Visual Performance of People With Albinism Assessed With Generalizable and Adaptive AIM and FInD Methods

Abstract

Purpose: People with albinism (PwA) are known to have visual impairments; however, little is known about whether these functions are disrupted across earlier and later stages of the visual pathway. We investigated distinct perceptual functions and fixation stability within each observer and compared the data with age- (±5 years) and sex-matched controls.

Methods: Twenty-one self-reported PwA and twenty-one controls were recruited. Angular-indication measurement (AIM) and foraging-interactive-D-prime (FInD) psychophysical methods were deployed to measure OS, OD, and OU near visual acuity, spatial contrast sensitivity function (CSF), temporal contrast sensitivity (tCS; 0.5 c/°; horizontal grating: 0, 1, 2, 4, and 8 Hz), OU glare acuity, threshold-versus-contrast (2c/° vertical grating), long, medium, and short wavelength cone-isolated color detection, color discrimination, stereoacuity across spatial frequencies (1c/°, 2c/°, 4c/°, 8c/°), horizontal, circular, radial pattern and motion coherence, and equivalent-noise motion detection. Thresholds were determined by AIM and FInD and compared using N-ANOVAs, t-tests, planned multi-comparisons, correlations, and unsupervised, agglomerative hierarchical cluster analysis for each group.

Results: We found significant differences between groups for most visual functions except for simple and complex form-coherence (two way-ANOVAs, P > 0.05) and complex motion coherence. Correlations between outcomes revealed more significant correlations for PwA and differences in the specific correlates between groups. Unsupervised hierarchical clustering revealed different functional clusters between groups.

Conclusions: AIM and FInD successfully interrogated visual deficits in PwA. Overall, PwA showed impaired performance in achromatic, chromatic, temporal, and binocular functions, and had higher intrinsic noise levels. Midlevel vision was comparable between groups. Unsupervised cluster analysis and correlation between outcomes revealed a difference in functional outcome clusters between groups. The results may help to increase the efficiency of screening and identify target deficits for rehabilitation.

Albinism is a genetic condition that results in reduced or absent biosynthesis of one or multiple forms of melanin (eumelanin, pheomelanin, neuromelanin), all of which have a role in visual development.^1,2 As a result, aberrant visual development occurs at the level of the retina and visual cortical pathways. Approximately 1:17,000 people worldwide have albinism either in isolation or as part of a genetic syndrome, with greater prevalence (1:3000) in some regions (e.g., Puerto Rico, Fiji).³ There are several distinct genotypes of albinism, with different inheritance patterns, including those with overall melanin deficiency, resulting in oculocutaneous albinism, as well specific melanin deficiencies, resulting in ocular-only albinism.⁴ The following ocular findings are known to be associated with albinism, including iris transillumination,¹ hyperopic refractive error,^5–7 foveal hypoplasia, nasotemporal shift in retinal ganglion cell fibers secondary to a greater percentage of uncrossed fibers at the optic chiasm,⁸ nystagmus,^9,10 and strabismus.^11,12 Misrouting of the optic nerves at the chiasm may vary in its extent in PwA,¹³ resulting in abnormal visual field representation in the visual brain.¹⁴ PwA often have visual deficits including reduced visual acuity,^6,12,15 an elevated subjective sense of glare/photosensitivity,¹⁶ atypical color discrimination,¹⁷ reduced contrast sensitivity,^10,18,19 reduced or absent stereoacuity,^7,9,20 and interocular suppression.⁷ A retrospective review revealed that the PwA without nystagmus had improved stereoacuities compared to PwA with nystagmus and that individuals with stereopsis had better binocular visual acuities than those without stereopsis.²¹ Another study reported increased internal noise in motion discrimination and motion coherence for simple unidirectional motions in the presence of nystagmus but no difference without nystagmus.⁹ Quality of life and vision-specific quality of life are affected as a consequence.^22,23 

There has been limited investigation of the complete contrast sensitivity function and no investigation of supra-threshold contrast processing in PwA. Information regarding spatiotemporal visual performance is also lacking, and there is limited investigation of mid- and higher-level perception, including form and complex motion coherence perception, and object or facial recognition. Many studies have also been limited by the use of clinical tests with a fixed range of visual stimuli, such as booklets or chart-based tests.^12,24,25 Recently, the generalizable FInD (Foraging Interactive D prime)^24,26,27 and AIM (Angular Indication Measurement)^24,28 methods were developed to deploy charts of stimuli that adaptively target individual's threshold values, thereby efficiently building a personalized model of visual performance for each visual function (Fig. 1). 

Figure 1.

