Optical and Visual Impact of Tear Break-up in Human Eyes

Abstract

purpose. The purpose of this study was to examine the optical and visual impact of tear break-up.

methods. Optical quality of the eye was assessed during periods of nonblinking by quantifying vessel contrast in the fundus image and by monitoring the psychophysical contrast sensitivity and the spatial distribution of tear thickness changes by retroillumination. All measures were obtained from three eyes either with or without a soft contact lens.

results. A noticeable decrease in retinal vessel contrast and contrast sensitivity were observed soon after a blink. Both of these measures of optical quality of the eye showed a similar pattern of image degradation both with and without a soft contact lens. Although trial-to-trial variability was considerable, sample means show that image contrast in the low spatial frequency range can drop to between 20% and 40% of initial values after 60 seconds of nonblinking. Retroillumination of the tear film showed local intensity fluctuations that progressively spread across the pupil with increasing time after the blink.

conclusions. Optical aberrations created by tear break-up contribute to the decline in image quality observed objectively and psychophysically. The decline in image quality that accompanies tear break-up may be a direct cause of the blurry vision complaints commonly encountered in dry-eye patients.

The front surface of the precorneal tear film is the most anterior optical surface of the eye. The large refractive index step from air to the tears provides this surface with the greatest optical power of any ocular surface. ¹ If the thickness of the tear film is uniform, the cornea/tear combination will have almost exactly the same power as the cornea alone, ¹ which leads to the common belief that the human precorneal tear film has little optical impact. ² However, this conclusion is true only if the tear film remains uniformly thick. 

Many clinical studies have shown that, during periods between blinks, the tear film does not remain uniform on the surface of the eye. ^{3 4 5 6} Instead it appears to locally disrupt, a phenomenon that is clinically termed tear film break-up. Any local changes in tear film thickness will result in an irregular air/tear interface, thus introducing aberrations into the eye’s optical system. Complete breaks in the tear film can also expose the irregular corneal epithelial surface, ² which may increase optical scatter. Aberrations and scatter produced in this manner have the potential to degrade retinal image quality. ⁷ Therefore, maintenance of an intact tear film may be essential for achieving high-quality retinal images. 

In spite of the widespread acceptance that tear break-up reflects a local change in tear film thickness, ⁸ little is known about its effect on retinal image quality. Nevertheless, the available evidence supports the hypothesis that tear break-up degrades retinal image quality. Using a double-pass optical method, Albarran et al. ¹ measured the eye’s point spread function before and after tear break-up in seven eyes. They observed a significant reduction in image quality with and without soft contact lenses, but the reduction was greater when soft contact lenses were worn. Timberlake et al. ⁹ measured visual acuity for low-contrast letters during periods of nonblinking and found a significant loss in acuity with soft contact lens wearing eyes, but no loss of acuity with hard lenses or for exposed corneas. More recently, Thibos and coworkers ^{10 11} have shown that optical aberrations and scatter increase after prolonged periods of nonblinking. Other investigators ¹² attributed increased optical scatter observed with soft contact lenses to poor tear film quality. 

Several clinical studies also support the hypothesis that tear film disruption leads to reduced retinal image quality. Reiger ¹³ found that threshold perimetry improved in dry eye patients after instillation of artificial tears, suggesting that tear supplements stabilized the tear film and improved the optical quality of the eye. Dry eye patients, who typically demonstrate faster tear break-up, ^{14 15 16} showed reduced visual acuity after holding the eye open for 10 seconds. ¹⁷ Lee and Tseng ¹⁸ found that 8% of dry eye patients reported “fluctuating blurry vision” that improved with blinking. In patients with primary Sjögren syndrome, other researchers have reported that 42% to 80% experience “disturbed vision.” ^{16 19} Twenty-two percent of dry eye patients in Japan reported having “blurry vision.” ²⁰ In a recent survey of contact lens wearers, Begley et al. ²¹ found that changeable or blurry vision was a common symptom, with the majority of respondents ranking it as an “infrequent,” but “noticeable” occurrence. 

To quantify the optical and visual impact of tear film disruption, we have devised a novel, single-pass objective method for quantifying changes in optical quality of the eye. These measurements were compared with observation of the tear film during retroillumination (RI) and a measurement of contrast sensitivity to evaluate the visual impact of tear film break-up. All three methods were applied continuously during extended periods of nonblinking both with and without soft contact lenses on the eye. 

