Purpose. To compare measurements of murine ocular axial lengths (ALs) made with 780 nm partial coherence interferometry (PCI) and 1310 nm spectral domain-optical coherence tomography (SD-OCT).
Methods. AL was measured at postnatal day (P) 58 in C57BL/6J mice. Repeated AL measurements were taken using a custom-made 780 nm PCI and a commercial 1310 nm SD-OCT. Intra- and interuser variability was assessed along the central optical axis and 2-degree off-axes angles with the SD-OCT. Data were collected and analyzed using Cronbach alpha (α), Bland-Altman coefficient of repeatability, agreement plots, and intraclass correlation coefficients (ICC).
Results. AL measurements agreed well between the two instruments (3.262 ± 0.042 mm for PCI; 3.264 ± 0.047 mm for SD-OCT; n = 20 eyes). The ICC for PCI compared with SD-OCT was 0.92, confirming high agreement between the two instruments. Intrauser ICC for the PCI and SD-OCT were 0.814 and 0.995, respectively. Similarly, interuser ICC for PCI and SD-OCT were 0.970 and 0.943, respectively. Using SD-OCT, a 2-degree misalignment of the eye along the horizontal meridian produced mean differences in AL of −0.002 ± 0.017 mm relative to the centrally aligned images, whereas similar misalignment along the vertical meridian created 0.005 ± 0.018 mm differences in AL measurements.
Conclusions. AL measurements from the 780 nm PCI and 1310 nm SD-OCT correlate well. Multiple statistical indices indicate that both instruments have good precision and agreement for measuring murine ocular AL in vivo. Although the vertical meridian had the greater variability in AL in the small mouse eye; 2-degree off-axes differences were within the SD of centrally aligned AL.
Department of Ophthalmology, School of Medicine, Emory University, Atlanta, Georgia (HP, YQ, CT, SBJ, GS, MTP), and Rehabilitation Research and Development Center of Excellence, Atlanta VA Medical Center, Decatur, Georgia (YC, MTP).
Received August 22, 2011; accepted November 16, 2011.
Machelle T. Pardue Research Service (151 Oph) Atlanta Veterans Affairs Medical Center 1670 Clairmont Road Decatur, Georgia 30033 e-mail: firstname.lastname@example.org
Myopia is a significant healthcare burden, affecting 41.5% of the adult population in the United States1 and >80% of children aged 16 to 18 years in countries such as Taiwan2 and Singapore.3,4 Most juvenile-onset myopia in children and induced myopia in animal studies have found that myopia is caused by excessive axial length (AL) that mismatches the optical power of the eye.5,6 Many studies suggest that myopia is caused by both genetic and environmental factors.7 To elucidate the mechanisms underlying myopia, further studies are needed in which both genes and visual input are manipulated. The mouse offers such a model where many genetic mutants are currently available and the visual environment can be easily altered.8–12
Although the mouse model offers many opportunities for myopia research, the small size of the eye, ∼3 mm in diameter, represents a challenge in many aspects. Using a paraxial schematic eye model, Schmucker and Schaeffel13 have shown that a 5.4 to 6.5 μm change in AL corresponds to a 1 diopter (D) change in refractive error. Thus, to monitor experimental myopia development in the various mutant mouse models, highly sensitive and non-invasive biometric techniques are needed for measuring AL. Recently, a high-resolution laser micrometer has been described with a resolution of 0.7679 μm.14 However, this micrometer can only be used on enucleated eyes. In this study, we compare two in vivo methods for measuring mouse AL: 780 nm partial coherence interferometry (PCI) and 1310 nm spectral domain-optical coherence tomography (SD-OCT). The PCI is a custom-built device15 and uses a principle that is similar to that of the commercial AC Master (Carl Zeiss Meditec AG, Jena, Germany). Although the SD-OCT technology has been used with an advancing stepper motor to record AL in mice,11 the 1310 nm SD-OCT used here offers the ability to visualize the entire globe in one image frame. We evaluated the agreement and precision of the two methods in measuring murine AL by determining the Bland-Altman coefficient of repeatability (CR) and the intraclass correlation coefficient (ICC) as statistical indicators. In addition, we investigated the effect of 2 degrees of eye misalignment on AL measurements using the SD-OCT. Our results indicate that both instruments offer good agreement in measuring AL in mice, and that relative to the central alignment of the eye with the SD-OCT, vertical misalignment created greater differences in measurements than horizontal misalignment of the eye.
