Effects of brightness variations on a smartphone-based version of Radner reading charts

Abstract

Objective

The purpose of this study was to evaluate the equivalence of smartphone-based measurements of near visual acuity under different screen brightness conditions with a standard near visual acuity test.

Methods

On a sample of 85 participants, we have evaluated near visual acuity with a smartphone-based version of the Radner reading chart at three distinct screen brightness levels. Results have been compared with those obtained with classical Radner paper charts.

Results

We have found that, when a sufficient screen brightness is employed, the smartphone-based version of the Radner reading chart produces results that are in line with the paper Radner charts while low brightness levels lead to a significant underestimation of reading acuities. This result was consistent across different refractive conditions.

Conclusions

In conclusion, we have shown that handheld devices, such as smartphones, can be potentially exploited for remote measurements of near visual acuity provided a correct control of brightness screen is employed.

Introduction

Near vision activities dominates our lives and are practiced in several environments such as education, work, and spare time. It is estimated that children spend much of their academic time in near vision activities and most of the adults perform jobs requiring a large proportion of time employed in near tasks [1].

To note, the advent of digital era has critically changed our near vision habits. The large-scale introduction of electronic devices, on the one hand, has increased the percentage of time spent in near vision activities and, on the other hand, has modified brightness and contrast parameters to which we are exposed while engaged in near vision tasks providing users with a sharply different visual experience. Indeed, not only the extensive use of these devices has been associated with symptoms of visual discomfort [2–4] but also with modification of various physiological parameters of vision. For instance, Miranda and colleagues have reported a 20% pupil size reduction while reading on electronic devices compared to print materials, a result likely due to the higher level of brightness emitted by the screen [5].

On this scenario, it has become increasingly important that optometric examinations could be progressively implemented also on such devices. In the growing field of teleoptometry, various visual tests have been adapted to be performed on electronic devices. For instance, distance visual acuity can now be easily self-administered with a reasonably low test-retest variability by means of a computer [6, 7] and an increasing number of smartphone’s applications have been developed to monitor and assess various visual parameters such as reading distance and pupil diameter [8, 9]. However, given the effect produced by brightness and contrast on visual efficiency [10], it is extremely important to clearly parametrize the effects of such variations. For instance, a previous study found that a popular optometric application produced a systematic overestimation of visual acuity which is likely attributable to the enhanced brightness and contrast levels used by such application [11].

In the current study we have developed a smartphone-based version of the Italian Radner reading chart [12, 13]. The scope was twofold. On the one hand, we have aimed at testing the feasibility of assessing near visual acuity on electronic devices. On the other hand, we have aimed at quantifying the influence of varying brightness levels on reading acuities. This manipulation has been obtained by controlled variations of background brightness level of the smartphone’s screen.

Materials and methods

Participants

Ninety participants with no subjectively reported ocular pathologies and distance visual acuity equal or greater than 0.1 LogMAR have taken part in the study. Sample size has been determined in agreement with previous studies having similar aims and using similar methodologies [11, 14].

All participants have provided written informed consent. The research has been approved by the local ethics committee (“Comitato Etico Area Vasta Centro”; protocol n. 22792) and has been conducted in accordance with the principles of the Declaration of Helsinki.

Apparatus and Stimuli

The experiment has been conducted in a sound attenuated room with participants seated in front of a 5.8 inches smartphone OLED screen with a resolution of 1440 × 2960 pixels. The smartphone has been arranged to form an angle that simulated a typical viewing condition with distance kept fixed at 40 cm by means of a smartphone holder and a chinrest (see Fig. 1).

Fig. 1. Schematic illustration of the apparatus used.

Participants were asked to read a smartphone-based version of the Italian Radner charts with varying levels of background brightness. Smartphone was arranged to simulate a typical viewing position and distance was kept fixed at 40 cm by a smartphone holder and a chinrest. Figure depicts one of the tables used in the experiment which states “The following morning Adele called the carpenter who was supposed to cut her some new shelves”.

Room’s illumination has been kept fixed at 300 lux by using only artificial lights while smartphone’s screen brightness has been varied across three distinct levels (High: 37 cd/m², Medium: 16 cd/m², Low: 8 cd/m²) by means of a dedicated application enabling a customizable control of this parameter [15]. We have chosen to vary smartphone’s screen brightness rather than characters’ contrast, as usually employed in clinical practice [16], to mimic natural reading conditions on a handheld device in which variation are employed on the brightness of the background while characters’ contrast is kept mostly unchanged.