Methods

Three subjects, one woman and two men, aged 42 years, participated in this study. The study followed the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board at Indiana University. Informed consent was obtained from all subjects. During each experiment, the left eye of each subject was cyclopleged (1% cyclopentolate) to prevent accommodation during testing and to allow retroillumination of a large area of the corneal surface and tear film. Ocular health was in the normal range for all subjects. Fluorescein tear break-up time ranged from 8 to 18 seconds among subjects. 

The tear film was monitored under two experimental conditions: with a soft contact lens (SCL) and without a soft contact lens (NSCL) on the eye. For SCL trials, subjects were fitted with a CIBA Focus Monthly (CIBA Vision Corporation, Atlanta, GA) contact lens of the correct prescription. During each trial, subjects were asked to hold their eyes wide open as long as possible. In the NSCL case, anesthesia was used before testing (1 drop, 0.5% proparacaine) to minimize discomfort and prevent reflex tearing and blinking. SCL testing required no anesthesia because subjects felt little or no discomfort during prolonged periods of nonblinking, presumably because the soft contact lens protected the cornea from drying. 

Continuous monitoring of the effects of tear break-up was achieved using three independent methods. ¹ Objective image quality was determined by measuring retinal vessel contrast (RVC) in fundus images. ² Subjective image quality was evaluated psychophysically using contrast sensitivity (CS). ³ The spatial distributions of tear film thickness changes were monitored using RI of the tear film. Although simultaneous acquisition of these three measures would have been ideal, this was not possible with our experimental apparatus. Therefore, each test was performed independently in a random order. 

To minimize cumulative changes to the cornea and/or tear film from successive periods of nonblinking, the three experiments were performed either on separate days or after a prolonged recovery period. Repeat testing after approximately 20 minutes gave no indication of a cumulative effect on either initial image quality or the rate of decline in image quality during periods of nonblinking. All tests were run in the temperature and humidity controlled Borish Center for Ophthalmic Research at Indiana University. 

Objective Image Quality

RVC was derived from fundus images acquired using a video biomicroscope and a clear +90 diopter (D) fundus lens. The magnification of the biomicroscope was set at 20× for all fundus image acquisitions. The +90 D lens was aligned and stabilized using a standard lens holder. The retinal area imaged by the biomicroscope was approximately 15 × 20° centered on the vessel-rich area surrounding and within the optic disc. Illumination was adjusted to provide a balance between brightness and contrast, while ensuring subjective comfort to prevent reflexive tearing. Care was taken to stabilize the subject’s head using a standard biomicroscope head and chin rest. Before each trial, the subject was instructed to blink three to four times. Subjects fixated using the nontested eye on a distant fixation target and suspended blinking as long as possible. 

Fundus images were captured with a high resolution CCD video camera (Sony DXC-107AP; Tokyo, Japan) and recorded at 30 frames/sec (standard NTSC signal) on an S-VHS video recorder (Mitsubishi HS-U69; Mitsubishi Electronics America, Cypress, CA) with a video image resolution of 720 × 486 pixels. Fundus video images were digitized using Avid VideoShop 3.0.2 on an Apple 7500/100 PowerMac (Cupertino, CA) into sequential 640 × 480 pixel arrays, and one frame/sec was selected for analysis. Using a public domain image processing program (NIH-Image version 1.60, available through: http://rsb.info.nih.gov/nih-image/), the sequence of images from each experimental trial was spatially aligned (to correct for eye movements), and intensity profiles were generated (Fig. 1) for a selected high contrast and medium-size vessel (diameter, 18 arc minutes or 10 pixels). Intensity profiles across the selected vessel were averaged over a 40-pixel length along the vessel. As image quality deteriorates, the image of the vessel is spread out and image contrast is reduced. To quantify the image quality changes that occur during periods of nonblinking, we calculated image contrast (ΔL/L) from each intensity profile (Fig. 1) . We refer to this metric as RVC. 