All animals were raised and treated in accordance to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and the institutional animal care and use committee. C57BL/6J breeder pairs were ordered from the Jackson Laboratory (Bar Harbor, ME) and bred in Atlanta Veterans Affairs Medical Center Animal Facility. The mice were raised in 12/12 light-dark cycles (∼10 lux) and provided chow and water ad libitum. Four independent C57BL/6J wild type litters for a total of 17 mice were measured at postnatal day 56 to 60 (P56–60; reported as P58). To determine agreement between instruments in measuring AL, measurements from each eye of an animal were treated as independent samples. The number of measurements that were analyzed for each eye is reported in the results.
Experiment Design and General Procedures
The PCI is based on the principle of a dual-beam Michelson interferometer with a partly coherent, near-infrared multimode diode light source that emits with a maximum emission at 780 nm and a coherence length of 120 μm. The output power is set at 175 μW. The movable mirror is adjusted such that the difference in the optical path length between the beams reflected from the movable mirror and stationary mirror in the interferometer matches the optical path length from the cornea to the retina to within the coherence length of the light source, at which point a concentric interference pattern is formed in front of the eye as shown on the display screen (Fig. 1A). For a scan, the movable mirror is translated at a constant speed whereas the intensity of the interference pattern is measured by a photodetector and plotted as a function of the mirror position. Peaks in the resulting plot correspond to reflections from retinal surfaces (Fig. 1B). The posterior peak of the eye was found to correspond to a reflection from the retinal pigment epithelial (RPE)-Bruch membrane interface.15,16 Because the anterior corneal surface acts as a reference reflector in this setup, peak locations equal optical distances from the cornea. Thus, AL measured with PCI is defined as the distance from the anterior corneal surface to the RPE-Bruch membrane.
For PCI measurements of AL, we used a custom-built PCI15 that had been modified for the mouse eye. Each mouse in this study received a head pedestal at P28 as described previously.17 At P58, mice were weighed and their eyes were dilated with 1% topical tropicamide (Bausch & Lomb, Tampa, FL) to cause pupil dilation; mice have a small ciliary muscle that likely does not contribute to accommodation.10,18 The awake mice were held in a plastic cylinder with the head immobilized by clipping the head pedestal to an adjustable stage. The optical axis of the mouse eye to be measured was carefully aligned along the collimated PCI measurement beam such that the first Purkinje image produced by the reflection from the anterior cornea was positioned in the center of the pupil (Fig. 1A). On the same eye, two users performed two repeated trials where 20 scans were taken per trial. Not every scan produces a peak corresponding to a reflection from the RPE because of eye movement or deterioration of the tear film. At least 10 scans that displayed RPE peaks were selected and averaged to determine the mean optical AL. Custom software converted optical AL to geometric AL through division by an average refractive index of 1.433.16
We used a 1310 nm SD-OCT (Bioptigen, Durham, NC) to measure AL. Similar to the PCI described above, OCT uses the principle of PCI to display the location of light-reflecting surfaces with respect to a reference mirror. Unlike PCI, where a scanning mirror matches the optical path length difference in the interferometer with optical distances between the cornea and retinal surfaces and a detector measures the intensity of the interference signal as a function of the mirror position, SD-OCT uses a stationary reference mirror, and a spectrometer measures the interference signal as a function of the wavelength of the reflected light reflected from ocular surfaces. Fourier transformation into time domain then produces the depth profile of ocular surfaces. A two-dimensional image is then created by combining a series of lateral depth scans (Fig. 2A to C). This particular SD-OCT system has an axial resolution of 8 μm with a hardware-limited imaging depth of 4 mm and a loss-limited imaging depth of 1.5 to 3 mm (information provided by the manufacturer). Because of the size limitations of the instrument's image window, the 4 mm image was viewed by advancing the anterior segment such that it inverts and both the anterior and posterior segments become overlapped within the same window (Fig. 2B). That is, the reference arm length was set to approximately the midposition in the crystalline lens so that both “halves” of the eye appeared within the available range of the instrument, taking advantage of the “mirror artifact.”19,20 The “unfolded” image was created using Adobe Photoshop to illustrate the true eye image (Fig. 2C).