 

Example screenshots of AIM Acuity (A) and FInD color detection for long wavelength cone–isolating targets (B). (C) An example psychometric error function resulting from the AIM method and its output parameters: threshold (midpoint between minimum and random angular error), slope of the function, and the range of stimulus detectability improvement (ROSDI), (i.e., distance between threshold and 0.05s above minimum angular error level). Depicted also is an illustration of the measurement principal of AIM applied here to visual acuity, which is the difference between true and indicated orientation as angular error in degrees here a C optotype, but also applicable for grating orientation (CSF), or motion direction (equivalent noise − motion). The gap width was fixed at the stroke width of the C. (D) An example FInD psychometric function: Blue data show the probability that the observer reported the presence of a stimulus as a function of stimulus intensity, here long wavelength cone-isolating contrasts, vertical lines show binomial standard deviation, red curve shows the probability of a “Yes” response for the d′ sensitivity function, and its upper and lower 95% confidence intervals (dashed blue lines).

FInD is a generalizable and adaptive psychophysical method that can measure many visual functions. Each of three charts (one initial and two adaptive charts) contains 4 × 4 cells, a randomly allocated subset of cells contains signal stimuli (with intensities adaptively spanning easy to difficult), and the remaining cells contain null stimuli. Stimuli can be defined by color, contrast, motion, and more, and the task in all cases is to click on each cell that contains a target stimulus. Thresholds are estimated from d’ analysis of hits, misses, false alarms, and correct rejection responses. AIM uses an orientation judgement approach in which an observer indicates the direction of the gap in a C target (AIM Acuity and AIM Glare Acuity), grating orientation (AIM CSF+), or motion direction (AIM motion equivalent noise) by clicking on the corresponding angle in surrounding response ring (Fig. 1C). Then a sector appears (Figs. 1A and 1C) that corresponds with the reported orientation of the respective stimulus and can be changed repeatedly before moving on to the next chart. This procedure enables the assessment of continuous report error as a function of stimulus intensity (e.g., stimulus size for visual acuity) and thereby builds personalized performance models of visual functions (e.g., Supplementary Fig. S3). Each chart used in the current study contained 4 × 4 cells with target stimuli of randomized orientation; the intensity levels are adaptively personalized and randomly assigned to cells. Performance is estimated from the angular error between the true and reported stimulus orientation, and AIM thresholds are estimated from the psychometric function mid-point between minimum angular error and random responses (±90° angular error for C optotypes and motion directions and ±45° for AIM CSF grating orientation). 

To address the above-mentioned gaps of knowledge and limitations, we had the overarching AIM to investigate visual performance within a group of PwA compared to age- and sex-matched controls across a range of visual functions, including chromatic, achromatic, binocular, and mid-level visual functions, as well tests of fixation stability/oculomotor function. The second goal was to determine which of these visual functions are critical in assessing PwA's vision. We applied cluster analysis machine learning to determine the key clusters. 

Methods

Participants

Twenty-one participants with a known diagnosis of ocular or oculocutaneous albinism were recruited from the Janet LaBreck Center for Vision Rehabilitation at the New England College of Optometry Center for Eye Care. An additional 21 normally-sighted participants were recruited as controls, matched ±5 years in age and sex to albinism patients. All participants had an eye examination within one year a normal score on the Mini Mental State Examination upon enrollment and wore optical correction if required (see Supplementary Table S15 for details). Participants with albinism who had additional ocular pathologies (e.g., glaucoma, macular disease) were excluded except one albinism participant who had cataracts. This patient was matched with a control subject who also had cataracts of similar optical density. Before completing computer-based psychophysical testing, Early Treatment Diabetic Retinopathy Study (ETDRS) visual acuity (logMAR), Pelli-Robson contrast sensitivity (logCS), near-ETDRS visual acuity, and MNRead critical and threshold print size tasks were performed under the supervision of trained optometrists. Ocular alignment and ocular motor function were also reviewed from electronic medical records. Nystagmus and strabismus direction and amplitude were assessed by a qualified optometrist. Because no eye-dominance test was performed, AIM Visual Acuity results were used as proxy eye dominance measures (i.e., better acuity determined dominant, worse acuity non-dominant eye, and equal dominance when the acuities were equal). Additionally, details regarding the strabismus in our albinism participants are provided in the supplement, including details of alternate fixation, whether the strabismus is intermittent, or whether a constant turn was present. This study was approved by the institutional review boards at Northeastern University in Boston, Massachusetts, and the New England College of Optometry, Boston. 

Participant Characteristics

Table 1 provides demographic, mental ability (Mini-Mental State Examination), and optometric information. Acuity (1 m and 4 m ETDRS), contrast sensitivity (1 m Pelli-Robson) and reading (Minnesota Low Vision Reading Test) were compared with two sample t-tests. A trained optometrist (coauthor N.C.R.) assessed the presence and direction of nystagmus. The majority of PwA exhibited a horizontal congenital nystagmus of moderate amplitude. All PwA also had strabismus. Other demographic information and clinical measures for each observer are shown in the Supplementary Materials. 

Table 1.