Subjective Image Quality

Psychophysical CS data were collected using a Pelli-Robson chart ²² at a test distance of 2 m. This distance was selected to create a letter stroke width (17 arc minutes) that approximated the average vessel diameter (18 arc minutes) used in the fundus video image analysis described above. Each subject was provided with the appropriate over-refraction for the 2-m viewing distance. The nontested eye was covered with a contact eye patch. Before each timed trial, the lowest contrast for which all three letters within a group were recognizable was identified. Subjects were then instructed to blink three to four times and to hold their eyelids wide open as long as possible. During this period of nonblinking, subjects fixated on the letter group with the minimum visible contrast, reported loss of readability, and then fixated on the letter group with the next higher contrast. Times at which each contrast level became invisible were recorded. Each trial terminated upon the subject’s first blink. 

Spatial Distribution of Tear Film Disruption

The technique of RI of the tear film was used to monitor the optical aberrations produced by tear film break-up. ²³ In this experiment, the biomicroscope illumination was positioned to produce maximal RI of the tear film and corneal surface within the dilated pupil, without stimulating reflexive tearing. The subject’s head was positioned and stabilized using a standard biomicroscope headrest. The subject was directed to blink three to four times, fixate with the nontested eye on a distant light source, and hold their eyelids wide open as long as possible. During this period of nonblinking, images of the retroilluminated cornea were captured at video rates on high-definition videotape for later display and analysis. 

Results

Objective Image Quality

There was a decrement in RVC immediately or very soon after the suspension of blinking during all SCL and NSCL trials. A sample data series from a SCL eye is shown in Figure 2 . This figure shows the fundus image at t = 0, 30, 50, and 70 seconds after a blink. Accompanying each image is the intensity profile across the selected vessel. The decline in image quality can be readily seen in the fundus images and the accompanying intensity profiles. In this example, RVC is high (contrast, 30%) at t = 0, but after 70 seconds of nonblinking it had declined to approximately 3%. The overall decline in RVC showed a similar pattern from trial to trial, but the absolute amount of contrast reduction often varied significantly from trial to trial (typically n = 4 trials per eye), as shown for one eye in Figure 3A and 3B . 

Although the decline in image quality varied from trial to trial, the average data from each of our three subjects were quite similar. After 60 seconds, RVC declined to approximately 40% of its initial level in all three subjects during NSCL (Fig. 4A ) and SCL (Fig. 4B) trials. One subject repeatedly initiated a blink after 30 seconds without a CL, and thus data were not collected beyond this point. Averaged across subjects, the SCL and NSCL data are quite similar (see Fig. 6 ). These data, therefore, confirm the findings of Albarran and coworkers, ¹ who found a similar drop in image quality with and without a soft contact lens, but the decline was slightly greater with a contact lens. 

Subjective Image Quality

Summary CS data are plotted for all three subjects in Figures 5A and 5B . As with RVC data, the decrease in CS began almost immediately after a blink. A consistent decline in CS followed, reaching levels between 20% and 40% of the original CS after 60 seconds. 

To compare the impact of prolonged periods of nonblinking on RVC and CS, we have replotted RVC and CS data averaged across all three subjects in Figure 6 . Three of the functions were remarkably similar. Only the SCL CS data showed a more rapid decline and overall lower final value than the other three. 

Spatial Distribution of Tear Film Disruption

Although the RVC and CS tests both quantify the effects of tear film changes on image quality of the eye, neither test provides insight into the specific optical changes that occur during periods of nonblinking. Either aberrations or scatter or some combination of the two would all reduce contrast in the retinal image. Therefore, we recorded the optical changes in the tear film during prolonged periods of nonblinking using an RI technique (Fig. 7) . This method shows the spatial distributions of local optical discontinuities that are generated in the tear film during prolonged periods of nonblinking. 

The optics of RI are similar to those used to measure the aberration function of the eye. ^{11 24} In both techniques, the retina is illuminated by a narrow slit or spot of light, and this retinal image acted as a secondary source that uniformly irradiates the pupil plane with reflected light. If the exiting beam, which is approximately collimated in an emmetropic eye, passes through an optically smooth cornea/tear film, the pupil will appear uniformly illuminated (Figs. 7A 7B ; t = 0 seconds). Optical discontinuities resulting from local changes in tear film thickness will act to deviate rays exiting the eye at these points. Local ray bundles can be deviated toward or away from the entrance pupil of the slit lamp biomicroscope and therefore result in local patches brighter or darker than average in the RI image (Figs. 7A 7B ; t > 0 seconds). In this way, local variations in tear film thickness produce dark or light zones in the image of the pupil. We have shown elsewhere that these local perturbations in the RI image are spatially correlated with tear film break-up as measured by the standard fluorescein technique. ²³  

