A feasibility study for objective evaluation of visual acuity based on pattern-reversal visual evoked potentials and...

Area of Science:

• Ophthalmology
• Neuroscience
• Machine Learning

Background:

• Objective evaluation of visual acuity (VA) is crucial for diagnosing and managing vision disorders.
• Traditional VA assessment methods can be subjective and time-consuming.

Purpose of the Study:

• To develop and validate machine learning models for objective VA assessment.
• To explore the relationship between VA and electrophysiological/optical parameters.

Main Methods:

• Utilized pattern-reversal visual evoked potentials (PRVEPs) and other visual parameters from 24 volunteers (48 eyes).
• Analyzed correlations between VA, P100 (peak time and amplitude), contrast sensitivity, refractive error, aberrations, and visual field.
• Developed four machine learning algorithms and an ensemble model for VA assessment.
• Validated model efficacy using receiver operating characteristic (ROC) curves, repeated sampling, and ten-fold cross-validation.

Main Results:

• Significant correlations were found between VA and P100 peak time/amplitude across various check sizes (P < 0.05).
• Optimal correlation between VA and P100 parameters occurred at a 1° check size.
• VA positively correlated with contrast sensitivity and spherical equivalent (P < 0.001) and negatively with coma aberrations (P < 0.05).
• Machine learning models achieved high accuracy, with mean Area Under the ROC Curve (AUC) > 0.95 and > 0.84 in cross-validation.

Conclusions:

• Machine learning models, leveraging PRVEP and visual parameters, provide an effective tool for objective VA evaluation.
• These models can assist clinicians in objectively assessing visual acuity.