Mice were first measured awake with PCI and thereafter under anesthesia with SD-OCT using intraperitoneal injections of ketamine (80 mg/kg)/xylazine (16 mg/kg). The anesthetized mouse was placed in a heated cylindrical holder (36°C), which was attached to an X-Y-Z movable stage (Bioptigen, Durham, NC) in front of the SD-OCT light source. The cornea was hydrated with normal saline drops. The reference arm and focus dial were adjusted simultaneously to a point where all structures of the eye are in focus. Correct alignment was confirmed by viewing the radial image of the surface of the eye and adjusting the light source for the central reflection along the horizontal and vertical optical meridians (Fig. 2A). Four to five linear scans were acquired per eye; each scan contained an average of 10 images. To measure AL, calipers (calibrated at refractive index of 1.43316) were placed from the cornea to lens fold and RPE border to lens fold (Fig. 2B). The sum of the two caliper measurements was recorded as the geometric AL. Again, at P58, inter and intrauser data were obtained by two users performing two trials per eye. On completion of imaging, the mouse was injected with yohimbine (2.1 mg/kg) to reverse the effects of xylazine and placed on a heating pad to recover.
Effects of Eye Alignment on AL Measurements
To determine the influence of misalignment from the optical axis on AL measurements, an initially centrally aligned mouse eye was intentionally misaligned using degree markings on the SD-OCT animal rotary stage. The eye was misaligned by 2° either along the horizontal or vertical meridian, with the mouse repositioned, centrally aligned, and then misaligned for each measurement. Misalignment >2° was not possible because the central cornea was no longer visible in the image to measure AL accurately. Images (n = 6 eyes) were acquired and analyzed as described above.
Collected data were organized into agreement and Bland-Altman plots.21 They were analyzed for precision and validity by using two statistical indices: (1) Bland-Altman CR for precision and (2) ICC for validity of same measure. Two other commonly used statistical indices were used for a familiar comparison: (1) Cronbach alpha (α) for internal consistency and (2) Pearson correlation coefficient for agreement. All data analysis was performed by commercial software (PASW, ver 18.0; IBM SPSS, Sombers, NY). All measurement values were reported in mean ± SD.
Variability, Repeatability, and Reliability of PCI
The instrument variability of the PCI was calculated by using the SD of repeated measurements in one eye. Average SD in AL measurements was 0.021 ± 0.045 mm for C57BL/6J mice at P58 (n > 20 PCI scans/eye in 14 eyes of seven mice). Fig. 3A to D summarizes the data agreement plots and Bland-Altman plots for intra- and interuser AL measurements using the PCI. The reliability (accuracy) of intrauser AL measurements as determined by the ICC was 0.814 (n > 20 PCI scans/eye in14 eyes) for the PCI. Statistical analysis of interuser AL measurements showed a high ICC of 0.97 (n > 10 PCI scans/eye/user in 20 eyes). Other statistical parameters can be found in Table 1.
