An Adjustable Magnetic Levator Prosthesis for Customizable Eyelid Re-Animation in Severe Blepharoptosis: Design and Proof-of-Concept

Abstract

Purpose: Blepharoptosis is a common oculoplastic condition causing incomplete opening of the upper eyelid. Surgical approaches, the mainstay for correction, often fail to improve blink function. The purpose of this study was to develop a nonsurgical treatment option for severe ptosis that allows blink re-animation.

Methods: Magnetic force required to perform blink re-animation was characterized by evaluation of eye-opening and closing using inter-palpebral fissure (IPF) outcomes with various combinations of eyelid array and box magnets. Optimal size of the spectacle magnet that achieved forces required for optimal blink dynamics was selected using simulation. The adjustable magnetic levator prosthesis (aMLP) included an eyelid array magnet and an adjustable rotating spectacle magnet that allowed change in the magnetic direction, thus changing the net magnetic interactive force between the magnets. The clinical feasibility of aMLP in improving eye opening without limiting eye closing was evaluated in patients with ptosis through a proof-of-concept study using IPF and comfort outcomes.

Results: Optimal eye opening and closing was achieved by a magnet-array combination providing 45 grams of surface force (gF) in the tested ptosis population. The aMLP was able to modulate eye opening and closing with change in rotation of the spectacle magnet in two patients with ptotis. The best fitting of an aMLP improved IPF opening without limiting eye closing and with good comfort reported.

Conclusions: Preliminary results suggest that the an aMLP can correct ptosis without adversely affecting blink function. Further evaluation in a larger patient population is warranted.

Translational Relevance: A nonsurgical, proof of concept, adjustable magnetic treatment option for blink re-animation in patients with severe ptosis is presented.

Introduction

Blepharoptosis (ptosis) is a common ophthalmologic condition defined either as unilateral or bilateral incomplete opening of the upper eyelid (drooping of the eyelid).¹ It occurs due to abnormalities in the function or structure of the levator palpebrae superioris muscle responsible for lifting the eyelid. The etiology for ptosis can be divided into structural, restrictive, and paralytic. Paralysis of the levator muscle may result from disease of the third cranial nerve or dysfunction of the neuromuscular junction.² Ptosis is prevalent in all age groups and requires treatment when it is severe and the visual axis becomes blocked. If untreated, it may lead to obstruction of vision in adults, or maldevelopment (i.e. amblyopia) in children.³ 

Current treatment of severe paralytic ptosis is suboptimal. Most patients are treated with the frontalis sling,⁴ which uses the frontalis muscle to compensate for the loss of levator muscle function. This treatment requires conscious effort by the patient to blink and does not allow normal or complete blinking.⁵ This can lead to substantial risk for chronic exposure of the ocular surface (over-correction), which can be avoided by a more conservative approach, which typically results in insufficient eye opening. 

In most types of ptosis, although opening the eyelid is impaired, the neuromuscular complex for eye closure remains intact.⁴ In these cases, an artificial levator prosthesis may be used to restore normal eyelid motility. To this end, previous studies have shown that re-animation of the levator muscle may be achieved by using electrical stimulation,⁶ muscle transposition,⁷ or an artificial muscle.⁸ However, these are invasive procedures, irreversible, and often lead to suboptimal clinical outcomes. Alternatively, noninvasive approaches, such as taping of the eyelids open or propping the eyelid open with a wire on glasses (ptosis crutch)⁹ can help in the short-term, but may also inhibit the natural blink and lead to corneal damage. Moreover, the ptosis crutch approach requires continuous adjustment to maintain the eyelid elevation,¹⁰ which is cumbersome and can lead to corneal abrasion if not carefully performed. 

More recently, Houston et al. demonstrated successful implementation of a proof of concept, noninvasive, blink re-animation approach, using externally fixed static neodymium magnets to compensate for the loss of levator function.¹¹ The device, referred to as the magnetic levator prosthesis (MLP), consisted of a spectacle frame mounted magnet and an eyelid array magnet embedded in polydimethylsiloxane (PDMS) elastomer which was attached to the ptotic upper eyelid using Tegaderm film. Appropriate fitting for eye opening without adversely affecting closing was obtained by adjusting the spectacle frames’ nose pads, and hence the distance between the two magnets, by adding buffers (physical barriers) to limit the approach of the two magnets or using different size magnets on the frame. Although preliminary results were satisfactory, laborious adjustments were required to fine tune the function of the device and considerable changes were required to fit the device between various patients. Moreover, although the MLP allowed better eyelid closing during volitional blinking than the ptosis crutch, it failed to allow natural spontaneous blinking in most of the participants. 

To address the limitation of the MLP, we aimed to characterize the range of forces that would provide eye opening while not limiting spontaneous eye closing in the ptotic population. Following the characterization of the optimal range of forces for blink re-animation, optimal magnet dimensions to provide the desired forces over a range of distances that are feasible relative to MLP clinical use was determined. A multiphysics model was generated to translate the results to parameters of the adjustable force. To improve the functionality of finer titration of magnetic forces in the selected range of the MLP and allow spontaneous blinking, we developed a new design with an adjustable force dial on the side of the frame magnet casing. This allowed manual change of the magnetic axis of the spectacle magnet, thereby shifting of the direction of the magnetic lines and the interaction with the eyelid array. This arrangement is feasible and eliminates the need for electromagnets or electronics to alter the interactive force between the two magnets. Here, we use this new design to test the hypothesis that angular translation of the magnetic field of the spectacle magnet modulates the force on the eyelid magnet in a manner that allows both volitional and spontaneous blink in a clinically feasible and useful manner. A prototype adjustable MLP (aMLP) was produced with 3D printing, based on the magnet parameters determined in the force experiments and modeling, and tested with the help of two participants with severe paralytic ptosis. If our hypothesis was supported, it would demonstrate proof of concept of our novel angular translation approach while documenting the methods of prototyping the first ever adjustable force version of the MLP. 

