Development and Preliminary Validation of a Virtual Reality Approach for Measurement of Torsional Strabismus

Double Maddox rod (DMR), the gold-standard method for in-office measurement of cyclodeviation, requires an examiner and specialized equipment. The objective of this study was to develop a virtual reality (VR) technique for measurement of cyclodeviation and validate this against the DMR

Original Contribution Section Editors: Clare Fraser, MD Susan Mollan, MD Development and Preliminary Validation of a Virtual Reality Approach for Measurement of Torsional Strabismus Megha Bindiganavale, BS, David Buickians, BS, Scott R. Lambert, MD, Zachary M. Bodnar, MD, Heather E. Moss, MD, PhD Background: Double Maddox rod (DMR), the gold-standard method for in-office measurement of cyclodeviation, requires an examiner and specialized equipment. The objective of this study was to develop a virtual reality (VR) technique for measurement of cyclodeviation and validate this against the DMR. Methods: A VR-DMR was implemented using a smartphone and commercially available VR viewer. The app displayed a line to each eye and accepted touch inputs from the user to rotate the lines into perceived alignment. VR-DMR cyclodeviation measurements were compared with traditional DMR (T-DMR) cyclodeviation measurements in adults with and without strabismus and children without strabismus. Results: Thirty-one subjects were studied (age 5–88 years, 20 with strabismus). VR-DMR had similar test–retest reliability as T-DMR. VR-DMR was highly correlated with T-DMR (r2 = 0.94, linear regression slope 1.12) with a slight positive bias (linear regression y intercept 1°). VR-DMR was preferred by 54% of subjects with 29% having no preference. Conclusions: A VR method of ocular cyclodeviation measurement using sensory techniques was implemented using commercially available hardware. VR measurements compared favorably with gold-standard DMR measurements, and user feedback was positive. The VR methodology has application for in office and home use by nonexperts for purposes of strabismus monitoring. Journal of Neuro-Ophthalmology 2022;42:e248–e253 doi: 10.1097/WNO.0000000000001451 © 2021 by North American Neuro-Ophthalmology Society Department of Ophthalmology (MB, DB, SRL, HEM), Stanford University, Palo Alto, California; Meadows Eye Physicians and Surgeons (ZMB), Henderson, Nevada; and Department of Neurology and Neurological Sciences (HEM), Stanford University, Palo Alto, California. The authors report no conflicts of interest M. Bindiganavale, D. Buickians, Z. M. Bodnar, and H. E. Moss contributed equally to this work. Address correspondence to Heather Moss, MD, PhD, Spencer Center for Vision Research at Stanford 2370 Watson Court, Suite 200, MC 5353 Palo Alto, CA 94303 *SA; E-mail: [e-mail] e248 BACKGROUND M isalignment of the eyes (strabismus) interferes with binocular visual processing, which can impair visual pathway development in children and can cause diplopia or visual disturbance in adults. This increases risk of falls and injuries and decreases quality of life (1,2). Cyclodeviation, torsional misalignment of the eyes, is an important form of strabismus which can cause double vision with 1 image tilted. Early detection and treatment of strabismus is crucial, so that treatment can be pursued with the goals of minimizing symptoms, improving quality of life, and preventing permanent vision loss. Measurement of cyclodeviation is important because it informs treatment approach (3). Current gold-standard methods of testing for strabismus include alternating cover, cover/uncover, Maddox rod, double Maddox rod (DMR), synoptophore, the Lancaster red-green test (subjective), and the Hirschberg test (objective). The DMR test is subjective and is the most common in-office test for measurement of cyclodeviation. It is performed by placing lenses that refract a point light source into a line image in front of each eye. The patient is asked to rotate the lenses, so the lines are perceived to be parallel to each other (4). Measurements are imprecise and prone to administrator error (5). They also require operator expertise, special medical equipment, and cannot be performed independently by the patient. Self-administered smartphone-based tests may be an alternative to analog strabismus measurements in the office and facilitate at-home monitoring. In addition to eliminating the need for specialized medical equipment and operator expertise, the measurements have the potential to provide both more precise and more accurate data with which to improve diagnosis and monitoring of strabismus by patients and providers. Bindiganavale et al: J Neuro-Ophthalmol 2022; 42: e248-e253 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution Efforts to develop smartphone and desktop applications to measure strabismus have