Participants have been presented with a smartphone-based version of the Italian Radner charts [13, 17] while wearing their habitual near reading correction (if any). The range of reading acuity tested spanned from 0.9 to −0.1 LogRAD (i.e., 11 sentences). The selection of such specific range was based on physical and resolution constraints of the smartphone employed in the current study (Samsung Galaxy S8) since lower acuities would have needed a larger display while higher acuities would have needed a higher resolution. Each sentence has been displayed as a single image. On a binocular vision setting, participants have been instructed to read sentence aloud and to swipe the next image until reaching their individual reading threshold. Each of the three available charts [13] has been assigned to one of the three brightness level and the assignment has been randomised across participants (simple randomization). This procedure has allowed us to minimize potential across-charts differences in readability while also minimizing the influence of any prior knowledge of the sentence to be read.

A measure of reading acuity has also been obtained on paper Radner charts to check for consistency with results obtained on smartphone-based Radner charts. In a within-subject design, each participant has been tested both on the three varying brightness smartphone-based Radner reading charts and on the paper Radner chart.

Voice has been recorded to extract reading acuity defined as the LogRAD value at which at least 75% of the words have been correctly read. This value has been corrected for reading errors which were defined as the number of misspelled syllables (i.e, “n. errors”) multiplied by the LogRAD weight assigned to each syllable (i.e., 0.003). LogRAD weight was calculated as the between sentence LogRAD variation (i.e., 0.1 LogRAD) divided by the average number of syllables presented on each sentence (i.e., ~30 syllables).

$$ReadingAcuity=\log RAD+0.003\times n.errors$$

Refractive Measures

Measures of objective ocular refraction (Autorefractometer WAM 5500 Grand Seiko Co.Ltd) and habitual near correction (Essilor ALM 800 Digital Lensmeter) were also obtained. These measures have been used to derive individual accommodation amplitude which was calculated as follows:

$$AccommodationAmplitude=\frac{1}{PPA}+AVGSphEq-\left(AVGCorrSphEq+Add\right)$$

Where PPA was the proximal point of accommodation (in meters), AVGSphEq was the average spherical equivalent (i.e., RE + LE / 2) obtained from the autorefractometer, AVGCorrSphEq was the average spherical equivalent of the correction in use (if any) and Add was the addition of the correction in use (if any).

Objective ocular refraction has also been used to split the sample into three sub-groups. More specifically, participants having a spherical equivalent lower (or equal) than −0.50 D have been assigned to the myopic group, participants with a spherical equivalent larger (or equal) than +0.50 D have been assigned to the hyperopic group while participants with a spherical equivalent between −0.50 and +0.50 have been assigned to emmetropic group.

Statistical analyses

Bland-Altman plots were used to assess the assumptions necessary for computation of limits of agreement for measures of visual acuity made using different methods. Changes of visual acuities produced by the four reading conditions (Paper, High, Medium and Low) have been evaluated through non-parametric Friedman tests and Conover’s post-hoc comparisons have been adjusted using the Bonferroni correction. All statistical analyses have been conducted using JASP v0.14.1.0.

Results

Five participants have been excluded because of insufficient accommodation amplitude whose threshold was set at 2.5 D to meet the minimum accommodation necessary to accomplish tasks at the viewing distance employed in the current study (i.e., 40 cm). The final sample then comprised 85 participants (44 males and 41 females; all Caucasian; mean age: 28.2 years; median age: 26 years; std: 9.3 years; age range: 18–59).

Bland-Altman plot have showed the 95% limits of agreement range from −0.11 to 0.12 LogRAD and a constant across measurements variability of the differences between methods (Fig. 2A). Nevertheless, for Medium and Low brightness levels the 95% limits of agreements ranged from −0.07 to 0.15 (Medium Brightness; Fig. 2B) and from −0.07 to 0.20 (Low Brightness; Fig. 2C) revealing a constant underestimation of near visual acuities as compared to the paper Radner chart.

Fig. 2. Bland-Altman plots.

Bland-Altman plots comparing differences between AV obtained with Paper Radner and High Brightness (A), Medium Brightness (B) and Low Brightness (C) Smartphone Radner.