Methods

We created a bipartite system similar to that described by Houston et al. 2014 and 2018, consisting of an eyelid array magnet attached to the outer surface of the upper eyelid of the participant with ptosis, and a larger magnet attached to the spectacle near the eyebrow of the participant (spectacle magnet).^11,12 Eyelid array magnet was fabricated with 3 rectangular cube magnets of length 3 mm, width 2 mm, and height 1 mm (Fig. 1A) embedded in PDMS. There were two polarization orientations of the arrays tested (see Fig. 1A): type I eyelid magnets had polarization with north and south poles oriented through the width so that when worn by the participant the poles ran in the same direction as the gravitational plane, and type II eyelid magnets had north and south poles oriented through the height so that when they were worn the poles ran perpendicular to the gravitational plane. The encapsulated magnets, referred to as the “array,” were attached to IV 3000 Flexifit (Smith and Nephew, Watford, England, UK), similar to those described by Houston et al.,^11,12 with the exception that the PDMS array was bonded to the outside of the adhesive, rather than draping the adhesive over the array during application. 

Figure 1.

 

Experimental setup to study the effect of increasing magnetic force on blink dynamics (box experiment). (A) Eyelid array magnets are small (L × W × H: 3 × 2 × 1 mm) rectangular magnets embedded in PDMS that is bonded on IV 3000 skin adhesive film. As shown by the red arrows, type I arrays have the manufacturer-defined magnetization orientation along the height of the magnets, whereas type II have magnetization orientation through the width (SM Magnetics, Pelham AL, USA), dot marker on denoted pole. (B) Boxes had a slot where rectangular static Neodymium grade 42 or 52 magnets (M1 to M9) were inserted to produce magnetic surface force ranging from 5 to 55 gF in 5 to 10 gF increments, verified with strain gauge. Shims were used to make minor adjustments in surface force as needed. All boxes (M1–M9) looked identical and were arbitraily labeled to mask the operator and participant to size and force. The north pole of the box magnet was oriented upward such that the north pole of the array was attracted to the bottom of the box. Black outlined rectangles represent the cross-sectional geometries of the various boxes. (C) To characterize eyelid dynamics in response to the different boxes, the eyelid array magnet was attached to the eyelid and the box magnets were positioned in front of the eyebrow in a randomized order. To attach the array on the participants eyelid, the white backing from the IV 3000 was removed and the array was pressed against the skin with the dots up so that the type I pole was in the gravitational plane and the type II pole was in the saggital plane. Video recordings were post-processed to measure the interpalpebral fissure. (D) Surface pull force for each box. The y-axis represents empirical measurements of surface pull force in grams (gF) exerted on the type I and II arrays, measured with a strain gauge.

To characterize blink dynamics, we aimed to test a range of forces that would vary from having no effect on the eyelid (very weak) to very strong (where the eye could not blink). Non-custom rectangular NdFeB magnets of different sizes in grades N42 or N52 were obtained from SM Magnetics (Pelham, AL) or K&J Magnetic (Pipersville, PA), based on the manufacturer’s pull force specifications. Nine magnets (M1–M9), spanning a range of 5 to 55 gf, with increments of 5 gf between each magnet were selected (Fig. 1B). To create more precise incremental forces without having to produce custom magnets, the magnets were inserted into 3D printed boxes (see Fig. 1A) and “shims” of 1.5 mm thickness were added until the desired surface full force was achieved, as measured with an array attached to a strain gauge. The nine boxes were produced to appear identical, to mask the experimenter and participant to the strength of the magnets to attract the eyelid array magnets. The difference between the boxes was the size of the magnet housed within the box or the distance between the magnet and the testing surface of the box (see Fig. 1B). The boxes were prototyped with standard greyscale resins from Formlabs (Somerville, MA) and Stereolithography printing technology to obtain precise geometry of the 3D-printed boxes. Magnets were inserted into the boxes and sealed to prevent unmasking and the boxes were numbered to facilitate randomization during testing. We heretofore refer to these as “box magnets.” Double masking (participant and experimenters) was also utilized for eyelid magnet polarization direction. The force range and increments were chosen based on unpublished preliminary tests which had the goal of providing a range beyond the physiological limits (beyond the force levels intended for clinical application). 

Box Experiment

The change in magnet force using box magnets (M1–M9) paired with eyelid array magnets (type I and type II) on blink dynamics was tested through a two-visit study design. In each visit, participant's baseline eye opening, and eye closing during volitional and spontaneous blinking, were measured without eyelid array magnets. Following baseline measurements, eyelid array magnets were placed on the ptotic eyelid of the participant after cleaning the eyelid with eyelid wipes. The experimenter held up different box magnets near the brow of the ptotic eye so that the surface of the box and arrays were in contact (see Fig. 1C). Next, participants were asked to blink normally (spontaneously) for approximately 15 seconds. Afterward, they were asked to tightly close and open their eyes widely (volitional blink), and then to flutter their eyes (fast volitional blink). Presentation of box magnets to each participant was counterbalanced using the Latin Square method. Videos recording various blink types were recorded. Blink dynamics (eye opening and eye closing during volitional and spontaneous blinking) with inter-magnet distance (distance between eyelid array magnet and box magnet) were measured with different box magnets and eyelid array magnets combinations (see Figs. 1A, 1B). Inter palpebral fissure measurements were taken from the videos of participants blinking with or without the magnets fitted. We applied mixed effect model analysis to the blink dynamics to evaluate the performance of box magnets for each eyelid array magnet type. 