included standardization of objective tests such as light reflection (6,7) and apps to obtain standardized photographs of the eye that can be qualitatively analyzed for eye movement limitation and alignment (8). Another strategy has been extraction of strabismus measurements based on color photographs, video, or infrared-eye movement recordings during binocular viewing and cover testing (9–11). Cyclodeviation measurements have been obtained by automated analysis of fundus photographs (objective) (12) and by having a patient manipulate red and green lines on a computer (13) or smartphone (14) screen while wearing red-green glasses to dissociate the eyes (subjective). We implemented a novel approach at sensory-based strabismus testing, specifically using a virtual reality (VR) headset to display different inputs to each eye while the patient manipulates these to achieve perceived alignment. In this study, we demonstrate proof of concept and initial validation using an iOS application to perform quantitative cyclodeviation measurements. METHODS FIG. 1. Hardware for virtual reality double Maddox rod. A smartphone is inserted into a commercial viewer that allows that user to view the screen through lenses. Touch inputs on the upper left and right of the viewer transmit downward pressure of the user’s finger to a touch input on the corner of the smartphone screen. red and allowed to rotate. However, the software provides settings to make both lines red and allows for manipulation of the left eye image only, right eye image only, or both. Virtual Reality Double Maddox Rod Device and Software The hardware consists of an iPhone 6 or above (Apple Inc, Cupertino, CA) and a commercial VR viewer that allows touch inputs (Merge AR/VR headset, Merge Labs Inc, San Antonio, CA) (Fig. 1). When the headset is worn, the user views the screen of the smartphone (60Hz refresh rate) through two 42-mm 20-diopter lenses with 96° field of view and adjustable interpupillary distance. The headset used does not accommodate glasses, although the size of the graphics is such that small distance refractive errors should not impact test measurements. Accommodation is relaxed by the lenses, so near refraction is not necessary. Software development and validation were conducted on the iOS platform using the Swift programming language. Similar to the traditional DMR (T-DMR) test, the iPhone application displays 2 lines: 1 line is displayed to the left eye, and 1 line is displayed to the right eye (Fig. 2). The patient is tasked with rotating the axis of 1 line by pressing the input buttons on the VR viewer (which generate a touch input on the iPhone screen) until the lines are perceived to be parallel. The relative cyclodeviation is calculated as the difference in the ending angle between the 2 lines and displayed on the screen such that the number is not visible to the patient in the VR viewer (positive = excyclotorsion and negative = incyclotorsion). For the purposes of this initial validation, the line displayed to the left eye was white and fixed horizontally, whereas the line displayed to the right eye was Bindiganavale et al: J Neuro-Ophthalmol 2022; 42: e248-e253 FIG. 2. Experimental set up for T-DMR and VR-DMR testing. A. Examination set up for T-DMR (left) and VRDMR (right). B. Subject point of view for T-DMR (left) and VR-DMR (right) at start (top) and end (bottom) of trial. Result shown for a subject without strabismus. T-DMR, traditional double Maddox rod; VR-DMR, virtual reality double Maddox rod. e249 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution TABLE 1. Characteristics of study participants Group Age in years, average (range) Female gender, n (%) Cyclodeviation, average (range)* Diagnosis Control (n = 11) Strabismus (n = 20) 26 (5–66) 60 (24–88) 6 (55%) 0.4 (0–1) 12 (60%) 3.9 (0–17) not applicable 11 esotropia 8 exotropia 1 fourth nerve palsy *Average of 3 traditional double Maddox rod trials. Subjects Subjects with strabismus were recruited from the adult strabismus clinic at Byers Eye Institute at Stanford. Control subjects were recruited from the Spencer Center for Vision Research at Stanford and the Byers Eye Institute at Stanford. Inclusion criteria for strabismus subjects were age $18 years with clinical diagnosis of strabismus with uncorrected near visual acuity of at least 20/100 with each eye. Inclusion criteria for healthy control subjects were age $18 years with lack of visual symptoms and no known ophthalmic or neurological disease. For these subjects, full extraocular movements were verified, but alignment was not measured. A small number of pediatric control subjects (age ,18 years) were studied to assess feasibility in younger-aged patients. For strabismus subjects, clinical diagnosis was extracted from the medical record. The research followed the tenets of