Luminance contrast preferences of people with a vision impairment for elements in the built environment

Abstract

BACKGROUND:

Standards for building elements recommend a minimum luminance contrast of 30%. The basis of this value and the metric originally used is not known.

OBJECTIVE:

To begin to provide an evidence base for the specification of minimum contrast in building elements.

METHODS:

Subjects with and without a vision impairment were characterized by visual acuity, contrast sensitivity, visual fields and color vision. On an iPad they rated the visibility, as a function of contrast, of simulated door frames, door handles, light switches and stair nosings as “not visible at all”, “poorly visible”, “easily visible” and “extremely easily visible”.

RESULTS:

The contrasts for each level of visibility were highly correlated with visual acuity and contrast sensitivity. A principal component analysis also verified the importance of visual acuity, contrast sensitivity and visual fields in rating visibility of simulations of building elements. The required contrast for door handles, light switches and stair nosings to attain the same ratings of visibility were very similar but less contrast was required for door frames.

CONCLUSIONS:

30% Michelson contrast for building elements renders building elements only poorly visible for those with severe vision impairments. 65% luminance contrast is necessary for all elements to be “easily visible”. Some increase (not a decrease) on the present 30% requirement and encouragement to exceed this requirement would seem appropriate. The use of simulated objects facilitates a systematic examination of the effect of contrast, but the applicability of the results to real-life remains to be demonstrated.

Keywords

Accessibility standards safe navigation visibility assessment visual detection universal design

1 Introduction

There are several issues in the provision of a visual environment with sufficient contrast for people with a vision impairment. These have been expressed as [10]:

Several contrast calculation methods are used, the relationship between them is not straightforward and their appropriateness has not been addressed;

Several different contrast calculation methods are used, but a 30% requirement is frequently set;

The origin of the 30% and the contrast measurement scale in which it is stated [10] are not known; and

Methods of measurement are ill-defined in the standards and give rise to different answers when the contrasting elements have fine color or physical texture, are translucent or are highly glossy (metallic being the highest form of gloss).

In a previous study we addressed the first two matters [10] and in this study we look at the third. The requirement of a minimum luminance contrast of 30% is a consistent requirement for building elements in standards for accessibility and integrated or tile format tactile ground surface indicators [25]. However, the method of calculating luminance contrast varies between standards so that the consistency of the 30% requirement does not mean that the requirements are all met with the same color combinations [25]. A basis for this, frequently used, minimum value of contrast of 30% cannot be found in the literature, even in a history of the development of the requirement [14]. There is some basis for a minimum Michelson contrast of 30% found in the work of Legge [23, 28] for reading at half peak speed. However, the subjects were extremely idiosyncratic, one needed a Michelson contrast as low as 10% before their peak reading speed was halved while another’s speed was halved at 80% contrast. There is also uncertainty in the validity of extrapolating the task of reading text, not signs, that contain a combination of high and low spatial frequencies, when word shape recognition is considered, with the task of detecting edges of larger objects such as integrated or tile format tactile ground surface indicators or contrast strip [24]. Another study, using tactile ground surface indicators, also showed extreme individual variations [25]. The issues of specifying contrast in signs for those with a vision impairment is additionally complicated by the effects of character size and font type [4] and the spatial area that requires the luminance contrast.

Building standards require luminance contrast for a range of building elements of varying shapes and sizes. However, a ubiquitous 30% luminance contrast prevails and provides no recognition of visual detectability by people with varying vision abilities and what is an effective size of the element area that aims to achieve appropriate level of luminance contrast.

Previous studies have found high levels of individual variability in required contrast [25], which may be due to the high numbers of experimental variables making carrying out of full-size scale experiments including lighting, weather and color of elements very complex. In this study, the number of experimental variables will be reduced by using a simulation of the building elements, with the expectation that the variability of the measures of visibility could be reduced and refined.

2 Objective

In this study, we will look at this third issue by examining four building elements and assessing the visibility of these elements as a function of contrast for people with normal vision and vision disabilities. These elements are door frames, door handles, light switches and stair nosings. The aim is to investigate the effects of luminance contrast while holding all other variables, most notably illumination, angular size (being related to viewing distance) and chromatic contrast constant.

3 Materials and method

3.1 Subjects

The participants were divided into two groups: those with normal vision and those with a visual impairment.

Vision impairment was defined according to definition of the World Health Organization (WHO). At the time of the study, this was defined as visual acuity (VA) of poorer than 0.5 logMAR (6/18 Snellen VA) in the better eye and visual field of less than 10° from the point of fixation which would correspond to a category of “moderate vision impairment” or worse under the present definitions from the WHO [31].

The normal vision group was defined as participants with a habitual visual acuity of 0.1 logMAR or better (equivalent to Snellen visual acuity of 6/7.5 or 20/25 and smaller; note that more negative logMAR values indicate better vision) and absence of any ocular disease when examined with direct ophthalmoscopy and slit lamp biomicroscopy.

3.2 Sample size

The sample size was calculated from the power and sample size program using the method of Dupont and Plummer [12] based on data from the first 10 participants, five with normal vision and five with vision impairment. This study was planned to model the relationship of preferred contrast (dependent variable) as a function of visual measures (independent variable). The standard deviation of the independent variable, contrast sensitivity, and the minimal contrast preference for building elements was calculated as x-variable and y-variable respectively, which were 0.64 and 19.80 with a slope estimate of –29.8 obtained. The power and sample size program showed that if the true slope of the line obtained by regressing the y-variable against the x-variable is –29.8, a minimum of 18 participants would be necessary for the study to be able to reject the null hypothesis that this slope equals zero with probability (power) 0.8. The type I error probability associated with this test of this null hypothesis is 0.05.