Participant Recruitment

All experiments were conducted in accordance with the Declaration of Helsinki and approved by the institutional review board of Mass General Brigham (formerly Partners Healthcare). All individuals who participated in this study were recommended by referral of their physician. Participants were screened, enrolled, and met the inclusion criteria of having paralytic ptosis which obscured the visual axis without frontalis drive and the ability to provide informed consent or assent. Informed consent was obtained from all participants following detailed explanation of the nature and consequences of the study. Selection was not affected by age or gender. The authors affirm that human research participants provided written informed consent for the publication of the images. 

COMSOL Modeling

To evaluate the concept of change in magnetic force through angular translation of the magnet and to select the magnet with operating range in the physiological range for clinical use, we developed a COMSOL model to perform electromagnetic simulations of the spectacle magnet and the eyelid array magnets. The model was validated experimentally for two conditions: change in inter-magnet distance between the eyelid array magnet and reference magnet and change in angular orientation of reference magnet with respect to the eyelid array magnet. 

Validation of the model prediction of the change in force exerted on the eyelid array magnet by the reference magnet at various distances was performed using a strain gauge. The reference magnet, a 12.7 mm × 12.7 mm cylindrical diametrically polarized N52 magnet (SM Magnetics, Pelham AL), was inserted in a rotatable housing similar to that in Figure 5A and mounted on a stand. A strain gauge (Grass Inc.) was mounted on a stand across from the reference magnet. A thin string attached the strain gauge to a three-magnet eyelid array (both types were tested). The eyelid array magnet was placed into the field of the reference magnet such that their polarizations were opposite to each other, meaning the north pole of the eyelid array magnets faced the south pole of the reference magnet (opposing poles), but the two were not allowed to touch. The array was suspended in the field so that the string was taut. Next, the stand with the strain gauge was adjusted until the inter-magnet separation was 1 mm, as measured with Vernier calipers. Using the same setup, the stand with the strain gauge was adjusted to vary the inter-magnet separation. Force was measured at a separation of 1 mm, 2.7 mm, 5.7 mm, and 10 mm. Similarly, the same environment with the eyelid array magnet and reference magnet was simulated in the COMSOL model and the inter-magnet distance was varied from 1 to 10 mm, while the force exerted on the surface of the eyelid array magnet was measured. The empirical and simulated measurements were compared for correlation using Pearson's correlation. 

Likewise, validation was also performed on the predicted change in orientation of the reference magnet with respect to the eyelid array magnet. The same reference magnet was simulated along with the eyelid array magnet with inter-magnet separation of 10 mm and the reference magnet orientation of 0 degrees (opposing poles). The reference magnet was rotated in 45 degrees increments from 0 to 180 while the force was measured from the surface of the eyelid array magnet. Empirical force measurements were also recorded with the strain gauge setup replicating the simulated conditions. The empirical and simulated measurements were compared for correlation using Pearson's correlation. 

Upon successful completion of the validation, we evaluated candidate spectacle magnets dimensions for use with the aMLP. We simulated three diametrically magnetized cylindrical magnets of diameters 12.7 mm (12.7 diameter [D]), 9.53 mm (9.53 D), and 6.35 mm (6.35 D) with the same length of 12.7 mm. All magnets were simulated with the type II eyelid array magnet. The simulations involved calculating the force on the surface of the eyelid array magnet with varying the inter-magnet distance from 3 to 22 mm and with orientation of the spectacle magnet from 0 to 180 degrees. 

Adjustable Magnetic Levator Prosthesis

Three prototypes were produced, tested, and refined in this testing process (Supplementary Fig. S2). The final design was produced and used in the proof-of-concept human studies. Solidworks CAD (Waltham, MA) was used to design a spectacle clip-on and magnet casing with compliant dial system that was 3D printed and attached to a frame (most visible in Fig. 5). The magnet case was a separate piece that slid onto a rail on the clip-on. The clip-on, most easily visible in Figure 5, had a base, which was designed to follow the contour of the frame for stability when attached onto the frame using clips and brackets. It consisted of two rails on either side of the frame nose bridge, with triangular ridges, onto which the magnet holder was mounted. The magnet holder was designed to rotate the cylindrical magnet by 360 degrees with 30 degrees steps using a simple rocker-pinion mechanism. It was mounted onto the clip-on base at the rails using the clamp. The clamp had a triangular ridge that meshed with the triangular ridges on the rails to restrict the motion of the magnet holder. The magnet holder had a cavity to house the circular spectacle magnet. The spectacle magnet was glued onto the holder dial that had a gear with its teeth 30 degrees apart from each other. The teeth of the gear meshed with the monolithic stopper in the magnet holder, which facilitated the rotation of the dial by 30 degrees steps. It restricted the rotation when force was not imparted onto the dial. We analyzed the design for stresses and flexibility that the rotation mechanism could withstand with various 3D printable materials using Solidworks (Waltham, MA). Following the analysis, we selected Polyamide 11 (Nylon 11) as the material to 3D print the clip-on, magnet holder, and holder dial using Selective Laser Sintering (SLS) printing technology.¹³ 

The effect of rotating the spectacle frame magnet in combination with eyelid array magnets (type I and type II) on blink dynamics was measured to show a proof of concept of an aMLP. Two participants with severe ptosis (subject S2 and S6) were fitted the eyelid array magnet (type II) and aMLP frame prototype. Video recordings of the participants natural blinking were used to measure eye opening, and the amount of eye closing during volitional and spontaneous blinking at baseline and at various aMLP spectacle magnet rotations (0, 30, 60, 90, and 180 degrees) through interpalpebral fissure measurements. Fifteen seconds of recording from each setting was processed and analyzed (resulting in 250 image stacks). Comfort ratings were also recorded on a Likert-type scale of 1 to 10, with 10 being the most comfortable. 