the Declaration of Helsinki and was approved by the institutional review board in the Stanford Office of Research Compliance. Written informed consent by subjects ($18 years) or subject’s guardians (for subjects aged ,18 years) in addition to assent for subjects ,18 years was obtained before testing. Validation Protocol A prospective cross-sectional study of adults with and without strabismus was used to validate the VR-DMR against the T-DMR. A small number of pediatric subjects were studied to assess feasibility in younger patients. Each subject completed 3 trials of T-DMR followed by 3 trials of VR-DMR (Fig. 2). T-DMR was performed using standard clinical techniques consisting of a Finoff transilluminator shone toward the subject who was wearing trial frames containing the Maddox prism lenses (white lens in front of the left eye and red lens in front of the right eye) while the subject manipulated the lenses using screws on the trial frames. VRDMR was performed with the VR viewer affixed to the subject’s head using an elastic and Velcro strap while the subject manipulated the display using touch inputs to e250 the phone achieved using the buttons on the viewer. Subjects were instructed to keep their heads in a neutral position, and this was monitored by the examiner during testing. For both devices, the left eye viewed a white line at 0° and the right eye viewed a red line. Trials differed by the starting angle of the red line (45°, 90°, or 135°). The starting angle order was randomized and was different for T-DMR and VR-DMR for a given subject. Subjects were asked to rotate the red line until the red and white lines appeared parallel. Once completed, the relative cyclodeviation between the eyes was recorded as the difference in angles between the lenses in the trial frame (T-DMR) or the difference in angles between the displayed lines in the app (VR-DMR). For this study, positive values represent relative excyclotorsion, whereas negative values represent relative incyclotorsion. Informal qualitative user feedback was obtained from the subjects regarding convenience, ease of use, speed, and comfort of 2 testing methods through nonstandardized interview. Statistical Analysis T-DMR and VR-DMR cyclodeviation measurements were compared within each subject. Test–retest reliability for each method as assessed using Bland–Altman plots of cyclodeviation measurements for Trial 1 and Trial 2. Validation of the VR-DMR against the T-DMR gold standard was performed using Bland–Altman plots of average cyclodeviation measurements using T-DMR and VR-DMR. Bias was calculated as the average of the differences between measurements across subjects. The upper and lower limits were calculated as the 95% confidence interval for the bias. Correlation between methods was studied using Pearson correlation and linear regression. RESULTS Twenty strabismus subjects (12 women, age 24–88 years), 9 control adult subjects (4 women, age 22–66 years), and 2 control minor subjects (2 girls, age 5 and 12 years) participated in the study. Of the strabismus subjects, 11 had intermittent esotropia, 8 had intermittent exotropia, and 1 had fourth nerve palsy (Table 1). Bland–Altman plots comparing trials using the same test methodology and average of trials using different testing methodologies are shown in Figure 3. Both T-DMR and VR-DMR trials showed good agreement without bias and each with 1 outlier. Variation was larger for VR-DMR than T-DMR. Variation was smaller for controls than for strabismus subjects. VR-DMR measurements were biased to be smaller than T-DMR, and this was more pronounced at higher levels of cyclodeviation. Bindiganavale et al: J Neuro-Ophthalmol 2022; 42: e248-e253 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution and for strabismus subjects only) with linear regression slopes and y-intercepts of 1.12 and 1.07° for strabismus subjects and 1.14 and 0.81° for all subjects. Approximately half of the subjects (17/31), including both minors, found VR-DMR to be easier to use compared with T-DMR. Five subjects preferred T-DMR; 2 reported challenges with the VR headset moving, 1 reported that the lines on the screen were more difficult to see than using the T-DMR, and the remaining subjects reported they needed more time to learn how to use the device. The remaining 9 subjects had no preference for testing method. All subjects expressed a promising future for the use of smartphone applications for testing. CONCLUSIONS FIG. 3. Bland–Altman plots comparing cyclodeviation measurements with T-DMR and VR-DMR protocols. A. Test– retest between T-DMR trials. B. Test–retest between VRDMR trials. C. Comparison between T-DMR and VR-DMR averages. For all plots, a point (o denotes strabismus subject, and x denotes control subject) is a single subject with the x value the average between the 