 

Demographic and Optometric Comparisons Between Groups Reporting Means, SD, and Other Outcomes

 

Equipment

Stimuli were presented on an LG HD 4K screen 32 inch and monitor system (LG Life Science, Seoul, Korea) with a framerate of 60 Hz with a pixel resolution of 3840/2160 at 60 cm distance, 1° included 63.5 pixels. Stimuli were generated using Matlab (R2021a; MathWorks, Inc., Natick, MA, USA) software in combination with The Psychophysics Toolbox.^29,30 A computer mouse was used as an indication device. The monitor was gamma-corrected using data color Spyder (Datacolor, Lucerne, Switzerland) photometer. A Gazepoint eye tracker (Gazepoint, Vancouver, Canada) (60 Hz) was used to investigate the fixation stability of each participant. 

Fixation Stability Procedure

A Gazepoint eye tracker (60 Hz) was used to measure OS and OD gaze while the participant fixated a central 0.32° × 0.32° square for 60 seconds (Supplementary Fig. S14). The task description was to look at the fixation spot without further indicating a specific location of that fixation spot. The spot size was chosen to ensure that PwA were able to identify the target. A five-point calibration was performed before the actual experiment. Seventy-six percent (n = 16) PwA had stable enough fixation to perform the task, and all participants were strabismic. We therefore only report the spread and not absolute direction of gaze. The stability of gaze was measured during a static fixation task using both eyes across testing time from the 95% bivariate contour ellipse area for each eye (i.e., a well-established, parametric gaussian approach to assess the spread of eye movements).³¹ 

Psychophysical Procedure

AIM and FInD psychophysical methods were deployed to measure near visual acuity at high luminance without and low luminance with a glare source, the spatial contrast sensitivity function (CSF), temporal contrast sensitivity, threshold-versus-contrast, contrast sensitivity, cone-isolated color detection, color discrimination, stereoacuity across spatial frequencies, horizontal, circular, and radial pattern and motion coherence, and equivalent-noise motion detection. A dark eyepatch was used to cover the untested eye. An overview of the methods, their targeted visual functions, as well as primary results, are shown in Table 2. Detailed descriptions for each test are shown in the Supplementary Materials and refer to other peer-reviewed publications.^24,26–28 All participants wore appropriate refractive correction for the 60 cm viewing distance. The test order was randomized. The eye conditions (i.e., OD, OS, OU) for FInD spatiotemporal CS, AIM CSF, and AIM Acuity were also randomized within each observer whereas all other tests were tested only binocularly (OU). Both AIM and FInD methods are adaptive in successive charts, narrowing in on an observer's threshold by presenting stimuli over a ±2 seconds logarithmic range around an estimated threshold from previous charts (see details in Supplementary Materials). Most tests used three charts (i.e., two adaptive steps, per condition), except for AIM Acuity and AIM Glare Low Luminance Acuity, which used two charts, based on estimates of their convergence. Short breaks between tests were given. The full session, including consent to participate, clinical vision screening, and psychophysical testing, took between two to four hours to complete mostly, broken down into two sessions. 

Table 2.

 

Overview of Methods Used, Target Visual Function, and Outcome Measures

 

Data Analysis

Customized Matlab (Version 2021a) programs were used to analyze the raw data, and standard functions were used for statistical analyses (i.e., ANOVAs [anovan], planned comparisons [multcompare], and two-sample t-tests [ttest2]) between groups. We then calculated for each source \(\eta_{\rm p}^2\) (partial eta squared) and report the effect sizes for ANOVAs and the median difference between t-tests using the meanEffectsize function. Boxplot graphs with superimposed swarm plots, and additional mean values were created using Matlab functions and are shown in the Supplementary Materials. We used Matlab's corrplot function and plotted the Kendall's rank linear correlation coefficients between all pairs of variables. We also investigated which correlations were significantly different from zero by conducting null-hypothesis tests. Results were then replotted in a customized color bar graph. In addition, we analyzed the correlation data also after a simple Bonferroni and the Holm-Bonferroni correction. To analyze eye dominance, we used visual acuity and in the event of equal acuities between eyes, the acuity of the right eye was allocated to the dominant and the left to the fellow eye in a counterbalanced order to have as few incomplete data sets as possible, which is important for the clustering analysis described below. 

Cluster Analysis

We investigated the relationship between the threshold outcomes using clustering analysis.^9,32 Hierarchal cluster analysis was performed using Matlab's linkage applying the ward method and was depicted using the dendrogram function. First, we connected all 41 outcomes (thresholds and other key results), and then we eliminated participants with incomplete data, resulting in 13 PwA and 14 controls because it is common practice to avoid introducing artificial biases. Moreover, we performed the same analysis using all participants’ data and replaced missing data points with median values to compare models regarding their similarity. Next, we normalized each group data set separately using z-score transformation to make different outcome values comparable. Then we estimated the Euclidean distance between each data point using pdist function, followed by cluster analysis using linkage, and used these outputs to calculate the cophenetic correlation coefficient c and the cophenetic distances D using the cophenet function. Depth perception results were excluded because most PwA were unable to see depth via stereoscopic disparity. 