3.3 Ethics approval

This study was approved by the Human Research Ethics Advisory Panel D: Biomedical of the University of New South Wales. All participants were given written and verbal explanations of the procedures and research methods. Written informed consent was obtained before recruiting participants in the study.

3.4 Assessing the independent factors

A brief explanation of each clinical measure has been provided in the Appendix.

3.4.1 Visual acuity

The monocular and binocular presenting visual acuity was assessed using a high contrast Bailey-Lovie logMAR visual acuity chart (see Appendix), “visual acuity”, “visual acuity chart”, “logMAR” and “Bailey-Lovie” at an initial testing distance of 3 m (10 ft) in the low vision clinic of the School of Optometry and Vision Science, University of New South Wales [5, 22]. If vision was too poor to read the letters from 3 m (10 ft), the testing distance was reduced. The World Health Organization (WHO) classification of vision impairment derived from the visual acuity measure for each participant with vision impairment is given in Table 1 [31].

Table 1
Categories of vision impairment with upper and lower limits

Category Presenting distance acuity Snellen (metric and imperial) (logMAR) Central visual field

0 No vision impairment ≥6/12 ≥20/40 ≤0.3 ≥10 degrees

1 Mild vision impairment <6/12 <20/40 >0.3

≥6/18 ≥20/60 ≤0.5

2 Moderate vision impairment ≤6/18 <20/60 >0.5

≥6/60 ≤20/200 ≤1.0

3 Severe vision impairment <6/60 <20/200 >1.0 <10 degrees

≥3/60 ≥20/400 ≤1.3 ≥5 degrees

4 Blindness <3/60 <20/400 >1.3 ≤5 degrees

≥1/60 ≥20/1200 ≤1.8

5 Blindness <1/60 <20/1200 >1.8

Light perception only

6 Blindness No light perception

9 Unspecified or undetermined

Category	Presenting distance acuity Snellen (metric and imperial) (logMAR)	Central visual field
0	No vision impairment	≥6/12	≥20/40	≤0.3	≥10 degrees
1	Mild vision impairment	<6/12	<20/40	>0.3
		≥6/18	≥20/60	≤0.5
2	Moderate vision impairment	≤6/18	<20/60	>0.5
		≥6/60	≤20/200	≤1.0
3	Severe vision impairment	<6/60	<20/200	>1.0	<10 degrees
		≥3/60	≥20/400	≤1.3	≥5 degrees
4	Blindness	<3/60	<20/400	>1.3	≤5 degrees
		≥1/60	≥20/1200	≤1.8
5	Blindness	<1/60	<20/1200	>1.8
		Light perception only
6	Blindness	No light perception
9		Unspecified or undetermined

3.4.2 Contrast sensitivity

Contrast sensitivity was measured using the Mars test. The Mars Letter Contrast Sensitivity test (Mars Perceptrix Corporation, Chappaqua, NY, USA) [26] evaluates the contrast necessary to identify letters with an angular subtense at the approximate peak of the contrast sensitivity function in people with normal vision. It has dimension of 23.0×35.5 cm and comprises 48 letters arranged in eight rows with six letters in each row. Each letter is 1.75 cm high and subtends an angle of 2 degree at the testing distance of 50 cm [3, 11]. The test incorporates contrast decrements of 0.04 log units.

3.4.3 Color vision

Color vision was tested binocularly using the Cambridge Colour Test (Metropsis Trivector, Cambridge Research Systems, Rochester, UK; https://www.crsltd.com/tools-for-vision-science/measuring-visual-functions/cambridge-colour-test/) [27].

This test uses Landolt C-shaped ring as target defined by two test colors to be discriminated on an achromatic background; the chromaticity of the target is varied from the background during testing. The background consists of grouped circles of varied diameters but no spatial structure which act as background noise. The gap in the target “C” subtends 1 degree at the viewing distance of 4 m. The subject needs to recognize and respond the orientation of the gap in the letter “C” by the 4-button infrared box [16, 29]. The chromaticities of the targets are controlled according to CIE LUV system. The test uses a staircase method of threshold for accuracy. This test is easy to perform and can be tested binocularly [29].

Although the standards for contrast in building elements specify luminance and not color contrast, color vision is generally considered to be more fragile than luminance contrast vision and may provide an earlier indicator of eye conditions, for instance [2].

3.4.4 Visual fields

Visual field was measured for each eye by using a Zeiss Humphrey Field Analyzer (Carl Zeiss Meditec, Inc., Oberkochen, Germany; https://www.manualslib.com/manual/1288596/Zeiss-Humphrey-Ii-I-Series.html#manual). The central 24-2 threshold test was used with Swedish Interactive Threshold Algorithm (SITA). It tests in 54 test point within the central 24 degree of visual field [9]. This threshold was chosen as this test is faster than other methods in an effort to prevent tiring our participants.

Binocular field estimation was calculated from the visual field of each individual eye according to a method by Crabb et al. [9]. In binocular viewing, every point in the monocular field of one eye has a corresponding point in the field of other. The greater sensitivity of either eye for any determined point in the visual field was regarded as the sensitivity of that point. The average of these points was calculated as the mean binocular sensitivity of the participant.

3.5 Assessing the dependent variables

An ascending and descending method of limits method was used to identify the range of luminance contrasts for each simulated building element that matched with a rating of (1) not visible, (2) poorly visible, (3) easily visible and (4) extremely easily visible for each participant. The average of the minimum of the estimates of the ranges from the ascending and descending methods became the minimum preferred luminance contrast value for a rating for a participant, which became the dependent values used in the subsequent regression analyses.