In Figure 7 , the image of the tear film is quite uniform immediately after a blink (t = 0), but it gradually develops local intensity fluctuations that progressively spread across the pupil with increasing time after the blink. In the two examples shown in Figure 7 , we can see that after 25 (Fig. 7B) and 40 (Fig. 7A) seconds, a large proportion of the pupil was covered by these local optical disturbances. The local disturbances were not distributed uniformly across the pupil. In some eyes we see tear film disruptions in the superior pupil, in others (Fig. 7B) the disruptions begin and remain predominantly in the inferior pupil, whereas in other eyes we observe tear film disruptions in both the superior and inferior pupil (e.g., Fig. 7A ). The specific patterns of these optical disturbances varied from eye to eye and from session to session. 

During the RVC and RI trials, the nontested eyes (all wearing a soft contact lens) were open and fixating a small point source of light, which provided an opportunity to record anecdotal observations. All subjects described the fixation point source as gradually increasing in size with a chromatic appearance around its edges. 

Discussion

We have used three experimental measures (RVC, CS, RI) and one anecdotal observation (subjective point spread function), all of which support the hypothesis that tear break-up introduces significant optical changes in the eye that can result in highly degraded retinal image quality and reduced spatial vision. These data confirm and extend the results of Albarran ¹ and may explain the fluctuating blurry vision reported by groups of clinical patients with dry eye. ^{16 18 19 20}  

It is well known that the cornea has a microscopically irregular surface that is neutralized or “optically smoothed” by the tears. ² Our results add further weight to this conclusion by demonstrating the loss of image quality that results when the tear film is disrupted. However, it is uncertain from our results whether the decrease in optical performance during a period of nonblinking is due to exposure of the underlying irregular corneal surface, nonuniformity of the refractive index of the tears, the irregular surface of the tear film during disruption, or a combination these factors. It is also possible that tear break-up leads to dehydration of the cornea, which in turn may produce structural and thus optical changes. Our results are quantitatively very different from those reported by Timberlake et al., ⁹ who found no measurable decline in low contrast acuity after several minutes of nonblinking when there was not a SCL covering the eye; however, our results do show the same trend as Timberlake in that psychophysical CS was affected more by prolonged periods of nonblinking with a SCL than without (Fig. 6) . This suggests that differences between studies may be due to inter-subject variability in tear break-up. 

In the case of contact lens wear, it has been suggested that psychophysical losses of low-contrast acuity ⁹ or light scattering ¹² associated with periods of nonblinking in eyes wearing SCL (Fig. 5) are due to changes in lens parameters (e.g., curvature, refractive index, transparency) secondary to lens dehydration. However, the irregular surface of the tear film during disruption must also be a contributing factor to reduced image quality in eyes wearing soft contact lenses. In either case, Begley et al. ²¹ have shown that blurry or changeable vision is a common symptom among contact lens wearers. 

Measures of image quality such as those used in this study (RVC, CS, subjective point spread function) cannot easily identify the optical cause of the image degradation accompanying tear break-up. In the introduction, we argue that tear break-up could produce either scatter, aberrations, or both, which may be responsible for the reductions in image quality seen in this and previous experiments. ^{1 9 12} Significant local aberrations are probable at the edge of a break in the tear film where surface slope is likely to be very different from the underlying cornea and surrounding intact tear film. Therefore, the question arises as to whether aberrations associated with tear break-up could be responsible for reducing image contrast at the spatial scale monitored in our experiment (vessels and character stroke widths of approximately 1/4 to 1/3 of a degree). Experiments have shown that even the relatively small-amplitude, higher-order, irregular aberrations present in normal eyes significantly attenuate image contrast at spatial frequencies below 15 cycles per degree. ²⁵ Therefore, it is likely that the increased aberrations and the optical scatter observed after tear break-up will further reduce image contrast at these low frequencies. ^{10 11}  

RI of the tear film showed that disruption of the optical surface that began very soon after blinking was suspended in some eyes (Fig. 7) . This was consistent with the observed decrements in RVC and CS (Figs. 4 5 6) . Initial foci of disruption developed into rod-shaped or branching rivulets or spots rapidly spreading across the RI image. The pupil center was not always involved in the initial disruption, and therefore, initial tear break-up in noncyclopleged eyes would not be expected to have any immediate optical impact on foveal image quality. However, we found that tear disruption eventually encroached upon the central pupil and thus degraded retinal image quality. Although we have argued that the RI data represent the spatial distribution of aberrations across the tear film, ²³ this method (unlike the Shack-Hartmann wavefront sensor ^{10 11} ) cannot yet specify these aberrations in quantitative terms (e.g., micrometers of wavefront deviation). 