Results of the Friedman Test have indicated that there was a differential rank ordered for the four reading conditions (Paper, High, Medium and Low; p < 0.001) which has been explained by lower reading acuities for Medium and Low brightness levels as compared to both High and Paper (all p < 0.001) which did not differ from each other (see Fig. 3). This result further corroborates the idea that, when sufficiently high brightness levels are employed, the smartphone-based version of Radner chart produces results that are in line with the classical paper Radner charts.

Fig. 3. Reading acuities.

 Reading acuities obtained in paper Radner reading chart (Paper) and in the three brightness levels (High, Medium, Low) used in the smartphone-based Radner chart. Error bars represent errors of the means.

As an exploratory analysis, we then tested the influence of initial refractive condition on reading acuities obtained at the three smartphone brightness levels employed. This experimental question was motivated by previous works showing that different refractive errors produce different pupil dynamics under low-luminance conditions [18, 19].

To test for any evidence of such relationship, the sample of participants has been then split into three subgroups based on spherical equivalent (see Refractive Measures section for further details). This procedure has led to the selection of 17 hyperopes (8 males and 9 females; mean age: 33.17 years; std: 12.4 years), 23 emmetropes (12 males and 11 females; mean age: 25.61 years; std: 6.4 years) and 45 myopes (21 males and 24 females; mean age: 27.58 years; std: 8.7 years). Results have revealed lowered reading acuities as a function of brightness while very similar acuities between the Paper and the High brightness conditions (see Table 1 for descriptive data). In addition, Hyperopes have showed lower reading acuities as compared to Emmetropes and Myopes.

Table 1.

Median visual acuities and Interquartile ranges (IQR; i.e., difference between the 75^th and 25^th percentile) expressed in LogRAD within the four experimental conditions (i.e., Paper Radner, High, Medium and Low -Brightness smartphone Radner) and the three experimental groups.

	Paper	High	Medium	Low
Hyperopes	Median: 0.00 IQR: 0.24	Median: 0.006 IQR: 0.22	Median: 0.006 IQR: 0.25	Median: 0.1 IQR: 0.30	Median: 0.086 IQR: 0.24
Emmetropes	Median: -0.091 IQR: 0.10	Median: -0.094 IQR: 0.10	Median: 0.000 IQR: 0.09	Median: 0.000 IQR: 0.10	Median: -0.035 IQR: 0.10
Myopes	Median: 0.000 IQR: 0.11	Median: 0.000 IQR: 0.11	Median: 0.000 IQR: 0.10	Median: 0.006 IQR: 0.10	Median: 0.003 IQR: 0.11
	Median: 0.000 IQR: 0.11	Median: 0.000 IQR: 0.10	Median: 0.000 IQR: 0.12	Median: 0.006 IQR: 0.10

This result suggests that, despite average difference in visual acuities, the three groups (i.e., hyperopes, emmetropes and myopes) have had a similar descending trend for visual acuities measured in the high, medium and low brightness conditions (Fig. 4).

Fig. 4. Reading acuities across different ametropic groups.

Reading acuities obtained with the High, Medium and Low brightness of the smartphone-based version of the Radner chart plotted separately for the group of Hyperopes (red plot), Emmetropes (green plot) and Myopes (blue plot). Shaded areas represent SEM.

Discussion

In the current study we have developed a smartphone-based version of the Radner reading charts. The aim was twofold. On the one hand we have tested the reproducibility of a standard near visual acuity test to be used on smartphone. On the other hand, we have exploited the in-built screen brightness modulation to investigate the effects produced by such modification on reading acuity.

Our results indicate that reading acuity obtained with the high brightness smartphone-based version of the Radner reading charts are in line with those obtained with paper Radner charts. Bland-Altman plots also revealed a good agreement between the two measures with a mean difference that was extremely close to zero. Nevertheless, with lower brightness values (i.e., 16 cd/m² and 8 cd/m²) the smartphone-based version of the Radner chart underestimated reading acuities. Bland-Altman plots have showed a significant disparity with an underestimation that was close to 0.04 LogRAD for the Medium brightness chart and close to 0.07 LogRAD for the Low brightness chart. Likewise, statistical analyses have showed no difference in visual acuities between paper chart and the high brightness smartphone-based chart while significant differences have emerged with both medium and low brightness charts.

