The Contrast Sensitivity Function (CSF) has been generally accepted as a better predictor of visual performance than high contrast Visual Acuity (VA). In fact, VA is usually considered as a measure of the clarity of vision, and it basically depends on the finest detail that an eye can resolve. On the other hand, the CSF is a more complete metric since it is a measure of the threshold contrast needed to see spatially varying stimuli.1 Indeed, the CSF is nowadays considered a routine clinical tool in optical quality assessment of the eye2,3 and in eye disease detection (e.g., cataracts,4 optic nerve pathologies,5,6 retinitis pigmentosa,7,8 glaucoma,9,10 etc.).

When CSF testing was initially introduced in clinical practice and clinical research, tests usually consisted of computer-generated visual images. However, those devices were typically costly, they needed a calibration and normative data that were not readily available. Consequently, chart-based methods for assessing CSF were developed in the early 1980s.11

In clinical practice, Contrast Sensitivity (CS) is generally measured by means of optotypes of different contrast, such as Pelli-Robson Chart12 or by means of sinusoidal gratings of different spatial frequency.13 The main difference between them is that an optotype contains a wide range of spatial frequencies whose relative weights depend on the letter and its size, while a sinusoidal grating evaluates the response of the visual system to a single spatial frequency.14

Today, the most popular commercial tests for measuring CSF by means of sinusoidal gratings are: Functional Acuity Contrast Test (FACT),15 and the Vector Vision CSV-1000 (VectorVision, Greenville, OH).16 These tests commonly use 9 patches for each spatial frequency but they differ in: the specific spatial frequencies evaluated, in the step contrast sizes and ranges, and in the psychophysical method to achieve the threshold.

Since tablets appeared, new applications (apps) have been proposed in the ophthalmology and optometry practice.17,18 The great advantages of these devices are that they offer the possibility to standardize vision screenings, and since there are many common models which share characteristics such as screen chromaticity and resolution, the chromatic properties of such devices might be assumed to be nearly the same. The aim of this study is to introduce a new App, called ClinicCSF19 to measure CSF with tablet devices and to compare it with other commercial device: the Optec Visual Function Analyzer (Stereooptical, Chicago)20 that contains the FACT.

No statistical differences were found in age between both groups of subjects (p=0.064) and median monocular visual acuities were 0logMAR (range, −0.2 to 0.3) in the Group 1 and 0logMAR (range, −0.2 to 0.2) in the Group 2 (p=0.570).

Median CS and range scores obtained at each spatial frequency are summarized for both groups in Table 1 and graphically represented by means of box plot whiskers in Fig. 3. The CSF median values were generally higher for the ClinicCSF.v1 than for the FACT test in Group 1 (Fig. 3A); the differences were statistically significant (p<0.001) for all frequencies except for 3cpd. However, the ClinicCSF.v2 gave similar scores than the FACT for all the evaluated spatial frequencies in subjects from Group 2 (p>0.05) (Fig. 3B).

Figure 3.

Box plot diagrams showing median contrast sensitivity values. (A) ClinicCSF.v1 and Functional Acuity Contrast Test (FACT) measured in Group 1 of subjects (B) ClinicCSF.v2 and FACT measured in Group 2 of subjects.

(0.18MB).

As can be seen in Table 2, both groups gave similar contrast sensitivities when the same FACT test was used to perform the exam (p>0.05). Even though both groups reported similar CSs with the FACT test (Fig. 4A), there existed significant differences between groups when they were measured with different versions of the ClinicCSF for spatial frequencies of 6 and 18 cpd (p<0.05) (Fig. 4B).

Figure 4.

Box plot diagrams showing median contrast sensitivity values. (A) Group 1 and Group 2 of subjects measured with the Functional Acuity Contrast Test (FACT). (B) Group 1 and Group 2 of subjects measured with the ClinicCSF.v1 and ClinicCSF.v2, respectively.

(0.18MB).

In Fig. 5, Bland–Altman plots are represented by means of the difference between the two methods [ClinicCSF.v1−FACT] against the mean [(ClinicCSF.v1+FACT)/2]. The same representation was also done for the ClinicCSF.v2 and the FACT in Fig. 6 by a direct comparison of each one of the spatial frequencies. It can be seen that the ClinicCSF.v1 overestimated the CS with respect to the FACT, and this overestimation was not found with the ClinicCSF.v2 (continuous lines in Figs. 5 and 6). It should also be noted that although we found less differences between the ClinicCSF.v2 vs. FACT than between the ClinicCSF.v1 vs. FACT, narrower agreement limits were obtained with the staircase psychophysical method of the ClinicCSF.v1; mainly for spatial frequencies of 6, 12, and 18cpd (dashed lines in Figs. 5 and 6).

Figure 5.

Bland–Altman plots. CS difference between methods versus mean of CS scores measured with FACT and ClinicCSF.v1 for spatial frequencies of 3cpd (top-left), 6cpd (top-right), 12cpd (bottom-left), and 18cpd (bottom-right). The solid lines represent the mean difference between the two instruments and the dashed lines correspond to the 95% confidence interval (mean±1.96SD). A linear fit was done for statistically significant correlations (p<0.05) and the Pearson coefficients (r) are reported for 3cpd (r=0.37), 12cpd (r=0.56), and 18cpd (r=0.34).

(0.34MB).

Figure 6.

Bland–Altman plots. CS difference between methods versus mean of CS scores measured with FACT and ClinicCSF.v2 for spatial frequencies of 3cpd (top-left), 6cpd (top-right), 12cpd (bottom-left), and 18cpd (bottom-right). The solid lines represent the mean difference between the two instruments and the dashed lines correspond to the 95% confidence interval (mean±1.96SD). A linear fit was done for statistically significant correlations (p<0.05) and the Pearson coefficients (r) are reported for 6cpd (r=0.44).