Although we observed significant trial-to-trial variability, image degradation progressed at similar rates for all eyes in SCL and NSCL trials. Albarran ¹ also observed only small inter-subject variability, and only slightly greater effects of nonblinking in eyes with SCLs when compared with exposed corneas. These results are in stark contrast to the data of Timberlake et al. ⁹ and Lohmann et al., ¹² who compared eyes wearing soft contact lenses and rigid gas permeable lenses (RGP) to the uncovered cornea. They found that eyes wearing soft contact lenses were affected more by nonblinking than were eyes with RGPs or without contact lenses, which were virtually unaffected by up to 5 minutes of nonblinking. The discrepancy between their studies and our data may be attributable to one of the following factors: First, in our study it was essential to use local anesthetic and control light levels to prevent reflexive tearing during extended periods of nonblinking. Methods to prevent reflex tearing were not reported in either previous study. Second, we have observed subjects in our laboratory who fail to show tear break-up up to 1 minute after a blink, and thus the inter-study difference may simple reflect inter-subject differences in the stability of tears. 

In this study, we monitored tear film changes with prolonged periods of nonblinking, either with or without a soft contact lens. Under natural conditions, it is likely that blinking would occur much more frequently. The average blink rate is approximately 12 blinks/min, ²⁶ although there is considerable variation between individuals and visual tasks. ^{27 28 29} Thus, under normal viewing conditions in nondiseased eyes, it is likely that the retinal image degradation observed in our experiments and the accompanying visual disturbance would be minimized. However, Patel et al. ²⁹ found that visual tasks requiring concentration, such as VDT use, caused a fivefold decrease in blink rate, from an average of 18.4 blinks/min to only 3.6 blinks/min. According to our data, this could be associated with a 20% to 40% loss in image contrast. In addition, patients with dry eye often have an inadequate or unstable tear film, which breaks up quickly over the surface of the eye. ^{14 15 16} If blinking does not occur rapidly enough retinal image quality should be compromised in these patients. This prediction was verified by Goto et al., ¹⁷ who demonstrated a drop in visual acuity of 0.3 when 16 patients with dry eye held their eyes open for only 10 seconds. Blurry or disturbed vision has also been reported in a number of clinical studies of patients with dry eye. ^{16 18 19 20} Therefore, it is likely that tear film disruption will cause some degree of retinal image degradation in patients with dry eye or normal individuals in everyday life, depending on the visual task. 

²  Present address: 2507 Kennedy Circle, Brooks AFB, TX 78235-5117.

 Supported by National Institutes of Health Grant RO1 EY05109 (LNT) and The Borish Center for Ophthalmic Research.

 Submitted for publication February 4, 2000; revised July 24, 2000; accepted August 2, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

Left: a sample fundus image with the chosen blood vessel area enclosed in a rectangle; right: the image intensity profile across this vessel was measured and plotted.

Figure 2.

 

Four fundus images from a single experimental session recorded immediately after the blink (t = 0) and then at 30, 50, and 70 seconds after the blink. Under each fundus image is the intensity profile across the same single vessel that traverses the center of the white rectangle in the t = 0 panel.

Figure 3.

 

Retinal vessel contrast (RVC) data from multiple trials with subject 1. Data from trials without (NSCL; A) and with (SCL; B) a soft contact lens.

Figure 4.

 

Averaged individual RVC data for all three subjects. (A) NSCL; (B) SCL. ▴, subject 1; ▪, subject 2; •, subject 3.

Figure 5.

 

Averaged contrast sensitivity (CS) data for three subjects (▴, subject 1; ▪, subject 2; •, subject 3) both without SCL (NSCL; A) and with SCL (B).

Figure 6.

 

Image degradation after a blink averaged across all three observers. Comparison of RVC (○ and – – –) with CS (□ and —) data. Filled symbols, eye with SCL; open symbols, eyes without SCL.

Figure 7.

 

Retroillumination (RI) images of the pupil at four times after a blink. (A) An eye without an SCL; (B) an eye with an SCL.