Data Processing and Statistical Analysis

Participants had interpalpebral fissure measurements during both experiments: box and aMLP experiments. In both experiments, participants were comfortably seated in a chair and videos of the participant's blinking while looking straight ahead were captured using an iPhone X camera placed 50 cm in front of the eyes. Interpalpebral fissure (IPF) height (see Fig. 1C) between the bottom eyelid margin to the top eyelid margin through primary corneal reflection, was measured using ImageJ from the 30 Hertz (Hz) iPhone video recordings with the eye open (eye opening) and at maximum closure during blinks (completeness of blink). IPF measurements were calibrated using the adult population norm 11.67 mm white-to-white scleral distance, also referred to as horizontal visible iris diameter (HVID). The HVID only varies approximately 0.26 mm with sex and race.¹⁴ Eye opening was separated from blinks using an algorithm that detected eye blink when the amount of visible sclera was below a preset whiteness threshold while the eye was open and determined blink rate. An observer validated the output by watching the videos and verifying that the number of visible blinks matched the output of the algorithm. 

All statistical analyses were performed in R and STATA. The outcome measures data were transformed into normalized eye opening, normalized volitional eye closing, and normalized spontaneous eye closing by subtracting the respective measures from mean baseline. The transformed distributions were inspected for normality and sphericity using Shapiro-Wilk test and Mauchly's test of sphericity, respectively. Three linear mixed effect models were fitted to the data, one for each of the outcome measures with force generated by each magnet box M1 to M9 as fixed effects for each eyelid array magnet type, with random intercept of each participant. Post hoc pairwise comparisons (t-tests) between box magnets were then performed, with the Bonferroni method to correct for multiple comparisons. Comparison between array types of box magnets M8 and M9 were done using a linear mixed effect model. Linear mixed effect model with fixed effects of magnets (M8–M9), arrays (type I or type II), the interaction between magnets and arrays, with random intercept and slope for each participant were fitted to the data. Post hoc comparisons of the interaction between magnets and arrays were made using the Bonferroni corrected t-tests. 

To compare various magnet forces (box magnets and eyelid array combinations) on blink dynamics (eye opening and closing) for each participant we computed optimality scores, which represents how well a combination of box magnet and eyelid array can open the upper eyelid while not limiting eye closing in both volitional and spontaneous blinking. Optimality score is computed for each box magnet by taking the difference between the amount of eye opening and volitional or spontaneous eye closing obtained through the box magnet and is normalized by the maximum optimality score achieved for each participant (Equation 1). Optimality scores help to compare the impact of the magnetic force between the box magnet force and the different types of eyelid arrays on blink dynamics for each participant. Optimality scores were computed for each box magnet and eyelid array magnet combinations in which eye opening, and volitional and spontaneous eye closing values were measured. Combinations that did not have at least one of the blink dynamic measurements were excluded from optimality computation.  

\begin{eqnarray} && Optimality\;\left( {of\;a\;box\;magnet} \right) \nonumber \\ && = \frac{{\begin{array}{@{}l@{}} \left( {Eye\;opening - Volitional\;Eye\;Closing} \right) \\ \quad + \left( {Eye\;opening - Spontaneous\;Eye\;Closing} \right)\end{array}}} {{{\rm{max}}\left( {Optimality\;scores\;of\;the\;participant} \right)}}\end{eqnarray}

(1)

 

Results

Participants

Fourteen participants met the inclusion criteria of paralytic ptosis which obscured the visual axis without frontalis drive and were enrolled after providing informed consent for a protocol approved by the institutional review board of Partners Healthcare. The study was conducted in accordance with the tenets of the Declaration of Helsinki. Three participants were excluded from analyses because they exited the study before data could be collected (2 due to early recovery and 1 due to scheduling conflicts and referral for a rigid scleral contact lens). Of the remaining 11 participants, median age was 55 years (interquartile range of 22) and 54% were women. Three had ptosis in the right eye, five in the left eye, and three had bilateral ptosis (see the Table). 

Table.

 

Characteristics of the Participants Investigated

 

Blink Dynamics and Optimality Scores From Box Magnet Experimentation

Linear mixed effect model analysis for normalized eye-opening (IPF) found a significant main effect of box magnet force with both type I (P < 0.001) and type II (P < 0.001) arrays; therefore, as expected, more force led to more eye opening. Bonferroni corrected post hoc t-tests revealed that M9 (55 grams of force [gF]) provided significantly more opening than M1 (approximately 5 gF), M2 (approximately 10 gF), and M3 (approximately 15 gF, P < 0.001, t-test, Fig. 2A), but not M4 to 8 (approximately 20–45 gF). This suggests that the opening effect may saturate at M4 (18 gF) for the type I array. Post hoc tests for type II eyelid array magnets revealed a difference in IPF at M8 (approximately 45 gF), being significantly greater than M1 to M7 (approximately 5-35 gF), all P < 0.001, but not different from M9 (55 gF). This suggests the opening effect may saturate at M8 (45 gF) for the type II array. At 45 gF (M8), the type II eyelid array magnet provided significantly better eye opening than type I (P = 0.011, t-test; Fig. 2B), but at 55 gF (M9), they were equivalent (P = 0.292, t-test; see Fig. 1E). Therefore, results suggest that type II provides maximum opening with lower force. 