Variability, Repeatability, and Reliability of SD-OCT
The instrument variability of the SD-OCT was calculated by using the SD of repeated measurements in one eye. Average SD in AL measurements was 0.010 ± 0.010 mm for C57BL/6J mice at P58 (n > 8 scans/eye in 16 eyes of eight mice). Fig. 4A, B presents the agreement and Bland-Altman identity plots for intrauser AL measurements using the SD-OCT with realignment and focus adjustments of the mouse eye preceding each measurement. The reliability (accuracy) of intrauser AL measurements as represented by the ICC was 0.995 (n > 8 scans/eye in 16 eyes) for the SD-OCT. Fig. 4C, D shows the agreement and Bland-Altman identity plots for intrauser AL measurements using the SD-OCT. ICC for interuser reliability of AL measurements with realignment of the mouse between each successive image was 0.943 (n > 4 scans/eye/user in 19 eyes). Table 1 reports CR and Cronbach alpha for internal consistency.
Agreement and Precision of PCI and SD-OCT AL Measurements
Average AL measurements made on four separate litters were 3.262 ± 0.042 mm for PCI and 3.264 ± 0.047 mm for SD-OCT (n = 20 eyes). Fig. 5A, B show the corresponding agreement and Bland-Altman identity plots for PCI and OCT murine ocular AL measurements. The two instruments had an ICC of 0.920 and CR of 0.048 mm. Additional statistical parameters can be found in Table 1.
Effects of Misalignment
Fig. 6 illustrates the effects of horizontal and vertical misalignment of the eye on AL measurements using the SD-OCT. Although there were no significant differences between the measurements obtained in the four regions, the mean difference in AL was greater in the vertical meridian (0.005 ± 0.018 mm) compared with the horizontal meridian (−0.002 ± 0.017 mm). As shown in Fig. 6, these differences were within the variability of the SD-OCT (0.010).
Methods for Ocular Biometry
Currently, the ocular biometric methods that have been tested and reported in the literature include A-scan ultrasonography,13,22 calipers and micrometers,14 image analysis of histological sections,23 magnetic resonance imaging and computed tomography scans,24 PCI also called optical low coherence interferometry (OLCI),15 and more recently, SD-OCT.11 Techniques performed in vitro or ex vivo (calipers, micrometers, histological sections, etc) are associated with changes in eye length secondary to desiccation and loss of turgidity that compromise their reliability in detecting and measuring small changes in AL.14 There is also limited resolution of these techniques to detect small changes in AL, although the use of a laser micrometer has somewhat circumvented some of these issues.14 With an axial resolution of about 40 to 50 μm, A-scan ultrasonography lacks the needed sensitivity for the small mouse eye in which a 10 D change in refractive error would be needed to detect a change in AL.13 Coherence interferometric techniques (OLCI and PCI) have the advantage of being able to image the mouse eye in vivo, however, the informational yield has been mainly limited to axial parameters: AL, corneal thickness, and anterior chamber depth.16 A 1310 nm SD-OCT with stepper motor focal plane advancement has been reported to measure various structures within the mouse eye11 but does not produce a cross-sectional image, as seen in Fig. 2.
In Vivo Ocular Measurements
The objective of this study was to determine the agreement and sensitivity of two instruments used to measure in vivo AL in the small mouse eye. Although in vivo mouse ocular AL has been reported previously,11,16 a comparison of methods has not been completed. As previously mentioned in the results, the AL were 3.262 ± 0.042 mm for PCI and 3.264 ± 0.047 mm for SD-OCT, respectively. These values are within the SDs of previously reported values during similar time points.11,16 The low Bland-Altman CR values (0.079 mm PCI and 0.018 mm for SD-OCT) indicate that the instruments have good reproducibility. Visually, the scatter shown in Bland-Altman identity plots of each instrument (Figs. 3, 4) suggest no specific trend or cluster differences between two instruments. The Bland-Altman plots indicated small mean differences of 0.003 to 0.038 mm between instruments, users, and trials. These two factors provide additional evidence that the AL measurements does not depend on any systemic errors from the instruments. The ICC values for inter- (>0.944) and intrausers (>0.814) indicate strong correlations. These comparisons were further strengthened by calculating the Cronbach alpha for internal consistency, which suggests high internal consistencies not only between trials but also between independent users.