The use of the CIE Metric Lightness difference, ΔL* as an appropriate single replacement for the multiple measures of contrast specified in various standards has been proposed [10, 30]. However, for the moment, since Michelson contrast was shown to be the most appropriate of the currently used measures, the data will be expressed in both measures [10].

3.6 Equipment

Simulated pictures of doorframe with light-switch and doorhandle and another picture of stairs with nosings were drawn using Microsoft Paint in grayscale and presented on an iPad Air, with a display of 2048×1536 pixels at a resolution of 264 pixels per inch. The contrast of the doorframe, light-switch, doorhandle against the wall and door and that of the nosing against the steps were created by using different values of RGB. The contrast was achromatic as R = G = B in each element. The dimensions of the doorframe, light-switch, doorhandle and door were sized so that they subtended the same angle in the eye as if they were located at a 6 m (20 ft) viewing distance, when viewed on the iPad screen at a 30 cm (1 ft) viewing distance. The dimensions of the stairs and nosings were designed to replicate eye height above a staircase looking down. The picture shapes were drawn by superimposition of shapes on real photographic images taken from these distances. For example, Australian Standards suggests the standard size of the door frame to accommodate a 2040×920 mm door with a 50 mm width frame or border. The picture of doorframe in the iPad was of dimensions 102×46 mm. The image, presented at 30 cm (1 ft), simulated the view of the doorframe at a viewing distance of 6 m (20 ft).

Samples of the pictures used in the study are given below. Figure 1 represents the sample image of doorframe and light-switch and doorhandle of RGB value 13 (L* = 6) against the wall of RGB value 236 (L* = 94) with a contrast of ΔL* = 88 and Michelson contrast = 98.4%. Similarly, Fig. 2 represents the sample of stair nosings used in the study. In this illustration, the steps and nosing have RGB value of 113 (L* = 49) and 118 (L* = 51) respectively with contrast of ΔL* = 2 and Michelson contrast = 4.5%. Note, the contrast values are given both as CIE ΔL* and Michelson contrast.

Fig. 1

Sample picture of wall with doorframe, doorhandle and light-switch.

Fig. 2

Sample picture of stairs with 75 mm width nosing strips (travelling down).

The stairs and 75 mm width nosing strips were drawn as they appear from the top, that is, as travelling down (See Fig. 2). Studies have found that there is higher risk of falls in stepping down than stepping up [6, 24].

These building elements are relatively smaller examples compared with whole doors and integrated tactile ground surface indicator tiles and represent the more challenging elements for people with a visual impairment.

The luminance of the display was measured as a function of RBG bit count and expressed as a percentage of peak. The CIE metric brightness, L* (being a perceptually uniform scale) was calculated for each contrast [8]. The relationship between RGB bit count and L* was found to be very linear (r = 0.997). The starting point of a mid-tone grey (L* = 50, R = G = B = 116) was set and the set of stimuli developed with increasing contrast in steps of ΔL* = 1, i.e., L* = 51 and 49, 52 and 48, etc.

3.7 Procedures

Written informed consent was obtained from all participants following an explanation of the study. Preliminary examination was done to collect the baseline data which would form the independent variables of visual acuity, contrast sensitivity and visual field. As part of the informed consent process, participants permitted researchers to access clinical data from their clinician regarding diagnosis of ocular or visual system condition.

During the rating experimental tasks, the participants were asked to observe the images of the doorframe, doorhandle, light-switch and stairs and nosings on an iPad at 30 cm with their habitual correction and near addition (if any). The screen was held perpendicular to the line of sight, the room was lit with subdued lighting and it was ensured that there were no visible reflections from the screen to dilute contrast. The stimuli were presented using the method of limits psychophysical technique from highest to lowest contrast and from lowest to highest contrast for each simulated building element. They were asked to rate the visibility of elements in each picture as:

Not visible at all

Poorly visible

Easily visible

Extremely easily visible

The average of the lowest contrast value of each rating in both directions was calculated as the threshold value for that rating.

These tasks were conducted by the same researcher (SM) in the same location under the same conditions under office lighting for different participants, at an optimum viewing angle for the participant that allowed minimal reflections.

3.8 Data analysis

The data were analyzed in Microsoft Excel 2013 and Statistical Package for Social Science (SPSS) version 22 (IBM Corp., Armonk, NY, USA). The dependent variables, contrast of the building elements, was correlated with the independent variables, visual function. Factor analysis was done using principal component analysis to evaluate if dimension reduction of the independent variable vision data could yield further insights; this was followed by linear regression.

4 Results

4.1 Demographics

A total of 28 participants were included in the study (exceeding the 18 indicated by the power calculation), 15 people had normal visual acuity and 13 people had a visual impairment.

The mean age (±standard deviation) of the control group was 28.5±6.0 years (range 20–39 years) whereas of the vision impaired group was 63.6±19.9 years (range 22–83 years). Out of the 28 participants, 12 were male and 16 were female.

4.1.1 Normal control subjects

The visual acuities and contrast sensitivities with age and sex distribution of control group are given in Table 2. All of them have no visual impairment according to the WHO rating [31]. All participants identified all but the last three letters on the Mars test chart and achieved a contrast sensitivity of 1.8.