Figure 2.

 

Results from characterizing eyelid dynamics with various box magnets and eyelid array type combinations. Bar plots of mean normalized values with standard error bars, where positive values represent increase in the interpalpebral fissure (IPF) height relative to baseline (without magnets) and vice versa. (A) Magnets of any strength increased IPF (all bars >0), and higher force box magnets provided more eye opening. For type I (red bars), the 55 gF magnet (M9) provided significantly more opening than lower force box magnets M1 to M3 (approximately 5–15 gF). For type II (blue bars), M8 (approximately 45 gF) and M9 (approximately 55F) provided significantly more opening than lower force box magnets (M1-M7, 5–35 gF). (B) Type II array achieved significantly more eye opening than type I using M8 (approximately 45 gF), without significant difference using M9 (approximately 55 gF). (C) Spontaneous blink was significantly impeded using M9 (55 gF) for type I (red bars). For type II (blue bars), there was no significant relationship to force, suggesting the threshold was above 55 gF. (D) Spontaneous closing was not significantly different between type I and II arrays, for box magnets M8 (approximately 45 gF) and M9 (approximately 55 gF). (E) None of the differences in volitional closing reached statistical significance; therefore, increasing the force as high as approximately 55 gF (M9) did not impede volitional blink for either array type.

Linear mixed effect model analysis for the normalized spontaneous closing IPF found significant main effect of box magnet force with type I arrays (P = 0.017), but not type II (P = 0.054). The lack of significance for type II suggests the spontaneous blink is not impeded at the range of forces tested (up to 55 gF). For type I, Bonferroni corrected post hoc t-tests revealed significant spontaneous blink impedance with M9 (approximately 55 gF, P < 0.05, t-test; Fig. 2C), with significantly worse blink IPF with M9 (approximately 55 gF) than M8 to M5 (approximately 45-25 gF). Pairwise comparisons of spontaneous closure IPF at M8 (approximately 45 gF) and M9 (approximately 55 gF) did not find significance for type I or type II arrays (Fig. 2F). 

Normalized volitional closing did not vary with increase in magnetic force. Analysis of the normalized volitional closing found no significant main effect for box magnets in type I (P = 0.072) or type II (P = 0.132) eyelid array magnets. Trends from both eyelid array magnet types show that with increased magnet forces volitional eye closing gets worse (Fig. 2E). Comparisons between type I and type II eyelid array magnets of M8 (approximately 45 gF, P = 1.00) and M9 (approximately 55 gF, P = 1.00) also show that they were not significantly different from each other (see Fig. 2F). These results suggest that all combinations of box magnets and eyelid array types allowed a similar amount of volitional closing, which was very similar to the baseline volitional closing. 

Data were normalized for each participant such that the best box magnet–eyelid array magnet type combination = 1. Optimality scores normalized for each participant showed different box magnets and eyelid array magnet type combinations being the best choice for various participants. Averaging the optimality scores across participants showed the combination of M8 with type II eyelid array magnets (45 gF) was the best overall option across participants (Fig. 3). This suggests that M8 (45 gF) with type II array provides the overall best eye opening while not limiting eye closing during the volitional or spontaneous blink. However, it also shows that having customizability of magnetic force can be beneficial, because the best combination varied widely across participants (see Fig. 3). 

Figure 3.

 

Optimality scores for each box magnet- eyelid array magnet combinations for all participants. Scores were normalized for each participant such that a score of 1 (red outline) was assigned to the box magnet eyelid array combination that provided the best eyelid dynamics for opening and closing (spontaneous and volitional). The blank areas in the chart represent missing data. Some participants could not complete all testing. The combination of box magnet M8 (45 gF) with eyelid array magnet type II had the highest frequency of optimality equal to 1 (n = 4). However, a wide variation was documented in the chart, suggesting the need for use of both array types and an adjustable spectacle magnet force from 18 gF to 55 gF (M4 to M9).

COMSOL Modeling

Based on the results of the box experiments, a range of 30 gF (M6) to 55 gF (M9), as measured at the box surface, could optimally fit >90% of the target population sample. This computed to 7 to 15 gF at a clinically feasible inter-magnet separation of 8 mm. We hypothesized that this range of force could be provided via angular translation of a static, diametrically polarized, cylindrical neodymium magnet of a size feasible to wear on a spectacle frame (e.g. < 15 mm × 15 mm). The adjustable force feature was realized using angular translation, that is, the magnet on the spectacles was manually rotated using a dial with predefined angular steps of 30 degrees.¹⁵ Using COMSOL, the design was simulated to verify empirical data using a strain gauge (see the Methods section, magnetic force testing), which also allowed further optimization of the system using COMSOL. 

The distance between the spectacle frame mounted magnet and type II eyelid array was systematically increased and the force acting on the eyelid array was measured. As can be seen in Figure 4A, empirical data confirm simulated results (Pearson's r = 0.995, P < 0.001), and the upper end of the necessary range (15 gF) was reached at a distance as high as 9 mm using a 12.7 × 12.7 mm diametrically polarized N52 cylinder (the reference magnet). 