Results

Psychophysics

This section reports comparisons of threshold results between groups as well as noteworthy other findings related to the psychometric analysis for each test. A complete overview including graphs and additional statistical reports for each parameter is reported in the Supplementary Materials. 

AIM Visual Acuity

A two-way ANOVA revealed a highly significant effect of group [F(1, 114) = 197.1, P < 0.001, \(\eta_{\rm p}^2\) = 0.63] because of lower acuities for PwA. No effect of eyes [F(2, 114) = 1.0, P > 0.05, \(\eta_{\rm p}^2\) = 0.02] nor interaction [F(2, 114) = 1.3, P > 0.05, \(\eta_{\rm p}^2\) = 0.02] was found (Supplementary Table S3; Supplementary Fig. S16). 

AIM Glare Visual Acuity

PwA showed significantly lower low-luminance near acuity (mean = 0.68 ± 0.22 seconds) than controls (mean = 0.04 ± 0.11 seconds under glare condition (t(29.9) = [11.8], P < 0.001], median effect size: 0.73 (0.55–0.87). An effect of glare condition was found for the subjective rating task [two-way ANOVA, F(1, 80) = 3.8, P < 0.001, \(\eta_{\rm p}^2\) = 0.16]. A planned multicomparison showed an increased subjective glare experience for PwA but not for controls (Supplementary Fig. S5). There was an interaction between AIM VAs for high-contrast, no glare vs glare [two-way ANOVA, F(1, 78) = 7.6, P < 0.01, \(\eta_{\rm p}^2\) = 0.09], which was due to significantly reduced acuities for PwA, with a larger effect of glare on acuity (Supplementary Materials S6). Other findings were significantly reduced: range of stimulus detectability improvement, steeper slopes, and increased min. angular error for PwA (Supplementary Materials). 

AIM+ Contrast Sensitivity Function

PwA had significantly lower CSF acuities [two-way ANOVA F(1, 112) = 102.1, P < 0.001, \(\eta_{\rm p}^2\) = 0.48] and AULCSF than controls [two-way ANOVA F(1, 112) = 120.5, P < 0.001, \(\eta_{\rm p}^2\) = 0.52] (Supplementary Table S2 & Supplementary Fig. S15). 

FInD Threshold Versus Contrast

PwA showed significantly higher intrinsic noise (mean = 0.04 ± 0.03 seconds) than controls (mean = 0.017 ± 0.002 seconds) (t(20.2) = [3.2], P < 0.01, median effect size: 0.009 (0.002, 0.02)), but no differences in Weber fractions (mean = 0.88 ± 0.85 seconds) compared to controls (mean = 0.74 ± 0.19 seconds) (t(21.9) = [0.7], P > 0.05, median effect size: −0.16 (−0.34, 0.09)) (Supplementary Fig. S8). 

FInD Spatiotemporal Contrast Sensitivity

A three-way ANOVA was performed (i.e. effects of group, flicker frequency, eyes) and found an effect of group [F(1, 473) = 45.3, P < 0.001, \(\eta_{\rm p}^2\) = 0.09], which was due to overall worse spatiotemporal thresholds (i.e., more positive numerical values) for the PwA cohort compared to controls. Also, an effect of temporal flicker frequency was found [F(4, 473) = 6.2, P < 0.001, \(\eta_{\rm p}^2\) = 0.05], which was due greater difference for higher temporal frequencies for PwA (Supplementary Fig. S22, Supplementary Table S8). 

FInD Color Detection

PwA showed lower sensitivity to cone-isolated colors than controls [two-way ANOVA F(1, 120) = 10.7, P = 0.001, \(\eta_{\rm p}^2\) = 0.08]. There was no effect of cone type [F(2, 120) = 1.1, P > 0.05, \(\eta_{\rm p}^2\) = 0.02] or an interaction [F(2, 120) = 0.06, P > 0.05, \(\eta_{\rm p}^2\) < 0.001] (Supplementary Fig. S10). 

FInD Color Discrimination

PwA showed larger color discrimination angles than controls (two-way ANOVA [F(1, 240) = 12.7, P < 0.001, \(\eta_{\rm p}^2\) = 0.05]), and an effect of color [F(5, 240) = 3.1, P < 0.01, \(\eta_{\rm p}^2\) = 0.06], but no interaction [F(5, 240) = 1.4, P > 0.05, \(\eta_{\rm p}^2\) = 0.01] (Supplementary Fig. S11). 

FInD Depth

Most PwA were unable to see stereoscopic depth regardless of the spatial frequency content of the stimulus (Supplementary Fig. S12). 