Table 2
Age, gender and visual acuity of subjects with no vision impairment

Age (years) Gender Visual acuity (logMAR)

24 F 0.04

39 M 0.00

26 M 0.00

37 M –0.12

32 F 0.02

24 F –0.06

27 F –0.12

23 F 0.00

22 M 0.00

31 F –0.06

35 F –0.06

26 M –0.08

20 M 0.00

36 M –0.12

25 F 0.00

Age (years)	Gender	Visual acuity (logMAR)
24	F	0.04
39	M	0.00
26	M	0.00
37	M	–0.12
32	F	0.02
24	F	–0.06
27	F	–0.12
23	F	0.00
22	M	0.00
31	F	–0.06
35	F	–0.06
26	M	–0.08
20	M	0.00
36	M	–0.12
25	F	0.00

4.1.2 Vision impaired subjects

Age-related macular degeneration (four participants) was the most common cause of vision impairment among the vision-impaired participants. Optic atrophy (two participants) and cone dystrophy, Stargardt’s disease, cytomegalovirus retinitis, angle closure glaucoma, central retinal vein occlusion and oculocutaneous albinism were present (one in each of the participants). The characteristics of the vision impaired participants are shown in Table 3. F indicates female, and M indicates male. LogMAR indicates logarithm of the minimum angle of resolution.

Table 3
Age, gender, cause of vision impairment, binocular visual acuity, binocular contrast sensitivity and WHO classification of the vision impairment group in order of increasing age. This participant had visual impairment but binocular vision was less impaired than monocular measures

Age Gender Cause of vision impairment Best corrected visual acuity Contrast sensitivity WHO category

22 F Cone dystrophy 1.38 0.72 2 Moderate

26 F Oculocutaneous albinism 0.96 1.52 4 Blindness

55 M Cytomegalovirus retinitis 1.34 0.24 4 Blindness

56 M Central retinal vein occlusion 0.30 1.44 2 Mild

65 F Optic atrophy 1.26 0.12 3 Severe

66 M Acute congestive glaucoma 0.56 1.04 2 Moderate

69 F Optic atrophy 1.90 0.40 5 Blindness

72 M Age-related macular degeneration 0.26 1.80 0 None

73 F Stargardt’s disease 1.46 0.68 4 Blindness

76 F Age-related macular degeneration 0.00 1.80* 0 None

77 F Age-related macular degeneration 1.08 0.96 3 Severe

83 F Central retinal vein occlusion 0.80 1.00 2 Moderate

87 M Age-related macular degeneration 1.38 0.72 4 Blindness

Age	Gender	Cause of vision impairment	Best corrected visual acuity	Contrast sensitivity	WHO category
22	F	Cone dystrophy	1.38	0.72	2 Moderate
26	F	Oculocutaneous albinism	0.96	1.52	4 Blindness
55	M	Cytomegalovirus retinitis	1.34	0.24	4 Blindness
56	M	Central retinal vein occlusion	0.30	1.44	2 Mild
65	F	Optic atrophy	1.26	0.12	3 Severe
66	M	Acute congestive glaucoma	0.56	1.04	2 Moderate
69	F	Optic atrophy	1.90	0.40	5 Blindness
72	M	Age-related macular degeneration	0.26	1.80	0 None
73	F	Stargardt’s disease	1.46	0.68	4 Blindness
76	F	Age-related macular degeneration	0.00*	1.80*	0 None
77	F	Age-related macular degeneration	1.08	0.96	3 Severe
83	F	Central retinal vein occlusion	0.80	1.00	2 Moderate
87	M	Age-related macular degeneration	1.38	0.72	4 Blindness

All participants with a visual impairment used one or more low vision devices in their daily lives. The most common optical devices used were reading glasses and handheld magnifier for near vision enhancement. Three of the participants used electronic vision enhancement assistive technology such as closed-circuit television (CCTV). Three of the participants used monocular telescope for distance vision, however they reported that they did not use these for recognizing building elements in the real world, suggesting that participants relied on their residual vision to recognize objects of interest before viewing them through a telescope. This is understandable as it is not possible to be physically moving while viewing through a telescope due to objects appearing closer than their real distance and speeds multiplied through a telescope. Five of them used a white cane as an aid for navigation and mobility.

4.1.3 Visual field

Visual fields were unable to be completed for both right and left eye by two participants with severe vision impairment in both eyes. Visual fields could not be completed for the left eye of two participants. In both scenarios, participants could not see the visual stimuli used in the test.

In the control group, the mean deviation (MD) and pattern standard deviation (PSD) were -1.89±2.67 and 1.79±1.02 for right eye and –2.09±2.65 and 1.87±1.21 for left eye respectively, whereas for people with vision impairment, MD and PSD were –13.40±7.83 and 8.90±4.10 for right eye, and –14.40±10.18 and 9.17±4.93 for left eye respectively. The MD and PSD measures are relative differences from the instrument’s database of age-matched participants for absolutely sensitivity and deviation in shape of the visual field after mean differences in absolute sensitivity are equalized respectively.

Binocular estimation of visual field was done as described in the methods, which gave the mean sensitivity of each participant for the overall field and inferior and superior field.

The binocular sensitivities of the vision impairment group are listed in Table 4.