Figure 4.

 

COMSOL model validation and simulated results for force modulation using angular translation, inter-magnet linear distance translation, and spectacle magnet diameter variation. (A) Line plot showing minimal difference between empirical and COMSOL simulated force measurements. Measurements represent forces of a diametrically magnetized N52 reference magnet (12.7 mm × 12.7 mm cylindrical magnet) on a type II eyelid array. The eyelid array reference magnets were placed at various predefined distances (1 mm–10 mm, x-axis) and force acting on the top surface of the eyelid array magnets was measured three times each with a strain gauge (y-axis). (B) Line plot showing minimal difference between COMSOL simulated and empirical force measurements of force variation (y-axis) with spectacle magnet angular translation (x-axis), validating the COMSOL model. Measurements represent the forces of a diametrically magnetized reference magnet with the type II eyelid array magnet separated at a distance of 10 mm. At 0 degrees, both magnet poles are aligned in the same direction (reference magnet south pole is aligned with eyelid array magnets north pole). The reference magnet was rotated by 45 degrees steps from 0 to 180 degrees and force was measured 3 times each with a strain gauge. (C) COMSOL model used for simulation of the environment of the spectacle frame magnet or reference magnet and the eyelid magnet array. The eyelid array was simulated as encapsulated in the PDMS substrate. The space between the eyelid array magnet and the spectacle frame magnet is simulated as air. (D) Line plots reporting the results of simulations. Three magnet sizes were simulated while varying the angle of the spectacle magnet (y-axis) and inter-magnet distance (x-axis). Rotation angles were simulated at 45 degrees steps. The shaded region represents the range of forces that can be exerted by the spectacle magnet on the type II eyelid array magnet due to a change in inter-magnet distance and angular rotation. Negative force values represent repulsion and positive force values represent attraction.

As can be seen in Figure 4B, empirical and simulated measurements were essentially identical for various angles (Pearson's r = 0.998, P < 0.001). The force exerted on the type II eyelid array magnet varied from −8 to +10 gF. As shown in R software version 2.1.1, the necessary upper range of 15 gF should be met by reducing the magnet separation from 10 to 9 mm, which could be easily accomplished in the clinic via nose pad adjustment. To confirm this and in the interest of making a device as small and lightweight as possible, we used the validated COMSOL model to vary size, distance, and angular orientation. 

Three N52 diametrically magnetized cylindrical magnets with three different diameters of 12.7 mm (12.7 D), 9.53 mm (9.53 D), and 6.35 mm (6.35 D) but with the same length of 12.7 mm were simulated. Based on the results from the COMSOL simulations, the peak attraction force (positive values of force) provided by the magnets was proportional to the size of the magnet. The 12.7 D magnet provided almost twice and three times the peak force compared to the 9.53 D magnet and the 6.35 D magnets. The range of force modulation possible with angular translation expanded dramatically when the magnets were closer together. For example, decreasing separation only from 8 to 7 mm increased the range 4 gF in the positive direction. The separation of the force lines can provide some guidance for the predicted fitting of the device. Namely, to get some potential effect of angular translation with the 9.5 D mm magnet, results suggest that the inter-magnet distance should be no more than 12 mm. Beyond this point there is very little (less than 1 gF) separation in the force lines (see Fig. 4D, plot lines for orientations 0 and 45 degrees). At a clinically feasible inter-magnet distance of 8 mm, the 12.7 D cylinder magnet provided 20 gF with orientation of 0 degrees and −18 gF for 90 degrees orientation, 9.53 D cylinder magnet provided +12 gF with orientation 0 degrees and −10 gF with orientation 90 degrees, and 6.35 D cylinder magnet provided 5 gF with 0 degrees and −5 gF with 90 degrees orientation (see Fig. 4D). 

Adjustable Magnetic Levator Prosthesis for Customizable Blink Re-Animation

Based on the simulation results, 9.5 × 12.7 mm (diameter × length) was selected for initial prototyping (comparison of force exerted on the eyelid array magnet by various magnets are shown in Supplementary Fig. S1). A prototype 3D printed frame clip-on was produced (see Methods, Adjustable Magnetic Levator Prosthesis and Fig. 5A). This clip-on had a magnet module attachment, which housed the spectacle mounted magnet and facilitated change in orientation of the spectacle mounted magnet through a gear and monolithic stopper at steps of 30 degrees. The range of force produced by the simulated magnet system ranged from 40 gF at 3.6 mm separation and 0 degrees angular orientation to −28 gF at 0 mm separation and 90 degrees of angular orientation. At a clinically feasible inter-magnetic distance of 8 mm, angular translation of the spectacle magnet could vary the force on the array by 22 gF (−10 gF to +12 gF). The prototype aMLP system, which had a magnet case plus buffer material of 3 mm thick, had a maximum surface force of 40 gF with type II array at orientation of 0 degrees. The box experiment suggested that the spontaneous blink would not begin to be impeded until forces reached 55 gF (see Fig. 2). 

Figure 5.

