Measurements of infant refraction and gaze position are central to understanding the development of clear, single vision and to understanding clinically significant abnormality (e.g., accommodative esotropia, intermittent exotropia, and amblyogenic factors such as high hyperopia and anisometropia).
The eccentric photorefraction technique can be used to make both of these measurements and is attractive for studying infants because it is fast, remote, and can record from both eyes simultaneously (see Fig. 1). When automated at video rates, its speed and potential to record data binocularly make it, in theory, preferable to conventional retinoscopy; infants’ short attention spans have less impact on data quality and anisometropia is easier to detect in a binocular measurement. Estimates of vergence, gaze position, and pupil size are made using a contrast detection algorithm to locate the pupils and first Purkinje images in each video frame, while the refraction estimate is made using the eccentric photorefraction principle.
FIGURE 1. An example PowerRefractor image from an infant. The distribution of light in the pupils provides simultaneous estimates of defocus in the two eyes, and the positions of the Purkinje images relative to the center of the pupils provide an estimate of gaze position. The image also demonstrates the technique used in this study for the relative validation. A positive-powered lens was held before one eye, producing a myopic crescent in the pupil, and the induced anisometropia was determined as a function of lens power. (The camera aperture size was kept in the recommended range of 5.6–8 throughout data collection.)
The current study concerns calibration of the refraction estimate. The eccentric photorefraction principle is based on an analysis of light reflected from the retina. Light from a source in the plane of the front of the camera lens forms an image on the retina and is then reflected back through the pupil into the camera. The estimate of the eye’s defocus is derived from the distribution of reflected light across the subject’s pupil in the camera image (see Fig. 1). The standard analytical description of this light distribution includes the following parameters: the refractive error of the eye, the distance of the subject to the camera, the pupil size, and the eccentricity of the light source from the edge of the aperture at the front of the camera lens (the photorefraction aperture). Knowing each of the other parameters, therefore, should reveal the eye’s refractive error. Unfortunately, this analysis of the technique is not a complete description of the system. There are additional factors that affect the calibration of these instruments.
These factors are mainly optical characteristics of the eye being measured and include:
1. Reflectance of the retina, which may affect the gradient of the light distribution in the pupil
2. The distance between the retinal structures that reflect the light and the photoreceptors that initiate the visual response, which may cause an absolute offset in the measurement and
3. Higher-order monochromatic aberrations, which may disrupt the light distribution in the pupil.
In a system with an extended light source, the slope of the light distribution in the pupil plane can be fit with a linear regression to estimate refractive error over a relatively wide operating range. The conversion of the slope of the linear regression into a refraction estimate has to be based on an empiric calibration rather than the theoretical analysis discussed previously, however, because of the additional factors.
Our goal was to test the validity of an adult empirical calibration for assessing infant subjects using one of the more commonly used instruments. The PowerRefractor (initially manufactured by Multichannel Systems but more recently by PlusOptix) is an example of a video-based eccentric photorefractor that is in principle well suited for use with pediatric populations. This instrument was calibrated using adult subjects. The results of the built-in calibration have been verified in uncyclopleged adults by other groups who have compared it with subjective refraction and autorefractors. It has also been found reliable, typically within 1 D of spherical equivalent, when compared with cycloplegic or noncycloplegic retinoscopy or autorefraction in populations extending down into childhood.
The results of these studies suggest that the PowerRefractor, like the eccentric photorefraction technique in general, holds promise as a clinical and research tool for use with infants. The goal of the current study was to assess the built-in adult PowerRefractor refraction estimate for use with uncyclopleged infants under naturalistic conditions, because this younger population would benefit the most from the rapid assessment method and has eyes that differ the most from the adults for which the instrument was calibrated (e.g., in terms of fundus reflectance, optical power of the eye, the small eye artifact, and higher-order aberrations). We wanted to determine whether the built-in adult calibration would be appropriate for use with free-viewing uncyclopleged infant subjects.
Other variants of the photorefraction technique and other eccentric photorefraction instruments have been calibrated directly for use as vision screening tools for infants, but to date the relatively widely used PowerRefractor system has not been validated for infants and we are not aware of an instance in which another adult defocus calibration has been validated for infants. Our study consisted of an absolute validation using the gold standard of retinoscopy and a relative validation using trial lenses to induce known amounts of defocus.
FIGURE 2. A demonstration of the absolute validation protocol. Retinoscopy was performed immediately adjacent to the PowerRefractor photorefraction aperture at a 1-m viewing distance.
DISCUSSION
We have performed an empirical analysis of photorefraction measurements provided by a commercial instrument. We have tested the validity of the calibration provided in the PowerRefractor software for infant subjects. We used two protocols, one that compared the measurements with simultaneous retinoscopy judgments, an absolute comparison, and another that compared the readings with known changes in defocus, a relative comparison. We performed these analyses empirically because the current theoretical descriptions of the eccentric photorefraction technique are not complete for predicting defocus.