2 measurements being compared and y value the difference between the 2 measurements. The solid line is the mean difference across all subjects, and the dashed lines are the 95% confidence interval for the mean. T-DMR, traditional double Maddox rod; VR-DMR, virtual reality double Maddox rod. Figure 4 displays a scatter plot of average cyclodeviation measurements for T-DMR vs VR-DMR for all subjects. Measurements were highly correlated (r2 = 0.94 overall Bindiganavale et al: J Neuro-Ophthalmol 2022; 42: e248-e253 We developed and implemented a novel methodology to measure ocular cyclodeviation using sensory measurements collected using a smartphone and VR headset. We demonstrate feasibility of use for adults with and without strabismus and children without strabismus. Preliminary validation shows similar measurements were obtained using the VR method as with the gold-standard T-DMR methodology. Test–retest reliability was comparable between the 2 methods. Notably, the 95% confidence interval for VR-DMR measurements was ,5° as previously recommended as a threshold to detect cyclotorsion change (13). Although there have been efforts to extract cyclodeviation from fundus images (15,16), sensory measurements using DMR remain the gold standard, comparable with synoptophore measurements which require a bulky desktop device (4). Sensory techniques have previously been used to obtain cyclodeviation measurements using a desktop display (16) or smartphone app (14) coupled with red/green glasses to isolate the inputs to each eye. Compared with these approaches, ours does not require specialty lenses. Our approach allows for 2 lines of the same color to be used when there is a concern of preferential fixation on the white over red line (17). Similar to other approaches, our methodology does require specialized equipment. Smartphones are ubiquitous, and smartphone VR holders are relatively inexpensive ($50 for the holder used in our study compared with w$140 for adult trial frames and 2 Maddox trial lenses) (www.good-lite.com). Our methodology serves as proof of concept for sensory-based measurements of torsional strabismus using a VR technique, which might be expanded to vertical and horizontal measurements. These portable sensory-based approaches complement image analysis (8) and motor testing approaches (10) to allow remote strabismus measurement and other external eye measurements (18) comparable with what is performed in the office setting. There are opportunities for improving and expanding our approach. The touch input mechanism inherent in the e251 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution work is needed to validate this methodology as an at-home monitoring system to support telehealth visits. Neither T-DMR nor VR-DMR incorporated a vertical offset in the images as is sometimes performed in clinical testing (19). A limitation of our validation study is the low sample size, particularly of subjects with large angles of cyclodeviation, and those with vertical strabismus in which cyclodeviations are most clinically relevant, as we were not able to recruit and run clinical examinations for research during the COVID-19 public health emergency. However, we believe we have presented sufficient data to show feasibility and promise. Subjective measurement of cyclodeviation remains 1 part of the ocular misalignment examination and may not be possible in patients with poor vision in 1 eye or suppression. In conclusion, we developed and performed preliminary validation studies for a VR-based ocular cyclodeviation measurement method that has promise for remote diagnosis and monitoring. The measurements obtained using the novel method are comparable with the T-DMR test. User feedback was positive. This technique complements other smartphone-based approaches for measurement of strabismus including publicly available iOS and android apps 9gaze (See Vision LLC) and iTorsion (University of Minnesota). FIG. 4. Correlation between T-DMR and VR-DMR test results. Strabismus subjects (o); control subjects (x). Each point represents the average for that subjects’ 3 trials for each testing protocol. Positive cyclotorsion values are relative excyclotorsion, whereas negative cyclotorsion values are relative incyclotorsion. Vertical error bars show the range of the 3 VR-DM trials for that subject, whereas horizontal error bars show the range of the 3 T-DM trials for that subject. Lower plot is enlargement of shaded area from upper plot. T-DMR, traditional double Maddox rod; VR-DMR, virtual reality double Maddox rod. headset we used limited the gestures we could use in the application. In addition, the headset could not accommodate glasses which may have compromised the visual experience and fixation for users. Other headsets might allow for wearing of glasses, and a Bluetooth remote