 

Conceptual and technical design of a prototype adjustable magnetic levator prosthesis and results of proof-of-concept testing for blink re-animation in two participants with severe ptosis. Adjustable magnetic levator prosthesis prototype generated for proof-of-concept testing in two participants with severe ptosis (S2 and S6). The prototype consisted of a 3D printed holder that attached to the spectacle frame (A). The spectacle frame magnet was mounted onto the clip and was manually adjustable laterally on a rail (outlined with dashed orange line), by the clinical study staff or participant, in order to position it above the eyelid array of the participant. The cylindrical 9.53 × 12.7 mm N52 spectacle frame magnet was housed inside the magnet module (A, inset) and was designed to allow rotation (angular translation) in 30-degree steps to change the force exerted on the eyelid array magnet. (B) Line plots represent various blink dynamic parameters and comfort scores given by the participants for various orientation settings of the prototype aMLP. Participants and the examiner together chose the best setting. The mechanical effect of angular translation on the array and eyelid can be seen in the participant photographs on the right.

Next, the prototype was tested with two participants (S2 and S6, see the Table) with ptosis who also participated in the box experiment, in a protocol approved by the institutional review board. The participants wore the aMLP and were monitored with video recording for spontaneous and volitional blinks. We measured the interpalpebral fissure and comfort ratings on a Likert-type scale of 1 to 10, with 10 being the most comfortable. Both the participant and the examiner fitting the aMLP were asked as to provide efficacy ratings on a Likert-type scale of 1 to 10, with 10 being the most effective, for overall functionality provided by each orientation setting of the spectacle frame mounted magnet of the aMLP. Here, we report the comfort and efficacy ratings for the best orientation setting chosen by the participants. 

For the first participant (S2, top panel of Fig. 5B), who wore the aMLP on the right eye and chose one of the lowest magnetic force settings (30 degrees) as the best setting, whereas the clinical study staff chose the 90 degrees orientation as the best setting. Although both settings provided similar levels of eye closing (spontaneous and volitional), the 90 degrees setting provided better eye opening than the 30 degrees setting. The participant rated the 30 degrees setting higher (8/10) for comfort compared to the 90 degrees setting (6/10). The inter-magnet separation with the best clinical staff fitting was 5 mm when the eye was closed and 3 mm when the eye was fully open, producing a COMSOL predicted force range of −6 gf to −20 gf. However, despite the negative values suggesting repulsion, we noted that the array rolled over, reoriented the poles to produce attraction (see photographs in Fig. 5B). The eye opening improved from 7.62 mm at baseline to 8.65 mm with the aMLP. The minimum spontaneous blink was 0.15 mm with aMLP compared to the 0 mm at baseline. 

The second participant (S6, see the bottom panel of Fig. 5B) was successfully fitted with a type II eyelid array magnet by the clinical study staff with the participant's input. They chose the 90 degrees orientation as the setting that provided adequate opening with a natural appearing and comfortable blink (8/10 rating; see Fig. 2F). The inter-magnet separation with this best fit was 6 mm when the eye was closed and 2 mm when the eye was fully open, producing a COMSOL predicted force range of −8 gf to −28 gF. Despite the negative values suggesting repulsion, we noted and it can be seen in the photographs in Figure 5B that the array actually rolled over, reorienting the poles to produce attraction. The eye opening improved from 4 mm at baseline, to 5.9 mm with the aMLP. The minimum IPF during spontaneous blink was nearly complete with the aMLP at 0.79 mm, possibly slightly better than the 1.6 mm measured at baseline. The participant gave this rotation position a comfort rating of 8 of 10 and efficacy rating of 7.5 of 10. 

Discussion

We aimed to develop a magnetic levator prosthesis with adjustable force to address the need for customization in the magnetic correction of severe paralytic blepharoptosis. The study results support the hypothesis that our novel approach of angular translation is clinically feasible and produces an adequate range of magnetic forces for real time titration of the force that controls the lid's position. Clinical empirical data were validated mathematically using COMSOL multi-physics analysis, which allowed further optimization of the parameters of the device in a laboratory setting. Proof of concept was subsequently confirmed by fabricating a prototype device that was used in two participants with ptosis and was able to change the interpalpebral fissure height and the amount of closing during spontaneous blinking by changing the spectacle magnet angular position (see Fig. 5). To our knowledge, this is the first demonstration of successful blink re-animation using a proof of concept manually adjustable MLP device. This adjustable force feature should help address issues of incomplete spontaneous blinking, as identified in our prior work¹² and provide improved fitting and ease of force titration. This addresses a major limitation of using fixed magnets for blink re-animation which are known to partially impede spontaneous lid closing. 

The masked box experiment with the box magnet is the first to systematically study the effect of magnet force on eyelid dynamics in severe ptosis during eyelid re-amination. Our results demonstrate saturation of eyelid opening around 55 gF for type I arrays and 45 gF for type II arrays (see Fig. 2A) without impeding volitional blink. Spontaneous blink was impeded at 45 gF for type I arrays and >55 gF for type II arrays. This apparent difference between arrays may be related to higher force at the magnet interface due to greater thickness of the type I array magnets in the polarized direction (2 mm in type I vs. 1 mm in type II). The force is also distributed over a wider area (6 mm² in type II vs. 3 mm² in type I). Despite this apparent difference, analysis did not find significance, that is, differences between eye closing during spontaneous blinking was not different for type I compared to the type II eyelid array magnet, in this small sample. Analysis of the optimal settings showed that both type I and type II arrays should be available for fitting in future studies. 

Although the aMLP provided good blink dynamics in these two cases, there are likely ways to make further improvements. We observed that in these two cases with unilateral ptosis, where the spectacle frame magnet was fitted only on one eye, the lightweight frame that we selected would sag onto one side due to the weight of the spectacle frame magnet. Substantial force changes occur with frame movements as small as 1 mm, therefore a more stable frame is paramount. While adding counterweights to the other side of the spectacle frame might be helpful to balance the weight distribution for the unilateral ptosis cases, distributing counterweight along a traditional frame might be challenging and would make the frame bulky and prone to sagging on the face. Using custom designed 3D printed spectacle frames, specifically designed for the individual user that provide better fit and allow uniform distribution of counterweight, as reported previously in a recent phase I study,¹⁶ would likely be a better alternative to using the frames used in this study. 

