Clinical applications of personalising the neural components of visual image quality metrics for individual eyes

Purpose: Eyecare is evolving increasingly personalised corrections and increasingly personalised evaluations of corrections on-eye. This report describes individualising optical and neural components of the VSX (visual Strehl) metric and evaluates personalisation using two clinical applications. (1) Better understanding visual experience: While VSX tracks visual performance in typical eyes, non-individualised metrics underestimated visual performance in highly aberrated eyes - could this be understood by personalising metrics? (2) Metric-optimised objective spherocylindrical refractions in typical and atypical eyes have used neural weighting functions of typical eyes - will personalisation affect the outcome in clinical 0.25D steps? Methods: Orientation-specific neural contrast sensitivity was measured in four typical myopic and astigmatic eyes and six eyes with keratoconus. Wavefront error was measured in all eyes while uncorrected and when the keratoconic eyes wore wavefront-guided scleral lenses. Total experiment duration was 24-28 h per subject. Two versions of VSX were calculated for each application: one weighted ocular optics with measured neural contrast sensitivity data from that eye, another weighted optics with a representative neural function of typical eyes. Wavefront-guided corrections were evaluated using the two metric values. Spherocylindrical corrections that optimised each metric were identified. Results: Metric values for keratoconic eyes improved by a mean factor of 1.99 (similar to 0.3 log units) when personalised. Applying this factor to a larger sample of eyes from a previous keratoconus study reconciled dissonances between the percentage of eyes reaching normative best-corrected metric levels and the percentages of eyes reaching normative levels of visual acuity and contrast sensitivity. Spherocylindrical corrections that optimised both versions of VSX were clinically equivalent (mean +/- SD Euclidean dioptric difference 0.13 +/- 0.18 D). Conclusions: Personalising visual image quality metrics is beneficial when actual metric values are used (evaluating ophthalmic corrections on-eye against norms) and when fine increments in visual quality are imparted (wavefront-guided corrections). However, partially individualised metrics appear adequate when metrics relatively rank spherocylindrical corrections in 0.25 D steps.