Absolute Validation
The point at which an eccentric photorefractor reads zero should not depend on an assumed conversion factor from slope of the light intensity distribution to diopters of defocus. The “slope” for zero defocus is zero and so a multiplicative factor has no effect on the result. The PowerRefractor should therefore read zero when the retinoscope is held in the same plane as the photorefraction aperture and the retinoscopy reflex is neutral. A nonzero reading from the photorefractor in this situation would imply a dioptric offset. Possible reasons for this might include any remaining longitudinal chromatic aberration resulting from the wavelength difference between the PowerRefractor and retinoscope light sources (with no correction, the PowerRefractor would read more hyperopia than the retinoscopy) or a compensation built into the PowerRefractor software for either tonic accommodation or the distance of the camera relative to infinity. Our results show that the adults and infants actually had similar offsets between the PowerRefractor and retinoscopy results. The PowerRefractor typically read a small amount of myopia when the retinoscopy reflex was judged to be neutral. The most simple explanation for this difference would be a compensation in the software for the viewing distance of the camera.
Two other studies have made simultaneous comparisons between the photorefractor reading and the focus of the eye. Allen et al. plotted PowerRefractor reading as a function of target fixation distance for five adults. For a fixation distance of -1 D, equivalent to the camera distance of 1 m, they found a PowerRefractor reading of approximately -1 D (Fig. 1). Seidemann and Schaeffel performed retinoscopy on adults while data were collected with the PowerRefractor. Their Figure 1 demonstrates that for a retinoscopy neutralization at -1 D (the 1-m camera distance), the PowerRefractor readings were between approximately 0 D and -1 D. The results of these two other studies are in good agreement.
Relative Validation
The slopes of the relative validation functions are close to one for the infants and adults. The data suggest that even the slopes at the extremes of each population lie within a factor of two of the ideal 1:1 line. The mean values also suggest that the infants had higher slopes than found for adults with approximately the same variability in the two groups. The increased slope of the infants’ induced anisometropia function must come from a factor that causes a multiplicative increase in calculated slope of the light distribution in the pupil. This could be caused by an increase in fundus reflectance, for example, the light intensity at each point in the pupil would be multiplied by a common factor. Whatever the factor causing this, it does not appear to be compensated for in a correction for mean image intensity such as that described by Schaeffel et al.
Using this induced anisometropia protocol to control for a change in infants’ accommodation depends on the assumption that infants’ accommodation is fully consensual. If this were not true, the apparent anisometropia would change with changes in accommodation; the anisometropia would either increase or decrease depending on the eye driving the response and the direction of the response. We saw no clear evidence of this behavior in the data and limited the range of mean values from the eye with no lens to prevent excessive changes in accommodation. We were therefore comfortable making this assumption and are not able to explain the difference in mean infant and adult slopes by assuming that infants’ accommodation is not consensual.
Clinical Estimation of Refractive Error
One of the most promising uses of the eccentric photorefraction technique is to examine children for the presence of high refractive errors, strabismus, or media abnormalities. A number of studies have compared screening data from other eccentric photorefractors with a full clinical examination of refractive error. Abrahamsson et al., Suryakumar and Bobier, and Schmidt et al. have all used the PowerRefractor to screen for refractive error. Abrahamsson et al. found that over 90% of their children from 6 months to 5 years of age had differences between PowerRefractor and autorefractor readings of less than 1 D, and Suryakumar and Bobier found that the PowerRefractor underestimated hyperopia less than a number of other similar screening instruments in uncyclopleged preschool children. Schmidt et al. found the PowerRefractor did not perform as well as a number of their other screening tools in the Vision in Preschoolers study.
We do not examine the ability of the instrument to detect refractive error, but look at its accuracy at measuring defocus in one meridian (the combination of refractive error and accommodative performance). The children that participated in the other studies were typically older than the infants tested here and so, under the assumption that children’s eyes are more mature than infants’, the children should only have more adult-like defocus calibrations.
An interesting observation has been made in a number of studies, that cycloplegia can actually make screening for refractive error less effective, particularly for assessing astigmatism and its axis. The aberrations introduced into the periphery of the image of a dilated pupil are thought to disrupt the gradient of the light-intensity profile. It is feasible that our defocus measurements would have suggested poorer performance of the instrument if we had conducted our protocols in less naturalistic conditions using cycloplegia.
The data collected here suggest that the PowerRefractor (Multichannel Systems version) is able to consistently detect large amounts of defocus in infants, as required in a vision screening environment. For more detailed analyses of absolute accommodative accuracy and small amounts of defocus, however, the absolute validation indicated an offset between the photorefraction and retinoscopy defocus estimates, and the relative protocol demonstrated different slopes for adults and infants with clear individual differences. It is therefore necessary to consider these factors when making detailed analyses using this instrument. Although this is true, the same requirement has also been noted by the developers of the instrument for work with adults.
Our experience in collecting these data suggests it will not be simple to do individual calibrations for infant subjects or to replicate the validation easily for individual instruments in different laboratories. These data were collected on different groups of infants, and the combined success rates imply that successful completion of both protocols on an individual would require excluding a large number of infants. The option of using cycloplegia would depend on the goals and design of the analysis.
Conclusions. The results suggest that the instrument is capable of detecting large amounts of defocus but needs individual calibration for detailed studies of accommodative accuracy and absolute levels of defocus, as has been recommended previously for adult subjects.
No comments:
Post a Comment