck. Also, in the TIR configuration, the guidance is optimal (guidance directly from the exit of the LED by an optimized curved surface and not a segmented one as in the M configurations).

A limitation of this comparison is that the TIR-2 device is not fully encapsulated. However, according to our experience, encapsulation generally improves slightly the uniformity and thus does not bias this comparison. For instance, for the M-3 circuit, the mean irradiance ± one standard deviation is 37.18 ± 8.16 µW/cm² before encapsulation and 40.64 ± 5.02 µW/cm² afterward. This is a general trend, observed on all the encapsulated samples we used. We also noticed a slight increase in the efficiency criteria (U1 increases from 0.77 to 0.82 and U2 from 0.34 to 0.55). This is probably due to the fact that these systems were designed to be encapsulated.

3.2. Reflection efficiency with 3 LEDs

The reflection efficiency is presented in Table 3 for the M-3 configuration and three LED powers (corresponding to the minimum, medium, and maximum level provided by adjusting the magnetic field strength delivered by the reader using the NFC Cockpit software). Figure 8(d) gives some views of the backward irradiance projected on a screen at 30 mm from a M-3 contact lens. Visualization 1 ^{(10.8MB, mp4)} shows in real-time the lighting variation remotely controlled by the eyewear.

Table 3. Measured flux for- and backward in the central area, depending on driving power.

Electric power in the primary coil (mW)	96	109	275
Optical power emitted frontward (µW)	18 (10%)	21 (9%)	34 (11%)
Optical power emitted backward (µW)	156 (90%)	208 (91%)	286 (89%)

We observe that the flux lost in the contact lens, or going backward is measured around 10% when it was about 30% in the simulations (Table 1). This is mainly because the simulation did not consider the external part of the contact lens.

The large difference between the supplied power and the optical power at the retina is due to the poor coupling between the two coils. Currently, the form factors vary greatly but a more adapted eyewear coil shape (circular rather than trapezoidal) would allow to reduce the amount of energy needed to activate the contact lens.

3.3. Impact of pupil variation on the optical efficiency:

The optical efficiency depends on the pupil size which varies with age and individuals and is impacted by dilatation or constriction of the pupil. As discussed in section 2, the effective pupil is determined here by the PCB (Fig. 2(b)) and the projection optics. In the considered examples, the projection optics (M and TIR) exceeds by more than 500 µm beyond the PCB pupil, reducing by an additional millimetre the eye pupil (Fig. 2(b),(c)). An alternative consists in reducing the annular section to few degrees as shown (Fig. 9(a)) by making a full 2D TIR guide. This solution presents many advantages, first to confine more light into the guide, while reducing the light hitting the iris, and second to make the device less sensitive to pupil variations (Fig. 9(c)), by extending the guide length, without impinging significantly the central vision (only a few % of the pupil surface). Furthermore, by adding on the TIR prismatic section, a 2-axis optics, as in Fig. 9(a),(b), the light can be steered on a given retina area (e.g., to exclude the fovea). This option will equip the future R-L contact lenses.

Fig. 9.

a) Annular TIR light guide with 4 projection optics before being cut and assembled, b) Detail of the TIR guide with one projection optics (a freeform), c) principle of the extended TIR solution.

3.4. Safety considerations:

A recent study [18] showed that current solutions are capable of damaging the retina and iris by causing unwanted photochemical reactions. Ostrin et al. [8] measured the optical power using an integrating sphere radiometer through a 7-mm diameter aperture, following ANSI Z136.1-2014, sections 3.2.3-3.2.4. and reported for the less powerful device (Future Vision): 0.624 mW.cm^-2 corneal irradiance and 0.08 W.cm^-2 retinal irradiance (Ø 2 mm pupil).

In our case, the situation is slightly different. Due to the source locations, the eye optical system is partially bypassed. With the current device, the emitted light flux of each LED can be electrically modulated from 21.7 µW to 162 µW, at 625 nm. The corresponding corneal and retinal irradiances have been calculated (Table 4), for a 4 mm pupil diameter.

Table 4. -Corneal and retinal irradiances for one single LED in the 3-LED contact lens configuration.

Flux (µW) Min | Avg | Max	$21.7$	$87.1$	$162$
Corneal Irradiance (mW.cm-2)	$1$	$4.03$	$7.05$
Retinal Irradiance (mW.cm-2)	$5.05\ast {10}^{-3}$	$2.03\ast {10}^{-2}$	$3.77\ast {10}^{-2}$

Due to the proximity of the LEDs to the cornea, the cornel irradiance is naturally higher than the retinal one and can become a limiting factor. To guarantee a red-light treatment efficiency, it is necessary to decorrelate the corneal irradiance from the retinal one. This can be done by increasing the number of LEDs (cf. Fig. 4(e)), or by introducing in the light path, a diffuser to homogenize and reshape the light distribution, enhancing at the same time its uniformity. This can be done easily with the TIR element, as shown in Fig. 10. A binary diffractive diffuser is deposited on the TIR surface (Fig. 10(a)), which is designed (Fig. 10(b)) to expand the beam, as illustrated in Fig. 10(c), with a 15° scattering angle in one direction. The result is first to reduce the corneal irradiance by almost a factor 1.5, second to mitigate the discrete source effect, providing a more corona-like lighting distribution. On the pictures, the diffuser is positioned to impact half of the LED beam, resulting in an elliptical beam expansion.