Table 4
Visual field of participants with a vision impairment

Age Binocular mean sensitivity (dB) Binocular superior field (dB) Binocular inferior field (dB)

65 18.9 19.3 19.2

26 25.3 25.5 25.9

72 23.2 27.8 19.4

76 30.3 30.6 31.0

83 10.4 2.89 18.9

66 16.9 18.1 16.3

56 17.0 18.9 15.6

22 9.8 7.8 12.1

73 10.4 19.9 1.36

87 17.3 14.3 20.9

77 24.3 21.5 28.0

Mean (±SD) 18.5 (±6.7) 18.8 (±8.2) 19.0 (±8.1)

Age	Binocular mean sensitivity (dB)	Binocular superior field (dB)	Binocular inferior field (dB)
65	18.9	19.3	19.2
26	25.3	25.5	25.9
72	23.2	27.8	19.4
76	30.3	30.6	31.0
83	10.4	2.89	18.9
66	16.9	18.1	16.3
56	17.0	18.9	15.6
22	9.8	7.8	12.1
73	10.4	19.9	1.36
87	17.3	14.3	20.9
77	24.3	21.5	28.0
Mean (±SD)	18.5 (±6.7)	18.8 (±8.2)	19.0 (±8.1)

4.1.4 Contrast sensitivity

Every participant from the control group had normal contrast sensitivity of 1.80 dB. This is consistent with a previous study [11]. The mean contrast sensitivity for the vision impairment group was 0.96±0.56 dB (range 0.12–1.8 dB).

4.2 Rating of building elements

The relationships between the rated levels and Michelson contrast for the door frame are set out in Fig. 3. The Pearson correlation coefficients are set out in Table 5. All values are significant, DF 27, p < 0.1.

Fig. 3

Relationship between the subject’s contrast sensitivity and the Michelson contrast (left) and CIE metric lightness (right) to achieve each of the rating levels fitted with a linear function and Pearson’s correlation coefficient provided. “Not visible at all” (♦ and unbroken line), “Poorly visible” (■ and dotted line), “Easily visible” (▴ and short dashed line) and “Extremely easily visible” (x and long dashed line).

Table 5

Correlation coefficients of Michelson contrast and contrast sensitivity and CIE metric lightness and contrast sensitivity for the door frame, door handle, light switch and stair nosings. p < 0.1% for all values (DF 27)

Michelson contrast	Not visible at all	Poorly visible	Easily visible	Extremely easily visible
Door frame and wall	0.61	0.68	0.87	0.79
Door handle and door	0.75	0.80	0.82	0.79
Light switch and wall	0.76	0.79	0.81	0.81
Stairs and nosings	0.75	0.83	0.85	0.77
CIE metric lightness
Door frame and wall	0.72	0.66	0.83	0.80
Door handle and door	0.82	0.79	0.78	0.72
Light switch and wall	0.76	0.77	0.69	0.80
Stairs and nosings	0.75	0.72	0.78	0.79

One effect of using the, perceptually uniform, difference in metric lightness compared with Michelson contrast may be seen by inspecting Fig. 4, where the spread of values when contrast is low is very much compressed when contrast is expressed as ΔY (on the right compared with the left) and somewhat expanded when contrast is high. The relative relationships between each of the ratings were, qualitatively, the same for the door frame, door handle, light switch and stair nosings. The regression lines for the “Extremely easily visible”, “Easily visible” and “Poorly visible” for the four features are summarized in Fig. 4.

Fig. 4

Regression lines for the poorly visible (lower group) and easily visible (upper group) of Michelson contrast (left) or difference in metric lightness (right) and contrast sensitivity for the door frame (short dashed lines), door handle (unbroken lines), light switch (dotted lines) and stair nosings (long dashed lines).

4.2.1 Relationship with visual acuity

The mean presenting visual acuity (binocular) of the total study population was 0.43±0.64 logMAR whereas for the control group and vision impairment group were –0.04 logMAR and 0.98 logMAR respectively. The relationships between the rated levels and Michelson contrast for the door frame are set out in Fig. 5.

Fig. 5

Relationship between the subject’s logMAR visual acuity and the Michelson contrast (left) and difference in metric lightness (right) to achieve each of the rating levels fitted with a linear function and Pearson’s correlation coefficient provided. “Not visible at all” (♦ and unbroken line), “Poorly visible” (■ and dotted line), “Easily visible” (▴ and short dashed line) and “Extremely easily visible” (x and long dashed line). The WHO classifications according to visual acuity are marked at the top of each graph.

The same reduction of spread of values at low contrast, as noted with contrast sensitivity, is also seen here with visual acuity. The relationships were, qualitatively, similar for the door frame, door handle, light switch and stair nosings. The regression lines are summarized in Fig. 6. The correlation coefficients are set out in Table 6 and were statistically significant (p < 0.1%, DF 27).

Fig. 6

Regression lines for the poorly visible (lower group) and easily visible (upper group) of Michelson contrast and logMAR visual acuity for the door frame (short dashed lines), door handle (unbroken lines), light switch (dotted lines) and stair nosings (long dashed lines). The WHO classifications according to visual acuity are marked at the top.

Table 6

Correlation coefficient of Michelson contrast and logMAR visual acuity and CIE metric lightness and logMAR visual acuity for the door frame, door handle, light switch and stair nosings. p < 0.1% for all values (DF 27)

Michelson contrast	Not visible at all	Poorly visible	Easily visible	Extremely easily visible
Door frame and wall	0.73	0.68	0.88	0.82
Door handle and door	0.82	0.84	0.86	0.81
Light switch and wall	0.80	0.81	0.83	0.84
Stairs and nosings	0.81	0.92	0.91	0.81
CIE metric lightness
Door frame and wall	0.77	0.65	0.78	0.83
Door handle and door	0.84	0.82	0.82	0.80
Light switch and wall	0.80	0.81	0.70	0.85
Stairs and nosings	0.81	0.82	0.81	0.87

4.2.2 Relationship with color vision

Color vision could not be assessed in 6 out of 13 participants with vision impairment due to a failure to recognize the Landolt C target even at the highest color contrast. This is consistent with the finding that, in disease, color vision is affected earlier and more profoundly than visual acuity [1]. Given that these subjects with a vision impairment have a considerable visual acuity loss, their poor color performance is not unexpected. The extent of the color loss and the limits of the CCT mean that color vision was not a useful measure in characterizing the chromatic visual loss of this group of participants. It is a timely reminder of why the standards relating to contrast requirements of building elements specify luminance contrast not color contrast [19 –21]. Furthermore, the stimuli were all grayscale, so ability of participants to detect color contrast would not be expected to have impact on findings, except for any positive associations between chromatic contrast deficits and other visual deficits due to shared anatomy and physiology.

