Three-Dimensional Finite Element Models of Corneal Electroretinogram Potentials in Rat

Multi-electrode electroretinography (meERG) measures biopotential signals at multiple locations simultaneously on the cornea following a visual stimulus. The recorded signals have the same waveform components of conventional ERG signals. Ultimately, the goal of meERG is to produce maps of electrophysiological retinal function. To predict local retinal function using meERG, a quantitative model relating retinal sources to corneal potentials is needed. Here, a three-dimensional, anatomically-accurate model of the rat eye was created, optimized and validated to recently available meERG measurements for normally-sighted and rats with experimental retinal lesions. The model predicted the spatial distributions of corneal ERG potentials in normal rat eyes with reasonable agreement, but did not reflect a slight asymmetry in measured potentials. Models incorporating hypotheses to explain the slight asymmetry were constructed, and based on quantitative agreement and an assessment of likelihood, it was determined that a slightly misaligned recording lens was the best explanation. Assessment of confidence in each tissue conductivity value derived from the literature led to optimization of the ciliary body conductivity. The optimized model yielded simulation results that agreed with measured values to within the measurement error (i.e. model error was less than the variability between meERG measured animals). The model was validated by predicting corneal potential profiles in rats with experimental peripheral retina lesions. Measured and simulated corneal potentials decreased in locations closest to the retinal lesion. Error measures between simulated and predicted profiles for experimental lesion rats were no greater than error of normally-sighted rats. The ability to predict the location of a retinal lesion from the corneal potential profile (inverse problem) was evaluated by iteratively solving the forward problem for a standard set of lesion models, and minimizing error between measured and simulated corneal potentials. There was a trade-off between classification sensitivity and specificity for detection of presence of lesion, and techniques described elsewhere for detecting presence of lesion were more robust. However, this method accurately predicted lesion location in seven out of eight experimental lesioned rats, suggesting its utility for producing maps of retinal function. This work demonstrates proof-of-concept for detecting local dysfunction in the peripheral retina from corneal potential maps, an area of the retina that is difficult to evaluate using other techniques yet is highly relevant to prevalent, potentially blinding eye diseases.