(0.31MB).

Correlations between differences versus mean scores obtained with tests were analyzed by the Pearson coefficient (r) and represented in the Bland–Altman plots by linear least squares fitting in case of being statistically significant (p<0.05). Therefore as can be seen in Fig. 5 for the comparison between ClinicCSF.v1 and FACT, the regression lines show positive correlations with the increment of mean CS for 3, 12 and 18cpd (r=0.37, 0.56 and 0.34, respectively). On the other hand, the correlation was significant only for 6cpd (r=0.44) in the comparison between ClinicCSF.v2 and FACT (Fig. 6).

The aim of this study was to present a new iPad app to measure CSF. Two versions (ClinicCSF.v1 and .v2) have been proposed and tested in comparison with other commercial test (FACT). Although two different groups of subjects were used in the evaluation of each one of the ClinicCSF versions, no statistical differences in visual acuity and age were found between both groups. Special attention was paid on age of participants considering that CSF could be influenced by this variable26. In fact, some commercially available tests, such as the CSV1000, have different normative ranks depending on age of the subjects.27

We found significant differences between ClinicCSF.v1 and FACT for all spatial frequencies except for 3cpd (Table 1). This lack of agreement can be attributed firstly to the fact that each test measures different CS levels (Fig. 2) and secondly to the different psychophysical method employed in each version. Other comparative studies also found discrepancies due to the similar reasons. Franco et al.13 compared the VCTS-6500 and the CSV-1000 and found mean differences of 0.30, 0.20, 0.08 and 0.18 for spatial frequencies of 3, 6, 12 and 18cpd, respectively, the differences being statistically significant for all spatial frequencies. Such differences are even higher than those obtained in the present study except for 12cpd (Fig. 5). As expected, the differences between the ClinicCSF.v2 and the FACT were very much lower due to both tests having the same CS levels and use the same thresholding technique (Fig. 6), unlike the ClinicCSF.v1 and FACT (Fig. 5).

Other researchers have sounded a note of caution with regard to the comparison of the same test with different configurations. For instance, FACT differs from the previous Vistech version in several characteristics: using smaller step sizes, a 3 alternative forced choice, “blurred” grating patch edges with the gratings smoothed, and a larger patch size to increase number of cycles at low spatial frequencies. Pesudovs et al.28 attributed dissimilar results between Vistech and FACT to the fact that this new version uses smaller step sizes with the same number of steps, and thus ranges of measurement are also smaller. As a consequence, they reported a ceiling effect in post-LASIK patients and a floor effect in cataract patients. Furthermore, Hitchcock et al.29 showed that not only step sizes could have influence on the CS since they found that although contrast levels were the same, results could be different depending on the way tests were presented.

Positive correlations between mean differences and the average of CS were found with the ClinicCSF.v1 for three spatial frequencies (Fig. 5). These correlations represent that the CS with ClinicCSF.v1 was higher than FACT mainly in subjects with poorer CS. Significance of this correlation disappeared with the ClinicCSF.v2 and curiously the only significant correlation between ClinicCSF.v2 and FACT was discovered at 6cpd (Fig. 6), spatial frequency that was not significant with ClinicCSF.v1. We consider that these correlations might mainly be due to step sizes with the same number of steps as was mentioned before when we discussed the conclusions of Pesudovs et al.28 This would appear to indicate that ceiling and floor effects might also appear for the comparison between ClinicCSF.v1 and FACT in cataract and refractive surgery patients.

We also found that mean differences confidence intervals were highly influenced by the psychophysical method used to achieve the CS threshold. Confidence intervals of the Bland–Altman plots for differences between ClinicCSF.v1 and FACT were narrowed by using a staircase method (Fig. 5). This fact underlines the importance of including a psychophysical method in iPad based screening tests instead of using it simply as an illuminated screen.30,31

The biggest differences in the CSFs between groups were found when we changed some properties of the test design (Table 2). This demonstrates just how important is it to use the same test in the comparison between groups of subjects. Consequently, clinical results in studies which implement different CS tests might also differ due to the configuration of tests used. In fact, the discrepancies in the comparison of several CS tests have been widely studied, mainly in order to obtain normative data for contrast sensitivity functions.32

One limitation of this study is that two different groups of subjects were used to compare each one of the ClinicCSF versions with the FACT. The reason is that ClinicCSF.v1 was first designed and evaluated clinically with one group of subjects. Lately ClinicCSF.v2 was developed as an improved version of ClinicCSF.v1 and it was not possible to measure the same group of subjects. A better statistical analysis of variance could have been done if we had measured the CSF with the three tests in the same group of subjects. Other limitation of our current proposal is related to the maximum brightness configuration of the iPad that might produce a glare effect in some patients, and a possible post-image after each answer. This issue should be considered in future versions of the app.

Our work led us to conclude that ClinicCSF app, designed for a given tablet device, can give similar results than FACT in CSF measurement in a normal population. Further experiments using the ClinicCSF app in different versions of the same device are required in order to extrapolate our results. The findings might not be generalized to all the iPad screens because the uniformity of colorimetric and photometric properties among all iPads has not yet been proven. We think that our new test could be useful to popularize the CSF measurement in centers that do not usually perform it, due to the high cost of current commercial equipment. Further experimental investigations are also needed to estimate normative ranges and ROC curves.

Rodríguez-Vallejo, M. has designed and programmed the ClinicCSF app which he currently distributes by the Apple Store with his own developer account. The other authors report no conflicts of interest and have no proprietary interest in any of the materials mentioned in this article.