4.2.3 Principal component analysis

Since the sample size of the study was small, it was not possible to perform multivariate regression analysis of all vision measures like visual acuity, contrast sensitivity (dB), chromatic contrast sensitivity and average visual field sensitivity (dB). So, a factor analysis, using the principal component analysis procedure in SPSS was performed to reduce the number of dimensions into factors or principals. The variables entered into the principal component analysis were visual acuity (logMAR where the more negative the value the better visual function), binocular visual field mean difference (dB, where the higher and more positive the value the better the visual function), contrast sensitivity (dB, where the higher and more positive the value the better the visual function), protan, deutan and tritan chromatic contrast sensitivity. Protan, deutan and tritan contrast sensitivity refer to the ability to discriminate between red and grey, green and grey, and purple and grey respectively along protan, deutan and tritan confusion lines using the Cambridge Colour Test.

Although binocular VA was highly correlated with binocular CS (R > 0.9), as the determinant was 0.01, there was no multicollinearity. The KMO test result was 0.83 indicating adequate sampling, and Bartlett’s measure indicated that factor analysis was appropriate. Principal component analysis was used and a scree plot was drawn which indicated 2 components. Component 1 had a total eigenvalue of 4.45 and component 2 had a total eigenvalue of 1.05. The % of variance if only the first component were considered was 74.2%. When only the second component is considered, the % of variance explained is 17.5%. When the two components are considered, this results in a cumulative % of variance explained of 91.7%. Oblimin rotation was used for the extraction as it was expected that the variables would be related. Protan, deutan and tritan sensitivity were heavily loaded onto component 1, indicating that higher values represented better chromatic contrast sensitivity. Poor binocular visual acuity (logMAR), poor visual field sensitivity (dB) and poor binocular contrast sensitivity loaded heavily onto component 2, indicating that higher values represent poorer luminance mediated contrast, sensitivity and ability to see fine detail (Fig. 7 shows normalized component plot rotated in space). Anderson Rubin factor scores were created for each component.

Fig. 7

Principal component analysis.

Linear regression analysis for each of the simulated building elements resulted in significant models when the principal components, PC1 and PC2, were entered into the model. The data showing the correlations between the easily visible ratings for the building elements and the components are presented here. The strongest relationship was between PC2 and the contrast of the building elements (standardized β ) was statistically significant (p < 0.05) as depicted in Table 7. The first principal component gave an insignificant effect for all elements except for the stair nosings where there was a weak negative significant relationship indicating that deficits in chromatic contrast sensitivity were related to requiring higher contrast levels for this building element for it to be seen as easily visible. For the other levels of visibility of other elements, PC1 also contributed weakly to the model.

Table 7

Regression analysis of principal components for the rating of “easily visible” for the ratings of “easily visible” and “poorly visible”. The regression equations are given in the form of Z’_y = –β ₁ *Z_X1 + β ₂ *Z_X2 where Z’_y is the predicted value of Y in Z scores; β ₁ represents the standardized partial regression coefficient for X ₁ (standardized score of VA, CS, VFs); β ₂ represents the standardized partial regression coefficient for X ₂ (standardized score of chromatic contrast sensitivity)

Building element	Rating	Adjusted R² value of the model	Regression equation	Significance of standardized β ₁ and standardized β ₂
Door frame	Easily visible	0.41	Z’_y = –0.02Z_X1 +0.67Z_X2	0.99	0.002
	Poorly visible	0.75	Z’_y = –0.56Z_X1 +0.40Z_X2	<0.0001	0.004
Stairs	Easily visible	0.83	Z’_y = –0.23Z_X1 +0.76Z_X2	0.03	<0.0001
	Poorly visible	0.80	Z’_y = –0.14Z_X1 +0.81Z_X2	0.21	<0.0001
Light switch	Easily visible	0.77	Z’_y = –0.15Z_X1 +0.79Z_X2	0.22	<0.0001
	Poorly visible	0.74	Z’_y = –0.10Z_X1 +0.81Z_X2	0.41	<0.0001
Door handle	Easily visible	0.73	Z’_y = –0.15Z_X1 +0.77Z_X2	0.25	<0.0001
	Poorly visible	0.78	Z’_y = –0.07Z_X1 +0.85Z_X2	0.54	<0.0001

5 Discussion

5.1 Contrast sensitivity and contrast

Figure 4 shows that there was linear relation between contrast sensitivity and the preferred contrast for recognition of the doorframe. The relationship is essentially the same whether Michelson Contrast and CIE Metric Lightness is used as the measure. The construction of the experiment meant that the mean luminance or mean CIE Metric Lightness remained constant. Differences between the two measures will only be evident when the mean luminance is varied. The required amount of contrast for the doorframe increased linearly with diminution of contrast sensitivity. It is clear that reduced contrast sensitivity has strong influence on the preferred contrast. While the criteria “not visible at all” and “extremely easily visible” were only used for context, there is a consistent pattern in all four criteria. From the data in Table 4, it may be seen that this is also true of the other building features. The contrast provided should be greater than that for “poorly visible” and need not exceed that necessary for the feature to be rated “easily visible”. The contrast preferred by the control group for the doorframe to be easily visible was 20% whereas for the low vision group, with poorest contrast sensitivity, it was about 60%. Figure 4 shows that much the same can be said about the door handle, light switch and stair nosings. These three features all require much the same contrast, 20% for the control group and just around 70% for those with a vision impairment, which is significantly more than for the door frame. The horizontally defined items (stair nosings and door handle) appeared to be more difficult to see than the elements that had more vertical elements. There is literature on this horizontal effect. In the real world, horizontal elements are not as salient as our visual system tends to discount horizontal lines as they are expected in the environment [13 , 17].