Fig. 10.

a) used of a binary diffractive diffuser on TIR basis, b) bottom view of the diffuser, c) beam anamorphic expansion observed at 2 mm.

In addition, to optimize the treatment effectiveness we need to be able to modulate the lighting power on the retina on a sufficiently large range while limiting the incoming power on the eye. Hence, a key parameter is the iris-to-retina efficiency ratio. In our case, this parameter is directly dependent on the mydriasis as shown (Fig. 11).

Fig. 11.

Iris to retina computed efficiency ratio as a function of the pupil size

The efficiency in Fig. 11 is defined as the ratio of the flux hitting the retina over the flux emitted by the tree sources in Zemax in a Liou-Brennan eye model. These values are important to ensure the device adheres to safety limits and to assess potential flow fluctuations caused by unwanted changes in pupil size, which could affect the effectiveness of the treatment. Hence, it could be interesting to measure in real-time the pupil size and to use this value to remotely modulate the LED power accordingly. This could be done at the eyewear level, using for instance a conventional eye-tracker [19] measuring the pupil size or at the contact lens itself [20] using a photodiode measuring the back propagated light. This would contribute to full control of the retinal lighting condition over time, guaranteeing full treatment effectiveness in real-time.

4. Discussion

In this paper, we demonstrated the principle, fabrication, and tests of a contact lens for controlled retinal lighting in R-L therapy. In particular, we designed, optimized, and compared three optical configurations to be inserted into a contact lens, to create uniform retinal illumination (M-3, TIR-3-1axis, TIR-3-2axes). Two configurations were manufactured and tested (M-3, TIR-3-1axis) and the M-3 configuration was fully encapsulated into a scleral lens.

As demonstrated, the advantage of this device is its ability to illuminate the retina while offering precise control of lighting parameters. Furthermore, it allows illuminating the retina over a large angle (130°), difficult to achieve when the light source is axial and far from the eye while preserving central vision. Among solutions tested, the TIR approach opens a new development field for future promising uses of contact lenses as medical devices in eye care and R-L therapy. The first radiometric measurements are encouraging, but need further investigations, in particular, using phantom eyes as in [21], to guarantee safe conditions of use and the setting of real treatment protocols. Furthermore, this prototype was designed to test the embedding of projection optics and should be improved be make them wearable by humans and then children. We achieved it already with a similar function for animals and adults [22]. We do not see particular obstacles in reshaping the device to make it wearable for children (7-11 years). The circuit curvature could be optimized by using flexible electronics for instance. The bulkiest component is the NXP transceiver that was not encapsulated here. However, we considered commercially available devices and a dedicated integrated circuit would solve this issue.

With regards to lens slippage, the use of a scleral lens, i.e., a lens that extends over the sclera, below the eyelids, ensures that the lens will be very stable on the eye [22]. An initial offset from the pupil due to gravity and eyelid interactions, will need to be anticipated and pre-compensated but even if there is a slight drift in time (e.g., 100 microns shift over an hour [23]) this should have little consequences for the considered application. Also, lens rotation is of no consequence if the source configuration is rotationally symmetric.

In this paper, the light intensity was tuned simply by the field intensity at a fixed frequency of 13.56 MHz. In the future, tuning the ON/OFF properties of the LEDs, by using a custom pulse-width modulated (PWM) signal to drive them, would be useful. We tested this capability using a smart sensor NHS3152 integrated circuit, from NXP Semiconductors, basically an NFC ISO 14443 type A interface tag with an ARM Cortex M0 + microcontroller. We chose this component for its size, low power consumption, ease of use and most of all the possibility to power it by NFC without any additional circuit. A prototype has been tested showing that it is possible to light the LEDs with a GPIO current ranging from 100 to 300 µA. This option will be useful to adapt the treatment to the patient and to balance the sources for better control of the light uniformity on the retina.

Finally, this preliminary study confirms the benefit of TIR optics to control retinal light exposure. More complex beam shaping optics could be designed to provide various projection options (e.g., full retinal lighting, excluding fovea, etc.) and optimize treatments in general, beyond read-light therapy in particular.

As far as clinical utility is concerned, there are a number of points to consider. Our device, in its current state, is not easily wearable and clinical practice shows that managing a scleral lens can be quite complex and impractical for application in young patients (children and adolescents).

Our aim was first to present a proof-of-principle by demonstrating a working prototype, which could easily be improved for wear, as we have already done for another device. Furthermore, pending ethical approval to perform tests on human subjects, and in the absence of an artificial eye reproducing the eye in all its complexity, bench-top simulations and measurements have long been a proven method for designing optical devices.

More fundamentally, it's clear that significant advances are needed before we arrive at a solution that is accepted and prescribed to children or adolescents. It's not certain that the benefits of the device can compensate for the difficulties associated with their maturity and ability to follow care and hygiene instructions, but, in our opinion, one of the best ways to clinically assess the true potential of RLRL is to use a device such as the one presented here.

In conclusion, we believe our solution is a good candidate to provide a safe and efficient implementation of red-light therapy. Current RLRL devices are not servo-controlled by eye movements, which leads to doubts about their real impact and raises safety issues. Our device, when fully compliant with safety and comfort for patients, could be beneficial for the clinical community.

Funding

Association Nationale de la Recherche et de la Technologie 10.13039/501100003032.

Disclosures

The authors declare that there are no conflicts of interest related to this article.

Data availability

The data generated and analyzed in the presented research is presented in the manuscript.