FInD Form Coherence

No significant effect for group [F(1, 117) = 0.5, P > 0.05, \(\eta_{\rm p}^2\) = 0.05], form type F(2, 117) = 1, P > 0.05, \(\eta_{\rm p}^2\) = 0.02], nor interaction [F(1, 117) = 0.1, P > 0.05, \(\eta_{\rm p}^2\) = 0.002] was found (Supplementary Fig. S13). 

FInD Motion Coherence

An interaction between group and motion type was found [F(2, 114) = 3.7, P < 0.05, \(\eta_{\rm p}^2\) = 0.06], which was due to significantly worse performance for horizontal motion coherence thresholds for PwA as identified via planned multi-comparison (Supplementary Fig. S14). 

AIM Equivalent Noise Motion

Relative to controls, PwA showed significantly higher intrinsic noise (mean = 49.9 ± 30 seconds) (mean = 30.2 ± 18 seconds) [t(2.5) = [28.8], P < 0.05], median effect size: 15.3 (1.2, 46.9)], reduced sampling efficiency (mean = 3.7 ± 4.5 seconds) (mean = 12.2 ± 11 s) [t(27) = [−3.2], P < 0.01], median effect size: −7.7 (−13.2, −1.4)], and increased minimum report error (mean = 35.4 ± 10 seconds) (mean = 16.6 ± 11 seconds) [t(38) = [5.8], P < 0.001], median effect size: 24.9 (15.9, 35.7)] (Supplementary Fig. S15). 

Fixation Stability

PwA exhibited larger bivariate contour ellipse areas compared to controls [two-way ANOVA, F(1, 99) = 10.8, P < 0.01, \(\eta_{\rm p}^2\) = 0.1] (Supplementary Table S13 and Supplementary Fig. S27). 

Correlation Analysis

Out of 820 correlations between different results for PwA, we found 116 statistically significant correlations, including 81 positive and 35 negative correlations (Fig. 2, lower half). For controls it resulted 77 significant correlations (40, 37, respectively) out of 820 (Fig. 2, upper half). Correlations between outcomes differed between groups substantially as shown in Figure 2, with a majority within the achromatic outcomes in the top left corner. We re-ran the analysis twice applying the simple Bonferroni-correction (α = 0.00006) and the Holm-Bonferroni correction, which resulted in eight and thirty-two significant correlations for PwA and six and ten for controls, respectively, with the same skew toward achromatic functions. 

Figure 2.

 

Correlations between psychophysical thresholds and other key outcomes generated with the battery of adaptive AIM and FInD methods for albinism (bottom left) and controls (top right). Increasing diameters depict significance levels from not different (P > 0.05) to significantly different (P < 0.05; P < 0.01; P < 0.001). Positive (gradient green, orange, yellow; i.e., values of variable y and values of variable x increase) or negative (cyan, blue, violet; i.e., values of variable y and values of variable x decrease) correlations are shown as gradient colors. AULCSF, area under the log contrast sensitivity function; CSFAcuity, contrast sensitivity function acuity; DE, dominant eye; FE, fellow eye; Glare VA, visual acuity with glare source; OU, both eyes; VA, visual acuity; TvC, threshold versus contrast: contrast sensitivity threshold for each pedestal step (0.01–0.32) measured with both eyes; TvC IN, internal noise parameter; TvC Weber, Weber fraction parameter; Temp, spatiotemporal contrast sensitivity thresholds for 0–8 Hz flicker frequency for both eyes; ColDetect, color detection thresholds for long, medium, and short wavelength stimuli; ColDiscrim, color discrimination thresholds for red, yellow, green, cyan, blue, magenta directions; Pattern, pattern coherence thresholds for expanding, horizontal, and circular forms; Motion, motion coherence thresholds for expanding, horizontal, and circular motion; EquiNM IN, equivalent noise-motion: internal noise results; EquiNM SE, equivalent noise-motion: sampling efficiency results.

Cluster Analysis

We applied an unsupervised machine learning approach to identify differences in outcomes for each cohort and used agglomerative hierarchical cluster analysis (Fig. 3). Overall, the control group showed a significantly different emergent clustering of outcomes compared to PwA. The dissimilarity of heights was verified using cophenetic distance correlation (c) and was 0.78 for controls and 0.77 for PwA. We also compared the data with clusters that used median values to replace missing values while including all 21 participants for each group (Supplementary Fig. S28) and show that for each group the emerging clusters of visual function are similar. Also, the cophenetic distance correlation was 0.77 for both controls and PwA and thus comparable to the findings with excluded participants numbers (Fig. 3). 

Figure 3.

 

Dendrograms for the control and albinism group. Depicted are all 41 perceptual outputs generated with the AIM and FInD methods.

