Abstract
Purpose:
To assess the performance of a cell phone based refracting device (NET). We identify the efficacy of NET confidence values by comparing refractive differences between NET and subjective refraction (SR).
Methods:
NET retrofits a high-resolution mobile phone (Samsung S4) by adding a pinhole mask and a lens onto the display. The device is bi-ocular and both eyes are constantly fogged (+6D). The subject aligns red and green lines with the translation on the screen proportional to refractive error. Besides refraction, the device computes a confidence value (from 0, poor, to 1, good). The hypothesis is that results under 0.4 require a retest. 89 subjects (mean + SD age 28.41 +/- 6.95) underwent SR and NET refraction. Auto-Refractor (AR) was used as a starting point for SR. Subjects were split according to NET's confidence value and compared. Group A (n=70) included high confidence readings (>0.4) and Group B (n=19) low confidence readings (<0.4). Subject's refractive error ranged from +1.50 to -8.25D (mean + SD refraction -1.37 +/- 1.90D). Exclusion criteria included: amblyopia, non-refractory pathologies, and 4 subjects in which NET yielded a better refraction than SR.
Results:
For all 89 subjects, the mean + SD Vector Power Difference (VDD) of NET-SR is 1.20 +/- 1.24D. Group A's mean + SD VDD is 0.95 +/- 0.60D (corr=95%). In comparison, the mean + SD VDD of AR-SR is 0.51 +/- 0.45D (note: SR is biased towards AR). NET, AR and SR yielded 20/25 or better vision on 92%, 96% and 98% respectively. Mean + SD visual acuity difference of NET-SR is -0.07 +/- 0.08 logMar. Group B's mean + SD VDD of NET-SR is 2.12 +/- 2.18D (corr=65%). NET, AR and SR yielded 20/25 or better for 61%, 72% and 89% respectively. Linear regression of spherical equivalent presented a slope of 0.86 and a y-intercept of -0.63D. Mean + SD of PD difference of NET-AR is -0.76 +/- 1.88mm.
Conclusions:
NET's confidence value can be used as a predictor for insufficient accuracy, triggering a retest or a discarding method. The low confidence group included the only 2 subjects with high-order aberrations and 5 subjects that required translation (a known language barrier for NET, since subjects need to understand the procedures). Group A included all subjects that could not reach 20/20 vision on the SR. The results show that NET has potential to be a used as an effective tool for rapidly estimating refractive errors and interpupillary distance measurements.