5.2 Visual acuity and contrast

Figure 5 shows that there was a very similar relationship between visual acuity and the preferred contrast for recognition of the doorframe. Again, the relationship is essentially the same whether Michelson Contrast and CIE Metric Lightness is used as the measure. In this study, contrast sensitivity and visual acuity are highly correlated (R = 0.93, DF 27, p < 0.01). As with contrast sensitivity the required amount of contrast for doorframe increased linearly with poorer visual acuity. It is seen that reduced visual acuity has more effect on the preferred contrast. Again, in Fig. 5, the “not visible at all” to “extremely easily visible” form a consistent pattern in all four criteria. From the data in Table 5, it may be seen that this is, again, also true of the other building features. Figure 6 shows that much the same can be said about the door handle, light switch and stair nosings. These three features all require much the same contrast, 20% for the control group and about 80% for those with a vision impairment which is significantly more than for the door frame.

The ranking by visual acuity shows more contrast is required by those with poorer visual acuity than ranking by contrast sensitivity. The ergonomic principle of accommodating a defined proportion of the population can be applied using this data. The current WHO categories are set out in Table 1 [31].

However, the functional vision of a person may be affected at 6/12 Snellen acuity (rated mild vision impairment) as people may have difficulty in driving, mobility etc. The WHO classification of low vision does not consider the role of contrast sensitivity.

To accommodate those with mild, moderate and severe impairment (upper limit of logMAR visual acuity 1.3), to be easily visible, a contrast of about 40% is indicated for door frames and 60% for the door handle, light switch and stair nosings. The present requirement of 30% would provide easily visible elements for people with a logMAR visual acuity≤0.35, which are those in the mild or no vision impairment categories. On the other hand, over 30% is better than “poorly visible” for those with a logMAR visual acuity≤1.2, so it does offer some assistance. It is also worth noting from Fig. 5 that the contrast of 30% is close to the “extremely easily visible” value for those with normal visual acuity (logMAR = 0.0). This may illustrate that people with normal visual acuity find it difficult to accept that what they see as “extremely easily visible” is, for those with a severe vision impairment, no more than “poorly visible”.

5.3 Color vision and contrast

From the factor analysis and regression analysis, it was seen that color vision had either a weak effect or no effect on the visibility ratings in the present study, and this effect was not unexpected as the stimuli were presented in grayscale in the present study. Studies have suggested that color contrast does help in the recognition of the elements in the visual space [18] and the PCA analysis suggests that deficits in chromatic contrast sensitivity can contribute to an increased luminance contrast requirement. The results support that the requirements for contrast are written as luminance contrast as it is the stronger relationship, and it is known that some causes of vision impairment include partial or complete loss of color vision. A total of six people could not perform the Cambridge Colour Test due to failure to detect the target at maximum color contrast.

6 Conclusions

Contrast sensitivity and visual acuity showed linear relationships with the contrast required for visibility of building elements. The PCA analysis also verified the importance of visual acuity, contrast sensitivity and visual fields to rating visibility of luminance contrast defined simulations of building elements. It may be concluded that 30% contrast for building elements will render building elements only poorly visible for those with severe vision impairment. A contrast level of around 65% is needed to render all small area building elements easily visible, which is over twice the currently required value. The aim to make all elements easily visible to anyone with a vision impairment might be argued as being over ambitious and beyond reasonable accommodation, but some increase (not a decrease) on the present, 30%, requirement and official encouragement to provide installations that exceed this requirement would seem appropriate.

The applicability to real-life situations of a study using an environment of simulated images is debatable. However, such studies are much simpler to carry out than in real-life and they also minimize the influence of co-morbidities, lighting variation and viewing angle on the findings and the results did correlate with previous real-life studies in both the controlled and uncontrolled experiments. This study demonstrates that simulation studies are possible and correlate well with the visual measures of visual acuity and contrast sensitivity and further studies are possible to advise the setting of minimum luminance contrast requirements in the future.

Conflict of interest

The authors declare that there are no conflicts of interest.

Funding

The authors do not have any funding to disclose.

Supplementary materials

The supplementary files are available at https://dx.doi.org/10.3233/WOR-210997.

References

Almog

, Nemet

. The correlation between visual acuity and color vision as an indicator of the cause of visual loss. Am J Ophthalmol. 2010;149:1000–4.

Arden

, Wolf

. Colour vision testing as an aid to diagnosis and management of age related maculopathy. Brit J Ophthalmol. 2004;88:1180–5.

Arditi

. Improving letter contrast sensitivity testing. Invest Ophthalmol Vis Sci. 2004;45:4583.

Arditi

. Rethinking ADA signage standards for low-vision accessibility. J Vis. 2017;17:1–20.