Discussion

Previous cohort case control studies have explored psychophysical outcomes in PwA⁹ and found absence of stereoscopic depth perception, reduced visual acuities,¹² and presence of nystagmus in the majority of participants.¹¹ One PwA had a mild cataract and was matched with a control that had the same medical grading of cataract. We reran the analysis without these individuals and found the same pattern of effects for all visual function findings. Both correlations and cluster analysis showed almost identical patterns. Stereovision is typically absent in PwA^1,9,12; however, one study reported that some individuals with less severe symptomology also showed coarse-to-fine stereo vision.²¹ Most of the PwAs in the current study were unable to perceive any stereoscopic depth regardless of the spatial frequency of the bandpass-filtered stimuli. The current study did not investigate interocular suppression and thus far only one study reports psychophysical data using a Worth 4 dot test.⁷ It is noteworthy that the researchers reported 71% (n = 25/35) of participants showed fusion either with, without, or regardless of spectacle correction, although 71% of participants were diagnosed with strabismus. The PwA group in this study showed almost no stereo vision. This is likely due to two reasons, namely dissociate dichoptic signals during the FInD Depth perception task due to strabismus in each participant and also due to misrouting of the optic nerves in albinism.¹³ 

Contrast sensitivity was significantly reduced (Supplementary Fig. S7), in agreement with previous work,¹⁸ and suprathreshold contrast sensitivity thresholds (Supplementary Fig. S8) were also significantly lower in PwA compared to controls. Interestingly, using low spatial frequency stimuli to measure contrast sensitivity with and without flicker showed that increasing flicker frequency lowered thresholds for PwA significantly compared to controls (Supplementary Fig. S9), suggesting temporal processing impairment in PwA. Further evidence for disrupted temporal processing in PwA comes from an electrophysiological study that demonstrated interhemispheric latency difference in PwA but not in controls.²⁰ Furthermore, that study reports the highest correlation, positive in direction, with foveal hypoplasia. Whether the temporal findings of the current study are explained by those retinal changes, cortical structural differences that lead to different activation patterns in early visual cortex,¹⁴ or a combination of both remains to be further investigated. 

Furthermore, Neveu and colleagues⁹ investigated motion processing with coherence and equivalent noise tasks in a psychophysical study with a cohort size of twelve participants. The researchers found that PwA with nystagmus had significantly worse motion coherence thresholds and increased internal noise, as well as impaired sampling efficiency compared to controls, but in a subset of two PwA without nystagmus, no significant difference in motion perception was found. Our current study also finds significantly higher (i.e., poorer horizontal motion coherence detection threshold values for PwA compared to controls) (Supplementary Materials, Supplementary Fig. S26), in agreement with the findings of Neveu et al.⁹ Moreover, our study complex motion perception (i.e., expanding or circular motion) in PwA and finds no significant group differences. The equivalent noise motion task in all 360° orientation deployed in the current study resulted in higher internal noise and lower sampling efficiency for PwA (Supplementary Materials, Supplementary Fig. S19). The study by Neveu and colleagues⁹ reported also higher internal noise but found similar sampling, which may be due to the methodological difference between their and the paradigm used in the current study. A recent study concerning ambiguous motion perception, reported significantly more horizontal compared to vertical motion percepts in PwA compared to a cohort without albinism but with nystagmus and to a normally-sighted control group.³³ Thus the greater proportions of horizontal percepts cannot be completely attributed to the presence of nystagmus. The authors reasoned that enhanced crossing of optic nerve fibers at the optic chiasm in PwA results in a mirror symmetrical overlay of opposing visual hemifields, consequently removing the interhemispheric information transfer and thus may be an explanation for the enhanced proportions of horizontal percepts. Their explanation would be plausible if the absence of interhemispheric information transfer removes a contributor to internal noise during information transfer and thereby improves the signal to noise ratio. This finding is in particular interesting given the poorer horizontal coherence thresholds and increased internal noise levels in PwA demonstrated here and elsewhere⁹ and may be due to the difference in viewing duration (i.e., short sequences of 150 ms vs. unlimited viewing time in the current study) and consequently the different impact of eye movements. Complex motions contain movement in all directions and may be affected differently by nystagmus than simple horizontal motion. Moreover, we did not find any significant differences in form coherence for any pattern type, which also suggests that processing at later stages of the visual pathways may be relatively spared in PwA. 

Poorer color discrimination has been reported in an early study of vision in a group of sixteen PwA.³⁴ In more recent studies, an increased number of color vision deficits and anomalous hue discrimination have been reported in PwA.^10,17,35 These findings may be partly explained by a lack of macula pigments reported in PwA.^36,37 None of the participants in the current study reported a color vision deficit; however, using cone-isolating contrasts for long, medium, and short wavelengths, we showed reduced chromatic sensitivity in PwA, but no wavelength-specific color defect, which may also be attributable to the lack of macula pigments in PwA^36,37; however, we did not measure macular pigmentation in our participants. We also measured color discrimination within Hue-Saturation-Value (HSV) color space and found greater color discrimination thresholds (angles in hue space) for PwA compared to controls. Given that the stimuli were large Gaussian blobs, these deficits are unlikely related simply to poor acuity. Therefore, while mid-level processing of form and motion may be relatively spared in PwA, mid-level processing of color is impaired. 

Sensitivity to bright light sources has been reported in albinism,^1,16 and we found increased subjective ratings of glare during the glare low luminance acuity task as well as significantly reduced acuities with glare compared to the control group. The glare source significantly reduced acuities for PwA for high-luminance acuity but did not affect the acuity for controls (Supplementary Materials, Supplementary Fig. S6). 

We deployed TvC³⁸ and equivalent noise³⁹ tasks, which indicate internal noise levels of observers via suprathreshold contrast discrimination (TvC) or averaging of motion directions (equivalent noise) and found increased noise for both tasks for PwA. In addition, the AIM method generated a min. angular error, which is the result of all external and intrinsic sources of variance. We show that AIM's minimum angular error was significantly elevated for acuity, glare acuity, and equivalent noise-motion (Supplementary Materials, Supplementary Figs. S16, S17, and S19). 

Psychophysical Data of the Control Group Compared to Previous Research

Visual Acuity

When compared with a population study regarding visual functions, mean binocular visual acuity was −0.15 logMAR ± 0.10,⁴⁰ which is comparable with our results (see Tables 1 and 2 for overview). Glare Low Luminance Acuity: One previous study measured acuity with a glare source⁴¹ and found that low luminance acuity in the control group was 2.00 MAR (i.e., 0.30 logMAR), which is lower than the data in our control group. This may be due to methodological differences because target contrast was varied whereas AIM Glare Acuity in the present study keeps the target contrast constant. 

Color Detection and Discrimination

Color perception results in a previous study using FInD Color in a young healthy control group were not significantly different to the control cohort data²⁷ for either cone-isolated color detection or color discrimination angle in HSV color space (Supplementary Fig. S28). Stereoacuity increased with spatial frequency consistent with a previous report.⁴² 

Temporal Contrast Sensitivity

Spatial and temporal contrast sensitivity in the current increased from 1Hz to 8Hz, consistent with a previous study using also 0.5 cycle/° stimulus.⁴³ 

Contrast Sensitivity Function

Our results indicate the expected inverted U-shaped tuning function for contrast sensitivity as a function of spatial frequency with an estimated high-spatial frequency limit (CSF-Acuity) of approximately 30 cycles/°, which has been reported previously.⁴³ 

Motion Coherence

A study by Merabet and colleagues²⁶ reported FInD mean circular motion coherence thresholds (27% ± 13%), which are comparable with the mean motion thresholds reported here (22% ±15). Horizontal motion coherence thresholds (mean 45% ± 18%) in a previous study using a standard alternative-forced choice task⁹ were higher than results in the current control group (mean 32% ± 27%). FInD tasks require a Yes/No report, which may be higher than those measured with forced choice procedures. Expanding motion thresholds in the current study (40% ± 25) were higher than those reported in a previous study (22% ± 2),⁴⁴ which again may be due to the differences in methods. 

Pattern Coherence

FInD Glass pattern coherence²⁶ mean thresholds (30% ± 13%) in a previous study were lower than the thresholds of the current cohort (52% ± 20%), which was may be due to the older cohort in the current study. Indeed, a study concerning form coherence across lifespan⁴⁵ showed that pattern coherence thresholds increase with age. Circular and concentric pattern coherence thresholds in previous studies range between around 15% to 45% and decrease with age. The current results (Table 3, expanding 40% ± 25% and circular 51% ± 20%) tend to be in broad agreement with those results. 

Table 3.

 

Summary of Adaptive Psychophysical Results Using AIM and FInD Methods and the Fixation Stability Method

 

Equivalent Noise Motion

Our equivalent noise paradigm presented translational motion signals in directions randomized across 360° and with the median estimate of internal noise being 28° ± 18° and sampling efficiency of 10 ± 11. A previous investigation by Neveu et al.⁹ measured equivalent noise for horizontal and vertical directions and reported individual noise and sampling efficiency results ranging from 3.7° to 12.9° and 1.1 to 2.5, respectively. For vertical motion directions, the findings ranged from 2.8° to 23° and 1.1 to 12.7. Next to difference in methods, the elevated internal noise may be due to the inclusion of oblique directions in the current method.⁴⁶ 

Cluster Analysis

We applied an unsupervised machine learning cluster method to determine which visual functions are predictive of albinism and are different from those of a control group and found fewer clusters for PwA than controls during an additional analysis of unique clusters for each group (Supplementary Materials, Supplementary Table S14). The dissimilarity measure was comparable between groups. According to our sample, CSF is the most predictive outcome measure for PwA as its outcomes cluster together (Fig. 3); however, given the heterogenicity of PwA subtypes, this analysis serves as proof-of-concept that an unsupervised clustering approach in combination with the psychophysical outcomes is capable of identifying key measures. 

Correlations

When testing for linear relationships between outcomes, more significant outcome correlations and thus fewer uncorrelated results were found for PwA compared to controls (Fig. 2). The majority of those PwA correlations were positive in direction whereas the control group's correlations almost evenly distributed in either direction. The emerging patterns of correlations within each group were also different between groups. Achromatic results (CSF, acuity, and TvC) tended to correlate more with one another for the PwA group than the control group. However, there was an interesting cluster of negative correlations for color discrimination angles and CSF outcomes for controls consistent with a general sensitivity relationship in contrast and color perception. A type one statistical error may be confounding this analysis because of the many independent statistical tests performed, but it is unlikely that the pattern of result is explained by it because a 5% of the variance chance (P = 0.05) would result in 41 false-positive results randomly distributed across results. However, we found 116 and 77 significant correlations for PwA and controls, respectively, clustered mainly for achromatic visual functions. PwA correlations were predominantly positive in direction whereas controls were more evenly distributed. We also applied a Bonferroni correction, which resulted in fewer correlations, but with the same skew toward achromatic visual functions. Interestingly, mid-level outcomes for form and motion tests tended not to correlate with achromatic or chromatic outcomes in either group. Eye dominance, (as determined by visual acuity), correlated with CSF results, but otherwise had no correlates. Together with results from the cluster analysis, the correlation results show that CSF outcomes appear to be a major predictive factor for early-stage vision but may not be predictive of mid-level vision. A recent population study investigated which visual functions predict general visual performance.³² The researchers measured 25 performance measures and showed that contrast sensitivity for medium and high spatial frequencies were among the most predictive functions. Similarly in another study in a general population testing seven visual functions, gain in contrast sensitivity was the most predictive outcome for visual performance ability.⁴⁷ 

Limitations and Future Direction

Although each test is independent from the other tests performed during this study, we cannot rule out type one errors that confounded one or more of our results. However, due to the overall one-sided skew of impaired visual performance in PwA, the potential effect of that confounder is unlikely to have significantly influenced the overall findings of the study. The current study was performed with 21 PwA. Future investigations may use this study's results to estimate sample size using statistical power analysis. The majority of the group effects were very significant with large associated effect sizes; however, the correlation and cluster analyses would benefit from larger data sets in terms of robustness and encourage replication with larger cohort sizes. Also, only FInD Color Detection and FInD Color Discrimination.²⁷ FInD Depth,²⁴ and AIM Visual Acuity²⁸ have been validated including test-retest and inter-test reliability analyses to assess the robustness of the new methods used to obtain the findings of this study. Future studies may incorporate these additional measures and evidence. Because of the age heterogeneity of the cohorts and the difference in retinal structure between groups, we decided to use 10-minute light adaptation before AIM Low Luminance Glare task. All tasks, including AIM Glare Low Luminance Acuity, are central vision tasks that primarily target rapid-adapting cones. However, light adaptation may be more variable in PwA because of structural (i.e., lack of fovea and differences in photoreceptor density) and oculomotor differences (i.e., strabismus and nystagmus). Future studies may therefore investigate the effect of prior light adaptation duration on low luminance acuity or other approaches to measure dark adaptation in rods and cones in greater detail in PwA. Additionally, albinism manifests in a number of different genotypes,^1,4 but genetic data was not available for our analyses. Thus we were not able to explore differences in visual phenotypes between genotypic groups. Future studies may investigate subtypes and their ability to perform a range of visual perceptual tasks including higher level tasks for facial or object-detection or discrimination abilities. Future direction may investigate those aspects of visual cognition in albinism. 

Conclusions

The current study investigated vision in people with albinism compared with age- and sex-matched controls. We deployed a range of chromatic, achromatic, binocular, mid-level psychophysical tests and an oculo-motor test and found that many visual functions in PwA showed a deficit with the exception of some mid-level vision functions. This suggests that despite the differences in visual pathway development, some midlevel visual functions may develop normally in albinism. Simple motion and temporal processing may be impaired due to the presence of nystagmus in our cohort of PwA. The number and patterns of correlations among outcomes were different in PwA, particularly for achromatic visual functions. Unsupervised hierarchal cluster analysis demonstrated different clusters of outcomes for the albinism group. Of the outcome measures included in the cluster analysis, CSF results were most predictive of other visual function deficits in albinism. The results shown here may have important implications for low vision rehabilitation of PwA. Current low vision rehabilitation interventions use magnification to address reduced visual acuity. Our current study indicates that chromatic, contrast, and temporal properties may also require aid. 

Acknowledgments

Supported by NIH grant EY029713. K.S. and O.W. were funded by a NIH- T-35 training grant, 5 T35 EY007149-24. J.S. and P.J.B. are founders of PerZeption Inc. FInD is patented, and AIM is provisionally patented and owned by Northeastern University, Boston, USA, which granted an exclusive license to PerZeption Inc. 

Disclosure: J. Skerswetat, (P); N.C. Ross, None; C. Idman-Rait, None; K. Sun, None; O. Wynn, None; P.J. Bex, (P)