These results are likely explained by brightness-mediated variation in pupil size. It is well established that pupil changes its diameter in response to changes in the incoming brightness. The effect of smartphone usage on such parameter has been described by Miranda and colleagues [5] who reported a 20% decrease in pupil diameter when participants were engaged on a reading task on their smartphone as compared to when they were reading a paper book. A decrease in pupil diameter produces, in turn, an increased in the depth of focus which contributes to the luminance-mediated visual acuity enhancement [20, 21]. In this view, the same visual task employed on a bright smartphone screen would, theoretically, produce higher visual acuity levels. This is exactly what has been found by Tofigh and colleagues who tested a popular smartphone application in estimating near visual acuities by comparing results with classical Snellen near vision charts [11]. The study has employed a smartphone in full brightness mode and showed a consistent overestimation of visual acuities of around 0.1 LogMAR. Our results, complement these findings showing that when brightness level is correctly adjusted, near visual acuity can be accurately estimated even on electronic devices. On the contrary, an excessive dampened smartphone brightness produces the opposite effect, that is an underestimation of visual acuities. To note, we here report statistical differences between different experimental conditions showing reduced visual acuities as screen brightness is dampened. As expected, the largest difference has been evident between the High and the Low brightness conditions where we have highlighted a consistent reduction in visual acuity. Previous evidence have reported that the coefficient of repeatability of near vision reading acuity tests is around 0.05 LogMAR units [22]. In this view, we can speculate that the transition from High to Low brightness levels has produced both statistically and clinically relevant changes in visual acuities (i.e., ~0.07 LogRAD) while the transition from High to Medium brightness levels has produced statistically relevant changes that has to be considered negligible from the clinical point of view (i.e., ~0.04 LogRAD).

For exploratory purposes, we have also examined whether different refractive conditions could be differently influenced by brightness variations. Previous studies have reported evidence that hyperopes exhibit smaller pupil diameters under low-luminance conditions [18, 19]. Small pupil diameter can lower retinal illuminance producing a decrease in contrast sensitivity which, in turn, can lead to impaired visual performances at low light levels [23]. We have not found any evidence of such trend as all the ametropic groups have shown a similar trend in terms of visual acuity decrements to changes in brightness levels. However, we acknowledge that the current study was not methodologically developed to fully investigate this point and further studies with the use of dedicated paradigms and a more balanced across groups participants recruitments are needed. For this reason, we want to stress that results arising from our exploratory analysis need to be considered preliminary results deserving a more in-depth hypothesis testing.

To conclude, in the current work we have showed that visual parameters such as reading acuity can be reliably estimated on electronic devices. This result pairs with previous evidence showing the potential offered by smartphone-based measurements in estimating users’ accommodative condition through a simple smartphone-based data acquisition of habitual viewing distance and character size [14, 24, 25]. The importance of proving reproducibility of a standard near visual acuity test on smartphones dwells in the possibility to employ remote measurements providing inferences of ocular refractive status from users’ visual performances and habits. The use of daily employed devices such as smartphone and/or tablet has the potential to provide an easy way to characterize visual performances at varying conditions in a daily-life environment.

Summary

What was known before:

• An increasing number of optometric examinations have been adapted to be performed remotely by means of digital devices Previous evidence showed smartphones-based measurements of visual acuity lead to an overestimation of such parameter which is probably due to the brightness level emitted by the screen (Tofigh et al. 2015, Eye)

What this study adds:

• We developed a smartphone-based version of the Radner near vision chart. We showed that smartphones can be exploited for accurate measurements of near visual acuity provided a correct control of brightness screen is employed

Author contributions

PAG: Project Administration, Conceptualization, Methodology, Data Collection, Data Analysis, Writing; MG: Supervision, Review & Editing; LB: Supervision, Methodology, Writing, Review & Editing.

Funding

The study was conducted with the contribution of the researcher Paolo Antonino Grasso with a research contract co-funded by the European Union - PON Research and Innovation 2014–2020 in accordance with Article 24, paragraph 3a, of Law No. 240 of December 30, 2010, as amended, and Ministerial Decree No. 1062 of August 10, 2021.

Data availability

The raw behavioral data have been deposited at Zenodo repository and are publicly available at the following address: 10.5281/zenodo.8158654.

Competing interests

The authors declare no competing interests.