An interesting observation in this proof-of-concept study was the dynamic rotational behavior of the eyelid array when the spectacle frame magnet was set at 90 degrees and 180 degrees rotation. From COMSOL modeling, we expected that these orientations of the spectacle frame magnet would repel the eyelid array magnet as both magnets were fixed in the simulated COMSOL environment. However, in actual testing with the two cases reported, the eyelid array magnet rolled on the participant's coronal axis, folding into the eyelid skin, effectively flipping the orientation of the magnetic poles so they aligned for attraction (0 degrees rotation). The eyelid skin thickness likely affected the amount of rolling. Further investigation is necessary to determine inter-subject variability in this rotational response and how this characteristic may be harnessed to improve performance and cosmesis. For example, rolling of the array had the effect of hiding it within the lid skin, and this is cosmetically preferable. In addition, rolling of the skin around the array appears to improve lid response to magnetic pull, and acts similarly to blepharoplasty, where the lid is shortened to correct ptosis. One possible method to control the rolling could be to increase or decrease the amount of PDMS flange on the surface of the IV 3000. To this end, a COMSOL model should be developed to account for changes in orientation with resistance factor, such as degree of skin laxity and thickness. This will accelerate prototyping, reduce cost, and remove the burden of testing with participants. 

Our study has some limitations. In the box experiment, boxes were held in place by the experimenter. Although this was a rapid, and therefore, clinically feasible way to test the range of magnet strengths, it may have allowed the experimenter to realize the size of the magnet in the box or help the patient blink by unintentionally moving the box downward. Future studies should use an interchangeable holder on a sturdy spectacle frame to eliminate this. Precise positioning of the box using landmarks, such as the eyebrow, should be used so that placement is consistent between and within subjects. Characterization of spontaneous versus volitional blinks was a potential problem in the study and is currently our best explanation for the bimodal distribution seen in the box experiment data. We used verbal instruction (blink rapidly = spontaneous, and close tightly and open as volitional). In fact, the blinks classified as spontaneous may not represent true spontaneous blinks. If this is the case, our results may underestimate the force where true spontaneous blink would be affected. Video recording for longer durations without specific instruction to blink may have allowed us to capture true spontaneous blinks, and, if so, the results could be different. We only chose box magnets in the range of 5 to 55 gF due to the lack of availability of off-the-shelf rectangular magnets. To know the exact upper limit for the optimal range of magnetic force, higher magnetic force testing would be required. With higher magnetic forces we expect that eye opening would reach a ceiling while it would eventually completely prevent eye closing (volitional and spontaneous). This would be painful, however, and may not be ethical (or necessary). Despite these limitations, the results provide an important proof of concept on the relevance of angular translation of the spectacle magnet to provide the opportunity for better fitting and ease of customization of the aMLP for blink re-animation in participants with ptosis. 

The aMLP is currently a working prototype, which needs to be further evaluated with a larger patient population to determine the efficacy. As such, it is not available as an off-the-shelf product and can only be procured from the clinicians involved in this study. The aMLP must be used under the supervision of a clinician. Furthermore, as a research device, it has not been extensively validated to be used outside the research settings. However, the authors aim to further evaluate the aMLP in future studies as a take-home device. As such, there need to be guidelines for the usage of the aMLP. The functionality of the aMLP can be interfered by other strong magnetic forces, which can be caused by devices that provide strong magnetic fields, such as magnetic resonance imaging (MRI), metal detectors, electric motors, metallic implants and prostheses closer to the eyes, and any other ferromagnetic substances. Hence, we recommend the users to steer away from such devices. Moreover, as the aMLP have magnetic substances and rest on the head, any activities that would involve sudden movement should not be performed when using the aMLP. 

Conclusions

The aMLP prototype developed addresses the need for customization in correction of severe paralytic blepharoptosis by providing eye opening while not limiting eye closing. The study results support the hypothesis that the novel approach of angular translation produces adequate range of forces for custom titration of force at clinically feasible magnet sizes and distances. The proof of concept was confirmed with a prototype device in vivo with two participants with ptosis, demonstrating changes in interpalpebral fissure when changing the spectacle magnet angular position. This adjustable force feature addresses the limitation of using a fixed magnet on the MLP, which partially impeded effective spontaneous lid closing. These results warrant further evaluation in a larger patient population to determine the efficacy of the aMLP in restoring eyelid opening with maintenance of the blink. 

Acknowledgments

The authors thank Renita Sebastin, OD, for participating in data collection of the box experiment and aMLP experiment. 

Supported in part by research project grant (National Institute of Health [NIH], R01EY029437, Department of Defense (W81XWH2210774 and W81XWH2010916), and the research project grant (NIH, R01EY013124-21). 

Disclosure: N.M. Kurukuti, None; M. Nadeau, None; E.I. Paschalis and K.E. Houston are named on a US patent application assigned to Schepens Eye Research Institute for the technology that is the topic of this research. The institution has reviewed this potential conflict and the authors have followed any requirements for management. The IP is not currently licensed and the device is not commercially available. E.I. Paschalis is also a paid consultant for Strategic Intelligence Inc. but is not a conflict to this work