Bailey

, Lovie-Kitchin

. Visual acuity testing. From the laboratory to the clinic. Vision Res. 2013;90:2–9.

Bochsler

, Legge

, Gage

, Kallie

. Recognition of ramps and steps by people with low vision. Invest Ophthalmol Vis Sci. 2013;54:288–94.

CIE S 017. ILV: International Lighting Vocabulary, Commission Internationale de l’Éclairage, Vienna, 2011.

CIE 15, Colorimetry, 4th Ed., Commission Internationale de l’Éclairage, Vienna, 2018.

Crabb

, Viswanathan

, McNaught

, Poinoosawmy

, Fitzke

, Hitchings

. Simulating binocular visual field status in glaucoma. B J Ophthalmol. 1998;82:1236–41.

10.

Dain

, Bridge

, Relf

, Lukman

, Manandhar

, Boon

M-Y

. Issues in specifying contrast in building elements for people with a visual disability. Work. 2021;70(4):1219-27.

11.

Dougherty

, Flom

, Bullimore

. An evaluation of the MARS letter contrast sensitivity test. Optom Vis Sci. 2005;82:970–5.

12.

Dupont

, Plummer

. Power and sample size calculations: a review and computer program, Control Clin Trials. 1990;11:116–28.

13.

Essock

, DeFord

, Hansen

, Sinai

. Oblique stimuli are seen best (not worst!) in naturalistic broad-band stimuli: a horizontal effect. Vision Res. 2003;43:1329–35.

14.

Galbraith

, Bowman

, Luminance contrast, the boy who could, and visionary pathfinders, in: 21st Congress of the International Ergonomics Association (IEA 2021). Kacprzyk

, ed., Springer, Vancouver, Canada, 2021, 154–161.

15.

Hansen

, Essock

. A horizontal bias in human visual processing of orientation and its correspondence to the structural components of natural scenes. J Vis. 2004;4:1044–60.

16.

Hasrod

, Rubin

. A review of the Cambridge Colour Test and other colour testing method. Afr Vis Eye Health. 2015;74.

17.

Haun

, Essock

. Contrast sensitivity for oriented patterns in 1/f noise: contrast response and the horizontal effect. J Vision. 2010;10.

18.

Haymes

, Johnston

, Heyes

. Relationship between vision impairment and ability to perform activities of daily living. Ophthalmic Physiol Opt. 2002;22.

19.

ISO TR9527 Building construction – Needs of disabled people in buildings – Design guidelines, International Standardization Organisation, Geneva, 1994.

20.

ISO 21542 Building construction – Accessibility and usability of the built environment, International Standardization Organisation, Geneva, 2011.

21.

ISO 23599 Assistive products for blind and vision-impaired persons — Tactile walking surface indicators, International Standardization Organisation, Geneva, 2019.

22.

Kaiser

. Prospective evaluation of visual acuity assessment: A comparison of Snellen versus ETDRS charts in clinical practice (An AOS thesis). Trans Am Ophthalmol Soc. 2009;107:311–24.

23.

Legge

, Rubin

, Luebker

. Psychophysics of reading-V. The role of contrast in normal vision. Vision Res. 1987;27:1165–7.

24.

Legge

, Yu

, Kalli

, Bochsler

, Gage

. Visual accessibility of ramps and steps. J Vision. 2010;10.

25.

Lukman

, Bridge

, Dain

, Boon

M-Y

. Luminance contrast of accessible tactile indicators for people with visual impairment. Ergon Des. 2020;28:4–15.

26.

MARS perceptrix, The Mars Letter Contrast Sensitivity Test: User manual, in, MARS Perceptrix Corporation, Cappaqua, New York., 2013.

27.

Mollon

, Regan

. Handbook of the Cambridge Colour Test., Cambridge Research Systems, London, UK., 2000.

28.

Rubin

, Legge

. Psychophysics of reading VI- The role of contrast in low vision. Vision Res. 1989;29:79–91.

29.

Shirato

, Inoue

, Fukushima

, Suzuki

. Clinical evaluation of SITA: a new family of perimetric testing strategies, Graefes Arch Clin Exp Ophthalmol. 1999;237:29–34.

30.

Tan

, Boon

, Dain

. The visibility of controls and labels on electronic devices and their suitability for people with impaired vision. Work. 2014;47:309–17.

31.

World Health Organization, 9D90 Vision impairment including blindness, in: ICD-11 for mortality and morbidity statistics, World Health Organization, Geneva, 2018.

Luminance contrast preferences of people with a vision impairment for elements in the built environment

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

1 Introduction

2 Objective

3 Materials and method

3.1 Subjects

3.2 Sample size

3.3 Ethics approval

3.4 Assessing the independent factors

3.4.1 Visual acuity

3.4.3 Color vision

3.4.4 Visual fields

3.5 Assessing the dependent variables

3.6 Equipment

3.8 Data analysis

4 Results

4.1 Demographics

4.1.1 Normal control subjects

Table 2 Age, gender and visual acuity of subjects with no vision impairment Age (years) Gender Visual acuity (logMAR) 24 F 0.04 39 M 0.00 26 M 0.00 37 M –0.12 32 F 0.02 24 F –0.06 27 F –0.12 23 F 0.00 22 M 0.00 31 F –0.06 35 F –0.06 26 M –0.08 20 M 0.00 36 M –0.12 25 F 0.00

4.2 Rating of building elements

4.2.3 Principal component analysis

5.1 Contrast sensitivity and contrast

5.2 Visual acuity and contrast

5.3 Color vision and contrast

6 Conclusions

Conflict of interest

Funding

Supplementary materials

References