would be an option for increased input options. Although our methodology is feasible for home measurement, our validation study was performed in a clinical study with an examiner. Study of VR-DMR allowing for manipulation of both images and with the images of same color (to minimize subject bias) is an important area of future study. Future e252 STATEMENT OF AUTHORSHIP Category 1: a. Conception and design: H. E. Moss, Z. M. Bodnar, D. Buickians, and M. Bindiganavale; b. Acquisition of data: M. Bindiganavale and S. R. Lambert; c. Analysis and interpretation of data: M. Bindiganavale and H. E. Moss; Category 2: a. Drafting the manuscript: H. E. Moss, Z. M. Bodnar, M. Bindiganavale, D. Buickians, and S. R. Lambert; b. Revising it for intellectual content: H. E. Moss, Z. M. Bodnar, M. Bindiganavale, D. Buickians, and S. R. Lambert. Category 3: a. Final approval of the completed manuscript: H. E. Moss, Z. M. Bodnar, M. Bindiganavale, D. Buickians, and S. R. Lambert. ACKNOWLEDGMENTS Funding from the National Eye Institute (NIH 30 026877), unrestricted grant from Research to Prevent Blindness to the Stanford Department of Ophthalmology, Stanford Clinical and Translational Science Award (CTSA) to Spectrum. The CTSA program is led by the National Center for Advancing Translational Sciences (NCATS) at the National Institutes of Health. REFERENCES 1. Pineles SL, Repka MX, Yu F, Lum F, Coleman AL. Risk of musculoskeletal injuries, fractures, and falls in medicare beneficiaries with disorders of binocular vision. JAMA Ophthalmol. 2015;133:60–65. 2. Nilsson J. The burden of amblyopia and strabismus: justification of treatment and screening revisited. Arch Ophthalmol. 2008;126:143–146; author reply 145-146. Bindiganavale et al: J Neuro-Ophthalmol 2022; 42: e248-e253 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution 3. Lemos J, Eggenberger E. Clinical utility and assessment of cyclodeviation. Curr Opin Ophthalmol. 2013;24:558–565. 4. Liebermann L, Hatt SR, Leske DA, Klaehn LD, Kramer AM, Holmes JM. Comparison of methods for measuring cyclodeviation. Am J Ophthalmol. 2020;224:332–342. 5. Marsh JD, Durkin MW, Hack AE, Markowitz BB, Cheeseman EW. Accuracy of double Maddox rod with induced hypertropia in normal subjects. Am Orthopt J. 2014;64:76–80. 6. Hunter DG, Guyton DL. Vertical location of the corneal light reflex in strabismus photography. Arch Ophthalmol. 1998;116:767–771. 7. Maor R, Holland J, Tailor V, Banteka M, Khandelwal P, Glaze S, Barnard S, Yashiv Y, Dahlmann-Noor AH. Rate of strabismus detection on digital photographs increases by using off-center near target. J Pediatr Ophthalmol Strabismus. 2017;54:90–96. 8. Phanphruk W, Liu Y, Morley K, Gavin J, Shah AS, Hunter DG. Validation of StrabisPIX, a mobile application for home measurement of ocular alignment. Translational Vis Sci Tech. 2019;8:9. 9. Luo G, Pundlik S, Tomasi M, Houston K. Using an automated Hirschberg test app to evaluate ocular alignment. J Vis Exp. 2020;157. doi: 10.3791/60908. 10. Pundlik S, Tomasi M, Liu R, Houston K, Luo G. Development and preliminary evaluation of a smartphone app for measuring eye alignment. Translational Vis Sci Tech. 2019;8:19. 11. Weber KP, Rappoport D, Dysli M, Schmückle Meier T, Marks GB, Bockisch CJ, Landau K, MacDougall HG. Strabismus measurements with novel video goggles. Ophthalmology. 2017;124:1849–1856. Bindiganavale et al: J Neuro-Ophthalmol 2022; 42: e248-e253 12. Simiera J, Ordon AJ, Loba P. Objective cyclodeviation measurement in normal subjects by means of Cyclocheck(Ò) application. Eur J Ophthalmol. 2021;31:704–708. 13. Liebermann L, Leske DA, Hatt SR, Holmes JM. Test-retest variability of cyclodeviations measured using the double Maddox rod test. J Aapos. 2018;22:146–148.e1. 14. Lee M, Koller M, Schweigert A, Marrill K, May L, Ismaiel N, Lor M, Mcclelland C, Lee S. Measuring torsion with a mobile app compared to double Maddox rod. North American NeuroOphthalmology Society Annual Meeting; Amelia Island, FL2020. 15. Yamadera K, Ishikawa H, Imai A, Okamoto M, Kimura A, Mimura O, Gomi F. A novel method for evaluation of ocular torsion angle by optical coherence tomography. Transl Vis Sci Technol. 2020;9:27. 16. Murata N, Toda H, Amaki H, Suzuki K, Nagai Y, Omiya Y, Kurashima T, Udagawa S, Ohkubo S. Assessment of objective ocular cyclodeviation under monocular condition and binocular condition using fundus photography. BMJ Open Ophthalmol. 2020;5:e000595. 17. Simons K, Arnoldi K, Brown MH. Color dissociation artifacts in double Maddox rod cyclodeviation testing. Ophthalmology. 1994;101:1897–1901. 18. Bodnar ZM, Neimkin M, Holds JB. Automated ptosis measurements from facial photographs. JAMA Ophthalmol. 2016;134:146–150. 19. Guyton DL. Clinical assessment of ocular torsion. Am Orthoptic J. 1983;33:7–15. e253 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited.