Archetypal Analysis Reveals Quantifiable Patterns of Visual Field Loss in Optic Neuritis

Abstract

Purpose: Identifying and monitoring visual field (VF) defects due to optic neuritis (ON) relies on qualitative clinician interpretation. Archetypal analysis (AA), a form of unsupervised machine learning, is used to quantify VF defects in glaucoma. We hypothesized that AA can identify quantifiable, ON-specific patterns (as archetypes [ATs]) of VF loss that resemble known ON VF defects.

Methods: We applied AA to a dataset of 3892 VFs prospectively collected from 456 eyes in the Optic Neuritis Treatment Trial (ONTT), and decomposed each VF into component ATs (total weight = 100%). AA of 568 VFs from 61 control eyes was used to define a minimum meaningful (≤7%) AT weight and weight change. We correlated baseline ON AT weights with global VF indices, visual acuity, and contrast sensitivity. For eyes with a dominant AT (weight ≥50%), we compared the ONTT VF classification with the AT pattern.

Results: AA generated a set of 16 ATs containing patterns seen in the ONTT. These were distinct from control ATs. Baseline study eye VFs were decomposed into 2.9 ± 1.5 ATs. AT2, a global dysfunction pattern, had the highest mean weight at baseline (36%; 95% confidence interval, 33%-40%), and showed the strongest correlation with MD (r = -0.91; P < 0.001), visual acuity (r = 0.70; P < 0.001), and contrast sensitivity (r = -0.77; P < 0.001). Of 191 baseline VFs with a dominant AT, 81% matched the descriptive classifications.

Conclusions: AA identifies and quantifies archetypal, ON-specific patterns of VF loss.

Translational relevance: AA is a quantitative, objective method for demonstrating and monitoring change in regional VF deficits in ON.

Trial registration: ClinicalTrials.gov NCT00000146.

Conflict of interest statement

Disclosure: H. Doshi, None; E. Solli, None; T. Elze, None; L.R. Pasquale, None; M. Wall, None; M. Kupersmith, None

Figures

Figure 1.

Residual sum of squares (RSS) plot generated during AA for the purposes of selecting the number of ATs. RSS values were normalized based on sample size. The final number of ATs for our model was selected based on the point at which this curve begins to flatten, to avoid overfitting; resulting in the selection of a 16-AT model.

Figure 2.

Map outlining the 16 ON-specific ATs contained within our model. The varying shades of red within each AT denote TD values, and scale at the bottom denotes the TD values associated with each shade. Each AT is shown along with its corresponding average TD value (avgTD) and RW within the dataset. The ATs are numbered and displayed in order of RW. Note the color scale range from –35 dB to 10 dB.

Figure 3.

Example of VF decompositions from baseline to one month. The progressive changes in AT weighting and MD at each time point are displayed, along with the corresponding grayscale image from the Humphrey VF plot. AT weights not considered to be meaningful (

Figure 4.

Frequency of baseline study eye VFs containing listed number of ATs of meaningful weight (≥ 7%).

Figure 5.

Frequency of study eyes with AT weight ≥7% at baseline, for each AT.

Figure 6.

(A) Correlation between AT2 weight and MD (dB) at baseline (r = −0.91; P < 0.001), represented by the solid line. Dotted line represents the same correlation when eyes for which AT2 = 0% at baseline are eliminated (r = −0.94; P < 0.001). Note study eyes with an MD of <−20 dB that did not have high AT2 weight contained other ATs representative of severe VF loss. (B) Correlation between AT1 weight and MD (dB) at baseline (r = 0.63; P < 0.001). The trend line is skewed owing to the large number of study eyes with no AT1. The dotted line represents the same correlation when eyes for which AT1 = 0% at baseline are eliminated (r = 0.78; P < 0.001).

Figure 7.

(A) Correlation between AT2 weight and PSD (dB) at baseline (r = −0.53; P < 0.001). The dotted line represents the same correlation when eyes for which AT2 = 0% at baseline are eliminated (r = −0.79; P < 0.001). Note the wide range of abnormal PSD values, when diffuse severe VF loss AT2 is 0%. (B) Correlation between AT1 weight and PSD (dB) at baseline (r = −0.2; P < 0.001). The dotted line represents the same correlation when eyes for which AT1 = 0% at baseline are eliminated (r = −0.49; P < 0.001). Note the wide range of abnormal PSD values, when the normal VF AT1 is 0%.