Compared with PCI recordings from other species,15 the RPE/choroid peak in the mouse was relatively small and could be missed if not averaged. In the SD-OCT, some imprecision was generated by the subjective placing of the caliper on the RPE/choroid border, which appeared as a bright band (Fig. 2). However, even with these inconsistencies, the instruments still had precision of 0.010 ± 0.010 mm; corresponding to approximately a 2 to 4 D shift in refractive error.
One of the challenges in using the mouse eye is the absence of the fovea to align along the visual axis. Thus, alignment on the optical axis is based on visualization of Purkinje images.15 In this study, we showed that misalignment in the vertical meridian produced the greatest change in AL. However, these changes were not greater than the SD for centrally aligned instruments. Using the SD-OCT, we were only able to measure off-axis 2° in each direction. The trend in shorter AL measurements in the nasal direction may suggest that the AL of the mouse eye is not uniform around the optic axis, particularly if measured beyond 2°. Such regional differences in AL would be similar to those reported for primates, guinea pigs, and children.25–26 These results also indicate that greater care needs to be taken when aligning the mouse eye along the vertical vs. horizontal meridian when performing any optical-based technique such as photorefraction, PCI, or OCT.
In conclusion, these results support the use of both the PCI and SD-OCT for measuring AL in the mouse eye. The PCI has advantages in speed of aligning the instruments and collecting data whereas the OCT offers the ability to image several ocular structures along with AL. In addition, we were able to show that misalignments <2° from the central optical axis will not affect AL measurements in the mouse eye.
Machelle T. Pardue
Research Service (151 Oph)
Atlanta Veterans Affairs Medical Center
1670 Clairmont Road
Decatur, Georgia 30033
We thank Bill Del'Aune, PhD, for providing statistical advice and WonChan Sohn for technical assistance with Fig. 2.
This work was supported by National Eye Institute, National Institute of Health, grant NIH R01 EY016435, P30 EY006360, and Veterans Affairs Research Career Scientist Award (to MTP) and Departmental Award from Research to Prevent Blindness.
The authors have no financial interest in any instrument listed.
1. Vitale S, Sperduto RD, Ferris FL III. Increased prevalence of myopia in the United States between 1971–1972 and 1999–2004. Arch Ophthalmol 2009;127:1632–9.
2. Lin LL, Shih YF, Hsiao CK, Chen CJ, Lee LA, Hung PT. Epidemiologic study of the prevalence and severity of myopia among schoolchildren in Taiwan in 2000. J Formos Med Assoc 2001;100:684–91.
3. Wong TY, Foster PJ, Hee J, Ng TP, Tielsch JM, Chew SJ, Johnson GJ, Seah SK. Prevalence and risk factors for refractive errors in adult Chinese in Singapore. Invest Ophthalmol Vis Sci 2000;41:2486–94.
4. Dirani M, Tong L, Gazzard G, Zhang X, Chia A, Young TL, Rose KA, Mitchell P, Saw SM. Outdoor activity and myopia in Singapore teenage children. Br J Ophthalmol 2009;93:997–1000.
5. Siegwart JT Jr., Norton TT. Perspective: how might emmetropization and genetic factors produce myopia in normal eyes? Optom Vis Sci 2011;88:365–72.
6. Meng W, Butterworth J, Malecaze F, Calvas P. Axial length of myopia: a review of current research. Ophthalmologica 2011;225:127–34.
7. Feldkamper M, Schaeffel F. Interactions of genes and environment in myopia. Dev Ophthalmol 2003;37:34–49.
8. Pardue MT, Faulkner AE, Fernandes A, Yin H, Schaeffel F, Williams RW, Pozdeyev N, Iuvone PM. High susceptibility to experimental myopia in a mouse model with a retinal on pathway defect. Invest Ophthalmol Vis Sci 2008;49:706–12.
9. Barathi VA, Boopathi VG, Yap EP, Beuerman RW. Two models of experimental myopia in the mouse. Vision Res 2008;48:904–16.
10. Schaeffel F, Burkhardt E, Howland HC, Williams RW. Measurement of refractive state and deprivation myopia in two strains of mice. Optom Vis Sci 2004;81:99–110.
11. Zhou X, Shen M, Xie J, Wang J, Jiang L, Pan M, Qu J, Lu F. The development of the refractive status and ocular growth in C57BL/6 mice. Invest Ophthalmol Vis Sci 2008;49:5208–14.
12. Schippert R, Burkhardt E, Feldkaemper M, Schaeffel F. Relative axial myopia in Egr-1 (ZENK) knockout mice. Invest Ophthalmol Vis Sci 2007;48:11–7.
13. Schmucker C, Schaeffel F. A paraxial schematic eye model for the growing C57BL/6 mouse. Vision Res 2004;44:1857–67.
14. Wisard J, Chrenek MA, Wright C, Dalal N, Pardue MT, Boatright JH, Nickerson JM. Non-contact measurement of linear external dimensions of the mouse eye. J Neurosci Methods 2010;187:156–66.
15. Schmid GF, Papastergiou GI, Nickla DL, Riva CE, Lin T, Stone RA, Laties AM. Validation of laser Doppler interferometric measurements in vivo of axial eye length and thickness of fundus layers in chicks. Curr Eye Res 1996;15:691–6.
16. Schmucker C, Schaeffel F. In vivo biometry in the mouse eye with low coherence interferometry. Vision Res 2004;44:2445–56.
17. Faulkner AE, Kim MK, Iuvone PM, Pardue MT. Head-mounted goggles for murine form deprivation myopia. J Neurosci Methods 2007;161:96–100.
18. Artal P, Herreros de Tejada P, Munoz Tedo C, Green DG. Retinal image quality in the rodent eye. Vis Neurosci 1998;15:597–605.
19. Wojtkowski M, Kowalczyk A, Leitgeb R, Fercher AF. Full range complex spectral optical coherence tomography technique in eye imaging. Opt Lett 2002;27:1415–7.
20. Ho J, Castro DP, Castro LC, Chen Y, Liu J, Mattox C, Krishnan C, Fujimoto JG, Schuman JS, Duker JS. Clinical assessment of mirror artifacts in spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci 2010;51:3714–20.
21. Altman DG, Bland JM. Measurement in medicine: the analysis of method comparison studies. Statistician 1983;32:307–17.
22. Tsinopoulos IT, Tsaousis KT, Symeonidis C, Chalvatzis N, Dimitrakos SA. Repeatability and reproducibility of a-scan biometry quantitative findings. Curr Eye Res 2009;34:447–53.
23. Tejedor J, de la Villa P. Refractive changes induced by form deprivation in the mouse eye. Invest Ophthalmol Vis Sci 2003;44:32–6.
24. Tkatchenko TV, Shen Y, Tkatchenko AV. Analysis of postnatal eye development in the mouse with high-resolution small animal magnetic resonance imaging. Invest Ophthalmol Vis Sci 2010;51:21–7.
25. Smith EL III, Hung LF, Huang J, Blasdel TL, Humbird TL, Bockhorst KH. Effects of optical defocus on refractive development in monkeys: evidence for local, regionally selective mechanisms. Invest Ophthalmol Vis Sci 2010;51:3864–73.
26. Schmid GF. Association between retinal steepness and central myopic shift in children. Optom Vis Sci 2011;88:684–90.
spectral domain-optical coherence tomography; partial coherence interferometry; low coherence interferometry; axial length; ocular biometry; myopia; refractive development© 2012 American Academy of Optometry